1
1

Compare commits

...

463 Commits

Author SHA1 Message Date
Julian Eisel
7ec3023926 Rename internal VAMR files (mainly removes VAMR_ prefix)
Also rename VAMR_internal.h to utils.h and move OpenXR include out of
it.
2019-08-31 14:39:56 +02:00
Julian Eisel
37fc5bcd8c Use VAMR namespace for internal types 2019-08-31 14:20:57 +02:00
Julian Eisel
88ef625c0b Remove unnecessary WITH_OPENXR definition in BKE 2019-08-31 13:10:52 +02:00
Julian Eisel
34430ac5f5 Remove Ghost-XR code
Also fixes remaining call to it.
2019-08-31 13:10:37 +02:00
Julian Eisel
00842f9efd Call VAMR functions in window-manager, not GHOST_Xr
This gives linking errors currently because of conflicting symbols, but
that is to be expected and will be fixed in a followup.
2019-08-31 13:00:13 +02:00
Julian Eisel
04eada72fa Add missing C-API source file
Also add missing license header.
2019-08-31 12:59:18 +02:00
Julian Eisel
17dda0b3e6 Rename types and functions for VAMR 2019-08-30 17:55:44 +02:00
Julian Eisel
b0a333e755 Move Ghost-XR files into new VAMR-module and rename them 2019-08-30 16:44:02 +02:00
Julian Eisel
bc8186b5d1 Merge branch 'master' into soc-2019-openxr 2019-08-30 14:22:04 +02:00
4608220c9b Fix (unreported) 'Duplicate Particle System' operator not ensuring
unique name

This was leading to equally named particle systems, causing problems
later on.
Spotted while looking into T67958.

Reviewers: brecht

Differential Revision: https://developer.blender.org/D5632
2019-08-30 14:07:59 +02:00
b9ed30c25c Cycles: OpenCL Separate Compilation Debug Flag
OpenCL Parallel compilation only works inside Blender. When using cycles in a different setup (standaline or other software) it failed compiling kernels as they don't have the appropriate Python API and command line arguments.

This change introduces a `running_inside_blender` debug flag, that triggers out of process compilation of the kernels. Compilation still happens in subthread that enabled the preview kernels and compilation of the kernels during BVH building

Reviewed By: brecht

Differential Revision: https://developer.blender.org/D5439
2019-08-30 13:53:23 +02:00
040d075f17 GPencil: Fix unreported problem with Move to Layer popup
As the operator was designed to use only as popup, when it was used as menu option, the popup was displayed again.

Reviewers: mendio, pepeland

Differential Revision: https://developer.blender.org/D5630
2019-08-30 13:00:41 +02:00
44ace647a8 Fix interger number buttons increment/decrement arrows not working in
some cases

Since rB78b56fa7d973 defining a button with 'but->a1 = 0' wasnt
automatically setting its step value to 1 in 'ui_do_but_NUM' anymore.
Instead of handling this more forgiving in 'ui_do_but_NUM', went over
the remaining cases of buttons and and defined them with a step value of
1.

Fixes T69305

Reviewed By: brecht

Maniphest Tasks: T69305

Differential Revision: https://developer.blender.org/D5631
2019-08-30 12:45:38 +02:00
f033fa0fbc GPencil: Set active layer when click in Dopesheet keyframe area
Now, when clicking in the keys area, the layer is synchronized, not only in the left area with the names. This is one of the most requested feature by artists.

Also, removed some old comments.

Reviewers: brecht, angavrilov

Reviewed By: brecht

Subscribers: pepeland, mendio

Differential Revision: https://developer.blender.org/D5627
2019-08-30 11:58:02 +02:00
e33b4572af GPencil: Make visible active layer when hide others
To make more consistent the UI, when hide all other layers, the active layer is always set to visble.
2019-08-30 11:26:18 +02:00
cc49644551 Build: check necessary software is installed for make deps on macOS and Linux
To avoid errors deep into the build process. Fixes T69297.
2019-08-30 11:21:23 +02:00
a771fdb5dc Cleanup: clang-format, spelling 2019-08-30 10:55:31 +10:00
2fca31a5d6 Cleanup: no need to zero the area in beautify calculation
Also correct comment.
2019-08-30 10:50:06 +10:00
7c44182c4a GPencil: Minor Menu tweaks
- Added 'Hide Inactive Layers' operator and reordered 'Show/Hide' menu items
- Move 'New Layer' to bottom in 'Move to Layer' operator for consistency.
- Removed extra end points in some menu items (...)

Reviewers: antoniov, billreynish

Tags: #bf_blender, #grease_pencil

Differential Revision: https://developer.blender.org/D5626
2019-08-29 20:12:39 +02:00
33cd0535a0 Build: install_deps add lib64 paths for ldconfig
Building deps on some systems will install to opt/lib/library/lib64
directories rather than opt/lib/library/lib. This adds additional
lib64 paths for ldconfig to ensure the libraries are found at
runtime.
2019-08-29 11:15:43 -06:00
5c7852ef9c LibOverride: Create override operator: various fixes.
* `make_override_library_exec` was not properly cleaning `LIB_TAG_DOIT`
from all IDs in the Main DB.
* `BKE_override_library_create_from_tag` was doing dangerous things
(like iterating over a BMain listbase while adding items to it...).
* It would remap *all* local usages of overridden linked IDs to new
overriding local IDs, which was very inconvinient.

New handling of remapping now allows to only remap inside of the group
of IDs that is being overridden, in other words you can still have e.g.
other empties still instancing the same linked collection...
2019-08-29 18:11:12 +02:00
b9c400cee2 BKE_libblock_relink_ex: pass all remapping flags instead of a single boolean.
There is no reasons to limit access to remapping flags here, we may want
to use other options than only the ID_REMAP_SKIP_NEVER_NULL_USAGE one...
2019-08-29 18:11:12 +02:00
f4307d4bd9 LibOverride: Add entry to create a liboverride to object->relations menu. 2019-08-29 18:11:12 +02:00
497297850a Copy Scale: support copying arbitrary non-uniform scale as uniform.
Uniform scale is superior to non-uniform scale in that it works with
parenting without causing shear. Thus it is a valid desire in some
cases to turn arbitrary scale into guaranteed uniform scale.

Implementing this in the Copy Scale constraint allows one for instance
to 'inherit scale as uniform' by disabling Inherit Scale, and using
Copy Scale from parent with Offset and Make Uniform.

Reviewers: brecht

Differential Revision: https://developer.blender.org/D5614
2019-08-29 17:22:27 +03:00
c160853b04 UI: Correct Sequencer Text Alignment Tooltips
Pointed out in rBM5719
2019-08-29 10:06:09 -04:00
7745c6e35c Fix T56532: Boolean locks up Blender
Actual issue is with triangle beautify,
avoid precision error by scaling the epsilon
by the face area when it's over 1

The mesh in the report was very large (approx 2000 on each side),
causing precision issues with a fixed epsilon.
2019-08-29 23:11:18 +10:00
7f23c91478 LibOverride: Fix use-after-free error when freeing whole Main DB.
We do not want to touch to other ID pointers in that case, those might
have already been freed...
2019-08-29 14:49:54 +02:00
6b33bd1067 GPencil: For Stroke select don't display points
Now, when it's selected the stroke select mode, the points are not displayed and the stroke is fully selected automatically extending the selection.

Differential Revision: https://developer.blender.org/D5622
2019-08-29 10:44:24 +02:00
6b5e2f61b9 Cleanup: unused warning 2019-08-29 13:22:10 +10:00
c61a47d521 Cleanup: use custom data for object transform data 2019-08-29 13:22:10 +10:00
f4424727de Outliner: fix sync select on read file without loading UI
Tag for a complete sync when loading a file with load UI disabled.
2019-08-28 20:49:15 -06:00
5b96dc2381 Fix T69267: Missing Boids force field tooltip 2019-08-28 21:14:18 -04:00
749567e0b2 Move math and vector double routines into blenlib from delaunay code 2019-08-28 18:33:24 -06:00
07b1a5e05c Fix T69273: Rotate tool gizmo not showing incremental snapping tick marks 2019-08-28 20:36:56 -03:00
b1079ec3f2 Edit Mesh: AutoMerge and Split: Improve detection of the best split face 2019-08-28 20:07:43 -03:00
7aed8f69d1 Fix possible endless loop in AutoMerge and Split 2019-08-28 20:04:26 -03:00
a7ac8a22f8 GPencil: Animation menu updates
Animation menu updates for Edit Mode and Draw Mode (header and context menu):

- Rename menu items for a better understanding of what the operators really do.
- Change shortcuts (legacy from the old Grease pencil)

{F7708018}

Reviewers: antoniov, pepeland, billreynish

Tags: #bf_blender, #grease_pencil

Differential Revision: https://developer.blender.org/D5617
2019-08-28 22:33:44 +02:00
34ed58dcaf Fix T68971: Copy As New Driver from Material node creates a bad reference.
NodeTree structures of materials and some other data blocks are
effectively node group datablock objects that are contained inside
the parent block. Thus, direct references to them are only valid
while blender is running, and are lost on save.

Fix Copy As New Driver to create a reference that goes through
the owner datablock, by adding a new ID flag to mark private
pseudo-datablocks.

Also fix functions that return full paths to structures and
properties, e.g. used in python tooltips. Functions for paths
from ID to struct or property can't be changed because of
Animation Data related code.

Reviewers: mont29

Differential Revision: https://developer.blender.org/D5559
2019-08-28 21:52:54 +03:00
69a966aca0 Transform: option to transform parent objects without children
Supports parent/child chains with mixed selections.

Currently accessible from the pivot popover
(may be moved along with the other options here).
2019-08-29 01:53:07 +10:00
01bd22929f Fix T69221: inconsistent handling of filenames derived from image names.
We did not properly sanitize image data block name when using it to init
file name...
2019-08-28 17:51:40 +02:00
d5192d6fd6 Fix strict compiler warning in release mode 2019-08-28 16:00:40 +02:00
6f97855d06 CMake: Remove unsupported G++ strict flag
According to the documentation this flag is only supported
by C and Objective-C languages:

  https://gcc.gnu.org/onlinedocs/gcc/Warning-Options.html

Solves noisy output on every C++ file in the project when
using latest GCC-9.
2019-08-28 15:47:05 +02:00
f88022b96f Cleanup: make transform data in object mode flag public
This fits better with intended flag usage.
2019-08-28 13:44:21 +10:00
ed066f231d Cleanup: remove automerge BMesh operator
Move logic into EDBM_automerge since this is meant to run after
transform and isn't a generic editing operation.
2019-08-28 13:20:05 +10:00
f93b69c17a Cleanup: split automerge logic into own file
This isn't closely related to selection & this file was over 5k lines.
2019-08-28 13:19:55 +10:00
697fd86506 Cycles: Stitching of subdivided and displaced meshes
This patch stitches the vertices along patch edges so that cracks can
no longer form when applying subdivision or displacement a mesh.

Subpatches are now formed in a way that ensures vertex indices along
subpatch edges are equal for adjacent subpatches. A mapping of vertices
along patch edges is built to preform stitching. Overall performance is
roughly the same, some gains were made in splitting, but some was lost
in stitching.

This fixes:
- T49049 (cracks between patches from material and uv seams)
- T49048 (discontinuous normals with true displacement)

Reviewers: sergey, brecht

Differential Revision: https://developer.blender.org/D3692
2019-08-27 14:27:53 -04:00
7aef075ef6 Cleanup some outdated dead code and comments in Outliner. 2019-08-27 19:49:01 +02:00
9bda62c136 Fix T69176: Outliner->BlenderFile->Delete not working on some objects.
Object not referenced in any scene would not be deletable with previous
code... that was fine in 2.7x since it was not supposed to happen, but
now we can just use the nuclear `BKE_id_delete` for objects as well,
will take care of properly unlinking everything as needed.
2019-08-27 19:44:49 +02:00
490f966af5 Fix multires modifier using too much memory outside of sculpt mode
This reduce memory usage by about 25% in object mode for multires meshes.
2019-08-27 18:22:29 +02:00
c5b8d3b6d7 Cleanup: fix compiler warning 2019-08-27 18:22:29 +02:00
ec9e2dc3c0 Cleanup: spelling 2019-08-28 01:57:09 +10:00
179226977b Cleanup: clang-format 2019-08-28 01:57:09 +10:00
d5ed3de654 Fix Poly Build crash with empty meshes
Reviewed By: campbellbarton

Differential Revision: https://developer.blender.org/D5609
2019-08-27 17:44:36 +02:00
6d37654ce8 UI: Area Join Tolerance
Allow area joining even if the areas are not quite perfectly aligned.

Differential Revision: https://developer.blender.org/D5595

Reviewed by Brecht Van Lommel
2019-08-27 08:30:36 -07:00
a758dce0fa UI: Graph Editor Dot Scaling
Scale Graph Editor keyframe dots and handles based on Resolution Scale rather than Line Width.

Differential Revision: https://developer.blender.org/D5600

Reviewed by Brecht Van Lommel
2019-08-27 08:05:17 -07:00
ce47dc5e70 LibOverride: Fix broken Hook modifier case.
Hook modifier uses same kind of parent inverse matrix as regular object
parenting, so we need to bypass that matrix re-computation when setting
the overridding object here as well.

Also cleanup some minor glitches in object parents' override_apply func.
2019-08-27 16:31:03 +02:00
d8baafd693 Edit Mesh: Poly build tool improvements
This commit changes the functionality of the Poly build tool to make it more suitable for retopology tasks:
  - Click and drag from a boundary edge extrudes a new quad
  - Click and drag on vertices tweaks the position
  - Ctrl + click adds geometry. There is a geometry preview in the gizmo. It also can automatically convert triangles to quads.
  - Shift + click deletes mesh elements (faces or vertices)
  - Updated preselection code. Different mesh elements take priority depending on the selected action.

Reviewed By: campbellbarton

Differential Revision: https://developer.blender.org/D5573
2019-08-27 16:24:16 +02:00
2a2fd75ef9 Cleanup: Fix build error with MSVC
`BLI_cpu_support_sse41` needs BLI_System.h.

Reviewers: brecht
2019-08-27 08:13:23 -06:00
d667105ba9 Fix T69210: Crash on Making GPencil Materials zero user
The evaluated data must not copy the material if it's NULL.

Still the stroke exists in the file, so maybe need some thinking in how resolve this, but anyway this is a different topic not related to bug.
2019-08-27 15:58:58 +02:00
0067b3b09b GPencil: Draw Mode menu update
- Rename menu 'Stroke' to 'Draw': the menu includes not only strokes related things and the name follows the usage in other paint modes in Blender.
- Remove duplicate animation operators.
- Menu updates to include clean up and layers operators that don't need previous selection.

{F7705120}

Reviewers: antoniov, pepeland, billreynish

Tags: #bf_blender, #grease_pencil

Differential Revision: https://developer.blender.org/D5599
2019-08-27 15:47:43 +02:00
8e4f3b2bb0 Fix T68763: Smooth Brush not working in LookDev or Rendered Mode
Reviewed By: brecht

Differential Revision: https://developer.blender.org/D5585
2019-08-27 15:02:38 +02:00
87cafe92ce Sculpt/Paint: Brush world spacing option for sculpt mode
New brush option to calculate the distance to generate a new brush step using the cursor position over the mesh instead of the screen coordinates. This avoids artifacts when sculpting across curved surfaces.

Reviewed By: brecht

Differential Revision: https://developer.blender.org/D5432
2019-08-27 14:58:24 +02:00
Stefan Werner
cd890c4c25 Compositor: Reversed order of Normal and Albedo in Denoising Node. 2019-08-27 14:43:48 +02:00
Stefan Werner
d547f9d3d2 Fix T68944: Added check for SSE4.1 to denoising node.
Since OpenImageDenoise requires a CPU with SSE 4.1 or newer,
let the node act as passthrough on unsupported CPUs and display
a message in the node itself.
2019-08-27 14:03:49 +02:00
e39528b351 Fix marker name drawing without intended offset in some cases
Order of arguments matters here.
2019-08-27 13:04:43 +02:00
Yevgeny Makarov
ed92a1ff65 UI: fix jumping of play/pause buttons, add more space when showing subframes
Differential Revision: https://developer.blender.org/D5604
2019-08-27 12:52:42 +02:00
Robert Fornof
b19099346a UI: rename Energy to Power in context menu, for consistency
Differential Revision: https://developer.blender.org/D5310
2019-08-27 12:50:08 +02:00
Yevgeny Makarov
41b27a4def UI: add title to assign shortcut popup, tweak padding
Differential Revision: https://developer.blender.org/D5576
2019-08-27 12:50:08 +02:00
Alex Fuller
ee2bed7947 Fix undefined symbols for ccl::Node with some compilers and build options
Differential Revision: https://developer.blender.org/D5590
2019-08-27 12:50:08 +02:00
Yevgeny Makarov
9169d40175 UI: tweak warning messages style and layout in save/close dialog
Differential Revision: https://developer.blender.org/D5589
2019-08-27 12:32:42 +02:00
Yevgeny Makarov
1a40f089ea UI: improve layout of curve widget clipping options popup
Differential Revision: https://developer.blender.org/D5593
2019-08-27 12:32:35 +02:00
Yevgeny Makarov
811ac2965a UI: fix some menus and popups not taking into account UI scale for padding
Differential Revision: https://developer.blender.org/D5581
2019-08-27 12:27:23 +02:00
Yevgeny Makarov
9d5fb3491d UI: make layout in adjust last operator popup consistent with corner panel
Differential Revision: https://developer.blender.org/D5581
2019-08-27 12:25:52 +02:00
a7609598fa Cleanup: clang-format 2019-08-27 12:18:28 +02:00
Vinay Mulakkayala
af71be23c2 Build: run ldconfig for OpenImageDenoise in install_deps.sh
Differential Revision: https://developer.blender.org/D5580
2019-08-27 12:18:28 +02:00
1c21b79108 Fix T69165: wrong update function on Bone b-bone properties crashes.
Obvious fix suggested by @lichtwerk.
2019-08-27 12:52:06 +03:00
666be7d886 Fix T69166: built-in start-up file has unnecessary screen data 2019-08-27 19:20:38 +10:00
ebe7a407d7 Cleanup: use static set instead of list 2019-08-27 19:20:34 +10:00
90751d1835 Fix compiler error in Windows
This error was introduced in commit https://developer.blender.org/rB66700196074ad168f3322f2766846a0a07f7a00f
2019-08-27 11:14:36 +02:00
Stefan Werner
b91b9a8eca Compositor: Run only one instance of OIDN at a time.
Running multiple instances of OIDN simultaneously can use dozens
of GBs of memory. Since OIDN is multithreaded internally, we can run
only one instance at a time and should not lose much performance.
Fixing T69006
2019-08-27 11:06:48 +02:00
6670019607 Workbench: Specular Highlighting for MatCaps
With Blender 2.80 we introduced a more flexible matcap system. One
change we did was to multiply the matcap with the base color that was
shaded. As matcaps contains diffuse and specular lighting in a single
texture this lead to rendering artifacts. Artists were complaining that
everything looked to metalic.

We now support a separate `diffuse` and `specular` pass for matcaps.

`shaded_color = diffuse_light * base_color + specular_light`

For matcaps to support this feature they need to be multilayer openexr
files with 2 renderpasses (named `diffuse` and `specular`). In the future
we can change this to first pass/second pass in stead of this naming
convention.

Reviewed By: fclem, brecht

Differential Revision: https://developer.blender.org/D5335
2019-08-27 08:56:56 +02:00
bc51250c1f UV Editor: Remove Unimplemented Show Edges Option
Show Edges option in the UV editor does not do anything. Only RNA/DNA
code could be found, so it is save to be removed.

Differential Revision: https://developer.blender.org/D5446
2019-08-27 08:35:47 +02:00
21daeaa95a Cleanup: rename max_len -> alloc_len 2019-08-27 15:38:38 +10:00
13f54f65c4 Cleanup: struct member name wasn't descriptive 2019-08-27 15:07:51 +10:00
2b618b0f2d Cleanup: unused warning 2019-08-27 15:01:41 +10:00
e57b14feae Edit Mesh AutoMerge: Improve Split Edges & Faces option
Now all wire edges linked to the merged vertex are used for split faces.

Reviewers: campbellbarton

Differential Revision: https://developer.blender.org/D5603
2019-08-27 01:42:02 -03:00
60ba7d4539 Clean: style 2019-08-27 14:25:05 +10:00
2435850961 Fix crash transforming origins when no editable origins exist 2019-08-27 14:25:05 +10:00
963a991a20 Fix specific case where a vertex is ignored in snapping
Sometimes the index of the edge and the vertex may coincide.
2019-08-27 01:06:30 -03:00
3df04851cf Cycles/Windows: Fix over estimation of System RAM.
`system_physical_ram` reported 1024x more System RAM than what
was actually available.
2019-08-26 18:34:54 -06:00
d9be59e872 Fix T68687 Eevee: Modulo node behaves unexpectedly/inconsistently
This was a regression.
2019-08-27 01:50:01 +02:00
ce2368cf69 Fix T69122: Area Join Error on Invalid Cursor Position
Improved error handling of Join Area operator in cased it is passed cursor position that is not valid.

Differential Revision: https://developer.blender.org/D5598

Reviewed by Brecht Van Lommel
2019-08-26 13:39:25 -07:00
378a13483f Fix T69185: Cycles kernel OpenCL compile error after recent changes 2019-08-26 22:06:02 +02:00
c69d75e880 GPencil: Fix typo error in previous commit 2019-08-26 21:04:04 +02:00
e148873f3f GPencil: Temporary fix to keep running Cycles + GPencil object in background.
This is a temporary solution to keep the render running.
2019-08-26 21:00:20 +02:00
8c7cbad542 Gpencil: Splitting and rearranging Context Menu in Edit Mode
Following the changes in the Edit Mode menus on the header, this patch split the current context menus into Point and Stroke and also rearrange the operators.

{F7704949}

Reviewers: antoniov, pepeland, billreynish

Reviewed By: billreynish

Tags: #bf_blender, #grease_pencil

Differential Revision: https://developer.blender.org/D5596
2019-08-26 20:53:16 +02:00
7f840426fd GPU: add mechanism for splitting up big gpu_shader_material.glsl file
Compiling this big file for every Eevee material is bad for performance, and
now that we are adding more nodes it gets worse. This patch adds a simple
mechanism to split up that file, and use only the parts used by shader nodes.

When a function is used by GPU_link, we detect which GLSL file it came from
and use it in GLSL code generation automatically. Dependencies between GLSL
files are manually specified, and function names must be unique across all
GLSL files.

Most of the actual splitting up will be done in later commits.

Differential Revision: https://developer.blender.org/D5569
2019-08-26 19:18:34 +02:00
6b189d2bcf Edit Mesh: New option "Split Edges & Faces" to "AutoMerge"
Ref T66423

Differential revision: https://developer.blender.org/D5562
2019-08-26 14:15:25 -03:00
7273dbd47b Fix T69051 Vertex Paint: Selection not show in vertex select mode. 2019-08-26 18:43:53 +02:00
1a6491639a Fix T68954 UVEdit: Auto Smooth option conflicts with uvs display 2019-08-26 18:43:53 +02:00
d5f3f4410d Missed adding file in recent commit 2019-08-27 01:53:00 +10:00
6d87ad08a4 Keymap: Use Ctrl-Period to toggle origin editing
Add a key binding since this option is typically for quick adjustments.
2019-08-27 01:49:21 +10:00
7fee153bf5 3D View: show selected object axis when affect origins is enabled
This replaces temporarily enabling draw-axis.
2019-08-27 01:46:03 +10:00
487cd7237c DRW: check exact draw flags before getting wire-frame color
Only some options need this.
2019-08-27 01:46:03 +10:00
c32377da98 Cycles: support move semantics for device_memory
Ref D5363
2019-08-26 17:42:52 +02:00
f6da680946 Cycles: refactor of BVH building to prepare for Optix
Ref D5363
2019-08-26 17:39:57 +02:00
2dd1e6b376 GPencil: Add missing update of modifiers
This was missing in previous commit for T66294
2019-08-26 17:31:55 +02:00
8f8e6a7c7b Fix Cycles session being (re)created twice
In most cases this only makes a small startup time difference, but there is
no reason to do this.

Ref D5363
2019-08-26 17:12:45 +02:00
2b55f621fe Cleanup: fix compiler warnings 2019-08-26 16:08:17 +02:00
1ff939f3f7 Cleanup: avoid unnecessary primitive type lookup and variable initialization
Ref D5363
2019-08-26 16:07:01 +02:00
6055db084d Cycles: optimize Sobol loop
Ref D5363
2019-08-26 16:07:01 +02:00
030a023eb2 Cleanup: add zero bit counting functions 2019-08-26 16:07:01 +02:00
d133934ea4 Cycles: code to optionally zero initialize some structs in the kernel
This will be used by Optix to help the compiler figure out scoping. It is not
used by other devices currently, but worth testing if it helps there too.

Ref D5363
2019-08-26 16:07:01 +02:00
39439a3afe Fix T69156: Blender crash when baking rigid body world.
Issue was exposed by recent own rB03bf84db86b commit, but was actually
present in RNA API for PointCaches since (probably) ages: whole accessor
code here was assuming that owner ID was an Object, when it is actually
a scene for RigidBody simulations...

Had also to make `BKE_ptcache_id_find()` and friends a bit more
flexible, now they also accept a NULL object pointer parameter...
2019-08-26 16:03:58 +02:00
51d9f56f87 GPencil: Invert Paste operator and make Paste to Active default
Before there were two options: Paste to original layer called "Paste" and Paste to active   layer called "Paste & Merge"

Now, by default the paste is in active layer and the "Paste & Merge" has been renamed "Paste".

For old "Paste", now is called "Paste by Layer" and it's not the default value anymore.

Note: Minor edits to add icons not present in Differential revision.

Differential Revision: https://developer.blender.org/D5591
2019-08-26 15:49:16 +02:00
742119bcb5 Fix T68826 Eevee: Multi-Mat not working if switching from Solid shading
The correct fix would be to avoid all those hacks but this is needed if
we want to be able to parallelize object vbo extractions.

This is the fixed version of the hack. The issue was that the ibo ranges
were saved before the batch were reset and the IBO was discarded, leading
to a read after free error. All previous reported crash were tested and
are now not reproductible.
2019-08-26 15:43:28 +02:00
b9f61eb874 Cycles: Fix Architecture logging on x64.
x64 builds with WITH_CYCLES_OPTIMIZED_KERNEL_SSE2 not defined
since SSE2 is the lower bar for x64 cpus. Turning the architecture
logging related if into the last if in the architecture detection
chain, which will never execute unless you turn off all kernels
in de debug flags.

Reviewers: brecht

Differential Revision: https://developer.blender.org/D5579
2019-08-26 07:22:44 -06:00
16c1b10ed2 Fix T69146: Segment Fault using Undo for file with several scenes and script.
The specifc bug here came fro; some IDProperties ID pointer storing
references to workspaces.

But that was actually a main loophole in that 'unndoing data while
keeping same UI' process, as we never know who might store a pointer to
one of those datablocks that we want to keep the 'old' version off.

It might actually be ever more needed when we start undoing (changing)
only the IDs actually modified in an undo step...

Notes:
* While not ideal, I think we can afford an extra looping over the whole
Main DB here... Remapping process in itself is fairly cheap, thanks to
the hashes.
* This commit is considered rather risky (especially thanks to 'private'
IDs), think it should work fine for now, unless some IDPointers start
storing references to private IDs...

Once D5559 is in, we shall do another pass here, probably also forbids
assigning private IDs to IDProperties, etc.
2019-08-26 14:51:30 +02:00
81e2fdcd1d Cleanup: simplify perlin noise function definition
Ref D5363
2019-08-26 14:13:00 +02:00
7b150e8767 Cleanup: use saturate3() for float3 clamping
Ref D5363
2019-08-26 14:11:08 +02:00
b5daed9680 Cleanup: fix compiler warning 2019-08-26 13:47:00 +02:00
2778754925 Cleanp: fix compiler warnings 2019-08-26 11:27:04 +02:00
fd52dc58dd Cycles: GPU code generation optimizations for direct lighting
Use a single loop to iterate over all lights, reducing divergence and amount
of code to generate. Moving ray intersection calls out of conditionals will
also help the Optix compiler.

Ref D5363
2019-08-26 10:26:53 +02:00
f491c23f1e Cycles: inline more functions on the GPU
This makes little difference for CUDA and OpenCL, but will be helpful
for Optix.
2019-08-26 10:26:53 +02:00
2b999c6a68 Cycles: change svm node decoding for more efficient code generation on GPU
These functions no longer accept NULL. They were renamed for clarity and to
avoid hidden merge issues.

Ref D5363
2019-08-26 10:26:53 +02:00
db257e679a Cycles: remove workaround to pass ray by value
CUDA is working correct without it now, and it's more efficient not to do this.

Ref D5363
2019-08-26 10:26:53 +02:00
edbb755dfe Cycles: tweaks for better GPU code generation
Uninitialized variables are harder to handle for the compiler.

Ref D5363
2019-08-26 10:26:53 +02:00
b05e7ea719 Cycles: fixes for building kernel without certain features
Ref D5363
2019-08-26 10:10:35 +02:00
7a15b938bd Cleanup: fix use of wrong constant
Ref D5363
2019-08-26 10:10:35 +02:00
7b0cd21750 Cleanup: fix compiler warning 2019-08-26 10:10:35 +02:00
dc216c89e0 Fix render test Python error on crashes 2019-08-26 10:10:35 +02:00
b2608af702 Gizmo: match the DOF gizmo size with the non-gizmo indicator 2019-08-26 17:12:42 +10:00
c70e6a2e21 Fix T69153: Camera DOF gizmo fails
Regression from bc3139d792
2019-08-26 17:04:47 +10:00
abd1a2d07d Gizmo: report error when property isn't found
Also add assert so this doesn't go unnoticed, see: T69153
2019-08-26 16:58:26 +10:00
db851c78b4 Transform: support snapping origins onto the objects geometry
Normally it wouldn't make sense for the object to snap onto it's self,
when moving origins this is a common use-case.
2019-08-26 14:40:00 +10:00
7b61fe1638 Cleanup: use BLI_assert 2019-08-26 14:32:35 +10:00
8d03e04493 Fix assert restricting allowed operation 2019-08-26 01:18:40 -03:00
22822b0572 Cleanup: unused arg 2019-08-26 12:20:02 +10:00
af8e8be7b1 Cleanup: Move RNA Manual References to Modules
This file used to be stored in the Add-ons repository
to allow easy commit access for updating.
Since then, we have created an automated process,
the script can now easily be updated by anyone.

This also makes more logical sense to store the file here
as it is not an add-on.
2019-08-25 15:05:25 -04:00
fd9614de0c UI: Dim Disabled Menu Items on Hover
Dims the text and background of disabled menu items while mouse is hovering.

Differential Revision: https://developer.blender.org/D5575

Reviewed by William Reynish
2019-08-25 12:01:29 -07:00
b876fe3f3d GPencil: Show edit points in Sculpt only if mask is enabled
If the masks are disabled, the edit points must not visible.
2019-08-25 20:56:09 +02:00
de0fc96dfe cleanup: collada: removed unused public class variable 2019-08-25 19:38:12 +02:00
2f1604cef7 fix: Removed Debug statement from Collada CMakeLists. cmake now silently determines which collada library version it uses 2019-08-25 19:38:12 +02:00
f2c3dc94b1 Fix T69127: Node sockets location bug after a keyframe is inserted 2019-08-25 16:37:06 +01:00
63533a0c7d Cleanup: Clarify alloc comment text 2019-08-25 17:13:13 +02:00
a67f218b54 GPencil: Use evaluated data for Sculpt Brushes
As part of T66294 is needed to use the evaluated data for Sculpt brushes to make possible to Sculpt a transformed stroke.

Without this commit, it was impossible sculpt the stroke if the modifier moves away the stroke point from original position. Also, some calculation is done in order to determine the rotation to transform the brush effect too.
2019-08-25 16:04:23 +02:00
6917ac06a2 Fix error accessing a named UV layer when baking 2019-08-25 16:45:47 +10:00
755c11c3db Curve: remove tessface calculation when calculating modifiers
- All parts of the code that need tessface should calculate it on demand.
- The check for tessloopnormal mask isn't correct
  (since this is loop data, not tessface data).
2019-08-25 16:45:47 +10:00
6eadd40597 Cleanup: redundant struct declarations 2019-08-25 16:45:47 +10:00
f1ddc6ed0a Cleanup: clang-format 2019-08-25 16:45:47 +10:00
8df6673165 Cleanup: skip adding tessface loop & color layers
Also correct check in unused poll function
2019-08-25 16:45:47 +10:00
7585d47b36 Cleanup: remove UV name syncing function
MTFace's on the mesh are now only used for conversion.
There is no need to keep both UV layers in sync at once.
2019-08-25 16:45:47 +10:00
5572986aad Cleanup: rename mesh looptri/tessface functions
Use consistent terminology.
2019-08-25 16:45:47 +10:00
2626a6fca2 Cleanup: remove tessface check when entering edit-mode 2019-08-25 15:49:32 +10:00
bada34c766 Cleanup: remove BKE_mesh_calc_normals_tessface
This was used for versioning, now normals are calculated after
initializing MPoly data.
2019-08-25 15:49:26 +10:00
4ab50eb7fa Cleanup: remove unused ED_mesh_*_tessface functions 2019-08-25 15:06:12 +10:00
f8d7993eec Cleanup: remove unused DM_to_mesh function 2019-08-25 14:09:09 +10:00
19f27cf3e9 RNA: remove Mesh.update() calc_loop_triangles argument
This calculated tessfaces, not loop-triangles.

Remove this since they aren't accessible from RNA anymore.
2019-08-25 12:39:15 +10:00
17e1c97e43 Cleanup: remove USE_TESSFACE_DEFAULT define 2019-08-25 12:21:09 +10:00
d42a23b31a Cleanup: remove tessface loop from set-smooth function 2019-08-25 12:08:18 +10:00
Julian Eisel
e0bb3f9286 Merge branch 'master' into soc-2019-openxr 2019-08-24 23:21:33 +02:00
Julian Eisel
d0b4ec00f6 Don't drwa relationship lines in the VR view 2019-08-24 22:06:23 +02:00
6bb2912704 GPencil: Fix unreported unable to deselect when masking is OFF
When the mask is disabled, all select operators must be disabled, but the deselect all operator must work or it's impossible to deselect.
2019-08-24 17:59:32 +02:00
Julian Eisel
7b6514c222 Merge branch 'master' into soc-2019-openxr 2019-08-24 14:45:41 +02:00
a53ed1e049 Fix crash when snapping ruler 2019-08-24 09:21:13 -03:00
3437fcf2f1 BKE_bvhutils: implement hiden check
Accidentally unused in rB7c3bbe93aaa2
2019-08-24 09:20:55 -03:00
091e7979d3 GPencil: Use evaluated data in selection
Now the selection is using the position after evaluating the modifiers and makes possible to select a stroke point that has been moved from the original location.

Related to T66294
2019-08-24 13:46:00 +02:00
b1ed72f152 Cleanup: use eval as suffix
Follow conversion already used in most places.
2019-08-24 13:27:08 +02:00
8b806b8e64 Fix object origin transform with vertex snapping 2019-08-24 21:23:04 +10:00
d9d1c68809 UI: Add Meta strip icon to Sequencer sidebar
Patch by Peter Fog (tintwotin).

Differential Revision: https://developer.blender.org/D5567

Reviewers: Brecht Van Lommel (brecht), William Reynish (billreynish)
2019-08-24 12:25:51 +02:00
b7f786c0b1 GPencil: Fix typo error 2019-08-24 09:02:42 +02:00
ae0b855a32 Fix T68807: smoothing group issue
Showed in OBJ export.
Caused by comparison mistake in rB2e91fc39ac7.

Reviewers: mont29

Maniphest Task: T68807

Differential Revision: https://developer.blender.org/D5561
2019-08-24 08:50:58 +02:00
c9923baf84 Fix BKE_bvhtree_from_editmesh_get not setting cached value
The stored value was always false because of shadowing.

Also disable unused code.
2019-08-24 12:16:30 +10:00
fd43854fa3 Cleanup: use doxy sections for editfont 2019-08-24 12:06:58 +10:00
dc434bc273 Cleanup: correct use of term 'split' 2019-08-24 12:06:58 +10:00
67d9647ae4 Cleanup: use eval as a suffix
Follow conversion already used in most places.
2019-08-24 12:06:58 +10:00
Julian Eisel
c506acf2fc Add Oculus to the list of known runtimes
With this, all currently available runtimes are in the runtime map. So
removing the TODO comment.
2019-08-24 03:35:32 +02:00
Julian Eisel
31b8350b01 Cleanup: Refactor (hacky) swapchain image submission
Code now assumes that the view-draw callback left the OpenGL state for
ongoing use by Ghost-XR. So the state doesn't have to be set by the
swapchain image submission and thus it we don't need to pass the Ghost
OpenGL context to it in a hacky way.
Also, rename drawViewEnd (to submitToSwapchain) and remove
drawViewBegin.
2019-08-24 03:32:11 +02:00
Julian Eisel
d7a216704b Fix warning in release builds 2019-08-24 03:17:14 +02:00
Julian Eisel
4b67477732 Own FBO for VR viewport to get previous changes to work 2019-08-24 03:15:25 +02:00
a68e8ac993 BKE bvhutils: create and use new BKE_bvhtree_from_editmesh_get
With this function it is easier to create and have control over editmeshes `BHVtree`s.
2019-08-23 19:51:18 -03:00
4c353205ea UI: Various tooltip corrections and fixes
Patch by Yevgeny Makarov (jenkm)

Differential Revision: D5514
2019-08-24 00:45:21 +02:00
bf242362f6 UI: Sequencer: Move header spacer in to Preview modes
Patch by Peter Fog (tintwotin)

Differential Revision: D5511
2019-08-24 00:24:43 +02:00
beb025b3ab UI: Clip Editor: Move fractional zoom values into sub-menu
Patch by Peter Fog (tintwotin)

Differential Revision: D5525
2019-08-24 00:22:38 +02:00
Julian Eisel
4a039158e5 Avoid OpenGL context deactivation, just to reactivate it immediately 2019-08-24 00:13:36 +02:00
Julian Eisel
61014e1dd9 Remove hack to resize the usable default framebuffer region 2019-08-23 23:46:09 +02:00
Julian Eisel
3441314e40 Remove blitting from default framebuffer 2019-08-23 23:36:04 +02:00
acdb14d264 Transform: option to transform origins in object mode
Currently supports mesh, armature, lattice, curve & metaballs.

Access from pivot popover.
2019-08-24 07:34:43 +10:00
71c43e9225 Cleanup: unused args 2019-08-24 07:34:43 +10:00
e270b6f34c Cleanup: Fix PEP8 issues 2019-08-23 23:24:07 +02:00
ee4ec69b28 Fix T66924 : Move GPencil Modifiers evaluation to Depsgraph
Before, the evaluation of modifers were done in draw manager. The reason of the old design was grease pencil was designed before depsgraph was in place.

This commit moves this logic to depsgraph to follow general design and reduce Draw Manager complexity. Also, this is required in order to use modifiers in Edit modes.

Really, there is nothing really new in the creation of derived data, only the logic has been moved to depsgraph, but the main logic is the same. In order to get a reference to the original stroke and points, a pointer is added to Runtime data as part of the evaluated data. These pointers allow to know and use the original data.

As the modifiers now are evaluated in Depsgraph, the evaluated stroke is usable in Edit modes, so now it's possible to work with the evaluated version instead to use a "ghost" of the final image over the original geometry as work today.

Reviewed By: brecht

Differential Revision: https://developer.blender.org/D5470
2019-08-23 23:10:48 +02:00
Julian Eisel
f175dcc35f Add Monado to the list of known runtimes 2019-08-23 23:08:21 +02:00
d795dd1fa7 GPencil: Object data UI updates
Some tweaks for the Grease Pencil object data UI:
- Update main layers controls for consistency (Opacity, Blend mode).
- Move "Show only On Keyframed" checkbox to Display section.
- Change "Duplicate Layer" icon in Layer menu.

{F7695928}

Reviewers: antoniov, pepeland, billreynish

Reviewed By: billreynish

Tags: #bf_blender, #grease_pencil

Differential Revision: https://developer.blender.org/D5571
2019-08-23 23:01:37 +02:00
f16a28c4af Revert "Fix T68826 Eevee: Multi-Mat not working if switching from Solid shading"
This reverts commit d357e7b065. This caused crashes in UV editor drawing and
updates in lookdev mode.

Fixes T69087, T69083, T69088, T69096.
2019-08-23 20:32:51 +02:00
777ca2dcdb Snapping System: Improve drawing indicating Perpendicular snap
Since pependicular snap depends on `snapTarget` it is important to indicate where this target is so as not to confuse users.
So draw a pivot where the target is and a dotted line toward the perpendicular snap point.

Reviewers: campbellbarton, brecht, billreynish

Differential Revision: https://developer.blender.org/D5557
2019-08-23 15:20:37 -03:00
9574ac1c8d LibOverride: Disable the re-parenting of overriding objects to the instancing empty.
Does not seems to be that useful... And it's breaking the objects also
'parented' to an armature through a modifier...
2019-08-23 18:58:21 +02:00
Yevgeny Makarov
e33fc3c37b UI: layout padding changes to autorun dialog, to match save dialog
Differential Revision: https://developer.blender.org/D5568
2019-08-23 18:42:26 +02:00
c3e4a26816 UI: layout and text changes for unsaved changes dialog
Patch by Harley Acheson, with additional changes by Yevgeny Makarov.

Differential Revision: https://developer.blender.org/D5133
2019-08-23 18:36:30 +02:00
Julian Eisel
756b676076 Merge branch 'master' into soc-2019-openxr 2019-08-23 17:51:28 +02:00
Julian Eisel
7462fca737 Avoid redundant framebuffer resize and offscreen texture drawing 2019-08-23 17:46:43 +02:00
77f01748e7 GPencil: Fix unreported set origin from Python
The set origin was not working from python because the operator was checking if the stroke was valid in the console area.

As the stroke only can be valid for GP obects, this check is not needed.
2019-08-23 17:20:12 +02:00
Julian Eisel
7c9e18c193 Cleanup: Add comments 2019-08-23 17:00:30 +02:00
18f4182e98 LibOverride: Fixed bug in removing override op operator.
In case the property is a RNA pointer, `RNA_path_resolve()` will try to
resolve it and return that pointer, instead of returning expected
container... That is a bad inconsistency in the rNA path API, but no
proper way to solve it for now...
2019-08-23 15:55:00 +02:00
6d64da1e67 Fix potential issues when loading files with missing libraries
This is a continuation of rB39f005eae8eed8b939579aff8c9a05a4f50e5e38

Now all the fields where we check for object type in RNA (like
rna_Curve_object_poll) will have a safe guard for when this isn't the
case.  For example when loading files that has missing object libraries
and all missing objects are replaced with empties (placeholders).

Reviewed By: Brecht

Differential Revision: http://developer.blender.org/D5425
2019-08-23 15:54:11 +02:00
8eb2296872 UI: use consistent "Color Burn" name for blend mode
Previously it was named just "Burn" in some places.

Differential Revision: https://developer.blender.org/D5186
2019-08-23 13:14:15 +02:00
2f8ea1c11e Fix T67140: transforming bone does not update bone tab in properties editor 2019-08-23 13:04:54 +02:00
3d1cb78336 Fix T69075: OSL build with install_deps.sh failing at runtime
Now uses the same cmake flags as make deps.
2019-08-23 12:04:54 +02:00
162c320a75 LibOverride: Add some basic non-nodal material properties to override realm. 2019-08-23 10:27:22 +02:00
a1aa4a2597 RNA: Cleanup PointerRNA struct
The old layout of `PointerRNA` was confusing for historic reasons:
```
typedef struct PointerRNA {
  struct {
    void *data;
  } id;

  struct StructRNA *type;
  void *data;
} PointerRNA;
```

This patch updates it to:
```
typedef struct PointerRNA {
  struct ID *owner_id;
  struct StructRNA *type;
  void *data;
} PointerRNA;
```

Throughout the code base `id.data` was replaced with `owner_id`.
Furthermore, many explicit pointer type casts were added which
were implicit before. Some type casts to `ID *` were removed.

Reviewers: brecht, campbellbarton

Differential Revision: https://developer.blender.org/D5558
2019-08-23 09:52:12 +02:00
232049dd94 Armature: use BKE_armature_transform when applying transformation
Keep ED_armature_transform for RNA Armature.transform
since it operates on edit-bones in edit-mode.

Rename ED_armature_transform_bones to ED_armature_edit_transform
since it wasn't obviously an edit-mode function.
2019-08-23 12:03:11 +10:00
30582c59cc Armature: add BKE_armature_transform
ED_armature_transform uses edit-mode conversion which re-creates bones.
Needed for efficiently transforming object-data in object-mode.
2019-08-23 11:53:28 +10:00
adfb9ec87c Cleanup: const args, naming, doxy groups, clang-format 2019-08-23 09:09:38 +10:00
Julian Eisel
6b69b5349b Address and remove some TODOs marked in code 2019-08-22 23:24:05 +02:00
Julian Eisel
daba8140c2 Cleanup: Unused functions, add comments, sync to master 2019-08-22 23:02:01 +02:00
1de7717ed7 Outliner: new icons for sequences and contraints
Adds a new icon for the action constraint so the icon draws with the
constraints color. Also adds two new icons for sequencer meta strips
and duplicate strips for use in the outliner sequence display mode.

The meta strip icon could be used in the sequencer sidebar.
2019-08-22 14:40:33 -06:00
22ebc57987 UI: Changes to Area Options Menu
Adds more options to the context menu that pops up on area edges. Both Split types, Join, and Swap.

Differential Revision: https://developer.blender.org/D5459

Reviewed by Brecht Van Lommel
2019-08-22 13:19:11 -07:00
7c3bbe93aa Cleanup/Refactor: Simplify/deduplicate bvhutils code
This is a step that allow using `bvh_cache` for `EditMeshe`s.
2019-08-22 14:07:40 -03:00
d09b1ff1a6 Cleanup: undeclared variable warnings
Forward declare variables, or make them static.
2019-08-23 02:29:45 +10:00
4457c92fa3 GPencil: Fix missing variable due typo error 2019-08-22 18:28:19 +02:00
5888a2283e Cmake/MSVC: Enable Edit and Continue for debug builds.
This change switches the debug symbol format from /Zi to /ZI for
debug builds of blender, allowing Edit and Continue to work.

This allows limited [1] code changes in the debugger without
having to stop the process and recompile a new binary leading
to improved developer productivity.

All MSVC versions we support support this flag, Clang on
windows does not mind the /ZI flag, but doesn't currently
emit the required information to have this feature work.

[1] https://docs.microsoft.com/en-us/visualstudio/debugger/supported-code-changes-cpp
2019-08-22 10:20:40 -06:00
ec07340004 UI: Remove "Show Frame Indicator" option
This option was doing nothing in Blender 2.80.
I don't really see a reason for keeping it around.

Reviewers: campbellbarton

Differential Revision: https://developer.blender.org/D5552
2019-08-22 17:55:21 +02:00
e6f3d8b3e1 Revert "Fix T68971: Copy As New Driver from Material node creates a bad reference."
This reverts commits 54fd8176d7, 4c5becb6b1 and 8f578150e.

Those kind of commits must be reviewed and approved by project owners.

That one:
* Broke Collada building by not properly updating all calls to modified
function.
* Broke *whole* ID management by not properly updating library_query.c.

And in general, I am strongly against backward ID pointers, those are
*always* a serious PITA for ID management. Sometimes  they cannot be
avoided, but in general other ways to get that kind of info should be
investigated first.
2019-08-22 16:07:36 +02:00
eae9b86297 Fix crash using 'Mesh > Normals > Point to Target' from the header menu
This was working from Alt+N menu but was passing wrong ARegion
(alongside wrong mouse coords) to ED_view3d_win_to_3d_int when called
from the header menu.

Operator context INVOKE_REGION_WIN takes care of this.

This also fixes wrong behavior of 'Mesh > Normals > Rotate' when called
from the header menu.

part of T69019

Reviewers: billreynish, mont29

Maniphest Tasks: T69019

Differential Revision: https://developer.blender.org/D5555
2019-08-22 16:01:47 +02:00
54fd8176d7 NodeTree: also assign the owner pointer when copying. 2019-08-22 16:55:02 +03:00
33a287e5c0 Snapping System: Return element type in 'ED_transform_snap_object_project_view3d_ex' 2019-08-22 10:53:21 -03:00
8965a81271 Cleanup: Split some code out of rna_access.c
That file was getting out of control, now comparison/override RNA code is
in `rna_access_compare_override.c`. 1K lines of code for now, but that
area is likely to grow more in the future...

Note that we can probably split more out of `rna_access.c`, but for now
that will do.
2019-08-22 15:33:44 +02:00
Dalai Felinto
662d94e020 Multi-View user interface minor tweak
There is no reason to not expand this enum as we did for 2.79.
It is self explanatory enough and it has only two options.
2019-08-22 10:16:43 -03:00
4c5becb6b1 Fix missing argument.
Missed because of broken dependency tracking in msvc build process.
2019-08-22 16:14:48 +03:00
8f578150ea Fix T68971: Copy As New Driver from Material node creates a bad reference.
NodeTree structures of materials and some other data blocks are
effectively node group data block objects that are contained inside
the parent block. Thus, direct references to them are only valid
while blender is running, and are lost on save.

Fix Copy As New Driver to create a reference that goes through
the owner data block, by adding a new runtime field to bNodeTree.
2019-08-22 15:46:37 +03:00
443586f34d LibOverride: Enforce full override operations check on file save.
We try to avoid diffing too much things during edition, but when saving
a file we need to check all possible overridable IDs to ensure we have
all needed override operations...

Was pretty sure that was already in code, but apparently not (or it got
lost at some point...).
2019-08-22 14:34:09 +02:00
e3442c8864 LibOverride: Cleanup: tget rid of G.main usage when we can get it from context. 2019-08-22 14:33:56 +02:00
df5237ec5e LibOverride: Fix broken logic in code checking and adding new override ops.
When we wanted to force all overidable IDs to be checked, code would end
up checking the whole Main DB, instead of only overrideable ones.
2019-08-22 14:33:56 +02:00
OmarSquircleArt
08ab3cbcce Shading: Add object color to Object Info node.
The object color property is added as an additional output in
the Object Info node.

Reviewers: brecht

Differential Revision: https://developer.blender.org/D5554
2019-08-22 14:26:09 +02:00
a244384864 Fix T68969: current value '0' matches no enum in 'ToolSettings', '(null)', 'snap_node_element'
I don't think a versioning solution is necessary in this case.
But it can still be implemented.
2019-08-22 09:10:46 -03:00
75127b6be7 Revert "Fix grid snap value in recent snapping updates"
This reverts commit 48a7f979a5.

Another solution to come.
2019-08-22 09:10:46 -03:00
OmarSquircleArt
c6f8ea7b45 Fix T69044: OpenCL fail due to bad fract function.
The fract function in OpenCL does more than just return the fraction.
It also writes the floor to the second argument. Which wasn't put
in consideration.

Instead, we use a simple `a - floor(a)` like the Math node.

Reviewers: brecht

Differential Revision: https://developer.blender.org/D5553
2019-08-22 13:51:39 +02:00
Julian Eisel
599d0611b0 Improve batch file text for starting Blender with Oculus support 2019-08-22 13:33:06 +02:00
b208096538 Transform: Snap Perpendicular: Avoid snapping to snapTarget 2019-08-22 08:07:41 -03:00
3f9c0ab993 Transform: Snap Perpendicular: Use snapTarget instead of center_global 2019-08-22 08:00:53 -03:00
d13952e603 LibOverride: Cleanup: rename parameters in RNA apply code.
We cannot use local/reference here, that is very confusing, since at
that stage current local is kind of src of data for the future local ID,
that is currently a mere copy of the linked data... ;)

So we are much better with src/dst names here.
2019-08-22 12:21:25 +02:00
Charlie Jolly
2ba233a31f Nodes: Support for socket shapes other than circle
Previously there was already "draw_shape" property,
but it was doing nothing. This commit renames the
property to "display_shape". Furthermore, different
shapes like SQUARE and DIAMOND are supported now.

Currently, the shapes are drawn using the shader that also
draws keyframes. In the future we might want to separate
this.

The new shapes are not used anywhere yet, but they can
be used by addon developers and will probably be useful
when we want to support different kinds node systems later.
For example, different shapes can be used to distinguish
between data and control flow.

Differential Revision: https://developer.blender.org/D2829
2019-08-22 11:10:11 +02:00
0356c8f25b Cleanup: remove edit-mode check in vertex coordinate access
This makes the function more predictable since other object
types don't access edit-mode data.
2019-08-22 13:23:27 +10:00
c2442541a6 Cleanup: typo in last commit 2019-08-22 09:13:17 +10:00
189aa32a3a Cleanup: vertex coordinate access, naming & minor changes
This also splits vertex access and allocation so it's possible
to copy coordinates into an existing array without allocating it.
2019-08-22 06:48:10 +10:00
Julian Eisel
e88970db82 Don't create a DirectX swap-chain
We only render offscreen anyway. There's no point in creating a
swap-chain. This simplifies things quite a bit.
2019-08-21 20:58:59 +02:00
4a2d1953f3 Cleanup: use doxy groups 2019-08-22 04:56:18 +10:00
Julian Eisel
20b0b36479 DirectX: Create an own render-target, don't use swapchain for blitting 2019-08-21 20:40:13 +02:00
e6f8f18902 Cleanup: clang-format 2019-08-22 04:38:05 +10:00
Julian Eisel
2c77f2deac Remove GHOST API functions for OpenGL offscreen blitting
These were previously used to blit from a offscreen (non-OpenGL)
context, into an onscreen one. We do this differently, and only within
GHOST_XrGraphicsBinding now, so this can be removed.
2019-08-21 20:36:40 +02:00
OmarSquircleArt
e83f092201 Shading: Add Volume Info node.
The Volume Info node provides the Color, Desnity, Flame, and Temperature
of smoke domains.

Reviewers: brecht

Differential Revision: https://developer.blender.org/D5551
2019-08-21 20:22:24 +02:00
OmarSquircleArt
133dfdd704 Shading: Add White Noise node.
The White Noise node hashes the input and returns a random number in the
range [0, 1]. The input can be a 1D, 2D, 3D, or a 4D vector.

Reviewers: brecht, JacquesLucke

Differential Revision: https://developer.blender.org/D5550
2019-08-21 20:04:09 +02:00
Julian Eisel
fec7ac7c51 Refactor OpenGL/DirectX resource sharing for multiple shared resources
Rather than max. one shared resource per DirectX context, code can now
request a handle for a shared resource and ask the DirectX context to do
operations on that.
2019-08-21 19:44:08 +02:00
OmarSquircleArt
7f4a2fc437 Shading: Add more operators to Vector Math node.
Add Multiply, Divide, Project, Reflect, Distance, Length, Scale, Snap,
Floor, Ceil, Modulo, Fraction, Absolute, Minimum, and Maximum operators
to the Vector Math node. The Value output has been removed from operators
whose output is a vector, and the other way around. All of those removals
has been handled properly in versioning code.

The patch doesn't include tests for the new operators. Tests will be added
in a later patch.

Reviewers: brecht, JacquesLucke

Differential Revision: https://developer.blender.org/D5523
2019-08-21 19:36:33 +02:00
6785da095d Cleanup: remove old Windows scons installer, this is handled by CMake now 2019-08-21 18:44:22 +02:00
6086a6d939 Cleanup: Fix build error with MSVC
Previously eigens internal include order somehow implicitly provided
M_PI and friends. The recent eigen version bump broke this implicit
behaviour, better to be explicit that we need the math defines for MSVC.
2019-08-21 10:38:33 -06:00
34921e9e56 Timeline: refactor cache drawing
I did this mostly to get to know the point cache from this direction.
This should not change anything from the user perspective.
2019-08-21 17:21:50 +02:00
922da6dfb4 LibOverride: Fix inverted logic in RNA collection operation application.
Note to self: need to recheck the namings of whole override code,
currently this is way to confusing and inconsistent.
2019-08-21 17:13:51 +02:00
654fd58270 Cleanup: minor error in assert message in libOverride collection code. 2019-08-21 17:13:51 +02:00
d357e7b065 Fix T68826 Eevee: Multi-Mat not working if switching from Solid shading
The correct fix would be to avoid all those hacks but this is needed if
we want to be able to parallelize object vbo extractions.
2019-08-21 15:53:57 +02:00
8a11e70252 GPU: Fix Element index range calculation
Fix T68880 2D line display is broken
2019-08-21 14:47:57 +02:00
b19c437eff Update Eigen to 3.3.7
This is in preparation for the QuadriFlow remesher lib.

Reviewed By: Brecht

Differential Revision: http://developer.blender.org/D5549
2019-08-21 14:15:28 +02:00
1658fd1f7e Fix T68837 Eevee: Fix "GPU failed to find function math_max" message
Was a missing rename.
2019-08-21 13:39:09 +02:00
Julian Eisel
2a1ec8ec2a Fix missing CMake hint for OpenXR SDK path on linux 2019-08-21 13:24:16 +02:00
de232f8887 Update CUEW to latest version 2019-08-21 11:49:29 +02:00
ecfe020e6c Fix T68951: Incrementing int property causes overflow
This was probably introduced in rBfdef1a6712b.
2019-08-21 11:31:10 +02:00
8f50cdd7d5 Fix T68943: GPencil Time modifier gets strange value in offset parameter
This is due a limitation in the RNA property when the range is too extreme. As we don't need that, the value was set to SHRT_MAX frames as maximum offset.

Also fixed the same problem in other modules of Grease Pencil.
2019-08-21 11:08:10 +02:00
Julian Eisel
6e8a76c96a UI: Don't left align operator buttons in toolbars
This hack would make operator (push down) buttons without icons align
their text to the left in toolbars. Everywhere else in Blender, we
center it by default.

We barely use operator buttons in toolbars anymore. Even if we do and
there's a good reason to make text left aligned, it's better to use
`uiLayout.alignment = 'LEFT'` to achieve the same effect, but without
lowish-level hacks for a specific region type.

Differential Revision: https://developer.blender.org/D5527
2019-08-21 10:35:46 +02:00
Demeter Dzadik
f041d2f116 Fix T65671: Armature X-Mirror inconsistencies
This fixes bendy bone properties not being mirrored correctly

Reviewed By: Brecht

Differential Revision: http://developer.blender.org/D5408
2019-08-21 09:59:11 +02:00
3d8f158697 GPencil: add new filter by material to modifiers
This commit adds a new filter by material using the name and not only the index.

Reviewers: antoniov, pepeland

Differential Revision: https://developer.blender.org/D5544
2019-08-21 08:41:19 +02:00
8ffbe39834 Cleanup: pass lattice directly vertex coord utilities 2019-08-21 15:05:52 +10:00
666d74e4d6 Fix T68444: outliner bone visibility not updating
Tags for update when pose bone visibility is toggled in the outliner.
2019-08-20 22:30:32 -06:00
55c626770e Fix T68429: wrong pose bone hide icon in outliner
The hide icon was drawing instead of the disable in viewports icon.
2019-08-20 22:17:27 -06:00
50fdbf3ec0 Cleanup: use const arguments for vertex coords 2019-08-21 11:58:41 +10:00
0050a044fa Cleanup: remove unused derived-mesh API calls 2019-08-21 11:41:40 +10:00
48a7f979a5 Fix grid snap value in recent snapping updates
Increase the size of the snap mode so the flag can be used.
2019-08-21 11:41:40 +10:00
899d2461bf Fix T68911: "illegal recursive expansion of macro id"
I imagined that could be a problem for some drivers.
I should have avoided.
2019-08-20 21:56:39 -03:00
daba3e871f Fix conflict with "Edge Center" snap mode
`SCE_SNAP_MODE_GRID` is used for both nodes and objects.
I don't think it's worth versioning in this case.
2019-08-20 19:35:11 -03:00
dd08d68df8 Transform: New Snap Option: Edge Perpendicular
Part of T66420

Option for snapping to the nearest point of a reference coordinate.
The patch also adds Edge Center and Perpendicular snaps to the ruler.

{F7675906}

Reviewers: campbellbarton, brecht

Reviewed By: campbellbarton

Differential Revision: https://developer.blender.org/D5543
2019-08-20 19:18:50 -03:00
2db09212fc Fix T68869: outliner missing subversion bump
The recent commit of synced selection added an active color theme to
the outliner, and synced selection enabled by default. The subversion
bump was missed.
2019-08-20 15:59:22 -06:00
f39145b211 Cleanup: remove redundant operator data in keymap 2019-08-21 07:23:20 +10:00
2930fbd86c Keymap: Shift-Ctrl-Click didn't allow toggle selection
D5518 by @jenkm
2019-08-21 07:13:23 +10:00
b6b01c48fa GPencil: Hide self_overlap parameter for Texture strokes
This property is not supported in this mode and must be removed from UI.
2019-08-20 20:26:48 +02:00
58341242bf Editor Transform: New Snap Option Middle
Part of T66420

Reviewers: campbellbarton, brecht

Reviewed By: campbellbarton

Differential Revision: https://developer.blender.org/D5532
2019-08-20 15:07:48 -03:00
1876e18b1b Fix use of uninitialized variable in Cycles math node 2019-08-20 18:19:22 +02:00
a23e1c38f3 Fix T68877: GPencil - Missing remove unused slots in menu
The operator was missing in the menu for grease pencil materials.
2019-08-20 18:11:26 +02:00
14c4a217eb Liboverride: Add more missing ID RNA pointers as overridable. 2019-08-20 17:48:33 +02:00
207f2137dd Fix T68789: Do not use new python format strings thingy in UI code.
This is 100% incompatible with translation...
2019-08-20 17:48:32 +02:00
OmarSquircleArt
be002bb57f Cycles Tests: Fix Math tests when use_clamp is true.
The clamp option is implemented using graph expansion, where a Clamp node
named "clamp" is added and connected to the output. So the final result
is actually from the node "clamp".

Reviewers: brecht

Differential Revision: https://developer.blender.org/D5540
2019-08-20 17:21:40 +02:00
247507a85f Cleanup: remove undefined function 2019-08-20 11:47:50 -03:00
c48d1d54f1 Select Engine: Move some private members out of select context 2019-08-20 11:36:36 -03:00
0827110a35 Cleanup: Move select_engine function to corresponding header 2019-08-20 11:36:35 -03:00
1548682cde Windows/deps: Add/fix openxr_sdk dependency
There were a few typos here and there, and the openxr_sdk does not respect the cflags we give it hence a patch was added to work around this undesirable behaviour.
2019-08-20 08:30:29 -06:00
61ff578eab UI: add URL presets
This adds a url-preset operator to simplify opening re-usable links.

- Standard websites have their own tool-tip.
- De-duplicate logic to get URL's that include version information.
- Reporting bugs no longer needs to have all information
  included in the URL.

D5498 by @luisbg with edits.
2019-08-20 23:44:18 +10:00
e6a81ad61e Cleanup: comparison warning 2019-08-20 23:42:04 +10:00
8bee9af462 BLI: double version of some math functions. 2019-08-20 21:09:55 +08:00
Julian Eisel
588dea6751 Merge branch 'master' into soc-2019-openxr 2019-08-20 12:36:51 +02:00
Julian Eisel
ad4f9a4d98 Fix compiling with WITH_OPENXR disabled 2019-08-20 09:17:56 +02:00
Julian Eisel
84466c7d71 Blacklist VR add-on in Python tests if WITH_OPENXR is disabled 2019-08-20 00:55:21 +02:00
Julian Eisel
de037b031d Expose WITH_OPENXR build option to Python API 2019-08-20 00:12:59 +02:00
Julian Eisel
a21ae28f96 Prepare Toggle VR Session operator to be enabled by an add-on 2019-08-19 23:44:59 +02:00
Julian Eisel
b2424bd781 Windows: Print warning and disable WITH_OPENXR if SDK is not found 2019-08-19 15:41:27 +02:00
Julian Eisel
d0177fc541 Integrate the OpenXR SDK better into dependency management
Adds the SDK so that `make deps` works, on all platforms. On Windows,
the SDK headers and libraries are now assumed to be in the usual lib\
directory.
Also fixes a mistake in install_deps.sh

Note that none of this is tested as it requires an older Visual Studio
version than I have.
2019-08-18 13:06:48 +02:00
Julian Eisel
4cf8740790 Fix compile error on Linux 2019-08-17 23:59:39 +02:00
Julian Eisel
a53b90329a Merge branch 'master' into soc-2019-openxr 2019-08-17 23:58:09 +02:00
Julian Eisel
2c41ffa380 Fix compiling with external OpenXR SDK on Windows 2019-08-15 21:49:10 +02:00
Julian Eisel
cdc2768f65 Cleanup: Don't cast away const in GHOST C-API 2019-08-15 15:52:45 +02:00
Julian Eisel
eb477c67a1 Fix crash when no OpenXR runtime is found (or setup failed) 2019-08-15 15:50:01 +02:00
Julian Eisel
0dd3f3925b Remove bundled OpenXR-SDK sources from extern/
Removes the OPENXR_USE_BUNDLED_SRC option which allowed using the
OpenXR-SDK sources bundled in extern/. It would be too much hassle to
keept these updated.
Now these have to be provided on the system Blender is compiled on. I've
already added necessary bits to install_deps.sh and our Windows
maintainer is ready to provide builds with OpenXR-SDK linked.
Once this gets into master, platform maintainers will have to be
notified about the added dependency.

This also removes the JsonCpp dependency, it was only needed for the
OpenXR-SDK sources.
2019-08-15 15:42:23 +02:00
Julian Eisel
9e7c48575a Linux: Add OpenXR-SDK to install_deps.sh 2019-08-14 22:59:33 +02:00
Julian Eisel
8138099fa9 Silence warnings in OpenXR SDK API layer files 2019-08-14 17:44:02 +02:00
Julian Eisel
02de1b9860 Quiet warning 2019-08-14 16:11:14 +02:00
Julian Eisel
57712625f2 Merge branch 'master' into soc-2019-openxr 2019-08-14 16:01:30 +02:00
Julian Eisel
96a34d7f50 Cleanup: Unused includes, better structuring for wm_xr.c 2019-08-09 16:35:28 +02:00
Julian Eisel
873223f38d Cleanup: Remove unused support for non-OpenGL Windows
This was very useful for testing, but there's no current need for it.
Neither for VR code, nor for anything else in master.
2019-08-09 16:08:34 +02:00
Julian Eisel
716c6b0c01 Cleanup: Move secondary GHOST-context handle to XR surface data
This is used to refer to the DirectX context. For other (currently
non-existant) surfaces, this may not make much sense.
2019-08-09 13:29:08 +02:00
Julian Eisel
62bbd2e37d Don't swap buffers of offscreen surface OpenGL contexts
This does frame syncing to the regular screen, so hurts performance in
some cases.
2019-08-08 22:00:21 +02:00
Julian Eisel
ba223dfac8 Exit OpenXR context/session gracefully if the runtime reports loss pending
Handle XR_TYPE_EVENT_DATA_INSTANCE_LOSS_PENDING and
XR_SESSION_STATE_LOSS_PENDING according to the specification.
2019-08-07 23:58:28 +02:00
Julian Eisel
8557997e81 Disable specular highlights in the VR viewport 2019-08-07 23:05:24 +02:00
Julian Eisel
bcc8f7c6be Fix xrDestroy calls to data already implicitly destroyed by OpenXR
Would effectifely result in null-ops, since the loader does good sanity
checks. The failing functions would still lead to error prints though.
2019-08-07 22:17:55 +02:00
Julian Eisel
fe7b01a08f Correctly request session end according to OpenXR 1.0 specification 2019-08-07 22:13:35 +02:00
Julian Eisel
5350015d51 Fix dark VR rendering on Windows Mixed Reality by appyling SRGB OETF
Discussed this in length with @sobotka, and it seems WMR has an utterly
broken pixel color pipeline. So we apply a SRGB OETF for this specific
runtime to compensate.
2019-08-07 11:52:30 +02:00
Julian Eisel
3483aa57b3 Allow querying OpenXR runtime ID
Useful in case special treatment is needed for certain runtimes.
Currently only WMR gets its ID assigned correctly, others are TODO.
2019-08-07 03:22:33 +02:00
Julian Eisel
44a220f721 Merge branch 'master' into soc-2019-openxr 2019-08-07 02:20:22 +02:00
Julian Eisel
5274c5cce6 Merge branch 'master' into soc-2019-openxr 2019-08-05 12:52:57 +02:00
Julian Eisel
0d70afed19 Merge branch 'master' into soc-2019-openxr 2019-08-02 23:35:40 +02:00
Julian Eisel
03b09dbe32 Fix compile error after merge and warning 2019-08-01 21:53:57 +02:00
Julian Eisel
f04a5ad1e4 Merge branch 'master' into soc-2019-openxr 2019-08-01 21:27:08 +02:00
Julian Eisel
ab1455e972 Fix error in OpenGL version check for OpenXR 1.0 2019-08-01 01:56:35 +02:00
Julian Eisel
b94af38c31 Fix compile error on GCC 2019-08-01 01:39:27 +02:00
Julian Eisel
40db778de3 Add sources for OpenXR core API validation layer
Had to do quite some changes to CMakeLists.txt from the SDK this time.
2019-08-01 00:28:16 +02:00
Julian Eisel
22966f4d35 Address changes in OpenXR 1.0 to get rendering to work again 2019-07-31 02:19:35 +02:00
Julian Eisel
e8f66ff060 Update OpenXR to version 1.0
Only tested on Windows.

Updates loader sources from the OpenXR SDK to latest 1.0 SDK and updates
requiremed version to 1.0. The compile time generation of files is a
thing of the past now (although you can still force it).
1.0 got released yesterday. Only one line needed fixing in our OpenXR
code to get it to compile. Rendering is black though.

For now I tried to keep edits to CMakeList.txt files minimal. So now
there are OpenXR CMake options exposed (with bad names), CMake prints,
etc.
2019-07-31 02:18:28 +02:00
Julian Eisel
9ac33e56a1 Merge branch 'master' into soc-2019-openxr 2019-07-30 22:19:41 +02:00
Julian Eisel
091cc94379 Use fixed lighting, don't make it rotate with view rotation 2019-07-24 16:41:46 +02:00
Julian Eisel
fc31be5ab0 Fix broken VR viewport rotation and movement 2019-07-24 02:52:03 +02:00
Julian Eisel
aff49f607a Use draw manager offscreen context for the VR session
This way we avoid the big overhead of context switches. Makes frames
render about twice as fast here. For heavy Spring scenes I'm getting
around 20 FPS here, classroom scene is at 50 FPS.
This is great given that drawing itself still isn't optimized for dual
eye rendering.
2019-07-24 01:39:04 +02:00
Julian Eisel
57b77cd193 Print 8 Frame average FPS for --debug-xr-time
Much more useful than a per frame FPS estimation. 8 frames are used for
the Blender viewport FPS stats too.
Refactored drawing data storage a bit.
2019-07-23 19:16:39 +02:00
Julian Eisel
d58eb8d29d Add --debug-xr-time command line option for VR frame time info prints
Outputs frame render time in milliseconds and FPS this time would add up
to. We could average times so FPS is a bit more stable, but the
precision of un-averaged results may be helpful too.
2019-07-23 02:06:08 +02:00
Julian Eisel
a4310ba85f Merge branch 'master' into soc-2019-openxr 2019-07-23 00:18:08 +02:00
Julian Eisel
bd42740ef1 Silence GCC warning 2019-07-21 13:15:00 +02:00
Julian Eisel
679a4c34fc Fix compile error with bundled OpenXR sources 2019-07-21 03:18:02 +02:00
Julian Eisel
55362436d9 Fix troubling memory leak in offscreen viewport drawing
I'm not really sure why the leak happened - draw manager kept allocating
certain buffers - but I figured I could avoid it by not recreating
GPU_offscreen/GPU_viewport on every redraw.
This should also improve performance a bit.
2019-07-21 01:10:18 +02:00
Julian Eisel
c1d164070b Pure OpenGL backend works now, let it take priority over DirectX 2019-07-20 19:34:50 +02:00
Julian Eisel
3a4034a741 Make OpenGL-only session backend work
Previously, only DirectX HMD rendering would work (and still takes
priority).
2019-07-20 19:33:11 +02:00
Julian Eisel
a0be113d2e Fix DirectX context not freed on session exit
Also rename XrSurfaceData to wmXrSurfaceData
2019-07-17 15:18:27 +02:00
Julian Eisel
151fb129e7 Fix for previous commit, accidental call to xrDestroySwapchain 2019-07-17 14:51:21 +02:00
Julian Eisel
165c5a5259 Fix possible OpenXR swapchain leak by using new unique_oxr_ptr 2019-07-17 14:05:06 +02:00
Julian Eisel
74fc1db25b Add helper class to RAII manage OpenXR handles
Adds generic unique_oxr_ptr to wrap xrCreate and xrDestroy functions of
OpenXR handles into a unique_ptr like RAII interface.

While for most cases, OpenXR resources can be freed by their owning
object, sometimes errors may occor before final ownership is established.
E.g. swapchain ownership is only transfered to the session object once
its swapchain-images are created - which may fail. With this RAII
wrapper, the swapchain would be freed on error (as this triggers stack
unwinding through an exception), no matter who holds ownership to it
currently.
There are other solutions to this problem, e.g. by establishing final
ownership right after/upon creation, or by explicit freeing in case an
error is spotted; it's too easy to make mistakes here though. Plus, we
may want to experiment with using this API for all OpenXR resources, to
entirely avoid the possibility of them leaking.
2019-07-17 13:35:14 +02:00
Julian Eisel
03ff2a8dd5 Fix compile error with USE_FORCE_WINDOWED_SESSION enabled
At least get it to compile, this is still kinda broken though.
2019-07-17 11:29:34 +02:00
d99d15f48d Add helper batch file for testing on oculus rift. 2019-07-16 16:41:44 -06:00
Julian Eisel
6a998833cd Temporarily let DirectX take priority over OpenGL
OpenGL doesn't work yet. Shouldn't be too difficult to get working, but
I don't have a working OpenXR runtime with OpenGL support to test here.
2019-07-16 21:48:29 +02:00
Julian Eisel
fca4826c7f Merge branch 'master' into soc-2019-openxr 2019-07-16 21:09:30 +02:00
Julian Eisel
b0845fc133 Fix some issues, improve error message
Fixes:
* Destruct surface when destroying session on error. Fixes null pointer
  dereference when trying to draw the surface on next redraw.
* Fix trying to enable same extensions/API-layers multiple times due to
  static array usage not being cleared after error.
* Null pointer dereference with OpenGL drawing
2019-07-13 16:22:26 +02:00
Julian Eisel
70cf8bd9c6 Fix crash on OpenXR error before context is created fully 2019-07-13 01:19:58 +02:00
Julian Eisel
3a2e459f58 Show useful error reports in case of errors.
E.g. with an active OpenXR runtime installed, but no HMD plugged in,
Blender will now show: "Failed to get device information. Is a device
plugged in?".
In case of such errors, the VR-session will be cleanly exited, with
no side effects to the rest of Blender (at least if there are no bugs).

The GHOST_Xr API now allows setting a custom error handling callback
which may cleanly destroy all GHOST_Xr data.
2019-07-12 19:51:38 +02:00
Julian Eisel
330f062099 Safe and informative GHOST_Xr error handling
This wraps all functions that could fail into some proper (although
unfinished) error handling.

These were the requirements for the error handling strategy:
* If an error occurs, cleanly exit the VR session (or the context if the
  error happend during its set up), causing no resource leaks or side
  effects to the rest of Blender.
* Show a *useful* error message to the user.
* Don't impair readability of code too much with error handling.

After some back and forth I decided Ghost internal exceptions are the
best way to go about this. I get exceptions are a controversial feature,
IMHO that's because most people use them 'wrong' however. Here's the
rationale:
* Most alternatives require early exiting functions. This early exiting
  has to be 'bubbled up' the call stack to the point that performs error
  handling. So the code gets really impaired by error checking. Tried
  this first and wasn't happy with it at all. Even if error handling is
  wrapped into macros.
* All GHOST_Xr resources are managed via RAII. So stack unwinding will
  cause them to be released cleanly whenever an exception is thrown.
* GHOST_Xr has a clear boundary (the Ghost C-API) with only a handful of
  public functions. That is the only place we need to have try-catch
  blocks at.
  (Generally, try-catch blocks at kinda random places are a bad code
  smell IMHO. Module boundaries are a valid place to put them.)
* Exceptions allow us to pass multiple bits of error information through
  mulitple layers of the call stack. This information can also be made
  specific with a useful error message.
  As of now, they conain a user error message, the OpenXR error code (if
  any), as well as the exact source code location the error was caught
  at.

So if an error is caught inside GHOST_Xr code, an exception is thrown
with specific and hopefully useful information in it. In the Ghost C-API,
these exceptions are caught and passed on to a custom error handling
callback. This callback will be defined by the Blender window-manager
and output the error information via a usual user report popup (not done
yet). It can also ensure the entire session is shut down.

Note that the majority of errors OpenXR can return are for invalid API
usage. Assuming the API usage is valid though, most error messages
should never reach users and can be left a bit more vague. Maybe we can
add something like "This is probably a bug and should be reported" to
those.
2019-07-12 14:40:50 +02:00
Julian Eisel
09872df1c7 Check runtime DirectX requirements for graphics binding too 2019-07-11 21:05:13 +02:00
Julian Eisel
67d6399da6 Check runtime OpenGL requirements prior to creating graphics binding
DirectX will need these checks too.
Also fix warnings due to wrong printf formatting.
2019-07-11 20:36:54 +02:00
Julian Eisel
c29724912b Merge branch 'master' into soc-2019-openxr 2019-07-11 20:28:19 +02:00
Julian Eisel
99560d9657 Disable debug extension on Windows
Unfortunately, enabling XR_EXT_debug_utils crashes instance creation
with the Windows Mixed Reality runtime. Maybe I'm doing something wrong,
for now, just disable it.
2019-07-11 16:44:51 +02:00
Julian Eisel
b8a7b87873 General minor fixes and cleanup
* Initialize all class member variables
* Add version to runtime name printing
* Separate functionality code from debug prints
* Improve code structure using Doxygen groups
* Make accessors const functions
* Add (Doxygen) comments
* Naming
* Reorder functions
2019-07-11 16:36:04 +02:00
Julian Eisel
c9f6c1a054 Fix wrong return value causing OpenXR initialization to fail 2019-07-11 00:51:00 +02:00
Julian Eisel
dd0fcb50b4 Fix compile errors on windows and with bundled OpenXR SDK sources 2019-07-11 00:50:10 +02:00
Julian Eisel
c1e9cf00ac Refactor GHOST_XrContext into class + interface
Makes GHOST_Xr much more consistent with the rest of GHOST. Basically I
added a GHOST_IXrContext interface which can be called by the GHOST
C-API. The internal GHOST_XrContext class implements this.
Outside of GHOST only the opaque GHOST_XrContextHandle is accessible.

Kept GHOST_Xr_intern.h and GHOST_Xr.cpp for now. I'll probably remove
them soon. There's not much reason for both of them to be there.
2019-07-10 23:53:07 +02:00
Julian Eisel
84d609f67b Use OpenXR extension for extra debug messages in debug mode
Enables (or tries to) the XR_EXT_debug_utils extension which allows
setting a custom callback for additional debug prints. Note that I
haven't been able to test this really, as the Monado runtime appears to
not have this fully implemented (had to force sending a custom message
to find that out...). Will test with the Windows MR runtime in a bit. We
can also improve message quality then (it can print the exact function
name the message was sent from, print additional custom labels which
could indicate session state, etc.).
2019-07-10 15:08:52 +02:00
Julian Eisel
86178548d5 Only try to enable OpenXR API validation layer in debug mode 2019-07-10 15:08:10 +02:00
Julian Eisel
d711540a85 Cleanup: Rename XR_DEBUG_BEGIN -> XR_DEBUG_ONLY_BEGIN
Of course also renames XR_DEBUG_END accordingly.
2019-07-10 15:06:08 +02:00
Julian Eisel
6e3158cb00 Only print info messages with --debug-xr enabled 2019-07-09 17:49:54 +02:00
Julian Eisel
b216690505 Add --debug-xr commandline arg and pass on to GHOST XR contexts
Doesn't do anything yet.
2019-07-09 17:20:01 +02:00
Julian Eisel
b98896ddd9 Preparations to allow enabling OpenXR API validation layer
For this to work two environment variables have to be set:
XR_API_LAYER_PATH and LD_LIBRARY_PATH, both have to point to the API
layers built by the OpenXR SDK. It also requires you compile Blender
linked to your own SDK build (OPENXR_USE_BUNDLED_SRC disabled).
Further changes will make this set up unnecessary, so validation layers
can be enabled via some flag.
2019-07-09 16:45:21 +02:00
Julian Eisel
a05104d61f Fix warnings and compile error on GCC 2019-07-09 11:28:17 +02:00
Julian Eisel
7fe1cc8d24 Enable grid floor in VR session
Helps quite a bit orientating.
2019-07-09 10:56:37 +02:00
Julian Eisel
70e84a2c20 Merge branch 'master' into soc-2019-openxr 2019-07-08 14:18:08 +02:00
Julian Eisel
084a33ca94 Use base pose from Blender camera
I was trying to set the camera pose as OpenXR reference pose. But then I
got really strange poses for drawing back from OpenXR. Couldn't figure
out a way to solve this.
Now we just apply OpenXR's draw pose to the Blender camera pose. If we
later want to support moving around in the scene (e.g. via teleporting),
that has to be changed.

Also uses flipped drawing for DirectX surface to correct DirectX upside
down drawing (compared to OpenGL).
2019-07-08 14:13:14 +02:00
Julian Eisel
37f619aea3 Don't recreate reference local space on each redraw
Set this up on session start.
2019-07-07 13:07:51 +02:00
Julian Eisel
6904ab10f4 Add GHOST_XrSession class for cleaner & safer session management
Session code and data structures are now localized. But also, this
enables sessions to use deterministic destruction to end itself cleanly.
So whenever an error occurs, we can use stack unwinding which will cause
graceful ending of the session.
2019-07-07 12:45:37 +02:00
Julian Eisel
146ef0dd1e Properly end VR sessions fixing crash on next session start
Previously, you had to restart Blender to start another session.
2019-07-03 21:01:58 +02:00
Julian Eisel
5891e86420 Cleanup GHOST_Xr types and functions
* Move GHOST_Xr types to GHOST_Types.h
* Add GHOST_XrPose. We'll have to pass around pose data at multiple
  places.
* Don't require extra call to prepare session rendering, handle
everything through GHOST_XrSessionStart
* Naming
2019-07-01 19:04:54 +02:00
Julian Eisel
2cedad990a Initial HMD viewport rendering (DirectX only first)
Finally: This makes it possible to render a viewport to an HMD via
OpenXR. Pure OpenGL rendering will need some more tweaks to work.
To my great delight, performance is quite good for reasonably sized
scenes.

Had to do some hacks and marked some TODOs. Nothing too bad though.

Here are a couple of notes:
* Current initial pose is pretty useless, think it just looks downwards
  from world origin. Will change that soon.
* The rendered viewport has some issues: Too dark (bad lighting?), grid
  doesn't show up even though I told it to, lighting seems to change with
  view position/rotation, etc. Needs some polish.
* Ideally we'd just use the D3D11 Texture given to us via the OpenXR
  swapchain and blit the OpenGL framebuffer into that. However the
  NV_DX_interop extension fails doing this. Seems like this is a NVidia
  Optimus only issue, but I'm missing the hardware to confirm.
  So instead, we blit into the D3D11 back buffer first and then into the
  Texture.
* The draw-manager uses its own offscreen context so we have to get the
  render result from the draw-manager context to the VR session's
  context first. Luckily I've already added code to support blitting from
  one OpenGL context into another. But it requires blitting twice.
  Blitting should be very cheap, but still...
  Draw-manager could get a context to use passed instead.
2019-07-01 15:57:32 +02:00
Julian Eisel
109be29e42 Draw offscreen viewport in XR session surface callback
Not visible yet, but it should draw in the offscreen. The way this is
now, we don't depend on the Window->Workspace->bScreen->... chain. We
simply draw an offscreen viewport in the draw callback of the XR session
surface.

The drawing also uses view and projection matrices from OpenXR (or
calculated from OpenXR data).
2019-06-29 02:30:23 +02:00
Julian Eisel
13442da69f Move WM level XR functions to wm_xr.c
Operator is still in wm_operators.c, but only calls wm_xr.c functions.
2019-06-28 22:53:46 +02:00
Julian Eisel
231dbd53bf Support Window-less (offscreen) VR session on Windows
Uses the new wmSurface type (non-window drawable container) to manage
the OpenGL, DirectX and GPU module contexts. The draw callback of the XR
surface calls the GHOST_Xr session drawing routines.

What you should see when starting a VR session now (using the WMR
runtime): The Windows Mixed Reality Portal pops up, and a blue
background is drawn on the HMD. This is from the blue color clear call
we do in the drawing preparations of the GHOST_Xr session drawing.
2019-06-28 16:24:24 +02:00
Julian Eisel
d3e48fb096 Merge branch 'master' into soc-2019-openxr 2019-06-28 14:09:08 +02:00
Julian Eisel
d544a0b41c Avoid std::vector copy 2019-06-28 13:56:10 +02:00
Julian Eisel
cf4799e299 Add wmSurface for non-Window offscreen VR session drawing
Adds a wmSurface type which acts as a container for non-Window (offscreen)
draw surfaces. Ideally wmWindow would of course also just do C-style
inheritance from wmSurface, but I guess they can co-exist too.
For the VR session a surface is created on Linux and passed to the
graphics binding to use.

Note this is not used on Windows yet, it still opens a window there.
2019-06-28 13:38:34 +02:00
Julian Eisel
57d9f004aa Fix compile/CMake errors and warnings on Linux 2019-06-27 19:55:22 +02:00
Julian Eisel
62cde7a9ef Finish VR view drawing callback set up
* Allow passing custom data to session draw function, passed to the
  callback
* Actually call the callback
* Create and bind a WM level callback. Will later be used to draw the
  viewport.

Also, check if session is actually visible before drawing.
2019-06-27 19:17:12 +02:00
Julian Eisel
e4fcf25fc5 Don't use Microsoft::WRL::ComPtr<...> for COM types
Makes things difficult if you're not familar with COM.
2019-06-27 18:56:20 +02:00
Julian Eisel
215c919b33 Merge branch 'master' into soc-2019-openxr 2019-06-26 18:39:31 +02:00
Julian Eisel
da44a02d00 The first pixels pushed to Windows Mixed Reality HMDs!
Just a clear call for now, so all you see is a blue world. This blue
"world" is however drawn by Blender!

Also fixes use after destruction of compositor layer data.
2019-06-24 21:22:23 +02:00
Julian Eisel
867007d9d7 Set up OpenXR compositing layers for drawing
That should be one of the last steps to prepare pushing pixels to HMDs.
Fingers crossed!
2019-06-24 16:44:43 +02:00
Julian Eisel
e9ea814233 Merge branch 'master' into soc-2019-openxr 2019-06-24 12:52:37 +02:00
Julian Eisel
6b43c82cb5 Set up OpenXR views and spaces for drawing
Using a dummy identity pose at {0, 0, 0} to start with. This should
later use the current viewport position & orientation I guess.

Also adds function to set a draw callback. There's quite some OpenXR
related stuff to be done before and after drawing anyting, as well as
before and after drawing each view (eye). Quite some info would have to
be exposed to WM to let it manage drawing. So I think using a callback
called from GHOST_Xr to draw each eye instead is a good way to go.

VR session currently crashes after opening. Seems to be related to
blocking frame wait call. I'm not too worried about that though, might
disappear once OpenXR frame sync functions get proper timing info
passed.
2019-06-23 18:27:24 +02:00
Julian Eisel
afbb62fecd Bring back wm_xr.c for higher level XR functions 2019-06-23 15:42:18 +02:00
Julian Eisel
8a676e8fa9 Execute necessary OpenXR frame timing calls 2019-06-23 13:47:42 +02:00
Julian Eisel
c52111e983 Correctly destruct swapchain on shutdown 2019-06-23 02:46:25 +02:00
Julian Eisel
d749e8a2c4 Create graphics binding specific swapchain images
Following the OpenXR SDK's example code very closely here. We have to do
some nasty converting of graphics binding specific image vectors to
generalized base ones. The SDK's approach seems like a good way to go
about this.
2019-06-23 02:12:38 +02:00
Julian Eisel
8663aa80e6 Use abstract class/interface for generalized graphics binding operations
Rather than having switch case blocks throughout the session code, give
binding operations an own interface with dedicated implementations.
2019-06-23 00:37:17 +02:00
Julian Eisel
88b39444a0 Fix error causing sessions to not start correctly
With this, the Windows Mixed Reality Portal finally pops up when
starting the session. That is how it's supposed to work. After it's
initialization phase all you see is black. That's expected too as we
don't send anything to the device yet.
2019-06-22 22:08:50 +02:00
Julian Eisel
4cfc3caa09 OpenXR swapchain creation
Nothing special to say. Just calls needed OpenXR functions to create
swapchains. Swapchain images are not created yet.

Interestingly, the Windows Mixed Realtiy portal now pops up when closing
Blender after having created a VR session. So we're getting closer ;)
2019-06-22 21:25:58 +02:00
Julian Eisel
28428b3462 Merge branch 'master' into soc-2019-openxr 2019-06-21 22:31:01 +02:00
Julian Eisel
24146563a0 Cleanup: Use CamelCase for XR functions/types
Also silence warnings.
2019-06-21 22:27:40 +02:00
Julian Eisel
2ee9c60c20 Finish OpenXR GLX binding initialization
Monado now opens a window here when asking it to start a session. That
seems to be the case because it doesn't detect the HMD as direct mode
capable yet. But that shouldn't be an issue from our side.
2019-06-19 21:11:21 +02:00
Julian Eisel
fb64675209 Fix invalid delete operator use
Again a thing MSVC should warn about but didn't...
2019-06-19 19:09:40 +02:00
Julian Eisel
76385139cf Use friend class to access graphics data on Windows too
Also fix memory leak.
2019-06-19 18:30:19 +02:00
Julian Eisel
b18f5a6a81 Pass graphics context data to OpenXR graphics bindings on Linux too
Refactors function into a class, and make this class a friend of
GHOST_ContextGLX. That seems like a better way to access low level
graphics data for this specific case, rather than giving anyone access
via a getter.
2019-06-19 17:23:41 +02:00
Julian Eisel
9d29373b06 Init OpenXR graphics bindings with valid graphics context (Win only)
Will do this for Linux in a separate commit.

Also fixes a stupid mistake from previous commit on lazy-creation of XR
context.
2019-06-19 12:20:18 +02:00
Julian Eisel
f2f1e4dfed Remove CTX_wm_xr_context
Not really useful, plus XR-Context is now GHOST data.
2019-06-19 01:00:37 +02:00
Julian Eisel
0da0f29ade Rename of XR types (WM -> GHOST) 2019-06-19 00:57:50 +02:00
Julian Eisel
a8519dbac2 Lazy-create XR-context
Creating the context causes the OpenXR loader to try connect to a
runtime. That would involve reading the OS'es active_runtime.json and
dynamic linking based on that. So better avoid doing this on startup.
Also: don't pay for what you don't use!
2019-06-19 00:44:42 +02:00
Julian Eisel
d15d07c4f6 Rename wm_xr_ -> GHOST_XR 2019-06-19 00:28:37 +02:00
Julian Eisel
d356787a65 Merge branch 'soc-2019-openxr' into temp-ghost_openxr 2019-06-18 23:50:28 +02:00
Julian Eisel
a284fa3abb Merge branch 'master' into soc-2019-openxr 2019-06-18 23:49:54 +02:00
Julian Eisel
4eeb7523e9 Use some C++ features where useful
Also correct include guards.
2019-06-18 01:02:33 +02:00
Julian Eisel
f30fcd6c85 Initially move XR files to GHOST and compile in C++ 2019-06-17 23:45:28 +02:00
Julian Eisel
47e8133520 Merge branch 'master' into soc-2019-openxr 2019-06-17 22:24:51 +02:00
Julian Eisel
3fa7d59a04 Fail without crashing if no runtime is found
For now simply do nothing. Plan is to work on proper error handling and
reporting later.
2019-06-17 01:24:07 +02:00
Julian Eisel
df8dc43946 Fix compile errors and warnings on certain configs
Addresses:
* Compile error on Linux
* Compile error with WITH_OPENXR disabled
* Unused parameter warnings
2019-06-16 23:28:14 +02:00
Julian Eisel
3b0a2fecfc Fix missing return value... how did this work even?
Looking at the code this should've failed miserably. No idea how this
could still work fine... Maybe return vars used same stack location?
2019-06-16 23:12:13 +02:00
Julian Eisel
80af6d534d Merge branch 'master' into soc-2019-openxr 2019-06-16 22:58:56 +02:00
Julian Eisel
71c7d61808 Draw OpenGL framebuffer upside down in DirectX windows
DirectX is Y coordinates top to bottom, while OpenGL is the opposite.
For the final window manager on-screen drawing, draw to a texture first
and draw that upside down if needed.
2019-06-16 21:46:57 +02:00
Julian Eisel
637b803b97 Support drawing the VR view into a DirectX window!
Phew! That a fight. But this is also a pretty important feature as it
allows interfacing with Windows Mixed Reality OpenXR runtime and the
HMDs supported by it.

Important remaining issue: The rendered viewport is upside down :) That
is of course because DirectX has the opposite vertical direction than
OpenGL.

When creating a DirectX window, we also create an OpenGL offscreen
context to use for all drawing. Just before swapping framebuffers, the
DirectX window blits the framebuffer of the offscreen context into its
swapchain. This requires the WGL_NV_DX_interop and WGL_NV_DX_interop2
extensions.

For testing/dev purposes, also adds:
* Version of the offscreen to onscreen blitting that's OpenGL only. So
  it blits the default framebuffer of the offscreen context into the one
  of the onscreen context.
  This is disabled by a #define.
* Code to draw a colored triangle using DirectX, also for testing.
  Requires the D3DCompiler.lib to be available at compile time. This is
  also disabled using a #define.
2019-06-16 21:05:21 +02:00
Julian Eisel
314eef0df8 Merge branch 'master' into soc-2019-openxr 2019-06-14 20:41:09 +02:00
Julian Eisel
34fa0f8ac6 Set up DirectX window to support drawing
With this the VR window should open fine and get cleared in a red-ish
orange using Direct3D 11 calls (well, on Windows that is).

The window still draws a 3D view to an offscreen buffer. Where we
usually just swap the buffers, we now allow calling a GHOST function to
blit the offscreen OpenGL buffer to whatever type of graphics buffer the
window uses (DirectX here).
The nice thing about this approach is that all DirectX code stays in
GHOST_ContextD3D.cpp. And the entire compatibiliy code can go into a
single function higher level modules don't need to care about.

This also fixes a number of issues introduced in earlier commits.
2019-06-14 19:07:10 +02:00
Julian Eisel
16366724bd Draw into OpenGL offscreen context in the DirectX window.
The window doesn't show anything of course. However we draw (at least I
assume it does) as regular, just into a window offscreen context.

A valid 3D view is created in the window. It's not visible but you see
cursor changes as you move over the window. So handling works.
2019-06-14 02:41:59 +02:00
Julian Eisel
fc8127d070 (Disabled) code to open a DirectX window with the VR session
The window immediately crashes, hence keeping it disabled for now.

Not sure how much of this I'll leave in, for now this is mainly for
testing DirectX compatibility.
2019-06-13 23:24:30 +02:00
Julian Eisel
3ac37fb552 Fix compile errors with older MSVC/Win-SDK versions
Reported by @LazyDodo, thanks!
2019-06-13 18:22:26 +02:00
Julian Eisel
8e51a75772 Fix mistake causing compiler error on Windows
Stupid one...
2019-06-12 23:45:19 +02:00
Julian Eisel
ba2dab402d Merge branch 'master' into soc-2019-openxr 2019-06-12 21:28:40 +02:00
Julian Eisel
796994d222 Refactor graphics context binding to avoid memory leaks
* Retrieve graphics context to bind through callbacks so the XR code can
  manage their lifetime.
* Use static union to store system & chosen graphics lib specific
  graphics binding data.

Makes some things a bit cleaner, too.
2019-06-12 21:27:00 +02:00
Julian Eisel
66e90e9179 Fix compile error and warning on GCC 2019-06-12 19:28:51 +02:00
Julian Eisel
1404edbe3a Support DirectX Ghost context creation
Adds support for creating a DirectX 11 Ghost context. It's not used yet
and more stuff besides creation is needed. Also, other versions than 11
should probably be supported.
We need a DirectX context to support the Windows Mixed Reality OpenXR
Runtime, a rather important one to support. Idea is to use an extension
for OpenGL-DirectX interoperability for drawing the OpenGL offscreen
viewport render using DirectX.
2019-06-12 11:34:21 +02:00
Julian Eisel
2f68a76409 Split wm_xr.c into multiple files
Similar to gizmo/ and message_bus/, there's now a xr/ directory
containing header files and a intern/ directory for source and internal
header files.

Guess this is reasonable to do. And better to do early on to avoid
loosing much git history.
2019-06-11 17:52:04 +02:00
Julian Eisel
304d9a20fc Fix compile errors and crashes on Windows
* Include needed Windows/DirectX headers
* Disable zero sized arraz (GCC extension)
* Add missing break causing failling assert
* Add missing semicolon in Win only code :)
2019-06-11 16:35:31 +02:00
Julian Eisel
b169e327bf Fix compile error with system OpenXR SDK
Graphics binding types were disabled by a compiler flag then.
2019-06-11 14:18:35 +02:00
Julian Eisel
7238bbecc4 Cleanup: Rename graphics-lib to graphics-binding 2019-06-10 22:16:09 +02:00
Julian Eisel
a49b4ca402 Merge branch 'master' into soc-2019-openxr 2019-06-10 22:06:42 +02:00
Julian Eisel
eb0424c946 Proper management of OpenXR graphics bindings for session setup
OpenXR needs to interface with some graphics library (OpenGL, Vulkan,
DirectX, etc.). This is done through graphics binding extensions. The
OpenXR specification requires these to be properly set up before a
session is created.

Adds the following:
* Support priority list of multiple graphics binding extensions (e.g.
  check OpenGL extension availability first, DirectX on Windows second,
  etc.)
* Barebones for passing graphics library data to OpenXR session
  creation. This is highly system dependent, e.g. it requires GLX data
  for OpenGL on X11 systems (XrGraphicsBindingOpenGLXlibKHR). More work,
  including additions to GHOST, will be needed once I get to the more
  graphics related stuff.
* Create an own graphics context for the VR session. It's not doing
  anything useful yet. This is just to fool the Monado OpenXR runtime
  so that it actually attempts to create the OpenXR session.
* Had to add two CMake modules for platform dependent #define's required
  by the OpenXR specification.
2019-06-10 21:59:26 +02:00
Julian Eisel
5d6da57aa3 Cleanup: Move OpenXRData struct out of wmXRContext
Just to make access a bit more convenient.
2019-06-06 18:38:06 +02:00
Julian Eisel
e968bbc5d8 Fix wrong enum-type used for switch
Why doesn't MSVC warn about this, grr...
2019-06-06 17:37:18 +02:00
Julian Eisel
b1b0e05c54 Fix CMake errors after merge (?)
Not sure if this is caused by changes just merged in, or by a system
upgrade I just did. Either way, fixed it now.
2019-06-06 11:25:53 +02:00
Julian Eisel
e547026dfe Merge branch 'master' into soc-2019-openxr 2019-06-06 10:56:19 +02:00
Julian Eisel
f816380af5 Silence warning, fix mem-leak, don't destroy NULL session 2019-06-05 18:03:59 +02:00
Julian Eisel
5e642bfea1 Initial session state handling
To correctly start a session, a graphics extension specific object needs
to be passed to the OpenXR runtime. E.g. for the Windows Mixed Reality
runtime, XrGraphicsBindingD3D11KHR needs to be passed with a valid
DirectX device. Since we don't have any DirectX compatibility working,
I can't test this on Windows yet.
So to test this I finally need to get Monado to work on Linux and
correctly setup the OpenGL extension there.
2019-06-05 16:55:51 +02:00
Julian Eisel
0d1a22f74f Barebones for managing session changes
Adds initial OpenXR event querying so that session state change events
can be handled. Doesn't do any handling yet.
2019-06-05 10:20:51 +02:00
Julian Eisel
20a1af936d Enable graphics extensions if available
Check if available extensions includes the GL one, or the DirectX one on
Windows. Enable them if available.
2019-06-05 01:00:13 +02:00
Julian Eisel
c9c1b90532 Cleanup: Correct local variable name 2019-06-04 18:30:07 +02:00
Julian Eisel
4192281e14 Fix layer and extension count not stored as intended 2019-06-04 18:27:44 +02:00
Julian Eisel
329deadc9e Print OpenXR runtime name once connected 2019-06-04 17:39:12 +02:00
Julian Eisel
53845c554d Merge branch 'master' into soc-2019-openxr 2019-06-04 17:00:58 +02:00
Julian Eisel
35c9e3beed Initial VR-session starting/ending
Adds operator to toggle a VR session, exposed in the Window top-bar
menu. It triggers the needed calls for session creation and destruction.
Setting up the XR-system (a configuration of related devices) is also
done now.

Calling WMR runtime functions fails currently. Not sure why. So while
this executes required routines, it doesn't really work.
2019-06-04 16:42:42 +02:00
Julian Eisel
60f66768cf Fix GCC compile error, warnings and uninitialized use
extension_count could be used uninitialized if no valid OpenXR runtime
was run.
2019-06-02 22:04:52 +02:00
Julian Eisel
3e88974a80 Merge branch 'master' into soc-2019-openxr 2019-06-02 15:24:33 +02:00
Julian Eisel
09e7d6cd26 Correct previous commit
Intended to print out name, not address.
2019-06-02 02:24:04 +02:00
Julian Eisel
5d5ad5d6dd Gather available OpenXR extensions and API-layers on OpenXR setup
We don't actually enable any extension or layer yet. We just put their
names into arrays and print them. The printing can be disabled via a
compiler define, later we should put them behind a proper
logging/debugging mechanism (CLOG).

Also fixes a Visual Studio compile error.
2019-06-01 17:16:49 +02:00
Julian Eisel
e70894c360 Fix compiling with bundled OpenXR SDK sources
Also remove unused include.
2019-05-30 16:22:05 +02:00
Julian Eisel
4c11886c4e Corrections to previous commit
* Remove testing instance creation
* Rename function
2019-05-30 14:56:23 +02:00
Julian Eisel
41e2f2e75f Proper creation & destruction of OpenXR instances
Adds wm_xr.c for an XR management API and creates/destroys the OpenXR
instance through this. This is as planned in my proposal, to lock the
OpenXR calls behind an abstraction.
XR data will be stored in a (non-public) wmXRContext struct within the
window-manager.

For now, creates the OpenXR instance on startup. I think it's better to
lazy setup this, as in, only creating the instance once the user starts
the first XR session. Just to avoid costs for something that may not be
used (the OpenXR loader we use will try loading and parsing the system's
OpenXR active_runtime.json on instance creation). That's for later when
I introduce session management though.

Also added a context getter for the xr-context, which is unused but may
be handy later.
2019-05-30 14:48:13 +02:00
Julian Eisel
864abbd425 Support system installed OpenXR SDK
Adds OPENXR_USE_BUNDLED_SRC so that if disabled, CMake tries to find
the SDK headers and libraries in system paths or in specified root
directory.

I guess this is the way we'd want to do this in master. However for
people testing the branch the bundled sources are much more convenient
(should work out of the box, no need to compile the SDK manually).
2019-05-30 01:44:13 +02:00
Julian Eisel
f333e7eb56 Merge branch 'master' into soc-2019-openxr 2019-05-29 20:14:11 +02:00
Julian Eisel
a2ac2ec3c3 Merge branch 'master' into soc-2019-openxr 2019-05-29 15:17:41 +02:00
Julian Eisel
06d52afcda Linux: OpenXR Linking via OpenXR Loader
Calling OpenXR functions should now work on both Windows and Linux.

Also includes:
* Silence Linux only warnings
* Remove common_cmake_config.h and let CMake generate it when
  configuring (contents vary on system it's compiled on)
* Remove JsonCpp CMake install target
* Remove unnecessary CMakeLists for JsonCpp includes
* Remove unnecessary CMake commands
* Style cleanup
2019-05-28 21:58:18 +02:00
Julian Eisel
e65ba62c37 Windows: OpenXR Linking via OpenXR Loader
Adds needed headers for OpenXR, the loader from the OpenXR SDK and
JsonCpp into extern. Took a while to get this to compile/link, but on
Win10 it works fine now without patching #includes. Linux probably
needs more work.

Added a compile option WITH_OPENXR to toggle XR feature compiling.

Also does a dummy xrCreateInstance() call to test linking.
2019-05-28 01:21:25 +02:00
1091 changed files with 82387 additions and 25608 deletions

View File

@@ -224,7 +224,9 @@ ForEachMacros:
- LISTBASE_CIRCULAR_BACKWARD_BEGIN
- LISTBASE_CIRCULAR_FORWARD_BEGIN
- LISTBASE_FOREACH
- LISTBASE_FOREACH_BACKWARD
- LISTBASE_FOREACH_MUTABLE
- LISTBASE_FOREACH_BACKWARD_MUTABLE
- MAN2D_ITER_AXES_BEGIN
- MAN_ITER_AXES_BEGIN
- NODE_INSTANCE_HASH_ITER

View File

@@ -236,6 +236,7 @@ option(WITH_BULLET "Enable Bullet (Physics Engine)" ON)
option(WITH_SYSTEM_BULLET "Use the systems bullet library (currently unsupported due to missing features in upstream!)" )
mark_as_advanced(WITH_SYSTEM_BULLET)
option(WITH_OPENCOLORIO "Enable OpenColorIO color management" ${_init_OPENCOLORIO})
option(WITH_OPENXR "Enable VR features through the OpenXR specification" ON)
# Compositor
option(WITH_COMPOSITOR "Enable the tile based nodal compositor" ON)
@@ -1430,7 +1431,6 @@ if(CMAKE_COMPILER_IS_GNUCC)
ADD_CHECK_CXX_COMPILER_FLAG(CXX_WARNINGS CXX_WARN_NO_DIV_BY_ZERO -Wno-div-by-zero)
ADD_CHECK_CXX_COMPILER_FLAG(CXX_WARNINGS CXX_WARN_TYPE_LIMITS -Wtype-limits)
ADD_CHECK_CXX_COMPILER_FLAG(CXX_WARNINGS CXX_WARN_ERROR_RETURN_TYPE -Werror=return-type)
ADD_CHECK_CXX_COMPILER_FLAG(CXX_WARNINGS CXX_WARN_ERROR_IMPLICIT_FUNCTION_DECLARATION -Werror=implicit-function-declaration)
ADD_CHECK_CXX_COMPILER_FLAG(CXX_WARNINGS CXX_WARN_NO_CHAR_SUBSCRIPTS -Wno-char-subscripts)
ADD_CHECK_CXX_COMPILER_FLAG(CXX_WARNINGS CXX_WARN_NO_UNKNOWN_PRAGMAS -Wno-unknown-pragmas)
ADD_CHECK_CXX_COMPILER_FLAG(CXX_WARNINGS CXX_WARN_POINTER_ARITH -Wpointer-arith)
@@ -1763,6 +1763,7 @@ if(FIRST_RUN)
info_cfg_option(WITH_CYCLES)
info_cfg_option(WITH_FREESTYLE)
info_cfg_option(WITH_OPENCOLORIO)
info_cfg_option(WITH_OPENXR)
info_cfg_option(WITH_OPENIMAGEDENOISE)
info_cfg_option(WITH_OPENVDB)
info_cfg_option(WITH_ALEMBIC)

View File

@@ -43,6 +43,7 @@ project("BlenderDependencies")
cmake_minimum_required(VERSION 3.5)
include(ExternalProject)
include(cmake/check_software.cmake)
include(cmake/options.cmake)
include(cmake/versions.cmake)
@@ -98,6 +99,7 @@ else()
include(cmake/pugixml.cmake)
endif()
include(cmake/openimagedenoise.cmake)
include(cmake/openxr.cmake)
if(WITH_WEBP)
include(cmake/webp.cmake)

View File

@@ -0,0 +1,53 @@
# ***** BEGIN GPL LICENSE BLOCK *****
#
# This program is free software; you can redistribute it and/or
# modify it under the terms of the GNU General Public License
# as published by the Free Software Foundation; either version 2
# of the License, or (at your option) any later version.
#
# This program is distributed in the hope that it will be useful,
# but WITHOUT ANY WARRANTY; without even the implied warranty of
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
# GNU General Public License for more details.
#
# You should have received a copy of the GNU General Public License
# along with this program; if not, write to the Free Software Foundation,
# Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, USA.
#
# ***** END GPL LICENSE BLOCK *****
if(UNIX)
set(_required_software
autoconf
automake
libtoolize
nasm
yasm
tclsh
)
foreach(_software ${_required_software})
find_program(_software_find NAMES ${_software})
if(NOT _software_find)
set(_software_missing "${_software_missing}${_software} ")
endif()
unset(_software_find CACHE)
endforeach()
if(_software_missing)
message(
"\n"
"Missing software for building Blender dependencies:\n"
" ${_software_missing}\n"
"\n"
"On Debian and Ubuntu:\n"
" apt install autoconf automake libtool yasm nasm tcl\n"
"\n"
"On macOS (with homebrew):\n"
" brew install cmake autoconf automake libtool yasm nasm\n"
"\n"
"Other platforms:\n"
" Install equivalent packages.\n")
message(FATAL_ERROR "Install missing software before continuing")
endif()
endif()

View File

@@ -174,6 +174,8 @@ harvest(opensubdiv/include opensubdiv/include "*.h")
harvest(opensubdiv/lib opensubdiv/lib "*.a")
harvest(openvdb/include/openvdb openvdb/include/openvdb "*.h")
harvest(openvdb/lib openvdb/lib "*.a")
harvest(openxr_sdk/include/openxr openxr_sdk/include/openxr "*.h")
harvest(openxr_sdk/lib openxr_sdk/src/loader "*.a")
harvest(osl/bin osl/bin "oslc")
harvest(osl/include osl/include "*.h")
harvest(osl/lib osl/lib "*.a")

View File

@@ -0,0 +1,53 @@
# ***** BEGIN GPL LICENSE BLOCK *****
#
# This program is free software; you can redistribute it and/or
# modify it under the terms of the GNU General Public License
# as published by the Free Software Foundation; either version 2
# of the License, or (at your option) any later version.
#
# This program is distributed in the hope that it will be useful,
# but WITHOUT ANY WARRANTY; without even the implied warranty of
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
# GNU General Public License for more details.
#
# You should have received a copy of the GNU General Public License
# along with this program; if not, write to the Free Software Foundation,
# Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, USA.
#
# ***** END GPL LICENSE BLOCK *****
set(OPENXR_SDK_EXTRA_ARGS
-DBUILD_API_LAYERS=ON
-DBUILD_FORCE_GENERATION=ON
-DBUILD_LOADER=ON
-DBUILD_SPECIFICATION=OFF
-DBUILD_TESTS=OFF
-DDYNAMIC_LOADER=OFF
)
ExternalProject_Add(external_openxr_sdk
URL ${OPENXR_SDK_URI}
DOWNLOAD_DIR ${DOWNLOAD_DIR}
URL_HASH MD5=${OPENXR_SDK_HASH}
PREFIX ${BUILD_DIR}/openxr_sdk
CMAKE_ARGS -DCMAKE_INSTALL_PREFIX=${LIBDIR}/openxr_sdk ${DEFAULT_CMAKE_FLAGS} ${OPENXR_SDK_EXTRA_ARGS}
PATCH_COMMAND ${PATCH_CMD} --verbose -p 1 -N -d ${BUILD_DIR}/openxr_sdk/src/external_openxr_sdk < ${PATCH_DIR}/openxr_sdk.diff
INSTALL_DIR ${LIBDIR}/openxr_sdk
)
if(WIN32)
if(BUILD_MODE STREQUAL Release)
ExternalProject_Add_Step(external_openxr_sdk after_install
COMMAND ${CMAKE_COMMAND} -E copy_directory ${LIBDIR}/openxr_sdk/include/openxr ${HARVEST_TARGET}/openxr_sdk/include/openxr
COMMAND ${CMAKE_COMMAND} -E copy_directory ${LIBDIR}/openxr_sdk/lib ${HARVEST_TARGET}/openxr_sdk/lib
DEPENDEES install
)
endif()
if(BUILD_MODE STREQUAL Debug)
ExternalProject_Add_Step(external_openxr_sdk after_install
COMMAND ${CMAKE_COMMAND} -E copy ${LIBDIR}/openxr_sdk/lib/openxr_loader-1_0.lib ${HARVEST_TARGET}/openxr_sdk/lib/openxr_loader-1_0_d.lib
DEPENDEES install
)
endif()
endif()

View File

@@ -306,3 +306,7 @@ set(EMBREE_HASH 3d4a1147002ff43939d45140aa9d6fb8)
set(OIDN_VERSION 1.0.0)
set(OIDN_URI https://github.com/OpenImageDenoise/oidn/releases/download/v${OIDN_VERSION}/oidn-${OIDN_VERSION}.src.zip)
set(OIDN_HASH 19fe67b0164e8f020ac8a4f520defe60)
set(OPENXR_SDK_VERSION 1.0.0)
set(OPENXR_SDK_URI https://github.com/KhronosGroup/OpenXR-SDK-Source/archive/release-${OPENXR_SDK_VERSION}.tar.gz)
set(OPENXR_SDK_HASH 260bdc87b5a9b7ef35a540e39f875d79)

View File

@@ -27,16 +27,16 @@ getopt \
-o s:i:t:h \
--long source:,install:,tmp:,info:,threads:,help,show-deps,no-sudo,no-build,no-confirm,\
with-all,with-opencollada,with-jack,with-embree,with-oidn,\
ver-ocio:,ver-oiio:,ver-llvm:,ver-osl:,ver-osd:,ver-openvdb:,\
ver-ocio:,ver-oiio:,ver-llvm:,ver-osl:,ver-osd:,ver-openvdb:,ver-openxr\
force-all,force-python,force-numpy,force-boost,\
force-ocio,force-openexr,force-oiio,force-llvm,force-osl,force-osd,force-openvdb,\
force-ffmpeg,force-opencollada,force-alembic,force-embree,force-oidn,\
force-ffmpeg,force-opencollada,force-alembic,force-embree,force-oidn,foce-openxr\
build-all,build-python,build-numpy,build-boost,\
build-ocio,build-openexr,build-oiio,build-llvm,build-osl,build-osd,build-openvdb,\
build-ffmpeg,build-opencollada,build-alembic,build-embree,build-oidn,\
build-ffmpeg,build-opencollada,build-alembic,build-embree,build-oidn,build-openxr\
skip-python,skip-numpy,skip-boost,\
skip-ocio,skip-openexr,skip-oiio,skip-llvm,skip-osl,skip-osd,skip-openvdb,\
skip-ffmpeg,skip-opencollada,skip-alembic,skip-embree,skip-oidn \
skip-ffmpeg,skip-opencollada,skip-alembic,skip-embree,skip-oidn,skip-openxr \
-- "$@" \
)
@@ -144,6 +144,9 @@ ARGUMENTS_INFO="\"COMMAND LINE ARGUMENTS:
--ver-openvdb=<ver>
Force version of OpenVDB library.
--ver-openxr=<ver>
Force version of OpenXR-SDK.
Note about the --ver-foo options:
It may not always work as expected (some libs are actually checked out from a git rev...), yet it might help
to fix some build issues (like LLVM mismatch with the version used by your graphic system).
@@ -196,6 +199,9 @@ ARGUMENTS_INFO="\"COMMAND LINE ARGUMENTS:
--build-ffmpeg
Force the build of FFMpeg.
--build-openxr
Force the build of OpenXR-SDK.
Note about the --build-foo options:
* They force the script to prefer building dependencies rather than using available packages.
This may make things simpler and allow working around some distribution bugs, but on the other hand it will
@@ -254,6 +260,9 @@ ARGUMENTS_INFO="\"COMMAND LINE ARGUMENTS:
--force-ffmpeg
Force the rebuild of FFMpeg.
--force-openxr
Force the rebuild of OpenXR-SDK.
Note about the --force-foo options:
* They obviously only have an effect if those libraries are built by this script
(i.e. if there is no available and satisfactory package)!
@@ -303,7 +312,10 @@ ARGUMENTS_INFO="\"COMMAND LINE ARGUMENTS:
Unconditionally skip OpenImageDenoise installation/building.
--skip-ffmpeg
Unconditionally skip FFMpeg installation/building.\""
Unconditionally skip FFMpeg installation/building.
--skip-openxr
Unconditionally skip OpenXR-SDK installation/building.\""
##### Main Vars #####
@@ -416,6 +428,11 @@ FFMPEG_FORCE_REBUILD=false
FFMPEG_SKIP=false
_ffmpeg_list_sep=";"
OPENXR_VERSION="1.0.0"
OPENXR_FORCE_BUILD=false
OPENXR_FORCE_REBUILD=false
OPENXR_SKIP=false
# FFMPEG optional libs.
VORBIS_USE=false
VORBIS_DEV=""
@@ -581,6 +598,11 @@ while true; do
OPENVDB_VERSION_MIN=$OPENVDB_VERSION
shift; shift; continue
;;
--ver-openxr)
OPENXR_VERSION="$2"
OPENXR_VERSION_MIN=$OPENXR_VERSION
shift; shift; continue
;;
--build-all)
PYTHON_FORCE_BUILD=true
NUMPY_FORCE_BUILD=true
@@ -597,6 +619,7 @@ while true; do
OIDN_FORCE_BUILD=true
FFMPEG_FORCE_BUILD=true
ALEMBIC_FORCE_BUILD=true
OPENXR_FORCE_BUILD=true
shift; continue
;;
--build-python)
@@ -648,6 +671,9 @@ while true; do
--build-alembic)
ALEMBIC_FORCE_BUILD=true; shift; continue
;;
--build-openxr)
OPENXR_FORCE_BUILD=true; shift; continue
;;
--force-all)
PYTHON_FORCE_REBUILD=true
NUMPY_FORCE_REBUILD=true
@@ -664,6 +690,7 @@ while true; do
OIDN_FORCE_REBUILD=true
FFMPEG_FORCE_REBUILD=true
ALEMBIC_FORCE_REBUILD=true
OPENXR_FORCE_REBUILD=true
shift; continue
;;
--force-python)
@@ -713,6 +740,9 @@ while true; do
--force-alembic)
ALEMBIC_FORCE_REBUILD=true; shift; continue
;;
--force-openxr)
OPENXR_FORCE_REBUILD=true; shift; continue
;;
--skip-python)
PYTHON_SKIP=true; shift; continue
;;
@@ -758,6 +788,9 @@ while true; do
--skip-alembic)
ALEMBIC_SKIP=true; shift; continue
;;
--skip-openxr)
OPENXR_SKIP=true; shift; continue
;;
--)
# no more arguments to parse
break
@@ -884,6 +917,12 @@ OIDN_SOURCE=( "https://github.com/OpenImageDenoise/oidn/releases/download/v${OID
FFMPEG_SOURCE=( "http://ffmpeg.org/releases/ffmpeg-$FFMPEG_VERSION.tar.bz2" )
OPENXR_USE_REPO=false
OPENXR_SOURCE=("https://github.com/KhronosGroup/OpenXR-SDK-Source/archive/release-$OPENXR_VERSION.tar.gz")
#~ OPENXR_SOURCE_REPO=("https://github.com/KhronosGroup/OpenXR-SDK-Source.git")
#~ OPENXR_REPO_UID="348912bf9bfaf445ac2974bda19fd0d50496460b"
#~ OPENXR_REPO_BRANCH="master"
# C++11 is required now
CXXFLAGS_BACK=$CXXFLAGS
CXXFLAGS="$CXXFLAGS -std=c++11"
@@ -1107,13 +1146,14 @@ _create_inst_shortcut() {
# ldconfig
run_ldconfig() {
_lib_path="$INST/$1/lib"
_lib64_path="$INST/$1/lib64"
_ldconf_path="/etc/ld.so.conf.d/$1.conf"
PRINT ""
if [ ! $SUDO ]; then
WARNING "--no-sudo enabled, impossible to run ldconfig for $1, you'll have to do it yourself..."
else
INFO "Running ldconfig for $1..."
$SUDO sh -c "echo \"$_lib_path\" > $_ldconf_path"
$SUDO sh -c "echo -e \"$_lib_path\n$_lib64_path\" > $_ldconf_path"
$SUDO /sbin/ldconfig # XXX OpenSuse does not include sbin in command path with sudo!!!
fi
PRINT ""
@@ -1974,7 +2014,8 @@ compile_OSL() {
cmake_d="$cmake_d -D OSL_BUILD_PLUGINS=OFF"
cmake_d="$cmake_d -D OSL_BUILD_TESTS=OFF"
cmake_d="$cmake_d -D USE_SIMD=sse2"
cmake_d="$cmake_d -D OSL_BUILD_CPP11=1"
cmake_d="$cmake_d -D USE_LLVM_BITCODE=OFF"
cmake_d="$cmake_d -D USE_PARTIO=OFF"
#~ cmake_d="$cmake_d -D ILMBASE_VERSION=$ILMBASE_VERSION"
@@ -2682,6 +2723,8 @@ compile_OIDN() {
INFO "Own OpenImageDenoise-$OIDN_VERSION is up to date, nothing to do!"
INFO "If you want to force rebuild of this lib, use the --force-oidn option."
fi
run_ldconfig "oidn"
}
#### Build FFMPEG ####
@@ -2795,6 +2838,103 @@ compile_FFmpeg() {
}
#### Build OpenXR SDK ####
_init_openxr_sdk() {
_src=$SRC/OpenXR-SDK-$OPENXR_VERSION
_git=true
_inst=$INST/openxr-sdk-$OPENXR_VERSION
_inst_shortcut=$INST/openxr-sdk
}
clean_OpenXR_SDK() {
_init_openxr_sdk
_clean
}
compile_OpenXR_SDK() {
if [ "$NO_BUILD" = true ]; then
WARNING "--no-build enabled, OpenXR will not be compiled!"
return
fi
# To be changed each time we make edits that would modify the compiled result!
openxr_magic=0
_init_openxr_sdk
# Clean install if needed!
magic_compile_check openxr-$OPENXR_VERSION $openxr_magic
if [ $? -eq 1 -o "$OPENXR_FORCE_REBUILD" = true ]; then
clean_OpenXR_SDK
fi
if [ ! -d $_inst ]; then
INFO "Building OpenXR-SDK-$OPENXR_VERSION"
prepare_opt
if [ ! -d $_src ]; then
mkdir -p $SRC
if [ "$OPENXR_USE_REPO" = true ]; then
git clone $OPENXR_SOURCE_REPO $_src
else
download OPENXR_SOURCE[@] "$_src.tar.gz"
INFO "Unpacking OpenXR-SDK-$OPENXR_VERSION"
tar -C $SRC --transform "s,(.*/?)OpenXR-SDK-[^/]*(.*),\1OpenXR-SDK-$OPENXR_VERSION\2,x" \
-xf $_src.tar.gz
fi
fi
cd $_src
if [ "$OPENXR_USE_REPO" = true ]; then
git pull origin $OPENXR_REPO_BRANCH
# Stick to same rev as windows' libs...
git checkout $OPENXR_REPO_UID
git reset --hard
fi
# Always refresh the whole build!
if [ -d build ]; then
rm -rf build
fi
mkdir build
cd build
cmake_d="-D CMAKE_BUILD_TYPE=Release"
cmake_d="$cmake_d -D CMAKE_INSTALL_PREFIX=$_inst"
cmake_d="$cmake_d -D BUILD_API_LAYERS=ON"
cmake_d="$cmake_d -D BUILD_FORCE_GENERATION=ON"
cmake_d="$cmake_d -D BUILD_LOADER=ON"
cmake_d="$cmake_d -D BUILD_SPECIFICATION=OFF"
cmake_d="$cmake_d -D BUILD_TESTS=OFF"
cmake $cmake_d ..
make -j$THREADS && make install
make clean
if [ -d $_inst ]; then
_create_inst_shortcut
else
ERROR "OpenXR-SDK-$OPENXR_VERSION failed to compile, exiting"
exit 1
fi
magic_compile_set openxr-$OPENXR_VERSION $openxr_magic
cd $CWD
INFO "Done compiling OpenXR-SDK-$OPENXR_VERSION!"
else
INFO "Own OpenXR-SDK-$OPENXR_VERSION is up to date, nothing to do!"
INFO "If you want to force rebuild of this lib, use the --force-openxr option."
fi
run_ldconfig "openxr"
}
#### Install on DEB-like ####
get_package_version_DEB() {
dpkg-query -W -f '${Version}' $1 | sed -r 's/([0-9]+:)?(([0-9]+\.?)+([0-9]+)).*/\2/'
@@ -3319,6 +3459,18 @@ install_DEB() {
compile_FFmpeg
fi
fi
PRINT ""
if [ "$OPENXR_SKIP" = true ]; then
WARNING "Skipping OpenXR-SDK installation, as requested..."
elif [ "$OPENXR_FORCE_BUILD" = true ]; then
INFO "Forced OpenXR-SDK building, as requested..."
compile_OpenXR_SDK
else
# No package currently!
PRINT ""
compile_OpenXR_SDK
fi
}
@@ -3905,6 +4057,17 @@ install_RPM() {
compile_FFmpeg
fi
fi
PRINT ""
if [ "$OPENXR_SKIP" = true ]; then
WARNING "Skipping OpenXR-SDK installation, as requested..."
elif [ "$OPENXR_FORCE_BUILD" = true ]; then
INFO "Forced OpenXR-SDK building, as requested..."
compile_OpenXR_SDK
else
# No package currently!
compile_OpenXR_SDK
fi
}
@@ -4387,6 +4550,17 @@ install_ARCH() {
compile_FFmpeg
fi
fi
PRINT ""
if [ "$OPENXR_SKIP" = true ]; then
WARNING "Skipping OpenXR-SDK installation, as requested..."
elif [ "$OPENXR_FORCE_BUILD" = true ]; then
INFO "Forced OpenXR-SDK building, as requested..."
compile_OpenXR_SDK
else
# No package currently!
compile_OpenXR_SDK
fi
}
@@ -4583,6 +4757,17 @@ install_OTHER() {
INFO "Forced FFMpeg building, as requested..."
compile_FFmpeg
fi
PRINT ""
if [ "$OPENXR_SKIP" = true ]; then
WARNING "Skipping OpenXR-SDK installation, as requested..."
elif [ "$OPENXR_FORCE_BUILD" = true ]; then
INFO "Forced OpenXR-SDK building, as requested..."
compile_OpenXR_SDK
else
# No package currently!
compile_OpenXR_SDK
fi
}
#### Printing User Info ####
@@ -4840,6 +5025,17 @@ print_info() {
fi
fi
if [ "$OPENXR_SKIP" = false ]; then
_1="-D WITH_OPENXR=ON"
PRINT " $_1"
_buildargs="$_buildargs $_1"
if [ -d $INST/openxr-sdk ]; then
_1="-D OPENXR_ROOT_DIR=$INST/openxr-sdk"
PRINT " $_1"
_buildargs="$_buildargs $_1"
fi
fi
PRINT ""
PRINT "Or even simpler, just run (in your blender-source dir):"
PRINT " make -j$THREADS BUILD_CMAKE_ARGS=\"$_buildargs\""

View File

@@ -0,0 +1,28 @@
diff -Naur orig/src/loader/CMakeLists.txt external_openxr_sdk/src/loader/CMakeLists.txt
--- orig/src/loader/CMakeLists.txt 2019-07-29 07:06:59 -0600
+++ external_openxr_sdk/src/loader/CMakeLists.txt 2019-08-20 07:56:51 -0600
@@ -128,24 +128,6 @@
configure_file("openxr.pc.in" "openxr.pc" @ONLY)
install(FILES "${CMAKE_CURRENT_BINARY_DIR}/openxr.pc" DESTINATION "${CMAKE_INSTALL_LIBDIR}/pkgconfig")
elseif(CMAKE_SYSTEM_NAME STREQUAL "Windows")
- foreach(configuration in CMAKE_C_FLAGS_DEBUG
- CMAKE_C_FLAGS_RELEASE
- CMAKE_C_FLAGS_RELWITHDEBINFO
- CMAKE_CXX_FLAGS_DEBUG
- CMAKE_CXX_FLAGS_RELEASE
- CMAKE_CXX_FLAGS_RELWITHDEBINFO)
- # If building DLLs, force static CRT linkage
- if(DYNAMIC_LOADER)
- if (${configuration} MATCHES "/MD")
- string(REGEX REPLACE "/MD" "/MT" ${configuration} "${${configuration}}")
- endif()
- else() # Otherwise for static libs, link the CRT dynamically
- if (${configuration} MATCHES "/MT")
- string(REGEX REPLACE "/MT" "/MD" ${configuration} "${${configuration}}")
- endif()
- endif()
- endforeach()
-
target_link_libraries(${LOADER_NAME} shlwapi)
target_compile_options(${LOADER_NAME} PRIVATE)
generate_export_header(${LOADER_NAME})

View File

@@ -0,0 +1,68 @@
# - Find OpenXR-SDK library
# Find the native OpenXR-SDK includes and library
# This module defines
# OPENXR_SDK_INCLUDE_DIRS, where to find OpenXR-SDK headers, Set when
# OPENXR_SDK_INCLUDE_DIR is found.
# OPENXR_SDK_LIBRARIES, libraries to link against to use OpenXR-SDK.
# OPENXR_SDK_ROOT_DIR, the base directory to search for OpenXR-SDK.
# This can also be an environment variable.
# OPENXR_SDK_FOUND, if false, do not try to use OpenXR-SDK.
#
# also defined, but not for general use are
# OPENXR_LOADER_LIBRARY, where to find the OpenXR-SDK library.
#=============================================================================
# Distributed under the OSI-approved BSD License (the "License");
# see accompanying file Copyright.txt for details.
#
# This software is distributed WITHOUT ANY WARRANTY; without even the
# implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
# See the License for more information.
#=============================================================================
# If OPENXR_SDK_ROOT_DIR was defined in the environment, use it.
IF(NOT OPENXR_SDK_ROOT_DIR AND NOT $ENV{OPENXR_SDK_ROOT_DIR} STREQUAL "")
SET(OPENXR_SDK_ROOT_DIR $ENV{OPENXR_SDK_ROOT_DIR})
ENDIF()
SET(_openxr_sdk_SEARCH_DIRS
${OPENXR_SDK_ROOT_DIR}
/usr/local
/sw # Fink
/opt/local # DarwinPorts
/opt/lib/openxr-sdk
)
FIND_PATH(OPENXR_SDK_INCLUDE_DIR
NAMES
openxr/openxr.h
HINTS
${_openxr_sdk_SEARCH_DIRS}
PATH_SUFFIXES
include
)
FIND_LIBRARY(OPENXR_LOADER_LIBRARY
NAMES
openxr_loader
HINTS
${_openxr_sdk_SEARCH_DIRS}
PATH_SUFFIXES
lib64 lib
)
# handle the QUIETLY and REQUIRED arguments and set OPENXR_SDK_FOUND to TRUE if
# all listed variables are TRUE
INCLUDE(FindPackageHandleStandardArgs)
FIND_PACKAGE_HANDLE_STANDARD_ARGS(OPENXR_SDK DEFAULT_MSG
OPENXR_LOADER_LIBRARY OPENXR_SDK_INCLUDE_DIR)
IF(OPENXR_SDK_FOUND)
SET(OPENXR_SDK_LIBRARIES ${OPENXR_LOADER_LIBRARY})
SET(OPENXR_SDK_INCLUDE_DIRS ${OPENXR_SDK_INCLUDE_DIR})
ENDIF(OPENXR_SDK_FOUND)
MARK_AS_ADVANCED(
OPENXR_SDK_INCLUDE_DIR
OPENXR_LOADER_LIBRARY
)

View File

@@ -0,0 +1,115 @@
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
# Copied right from the OpenXR-SDK (src/cmake/presentation.cmake).
# Don't forget to add the license header above.
set(PRESENTATION_BACKENDS xlib xcb wayland)
set(PRESENTATION_BACKEND xlib CACHE STRING
"Presentation backend chosen at configure time")
set_property(CACHE PRESENTATION_BACKEND PROPERTY STRINGS
${PRESENTATION_BACKENDS})
list(FIND PRESENTATION_BACKENDS ${PRESENTATION_BACKEND} index)
if(index EQUAL -1)
message(FATAL_ERROR "Presentation backend must be one of
${PRESENTATION_BACKENDS}")
endif()
message(STATUS "Using presentation backend: ${PRESENTATION_BACKEND}")
if( PRESENTATION_BACKEND MATCHES "xlib" )
find_package(X11 REQUIRED)
if ((NOT X11_Xxf86vm_LIB) OR (NOT X11_Xrandr_LIB))
message(FATAL_ERROR "OpenXR xlib backend requires Xxf86vm and Xrandr")
endif()
add_definitions( -DSUPPORT_X )
add_definitions( -DOS_LINUX_XLIB )
set( XLIB_LIBRARIES
${X11_LIBRARIES}
${X11_Xxf86vm_LIB}
${X11_Xrandr_LIB} )
elseif( PRESENTATION_BACKEND MATCHES "xcb" )
find_package(PkgConfig REQUIRED)
# XCB + XCB GLX is limited to OpenGL 2.1
# add_definitions( -DOS_LINUX_XCB )
# XCB + Xlib GLX 1.3
add_definitions( -DOS_LINUX_XCB_GLX )
pkg_search_module(X11 REQUIRED x11)
pkg_search_module(XCB REQUIRED xcb)
pkg_search_module(XCB_RANDR REQUIRED xcb-randr)
pkg_search_module(XCB_KEYSYMS REQUIRED xcb-keysyms)
pkg_search_module(XCB_GLX REQUIRED xcb-glx)
pkg_search_module(XCB_DRI2 REQUIRED xcb-dri2)
pkg_search_module(XCB_ICCCM REQUIRED xcb-icccm)
set( XCB_LIBRARIES
${XCB_LIBRARIES}
${XCB_KEYSYMS_LIBRARIES}
${XCB_RANDR_LIBRARIES}
${XCB_GLX_LIBRARIES}
${XCB_DRI2_LIBRARIES}
${X11_LIBRARIES} )
elseif( PRESENTATION_BACKEND MATCHES "wayland" )
find_package(PkgConfig REQUIRED)
pkg_search_module(WAYLAND_CLIENT REQUIRED wayland-client)
pkg_search_module(WAYLAND_EGL REQUIRED wayland-egl)
pkg_search_module(WAYLAND_SCANNER REQUIRED wayland-scanner)
pkg_search_module(WAYLAND_PROTOCOLS REQUIRED wayland-protocols>=1.7)
pkg_search_module(EGL REQUIRED egl)
add_definitions( -DOS_LINUX_WAYLAND )
set( WAYLAND_LIBRARIES
${EGL_LIBRARIES}
${WAYLAND_CLIENT_LIBRARIES}
${WAYLAND_EGL_LIBRARIES} )
# generate wayland protocols
set(WAYLAND_PROTOCOLS_DIR ${CMAKE_SOURCE_DIR}/wayland-protocols/)
file(MAKE_DIRECTORY ${WAYLAND_PROTOCOLS_DIR})
pkg_get_variable(WAYLAND_PROTOCOLS_DATADIR wayland-protocols pkgdatadir)
pkg_get_variable(WAYLAND_SCANNER wayland-scanner wayland_scanner)
set(PROTOCOL xdg-shell-unstable-v6)
set(PROTOCOL_XML
${WAYLAND_PROTOCOLS_DATADIR}/unstable/xdg-shell/${PROTOCOL}.xml)
if( EXISTS ${PROTOCOL_XML} )
execute_process(COMMAND
${WAYLAND_SCANNER}
code
${PROTOCOL_XML}
${WAYLAND_PROTOCOLS_DIR}/${PROTOCOL}.c)
execute_process(COMMAND
${WAYLAND_SCANNER}
client-header
${PROTOCOL_XML}
${WAYLAND_PROTOCOLS_DIR}/${PROTOCOL}.h)
set( WAYLAND_PROTOCOL_SRC
${WAYLAND_PROTOCOLS_DIR}/${PROTOCOL}.c
${WAYLAND_PROTOCOLS_DIR}/${PROTOCOL}.h )
include_directories(${WAYLAND_PROTOCOLS_DIR})
else()
message(FATAL_ERROR
"xdg-shell-unstable-v6.xml not found in "
${WAYLAND_PROTOCOLS_DATADIR}
"\nYour wayland-protocols package does not "
"contain xdg-shell-unstable-v6.")
endif()
endif()

View File

@@ -0,0 +1,50 @@
# Copyright (c) 2017 The Khronos Group Inc.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
# These defines are required as per the OpenXR specification. We can
# just take them from the OpenXR-SDK's src/CMakeLists.txt. Not all of
# them are needed (e.g. XCB and Wayland), but we just copy them anyway.
if(WIN32)
add_definitions(-DXR_OS_WINDOWS)
elseif(CMAKE_SYSTEM_NAME STREQUAL "Linux")
add_definitions(-DXR_OS_LINUX)
endif()
# Determine the presentation backend for Linux systems.
# Use an include because the code is pretty big.
if(CMAKE_SYSTEM_NAME STREQUAL "Linux")
include(presentation)
endif()
# Several files use these compile-time platform switches
if(CMAKE_SYSTEM_NAME STREQUAL "Windows")
add_definitions( -DXR_USE_PLATFORM_WIN32 )
elseif( PRESENTATION_BACKEND MATCHES "xlib" )
add_definitions( -DXR_USE_PLATFORM_XLIB )
elseif( PRESENTATION_BACKEND MATCHES "xcb" )
add_definitions( -DXR_USE_PLATFORM_XCB )
elseif( PRESENTATION_BACKEND MATCHES "wayland" )
add_definitions( -DXR_USE_PLATFORM_WAYLAND )
endif()
add_definitions(-DXR_USE_GRAPHICS_API_OPENGL)
if(CMAKE_SYSTEM_NAME STREQUAL "Windows")
add_definitions(-DXR_USE_GRAPHICS_API_D3D)
add_definitions(-DXR_USE_GRAPHICS_API_D3D10)
add_definitions(-DXR_USE_GRAPHICS_API_D3D11)
add_definitions(-DXR_USE_GRAPHICS_API_D3D12)
endif()

View File

@@ -41,6 +41,7 @@ set(WITH_OPENAL ON CACHE BOOL "" FORCE)
set(WITH_OPENCOLLADA ON CACHE BOOL "" FORCE)
set(WITH_OPENCOLORIO ON CACHE BOOL "" FORCE)
set(WITH_OPENIMAGEDENOISE ON CACHE BOOL "" FORCE)
set(WITH_OPENXR ON CACHE BOOL "" FORCE)
set(WITH_OPENMP ON CACHE BOOL "" FORCE)
set(WITH_OPENSUBDIV ON CACHE BOOL "" FORCE)
set(WITH_OPENVDB ON CACHE BOOL "" FORCE)

View File

@@ -46,6 +46,7 @@ set(WITH_OPENAL OFF CACHE BOOL "" FORCE)
set(WITH_OPENCOLLADA OFF CACHE BOOL "" FORCE)
set(WITH_OPENCOLORIO OFF CACHE BOOL "" FORCE)
set(WITH_OPENIMAGEDENOISE OFF CACHE BOOL "" FORCE)
set(WITH_OPENXR OFF CACHE BOOL "" FORCE)
set(WITH_OPENIMAGEIO OFF CACHE BOOL "" FORCE)
set(WITH_OPENMP OFF CACHE BOOL "" FORCE)
set(WITH_OPENSUBDIV OFF CACHE BOOL "" FORCE)

View File

@@ -42,6 +42,7 @@ set(WITH_OPENAL ON CACHE BOOL "" FORCE)
set(WITH_OPENCOLLADA ON CACHE BOOL "" FORCE)
set(WITH_OPENCOLORIO ON CACHE BOOL "" FORCE)
set(WITH_OPENIMAGEDENOISE ON CACHE BOOL "" FORCE)
set(WITH_OPENXR ON CACHE BOOL "" FORCE)
set(WITH_OPENMP ON CACHE BOOL "" FORCE)
set(WITH_OPENSUBDIV ON CACHE BOOL "" FORCE)
set(WITH_OPENVDB ON CACHE BOOL "" FORCE)

View File

@@ -474,6 +474,9 @@ function(setup_liblinks
if(WITH_OPENSUBDIV)
target_link_libraries(${target} ${OPENSUBDIV_LIBRARIES})
endif()
if(WITH_OPENXR)
target_link_libraries(${target} ${OPENXR_SDK_LIBRARIES})
endif()
if(WITH_CYCLES_EMBREE)
target_link_libraries(${target} ${EMBREE_LIBRARIES})
endif()

View File

@@ -416,6 +416,14 @@ if(WITH_OPENMP)
endif()
endif()
if(WITH_OPENXR)
find_package(OpenXR-SDK)
if(NOT OPENXR_SDK_FOUND)
message(WARNING "OpenXR-SDK was not found, disabling WITH_OPENXR")
set(WITH_OPENXR OFF)
endif()
endif()
set(EXETYPE MACOSX_BUNDLE)
set(CMAKE_C_FLAGS_DEBUG "-fno-strict-aliasing -g")

View File

@@ -416,6 +416,15 @@ if(WITH_OPENSUBDIV)
endif()
endif()
if(WITH_OPENXR)
find_package(OpenXR-SDK)
if(NOT OPENXR_SDK_FOUND)
message(WARNING "OpenXR-SDK was not found, disabling WITH_OPENXR")
set(WITH_OPENXR OFF)
endif()
endif()
# OpenSuse needs lutil, ArchLinux not, for now keep, can avoid by using --as-needed
if(HAIKU)
list(APPEND PLATFORM_LINKLIBS -lnetwork)

View File

@@ -135,8 +135,8 @@ else()
set(CMAKE_C_FLAGS "${CMAKE_C_FLAGS} /nologo /J /Gd /MP /bigobj")
endif()
set(CMAKE_CXX_FLAGS_DEBUG "${CMAKE_CXX_FLAGS_DEBUG} /MTd")
set(CMAKE_C_FLAGS_DEBUG "${CMAKE_C_FLAGS_DEBUG} /MTd")
set(CMAKE_CXX_FLAGS_DEBUG "${CMAKE_CXX_FLAGS_DEBUG} /MTd /ZI")
set(CMAKE_C_FLAGS_DEBUG "${CMAKE_C_FLAGS_DEBUG} /MTd /ZI")
set(CMAKE_CXX_FLAGS_RELEASE "${CMAKE_CXX_FLAGS_RELEASE} /MT")
set(CMAKE_C_FLAGS_RELEASE "${CMAKE_C_FLAGS_RELEASE} /MT")
set(CMAKE_CXX_FLAGS_MINSIZEREL "${CMAKE_CXX_FLAGS_MINSIZEREL} /MT")
@@ -691,3 +691,15 @@ if(WINDOWS_PYTHON_DEBUG)
</Project>")
endif()
endif()
if(WITH_OPENXR)
if(EXISTS ${LIBDIR}/openxr_sdk)
set(OPENXR_SDK ${LIBDIR}/openxr_sdk)
set(OPENXR_SDK_LIBPATH ${LIBDIR}/openxr_sdk/lib)
set(OPENXR_SDK_INCLUDE_DIR ${OPENXR_SDK}/include)
set(OPENXR_SDK_LIBRARIES optimized ${OPENXR_SDK_LIBPATH}/openxr_loader-1_0.lib debug ${OPENXR_SDK_LIBPATH}/openxr_loader-1_0_d.lib)
else()
message(WARNING "OpenXR-SDK was not found, disabling WITH_OPENXR")
set(WITH_OPENXR OFF)
endif()
endif()

View File

@@ -1,7 +1,15 @@
// This file is part of Eigen, a lightweight C++ template library
// for linear algebra.
//
// This Source Code Form is subject to the terms of the Mozilla
// Public License v. 2.0. If a copy of the MPL was not distributed
// with this file, You can obtain one at http://mozilla.org/MPL/2.0/.
#ifndef EIGEN_CHOLESKY_MODULE_H
#define EIGEN_CHOLESKY_MODULE_H
#include "Core"
#include "Jacobi"
#include "src/Core/util/DisableStupidWarnings.h"
@@ -10,20 +18,26 @@
*
*
* This module provides two variants of the Cholesky decomposition for selfadjoint (hermitian) matrices.
* Those decompositions are accessible via the following MatrixBase methods:
* - MatrixBase::llt(),
* Those decompositions are also accessible via the following methods:
* - MatrixBase::llt()
* - MatrixBase::ldlt()
* - SelfAdjointView::llt()
* - SelfAdjointView::ldlt()
*
* \code
* #include <Eigen/Cholesky>
* \endcode
*/
#include "src/misc/Solve.h"
#include "src/Cholesky/LLT.h"
#include "src/Cholesky/LDLT.h"
#ifdef EIGEN_USE_LAPACKE
#include "src/Cholesky/LLT_MKL.h"
#ifdef EIGEN_USE_MKL
#include "mkl_lapacke.h"
#else
#include "src/misc/lapacke.h"
#endif
#include "src/Cholesky/LLT_LAPACKE.h"
#endif
#include "src/Core/util/ReenableStupidWarnings.h"

View File

@@ -1,3 +1,10 @@
// This file is part of Eigen, a lightweight C++ template library
// for linear algebra.
//
// This Source Code Form is subject to the terms of the Mozilla
// Public License v. 2.0. If a copy of the MPL was not distributed
// with this file, You can obtain one at http://mozilla.org/MPL/2.0/.
#ifndef EIGEN_CHOLMODSUPPORT_MODULE_H
#define EIGEN_CHOLMODSUPPORT_MODULE_H
@@ -12,7 +19,7 @@ extern "C" {
/** \ingroup Support_modules
* \defgroup CholmodSupport_Module CholmodSupport module
*
* This module provides an interface to the Cholmod library which is part of the <a href="http://www.cise.ufl.edu/research/sparse/SuiteSparse/">suitesparse</a> package.
* This module provides an interface to the Cholmod library which is part of the <a href="http://www.suitesparse.com">suitesparse</a> package.
* It provides the two following main factorization classes:
* - class CholmodSupernodalLLT: a supernodal LLT Cholesky factorization.
* - class CholmodDecomposiiton: a general L(D)LT Cholesky factorization with automatic or explicit runtime selection of the underlying factorization method (supernodal or simplicial).
@@ -33,12 +40,8 @@ extern "C" {
*
*/
#include "src/misc/Solve.h"
#include "src/misc/SparseSolve.h"
#include "src/CholmodSupport/CholmodSupport.h"
#include "src/Core/util/ReenableStupidWarnings.h"
#endif // EIGEN_CHOLMODSUPPORT_MODULE_H

View File

@@ -14,6 +14,74 @@
// first thing Eigen does: stop the compiler from committing suicide
#include "src/Core/util/DisableStupidWarnings.h"
#if defined(__CUDACC__) && !defined(EIGEN_NO_CUDA)
#define EIGEN_CUDACC __CUDACC__
#endif
#if defined(__CUDA_ARCH__) && !defined(EIGEN_NO_CUDA)
#define EIGEN_CUDA_ARCH __CUDA_ARCH__
#endif
#if defined(__CUDACC_VER_MAJOR__) && (__CUDACC_VER_MAJOR__ >= 9)
#define EIGEN_CUDACC_VER ((__CUDACC_VER_MAJOR__ * 10000) + (__CUDACC_VER_MINOR__ * 100))
#elif defined(__CUDACC_VER__)
#define EIGEN_CUDACC_VER __CUDACC_VER__
#else
#define EIGEN_CUDACC_VER 0
#endif
// Handle NVCC/CUDA/SYCL
#if defined(__CUDACC__) || defined(__SYCL_DEVICE_ONLY__)
// Do not try asserts on CUDA and SYCL!
#ifndef EIGEN_NO_DEBUG
#define EIGEN_NO_DEBUG
#endif
#ifdef EIGEN_INTERNAL_DEBUGGING
#undef EIGEN_INTERNAL_DEBUGGING
#endif
#ifdef EIGEN_EXCEPTIONS
#undef EIGEN_EXCEPTIONS
#endif
// All functions callable from CUDA code must be qualified with __device__
#ifdef __CUDACC__
// Do not try to vectorize on CUDA and SYCL!
#ifndef EIGEN_DONT_VECTORIZE
#define EIGEN_DONT_VECTORIZE
#endif
#define EIGEN_DEVICE_FUNC __host__ __device__
// We need cuda_runtime.h to ensure that that EIGEN_USING_STD_MATH macro
// works properly on the device side
#include <cuda_runtime.h>
#else
#define EIGEN_DEVICE_FUNC
#endif
#else
#define EIGEN_DEVICE_FUNC
#endif
// When compiling CUDA device code with NVCC, pull in math functions from the
// global namespace. In host mode, and when device doee with clang, use the
// std versions.
#if defined(__CUDA_ARCH__) && defined(__NVCC__)
#define EIGEN_USING_STD_MATH(FUNC) using ::FUNC;
#else
#define EIGEN_USING_STD_MATH(FUNC) using std::FUNC;
#endif
#if (defined(_CPPUNWIND) || defined(__EXCEPTIONS)) && !defined(__CUDA_ARCH__) && !defined(EIGEN_EXCEPTIONS) && !defined(EIGEN_USE_SYCL)
#define EIGEN_EXCEPTIONS
#endif
#ifdef EIGEN_EXCEPTIONS
#include <new>
#endif
// then include this file where all our macros are defined. It's really important to do it first because
// it's where we do all the alignment settings (platform detection and honoring the user's will if he
// defined e.g. EIGEN_DONT_ALIGN) so it needs to be done before we do anything with vectorization.
@@ -21,7 +89,7 @@
// Disable the ipa-cp-clone optimization flag with MinGW 6.x or newer (enabled by default with -O3)
// See http://eigen.tuxfamily.org/bz/show_bug.cgi?id=556 for details.
#if defined(__MINGW32__) && EIGEN_GNUC_AT_LEAST(4,6)
#if EIGEN_COMP_MINGW && EIGEN_GNUC_AT_LEAST(4,6)
#pragma GCC optimize ("-fno-ipa-cp-clone")
#endif
@@ -31,26 +99,26 @@
// and inclusion of their respective header files
#include "src/Core/util/MKL_support.h"
// if alignment is disabled, then disable vectorization. Note: EIGEN_ALIGN is the proper check, it takes into
// account both the user's will (EIGEN_DONT_ALIGN) and our own platform checks
#if !EIGEN_ALIGN
// if alignment is disabled, then disable vectorization. Note: EIGEN_MAX_ALIGN_BYTES is the proper check, it takes into
// account both the user's will (EIGEN_MAX_ALIGN_BYTES,EIGEN_DONT_ALIGN) and our own platform checks
#if EIGEN_MAX_ALIGN_BYTES==0
#ifndef EIGEN_DONT_VECTORIZE
#define EIGEN_DONT_VECTORIZE
#endif
#endif
#ifdef _MSC_VER
#if EIGEN_COMP_MSVC
#include <malloc.h> // for _aligned_malloc -- need it regardless of whether vectorization is enabled
#if (_MSC_VER >= 1500) // 2008 or later
#if (EIGEN_COMP_MSVC >= 1500) // 2008 or later
// Remember that usage of defined() in a #define is undefined by the standard.
// a user reported that in 64-bit mode, MSVC doesn't care to define _M_IX86_FP.
#if (defined(_M_IX86_FP) && (_M_IX86_FP >= 2)) || defined(_M_X64)
#if (defined(_M_IX86_FP) && (_M_IX86_FP >= 2)) || EIGEN_ARCH_x86_64
#define EIGEN_SSE2_ON_MSVC_2008_OR_LATER
#endif
#endif
#else
// Remember that usage of defined() in a #define is undefined by the standard
#if (defined __SSE2__) && ( (!defined __GNUC__) || (defined __INTEL_COMPILER) || EIGEN_GNUC_AT_LEAST(4,2) )
#if (defined __SSE2__) && ( (!EIGEN_COMP_GNUC) || EIGEN_COMP_ICC || EIGEN_GNUC_AT_LEAST(4,2) )
#define EIGEN_SSE2_ON_NON_MSVC_BUT_NOT_OLD_GCC
#endif
#endif
@@ -82,6 +150,31 @@
#ifdef __SSE4_2__
#define EIGEN_VECTORIZE_SSE4_2
#endif
#ifdef __AVX__
#define EIGEN_VECTORIZE_AVX
#define EIGEN_VECTORIZE_SSE3
#define EIGEN_VECTORIZE_SSSE3
#define EIGEN_VECTORIZE_SSE4_1
#define EIGEN_VECTORIZE_SSE4_2
#endif
#ifdef __AVX2__
#define EIGEN_VECTORIZE_AVX2
#endif
#ifdef __FMA__
#define EIGEN_VECTORIZE_FMA
#endif
#if defined(__AVX512F__) && defined(EIGEN_ENABLE_AVX512)
#define EIGEN_VECTORIZE_AVX512
#define EIGEN_VECTORIZE_AVX2
#define EIGEN_VECTORIZE_AVX
#define EIGEN_VECTORIZE_FMA
#ifdef __AVX512DQ__
#define EIGEN_VECTORIZE_AVX512DQ
#endif
#ifdef __AVX512ER__
#define EIGEN_VECTORIZE_AVX512ER
#endif
#endif
// include files
@@ -95,9 +188,10 @@
extern "C" {
// In theory we should only include immintrin.h and not the other *mmintrin.h header files directly.
// Doing so triggers some issues with ICC. However old gcc versions seems to not have this file, thus:
#if defined(__INTEL_COMPILER) && __INTEL_COMPILER >= 1110
#if EIGEN_COMP_ICC >= 1110
#include <immintrin.h>
#else
#include <mmintrin.h>
#include <emmintrin.h>
#include <xmmintrin.h>
#ifdef EIGEN_VECTORIZE_SSE3
@@ -112,8 +206,20 @@
#ifdef EIGEN_VECTORIZE_SSE4_2
#include <nmmintrin.h>
#endif
#if defined(EIGEN_VECTORIZE_AVX) || defined(EIGEN_VECTORIZE_AVX512)
#include <immintrin.h>
#endif
#endif
} // end extern "C"
#elif defined __VSX__
#define EIGEN_VECTORIZE
#define EIGEN_VECTORIZE_VSX
#include <altivec.h>
// We need to #undef all these ugly tokens defined in <altivec.h>
// => use __vector instead of vector
#undef bool
#undef vector
#undef pixel
#elif defined __ALTIVEC__
#define EIGEN_VECTORIZE
#define EIGEN_VECTORIZE_ALTIVEC
@@ -123,13 +229,35 @@
#undef bool
#undef vector
#undef pixel
#elif defined __ARM_NEON
#elif (defined __ARM_NEON) || (defined __ARM_NEON__)
#define EIGEN_VECTORIZE
#define EIGEN_VECTORIZE_NEON
#include <arm_neon.h>
#elif (defined __s390x__ && defined __VEC__)
#define EIGEN_VECTORIZE
#define EIGEN_VECTORIZE_ZVECTOR
#include <vecintrin.h>
#endif
#endif
#if defined(__F16C__) && !defined(EIGEN_COMP_CLANG)
// We can use the optimized fp16 to float and float to fp16 conversion routines
#define EIGEN_HAS_FP16_C
#endif
#if defined __CUDACC__
#define EIGEN_VECTORIZE_CUDA
#include <vector_types.h>
#if EIGEN_CUDACC_VER >= 70500
#define EIGEN_HAS_CUDA_FP16
#endif
#endif
#if defined EIGEN_HAS_CUDA_FP16
#include <host_defines.h>
#include <cuda_fp16.h>
#endif
#if (defined _OPENMP) && (!defined EIGEN_DONT_PARALLELIZE)
#define EIGEN_HAS_OPENMP
#endif
@@ -139,7 +267,7 @@
#endif
// MSVC for windows mobile does not have the errno.h file
#if !(defined(_MSC_VER) && defined(_WIN32_WCE)) && !defined(__ARMCC_VERSION)
#if !(EIGEN_COMP_MSVC && EIGEN_OS_WINCE) && !EIGEN_COMP_ARM
#define EIGEN_HAS_ERRNO
#endif
@@ -159,29 +287,30 @@
// for min/max:
#include <algorithm>
// for std::is_nothrow_move_assignable
#ifdef EIGEN_INCLUDE_TYPE_TRAITS
#include <type_traits>
#endif
// for outputting debug info
#ifdef EIGEN_DEBUG_ASSIGN
#include <iostream>
#endif
// required for __cpuid, needs to be included after cmath
#if defined(_MSC_VER) && (defined(_M_IX86)||defined(_M_X64)) && (!defined(_WIN32_WCE))
#if EIGEN_COMP_MSVC && EIGEN_ARCH_i386_OR_x86_64 && !EIGEN_OS_WINCE
#include <intrin.h>
#endif
#if defined(_CPPUNWIND) || defined(__EXCEPTIONS)
#define EIGEN_EXCEPTIONS
#endif
#ifdef EIGEN_EXCEPTIONS
#include <new>
#endif
/** \brief Namespace containing all symbols from the %Eigen library. */
namespace Eigen {
inline static const char *SimdInstructionSetsInUse(void) {
#if defined(EIGEN_VECTORIZE_SSE4_2)
#if defined(EIGEN_VECTORIZE_AVX512)
return "AVX512, FMA, AVX2, AVX, SSE, SSE2, SSE3, SSSE3, SSE4.1, SSE4.2";
#elif defined(EIGEN_VECTORIZE_AVX)
return "AVX SSE, SSE2, SSE3, SSSE3, SSE4.1, SSE4.2";
#elif defined(EIGEN_VECTORIZE_SSE4_2)
return "SSE, SSE2, SSE3, SSSE3, SSE4.1, SSE4.2";
#elif defined(EIGEN_VECTORIZE_SSE4_1)
return "SSE, SSE2, SSE3, SSSE3, SSE4.1";
@@ -193,8 +322,12 @@ inline static const char *SimdInstructionSetsInUse(void) {
return "SSE, SSE2";
#elif defined(EIGEN_VECTORIZE_ALTIVEC)
return "AltiVec";
#elif defined(EIGEN_VECTORIZE_VSX)
return "VSX";
#elif defined(EIGEN_VECTORIZE_NEON)
return "ARM NEON";
#elif defined(EIGEN_VECTORIZE_ZVECTOR)
return "S390X ZVECTOR";
#else
return "None";
#endif
@@ -202,42 +335,21 @@ inline static const char *SimdInstructionSetsInUse(void) {
} // end namespace Eigen
#define STAGE10_FULL_EIGEN2_API 10
#define STAGE20_RESOLVE_API_CONFLICTS 20
#define STAGE30_FULL_EIGEN3_API 30
#define STAGE40_FULL_EIGEN3_STRICTNESS 40
#define STAGE99_NO_EIGEN2_SUPPORT 99
#if defined EIGEN2_SUPPORT_STAGE40_FULL_EIGEN3_STRICTNESS
#define EIGEN2_SUPPORT
#define EIGEN2_SUPPORT_STAGE STAGE40_FULL_EIGEN3_STRICTNESS
#elif defined EIGEN2_SUPPORT_STAGE30_FULL_EIGEN3_API
#define EIGEN2_SUPPORT
#define EIGEN2_SUPPORT_STAGE STAGE30_FULL_EIGEN3_API
#elif defined EIGEN2_SUPPORT_STAGE20_RESOLVE_API_CONFLICTS
#define EIGEN2_SUPPORT
#define EIGEN2_SUPPORT_STAGE STAGE20_RESOLVE_API_CONFLICTS
#elif defined EIGEN2_SUPPORT_STAGE10_FULL_EIGEN2_API
#define EIGEN2_SUPPORT
#define EIGEN2_SUPPORT_STAGE STAGE10_FULL_EIGEN2_API
#elif defined EIGEN2_SUPPORT
// default to stage 3, that's what it's always meant
#define EIGEN2_SUPPORT_STAGE30_FULL_EIGEN3_API
#define EIGEN2_SUPPORT_STAGE STAGE30_FULL_EIGEN3_API
#else
#define EIGEN2_SUPPORT_STAGE STAGE99_NO_EIGEN2_SUPPORT
#if defined EIGEN2_SUPPORT_STAGE40_FULL_EIGEN3_STRICTNESS || defined EIGEN2_SUPPORT_STAGE30_FULL_EIGEN3_API || defined EIGEN2_SUPPORT_STAGE20_RESOLVE_API_CONFLICTS || defined EIGEN2_SUPPORT_STAGE10_FULL_EIGEN2_API || defined EIGEN2_SUPPORT
// This will generate an error message:
#error Eigen2-support is only available up to version 3.2. Please go to "http://eigen.tuxfamily.org/index.php?title=Eigen2" for further information
#endif
#ifdef EIGEN2_SUPPORT
#undef minor
#endif
namespace Eigen {
// we use size_t frequently and we'll never remember to prepend it with std:: everytime just to
// ensure QNX/QCC support
using std::size_t;
// gcc 4.6.0 wants std:: for ptrdiff_t
// gcc 4.6.0 wants std:: for ptrdiff_t
using std::ptrdiff_t;
}
/** \defgroup Core_Module Core module
* This is the main module of Eigen providing dense matrix and vector support
* (both fixed and dynamic size) with all the features corresponding to a BLAS library
@@ -249,8 +361,8 @@ using std::ptrdiff_t;
*/
#include "src/Core/util/Constants.h"
#include "src/Core/util/ForwardDeclarations.h"
#include "src/Core/util/Meta.h"
#include "src/Core/util/ForwardDeclarations.h"
#include "src/Core/util/StaticAssert.h"
#include "src/Core/util/XprHelper.h"
#include "src/Core/util/Memory.h"
@@ -258,41 +370,94 @@ using std::ptrdiff_t;
#include "src/Core/NumTraits.h"
#include "src/Core/MathFunctions.h"
#include "src/Core/GenericPacketMath.h"
#include "src/Core/MathFunctionsImpl.h"
#include "src/Core/arch/Default/ConjHelper.h"
#if defined EIGEN_VECTORIZE_SSE
#if defined EIGEN_VECTORIZE_AVX512
#include "src/Core/arch/SSE/PacketMath.h"
#include "src/Core/arch/AVX/PacketMath.h"
#include "src/Core/arch/AVX512/PacketMath.h"
#include "src/Core/arch/AVX512/MathFunctions.h"
#elif defined EIGEN_VECTORIZE_AVX
// Use AVX for floats and doubles, SSE for integers
#include "src/Core/arch/SSE/PacketMath.h"
#include "src/Core/arch/SSE/Complex.h"
#include "src/Core/arch/SSE/MathFunctions.h"
#include "src/Core/arch/AVX/PacketMath.h"
#include "src/Core/arch/AVX/MathFunctions.h"
#include "src/Core/arch/AVX/Complex.h"
#include "src/Core/arch/AVX/TypeCasting.h"
#include "src/Core/arch/SSE/TypeCasting.h"
#elif defined EIGEN_VECTORIZE_SSE
#include "src/Core/arch/SSE/PacketMath.h"
#include "src/Core/arch/SSE/MathFunctions.h"
#include "src/Core/arch/SSE/Complex.h"
#elif defined EIGEN_VECTORIZE_ALTIVEC
#include "src/Core/arch/SSE/TypeCasting.h"
#elif defined(EIGEN_VECTORIZE_ALTIVEC) || defined(EIGEN_VECTORIZE_VSX)
#include "src/Core/arch/AltiVec/PacketMath.h"
#include "src/Core/arch/AltiVec/MathFunctions.h"
#include "src/Core/arch/AltiVec/Complex.h"
#elif defined EIGEN_VECTORIZE_NEON
#include "src/Core/arch/NEON/PacketMath.h"
#include "src/Core/arch/NEON/MathFunctions.h"
#include "src/Core/arch/NEON/Complex.h"
#elif defined EIGEN_VECTORIZE_ZVECTOR
#include "src/Core/arch/ZVector/PacketMath.h"
#include "src/Core/arch/ZVector/MathFunctions.h"
#include "src/Core/arch/ZVector/Complex.h"
#endif
// Half float support
#include "src/Core/arch/CUDA/Half.h"
#include "src/Core/arch/CUDA/PacketMathHalf.h"
#include "src/Core/arch/CUDA/TypeCasting.h"
#if defined EIGEN_VECTORIZE_CUDA
#include "src/Core/arch/CUDA/PacketMath.h"
#include "src/Core/arch/CUDA/MathFunctions.h"
#endif
#include "src/Core/arch/Default/Settings.h"
#include "src/Core/Functors.h"
#include "src/Core/functors/TernaryFunctors.h"
#include "src/Core/functors/BinaryFunctors.h"
#include "src/Core/functors/UnaryFunctors.h"
#include "src/Core/functors/NullaryFunctors.h"
#include "src/Core/functors/StlFunctors.h"
#include "src/Core/functors/AssignmentFunctors.h"
// Specialized functors to enable the processing of complex numbers
// on CUDA devices
#include "src/Core/arch/CUDA/Complex.h"
#include "src/Core/IO.h"
#include "src/Core/DenseCoeffsBase.h"
#include "src/Core/DenseBase.h"
#include "src/Core/MatrixBase.h"
#include "src/Core/EigenBase.h"
#include "src/Core/Product.h"
#include "src/Core/CoreEvaluators.h"
#include "src/Core/AssignEvaluator.h"
#ifndef EIGEN_PARSED_BY_DOXYGEN // work around Doxygen bug triggered by Assign.h r814874
// at least confirmed with Doxygen 1.5.5 and 1.5.6
#include "src/Core/Assign.h"
#endif
#include "src/Core/ArrayBase.h"
#include "src/Core/util/BlasUtil.h"
#include "src/Core/DenseStorage.h"
#include "src/Core/NestByValue.h"
#include "src/Core/ForceAlignedAccess.h"
// #include "src/Core/ForceAlignedAccess.h"
#include "src/Core/ReturnByValue.h"
#include "src/Core/NoAlias.h"
#include "src/Core/PlainObjectBase.h"
#include "src/Core/Matrix.h"
#include "src/Core/Array.h"
#include "src/Core/CwiseTernaryOp.h"
#include "src/Core/CwiseBinaryOp.h"
#include "src/Core/CwiseUnaryOp.h"
#include "src/Core/CwiseNullaryOp.h"
@@ -300,32 +465,32 @@ using std::ptrdiff_t;
#include "src/Core/SelfCwiseBinaryOp.h"
#include "src/Core/Dot.h"
#include "src/Core/StableNorm.h"
#include "src/Core/MapBase.h"
#include "src/Core/Stride.h"
#include "src/Core/MapBase.h"
#include "src/Core/Map.h"
#include "src/Core/Ref.h"
#include "src/Core/Block.h"
#include "src/Core/VectorBlock.h"
#include "src/Core/Ref.h"
#include "src/Core/Transpose.h"
#include "src/Core/DiagonalMatrix.h"
#include "src/Core/Diagonal.h"
#include "src/Core/DiagonalProduct.h"
#include "src/Core/PermutationMatrix.h"
#include "src/Core/Transpositions.h"
#include "src/Core/Redux.h"
#include "src/Core/Visitor.h"
#include "src/Core/Fuzzy.h"
#include "src/Core/IO.h"
#include "src/Core/Swap.h"
#include "src/Core/CommaInitializer.h"
#include "src/Core/Flagged.h"
#include "src/Core/ProductBase.h"
#include "src/Core/GeneralProduct.h"
#include "src/Core/Solve.h"
#include "src/Core/Inverse.h"
#include "src/Core/SolverBase.h"
#include "src/Core/PermutationMatrix.h"
#include "src/Core/Transpositions.h"
#include "src/Core/TriangularMatrix.h"
#include "src/Core/SelfAdjointView.h"
#include "src/Core/products/GeneralBlockPanelKernel.h"
#include "src/Core/products/Parallelizer.h"
#include "src/Core/products/CoeffBasedProduct.h"
#include "src/Core/ProductEvaluators.h"
#include "src/Core/products/GeneralMatrixVector.h"
#include "src/Core/products/GeneralMatrixMatrix.h"
#include "src/Core/SolveTriangular.h"
@@ -340,6 +505,7 @@ using std::ptrdiff_t;
#include "src/Core/products/TriangularSolverVector.h"
#include "src/Core/BandMatrix.h"
#include "src/Core/CoreIterators.h"
#include "src/Core/ConditionEstimator.h"
#include "src/Core/BooleanRedux.h"
#include "src/Core/Select.h"
@@ -347,18 +513,17 @@ using std::ptrdiff_t;
#include "src/Core/Random.h"
#include "src/Core/Replicate.h"
#include "src/Core/Reverse.h"
#include "src/Core/ArrayBase.h"
#include "src/Core/ArrayWrapper.h"
#ifdef EIGEN_USE_BLAS
#include "src/Core/products/GeneralMatrixMatrix_MKL.h"
#include "src/Core/products/GeneralMatrixVector_MKL.h"
#include "src/Core/products/GeneralMatrixMatrixTriangular_MKL.h"
#include "src/Core/products/SelfadjointMatrixMatrix_MKL.h"
#include "src/Core/products/SelfadjointMatrixVector_MKL.h"
#include "src/Core/products/TriangularMatrixMatrix_MKL.h"
#include "src/Core/products/TriangularMatrixVector_MKL.h"
#include "src/Core/products/TriangularSolverMatrix_MKL.h"
#include "src/Core/products/GeneralMatrixMatrix_BLAS.h"
#include "src/Core/products/GeneralMatrixVector_BLAS.h"
#include "src/Core/products/GeneralMatrixMatrixTriangular_BLAS.h"
#include "src/Core/products/SelfadjointMatrixMatrix_BLAS.h"
#include "src/Core/products/SelfadjointMatrixVector_BLAS.h"
#include "src/Core/products/TriangularMatrixMatrix_BLAS.h"
#include "src/Core/products/TriangularMatrixVector_BLAS.h"
#include "src/Core/products/TriangularSolverMatrix_BLAS.h"
#endif // EIGEN_USE_BLAS
#ifdef EIGEN_USE_MKL_VML
@@ -369,8 +534,4 @@ using std::ptrdiff_t;
#include "src/Core/util/ReenableStupidWarnings.h"
#ifdef EIGEN2_SUPPORT
#include "Eigen2Support"
#endif
#endif // EIGEN_CORE_H

View File

@@ -1,2 +1,2 @@
#include "Dense"
//#include "Sparse"
#include "Sparse"

View File

@@ -1,3 +1,10 @@
// This file is part of Eigen, a lightweight C++ template library
// for linear algebra.
//
// This Source Code Form is subject to the terms of the Mozilla
// Public License v. 2.0. If a copy of the MPL was not distributed
// with this file, You can obtain one at http://mozilla.org/MPL/2.0/.
#ifndef EIGEN_EIGENVALUES_MODULE_H
#define EIGEN_EIGENVALUES_MODULE_H
@@ -25,6 +32,7 @@
* \endcode
*/
#include "src/misc/RealSvd2x2.h"
#include "src/Eigenvalues/Tridiagonalization.h"
#include "src/Eigenvalues/RealSchur.h"
#include "src/Eigenvalues/EigenSolver.h"
@@ -37,9 +45,14 @@
#include "src/Eigenvalues/GeneralizedEigenSolver.h"
#include "src/Eigenvalues/MatrixBaseEigenvalues.h"
#ifdef EIGEN_USE_LAPACKE
#include "src/Eigenvalues/RealSchur_MKL.h"
#include "src/Eigenvalues/ComplexSchur_MKL.h"
#include "src/Eigenvalues/SelfAdjointEigenSolver_MKL.h"
#ifdef EIGEN_USE_MKL
#include "mkl_lapacke.h"
#else
#include "src/misc/lapacke.h"
#endif
#include "src/Eigenvalues/RealSchur_LAPACKE.h"
#include "src/Eigenvalues/ComplexSchur_LAPACKE.h"
#include "src/Eigenvalues/SelfAdjointEigenSolver_LAPACKE.h"
#endif
#include "src/Core/util/ReenableStupidWarnings.h"

View File

@@ -1,3 +1,10 @@
// This file is part of Eigen, a lightweight C++ template library
// for linear algebra.
//
// This Source Code Form is subject to the terms of the Mozilla
// Public License v. 2.0. If a copy of the MPL was not distributed
// with this file, You can obtain one at http://mozilla.org/MPL/2.0/.
#ifndef EIGEN_GEOMETRY_MODULE_H
#define EIGEN_GEOMETRY_MODULE_H
@@ -9,21 +16,17 @@
#include "LU"
#include <limits>
#ifndef M_PI
#define M_PI 3.14159265358979323846
#endif
/** \defgroup Geometry_Module Geometry module
*
*
*
* This module provides support for:
* - fixed-size homogeneous transformations
* - translation, scaling, 2D and 3D rotations
* - quaternions
* - \ref MatrixBase::cross() "cross product"
* - \ref MatrixBase::unitOrthogonal() "orthognal vector generation"
* - some linear components: parametrized-lines and hyperplanes
* - \link Quaternion quaternions \endlink
* - cross products (\ref MatrixBase::cross, \ref MatrixBase::cross3)
* - orthognal vector generation (\ref MatrixBase::unitOrthogonal)
* - some linear components: \link ParametrizedLine parametrized-lines \endlink and \link Hyperplane hyperplanes \endlink
* - \link AlignedBox axis aligned bounding boxes \endlink
* - \link umeyama least-square transformation fitting \endlink
*
* \code
* #include <Eigen/Geometry>
@@ -33,27 +36,23 @@
#include "src/Geometry/OrthoMethods.h"
#include "src/Geometry/EulerAngles.h"
#if EIGEN2_SUPPORT_STAGE > STAGE20_RESOLVE_API_CONFLICTS
#include "src/Geometry/Homogeneous.h"
#include "src/Geometry/RotationBase.h"
#include "src/Geometry/Rotation2D.h"
#include "src/Geometry/Quaternion.h"
#include "src/Geometry/AngleAxis.h"
#include "src/Geometry/Transform.h"
#include "src/Geometry/Translation.h"
#include "src/Geometry/Scaling.h"
#include "src/Geometry/Hyperplane.h"
#include "src/Geometry/ParametrizedLine.h"
#include "src/Geometry/AlignedBox.h"
#include "src/Geometry/Umeyama.h"
#include "src/Geometry/Homogeneous.h"
#include "src/Geometry/RotationBase.h"
#include "src/Geometry/Rotation2D.h"
#include "src/Geometry/Quaternion.h"
#include "src/Geometry/AngleAxis.h"
#include "src/Geometry/Transform.h"
#include "src/Geometry/Translation.h"
#include "src/Geometry/Scaling.h"
#include "src/Geometry/Hyperplane.h"
#include "src/Geometry/ParametrizedLine.h"
#include "src/Geometry/AlignedBox.h"
#include "src/Geometry/Umeyama.h"
#if defined EIGEN_VECTORIZE_SSE
#include "src/Geometry/arch/Geometry_SSE.h"
#endif
#endif
#ifdef EIGEN2_SUPPORT
#include "src/Eigen2Support/Geometry/All.h"
// Use the SSE optimized version whenever possible. At the moment the
// SSE version doesn't compile when AVX is enabled
#if defined EIGEN_VECTORIZE_SSE && !defined EIGEN_VECTORIZE_AVX
#include "src/Geometry/arch/Geometry_SSE.h"
#endif
#include "src/Core/util/ReenableStupidWarnings.h"

View File

@@ -1,3 +1,10 @@
// This file is part of Eigen, a lightweight C++ template library
// for linear algebra.
//
// This Source Code Form is subject to the terms of the Mozilla
// Public License v. 2.0. If a copy of the MPL was not distributed
// with this file, You can obtain one at http://mozilla.org/MPL/2.0/.
#ifndef EIGEN_HOUSEHOLDER_MODULE_H
#define EIGEN_HOUSEHOLDER_MODULE_H

View File

@@ -1,3 +1,10 @@
// This file is part of Eigen, a lightweight C++ template library
// for linear algebra.
//
// This Source Code Form is subject to the terms of the Mozilla
// Public License v. 2.0. If a copy of the MPL was not distributed
// with this file, You can obtain one at http://mozilla.org/MPL/2.0/.
#ifndef EIGEN_ITERATIVELINEARSOLVERS_MODULE_H
#define EIGEN_ITERATIVELINEARSOLVERS_MODULE_H
@@ -12,28 +19,29 @@
* This module currently provides iterative methods to solve problems of the form \c A \c x = \c b, where \c A is a squared matrix, usually very large and sparse.
* Those solvers are accessible via the following classes:
* - ConjugateGradient for selfadjoint (hermitian) matrices,
* - LeastSquaresConjugateGradient for rectangular least-square problems,
* - BiCGSTAB for general square matrices.
*
* These iterative solvers are associated with some preconditioners:
* - IdentityPreconditioner - not really useful
* - DiagonalPreconditioner - also called JAcobi preconditioner, work very well on diagonal dominant matrices.
* - IncompleteILUT - incomplete LU factorization with dual thresholding
* - DiagonalPreconditioner - also called Jacobi preconditioner, work very well on diagonal dominant matrices.
* - IncompleteLUT - incomplete LU factorization with dual thresholding
*
* Such problems can also be solved using the direct sparse decomposition modules: SparseCholesky, CholmodSupport, UmfPackSupport, SuperLUSupport.
*
* \code
* #include <Eigen/IterativeLinearSolvers>
* \endcode
\code
#include <Eigen/IterativeLinearSolvers>
\endcode
*/
#include "src/misc/Solve.h"
#include "src/misc/SparseSolve.h"
#include "src/IterativeLinearSolvers/SolveWithGuess.h"
#include "src/IterativeLinearSolvers/IterativeSolverBase.h"
#include "src/IterativeLinearSolvers/BasicPreconditioners.h"
#include "src/IterativeLinearSolvers/ConjugateGradient.h"
#include "src/IterativeLinearSolvers/LeastSquareConjugateGradient.h"
#include "src/IterativeLinearSolvers/BiCGSTAB.h"
#include "src/IterativeLinearSolvers/IncompleteLUT.h"
#include "src/IterativeLinearSolvers/IncompleteCholesky.h"
#include "src/Core/util/ReenableStupidWarnings.h"

View File

@@ -1,3 +1,10 @@
// This file is part of Eigen, a lightweight C++ template library
// for linear algebra.
//
// This Source Code Form is subject to the terms of the Mozilla
// Public License v. 2.0. If a copy of the MPL was not distributed
// with this file, You can obtain one at http://mozilla.org/MPL/2.0/.
#ifndef EIGEN_JACOBI_MODULE_H
#define EIGEN_JACOBI_MODULE_H

View File

@@ -1,3 +1,10 @@
// This file is part of Eigen, a lightweight C++ template library
// for linear algebra.
//
// This Source Code Form is subject to the terms of the Mozilla
// Public License v. 2.0. If a copy of the MPL was not distributed
// with this file, You can obtain one at http://mozilla.org/MPL/2.0/.
#ifndef EIGEN_LU_MODULE_H
#define EIGEN_LU_MODULE_H
@@ -16,25 +23,27 @@
* \endcode
*/
#include "src/misc/Solve.h"
#include "src/misc/Kernel.h"
#include "src/misc/Image.h"
#include "src/LU/FullPivLU.h"
#include "src/LU/PartialPivLU.h"
#ifdef EIGEN_USE_LAPACKE
#include "src/LU/PartialPivLU_MKL.h"
#ifdef EIGEN_USE_MKL
#include "mkl_lapacke.h"
#else
#include "src/misc/lapacke.h"
#endif
#include "src/LU/PartialPivLU_LAPACKE.h"
#endif
#include "src/LU/Determinant.h"
#include "src/LU/Inverse.h"
#include "src/LU/InverseImpl.h"
#if defined EIGEN_VECTORIZE_SSE
// Use the SSE optimized version whenever possible. At the moment the
// SSE version doesn't compile when AVX is enabled
#if defined EIGEN_VECTORIZE_SSE && !defined EIGEN_VECTORIZE_AVX
#include "src/LU/arch/Inverse_SSE.h"
#endif
#ifdef EIGEN2_SUPPORT
#include "src/Eigen2Support/LU.h"
#endif
#include "src/Core/util/ReenableStupidWarnings.h"
#endif // EIGEN_LU_MODULE_H

View File

@@ -1,3 +1,10 @@
// This file is part of Eigen, a lightweight C++ template library
// for linear algebra.
//
// This Source Code Form is subject to the terms of the Mozilla
// Public License v. 2.0. If a copy of the MPL was not distributed
// with this file, You can obtain one at http://mozilla.org/MPL/2.0/.
#ifndef EIGEN_METISSUPPORT_MODULE_H
#define EIGEN_METISSUPPORT_MODULE_H

View File

@@ -1,3 +1,10 @@
// This file is part of Eigen, a lightweight C++ template library
// for linear algebra.
//
// This Source Code Form is subject to the terms of the Mozilla
// Public License v. 2.0. If a copy of the MPL was not distributed
// with this file, You can obtain one at http://mozilla.org/MPL/2.0/.
#ifndef EIGEN_ORDERINGMETHODS_MODULE_H
#define EIGEN_ORDERINGMETHODS_MODULE_H

View File

@@ -1,3 +1,10 @@
// This file is part of Eigen, a lightweight C++ template library
// for linear algebra.
//
// This Source Code Form is subject to the terms of the Mozilla
// Public License v. 2.0. If a copy of the MPL was not distributed
// with this file, You can obtain one at http://mozilla.org/MPL/2.0/.
#ifndef EIGEN_PASTIXSUPPORT_MODULE_H
#define EIGEN_PASTIXSUPPORT_MODULE_H
@@ -5,7 +12,6 @@
#include "src/Core/util/DisableStupidWarnings.h"
#include <complex.h>
extern "C" {
#include <pastix_nompi.h>
#include <pastix.h>
@@ -35,12 +41,8 @@ extern "C" {
*
*/
#include "src/misc/Solve.h"
#include "src/misc/SparseSolve.h"
#include "src/PaStiXSupport/PaStiXSupport.h"
#include "src/Core/util/ReenableStupidWarnings.h"
#endif // EIGEN_PASTIXSUPPORT_MODULE_H

View File

@@ -1,3 +1,10 @@
// This file is part of Eigen, a lightweight C++ template library
// for linear algebra.
//
// This Source Code Form is subject to the terms of the Mozilla
// Public License v. 2.0. If a copy of the MPL was not distributed
// with this file, You can obtain one at http://mozilla.org/MPL/2.0/.
#ifndef EIGEN_PARDISOSUPPORT_MODULE_H
#define EIGEN_PARDISOSUPPORT_MODULE_H
@@ -7,8 +14,6 @@
#include <mkl_pardiso.h>
#include <unsupported/Eigen/SparseExtra>
/** \ingroup Support_modules
* \defgroup PardisoSupport_Module PardisoSupport module
*

View File

@@ -1,3 +1,10 @@
// This file is part of Eigen, a lightweight C++ template library
// for linear algebra.
//
// This Source Code Form is subject to the terms of the Mozilla
// Public License v. 2.0. If a copy of the MPL was not distributed
// with this file, You can obtain one at http://mozilla.org/MPL/2.0/.
#ifndef EIGEN_QR_MODULE_H
#define EIGEN_QR_MODULE_H
@@ -15,31 +22,30 @@
*
* This module provides various QR decompositions
* This module also provides some MatrixBase methods, including:
* - MatrixBase::qr(),
* - MatrixBase::householderQr()
* - MatrixBase::colPivHouseholderQr()
* - MatrixBase::fullPivHouseholderQr()
*
* \code
* #include <Eigen/QR>
* \endcode
*/
#include "src/misc/Solve.h"
#include "src/QR/HouseholderQR.h"
#include "src/QR/FullPivHouseholderQR.h"
#include "src/QR/ColPivHouseholderQR.h"
#include "src/QR/CompleteOrthogonalDecomposition.h"
#ifdef EIGEN_USE_LAPACKE
#include "src/QR/HouseholderQR_MKL.h"
#include "src/QR/ColPivHouseholderQR_MKL.h"
#ifdef EIGEN_USE_MKL
#include "mkl_lapacke.h"
#else
#include "src/misc/lapacke.h"
#endif
#ifdef EIGEN2_SUPPORT
#include "src/Eigen2Support/QR.h"
#include "src/QR/HouseholderQR_LAPACKE.h"
#include "src/QR/ColPivHouseholderQR_LAPACKE.h"
#endif
#include "src/Core/util/ReenableStupidWarnings.h"
#ifdef EIGEN2_SUPPORT
#include "Eigenvalues"
#endif
#endif // EIGEN_QR_MODULE_H
/* vim: set filetype=cpp et sw=2 ts=2 ai: */

View File

@@ -1,3 +1,9 @@
// This file is part of Eigen, a lightweight C++ template library
// for linear algebra.
//
// This Source Code Form is subject to the terms of the Mozilla
// Public License v. 2.0. If a copy of the MPL was not distributed
// with this file, You can obtain one at http://mozilla.org/MPL/2.0/.
#ifndef EIGEN_QTMALLOC_MODULE_H
#define EIGEN_QTMALLOC_MODULE_H
@@ -8,7 +14,7 @@
#include "src/Core/util/DisableStupidWarnings.h"
void *qMalloc(size_t size)
void *qMalloc(std::size_t size)
{
return Eigen::internal::aligned_malloc(size);
}
@@ -18,10 +24,10 @@ void qFree(void *ptr)
Eigen::internal::aligned_free(ptr);
}
void *qRealloc(void *ptr, size_t size)
void *qRealloc(void *ptr, std::size_t size)
{
void* newPtr = Eigen::internal::aligned_malloc(size);
memcpy(newPtr, ptr, size);
std::memcpy(newPtr, ptr, size);
Eigen::internal::aligned_free(ptr);
return newPtr;
}

View File

@@ -1,3 +1,10 @@
// This file is part of Eigen, a lightweight C++ template library
// for linear algebra.
//
// This Source Code Form is subject to the terms of the Mozilla
// Public License v. 2.0. If a copy of the MPL was not distributed
// with this file, You can obtain one at http://mozilla.org/MPL/2.0/.
#ifndef EIGEN_SPQRSUPPORT_MODULE_H
#define EIGEN_SPQRSUPPORT_MODULE_H
@@ -10,7 +17,7 @@
/** \ingroup Support_modules
* \defgroup SPQRSupport_Module SuiteSparseQR module
*
* This module provides an interface to the SPQR library, which is part of the <a href="http://www.cise.ufl.edu/research/sparse/SuiteSparse/">suitesparse</a> package.
* This module provides an interface to the SPQR library, which is part of the <a href="http://www.suitesparse.com">suitesparse</a> package.
*
* \code
* #include <Eigen/SPQRSupport>
@@ -21,8 +28,6 @@
*
*/
#include "src/misc/Solve.h"
#include "src/misc/SparseSolve.h"
#include "src/CholmodSupport/CholmodSupport.h"
#include "src/SPQRSupport/SuiteSparseQRSupport.h"

View File

@@ -1,3 +1,10 @@
// This file is part of Eigen, a lightweight C++ template library
// for linear algebra.
//
// This Source Code Form is subject to the terms of the Mozilla
// Public License v. 2.0. If a copy of the MPL was not distributed
// with this file, You can obtain one at http://mozilla.org/MPL/2.0/.
#ifndef EIGEN_SVD_MODULE_H
#define EIGEN_SVD_MODULE_H
@@ -12,23 +19,30 @@
*
*
* This module provides SVD decomposition for matrices (both real and complex).
* This decomposition is accessible via the following MatrixBase method:
* Two decomposition algorithms are provided:
* - JacobiSVD implementing two-sided Jacobi iterations is numerically very accurate, fast for small matrices, but very slow for larger ones.
* - BDCSVD implementing a recursive divide & conquer strategy on top of an upper-bidiagonalization which remains fast for large problems.
* These decompositions are accessible via the respective classes and following MatrixBase methods:
* - MatrixBase::jacobiSvd()
* - MatrixBase::bdcSvd()
*
* \code
* #include <Eigen/SVD>
* \endcode
*/
#include "src/misc/Solve.h"
#include "src/SVD/JacobiSVD.h"
#if defined(EIGEN_USE_LAPACKE) && !defined(EIGEN_USE_LAPACKE_STRICT)
#include "src/SVD/JacobiSVD_MKL.h"
#endif
#include "src/misc/RealSvd2x2.h"
#include "src/SVD/UpperBidiagonalization.h"
#ifdef EIGEN2_SUPPORT
#include "src/Eigen2Support/SVD.h"
#include "src/SVD/SVDBase.h"
#include "src/SVD/JacobiSVD.h"
#include "src/SVD/BDCSVD.h"
#if defined(EIGEN_USE_LAPACKE) && !defined(EIGEN_USE_LAPACKE_STRICT)
#ifdef EIGEN_USE_MKL
#include "mkl_lapacke.h"
#else
#include "src/misc/lapacke.h"
#endif
#include "src/SVD/JacobiSVD_LAPACKE.h"
#endif
#include "src/Core/util/ReenableStupidWarnings.h"

View File

@@ -1,3 +1,10 @@
// This file is part of Eigen, a lightweight C++ template library
// for linear algebra.
//
// This Source Code Form is subject to the terms of the Mozilla
// Public License v. 2.0. If a copy of the MPL was not distributed
// with this file, You can obtain one at http://mozilla.org/MPL/2.0/.
#ifndef EIGEN_SPARSE_MODULE_H
#define EIGEN_SPARSE_MODULE_H
@@ -11,14 +18,16 @@
* - \ref SparseQR_Module
* - \ref IterativeLinearSolvers_Module
*
* \code
* #include <Eigen/Sparse>
* \endcode
\code
#include <Eigen/Sparse>
\endcode
*/
#include "SparseCore"
#include "OrderingMethods"
#ifndef EIGEN_MPL2_ONLY
#include "SparseCholesky"
#endif
#include "SparseLU"
#include "SparseQR"
#include "IterativeLinearSolvers"

View File

@@ -34,8 +34,6 @@
#error The SparseCholesky module has nothing to offer in MPL2 only mode
#endif
#include "src/misc/Solve.h"
#include "src/misc/SparseSolve.h"
#include "src/SparseCholesky/SimplicialCholesky.h"
#ifndef EIGEN_MPL2_ONLY

View File

@@ -1,3 +1,10 @@
// This file is part of Eigen, a lightweight C++ template library
// for linear algebra.
//
// This Source Code Form is subject to the terms of the Mozilla
// Public License v. 2.0. If a copy of the MPL was not distributed
// with this file, You can obtain one at http://mozilla.org/MPL/2.0/.
#ifndef EIGEN_SPARSECORE_MODULE_H
#define EIGEN_SPARSECORE_MODULE_H
@@ -26,37 +33,35 @@
* This module depends on: Core.
*/
namespace Eigen {
/** The type used to identify a general sparse storage. */
struct Sparse {};
}
#include "src/SparseCore/SparseUtil.h"
#include "src/SparseCore/SparseMatrixBase.h"
#include "src/SparseCore/SparseAssign.h"
#include "src/SparseCore/CompressedStorage.h"
#include "src/SparseCore/AmbiVector.h"
#include "src/SparseCore/SparseCompressedBase.h"
#include "src/SparseCore/SparseMatrix.h"
#include "src/SparseCore/SparseMap.h"
#include "src/SparseCore/MappedSparseMatrix.h"
#include "src/SparseCore/SparseVector.h"
#include "src/SparseCore/SparseBlock.h"
#include "src/SparseCore/SparseTranspose.h"
#include "src/SparseCore/SparseRef.h"
#include "src/SparseCore/SparseCwiseUnaryOp.h"
#include "src/SparseCore/SparseCwiseBinaryOp.h"
#include "src/SparseCore/SparseTranspose.h"
#include "src/SparseCore/SparseBlock.h"
#include "src/SparseCore/SparseDot.h"
#include "src/SparseCore/SparsePermutation.h"
#include "src/SparseCore/SparseRedux.h"
#include "src/SparseCore/SparseFuzzy.h"
#include "src/SparseCore/SparseView.h"
#include "src/SparseCore/SparseDiagonalProduct.h"
#include "src/SparseCore/ConservativeSparseSparseProduct.h"
#include "src/SparseCore/SparseSparseProductWithPruning.h"
#include "src/SparseCore/SparseProduct.h"
#include "src/SparseCore/SparseDenseProduct.h"
#include "src/SparseCore/SparseDiagonalProduct.h"
#include "src/SparseCore/SparseTriangularView.h"
#include "src/SparseCore/SparseSelfAdjointView.h"
#include "src/SparseCore/SparseTriangularView.h"
#include "src/SparseCore/TriangularSolver.h"
#include "src/SparseCore/SparseView.h"
#include "src/SparseCore/SparsePermutation.h"
#include "src/SparseCore/SparseFuzzy.h"
#include "src/SparseCore/SparseSolverBase.h"
#include "src/Core/util/ReenableStupidWarnings.h"

View File

@@ -20,9 +20,6 @@
* Please, see the documentation of the SparseLU class for more details.
*/
#include "src/misc/Solve.h"
#include "src/misc/SparseSolve.h"
// Ordering interface
#include "OrderingMethods"

View File

@@ -1,3 +1,10 @@
// This file is part of Eigen, a lightweight C++ template library
// for linear algebra.
//
// This Source Code Form is subject to the terms of the Mozilla
// Public License v. 2.0. If a copy of the MPL was not distributed
// with this file, You can obtain one at http://mozilla.org/MPL/2.0/.
#ifndef EIGEN_SPARSEQR_MODULE_H
#define EIGEN_SPARSEQR_MODULE_H
@@ -21,9 +28,6 @@
*
*/
#include "src/misc/Solve.h"
#include "src/misc/SparseSolve.h"
#include "OrderingMethods"
#include "src/SparseCore/SparseColEtree.h"
#include "src/SparseQR/SparseQR.h"

View File

@@ -14,7 +14,7 @@
#include "Core"
#include <deque>
#if (defined(_MSC_VER) && defined(_WIN64)) /* MSVC auto aligns in 64 bit builds */
#if EIGEN_COMP_MSVC && EIGEN_OS_WIN64 && (EIGEN_MAX_STATIC_ALIGN_BYTES<=16) /* MSVC auto aligns up to 16 bytes in 64 bit builds */
#define EIGEN_DEFINE_STL_DEQUE_SPECIALIZATION(...)

View File

@@ -13,7 +13,7 @@
#include "Core"
#include <list>
#if (defined(_MSC_VER) && defined(_WIN64)) /* MSVC auto aligns in 64 bit builds */
#if EIGEN_COMP_MSVC && EIGEN_OS_WIN64 && (EIGEN_MAX_STATIC_ALIGN_BYTES<=16) /* MSVC auto aligns up to 16 bytes in 64 bit builds */
#define EIGEN_DEFINE_STL_LIST_SPECIALIZATION(...)

View File

@@ -14,7 +14,7 @@
#include "Core"
#include <vector>
#if (defined(_MSC_VER) && defined(_WIN64)) /* MSVC auto aligns in 64 bit builds */
#if EIGEN_COMP_MSVC && EIGEN_OS_WIN64 && (EIGEN_MAX_STATIC_ALIGN_BYTES<=16) /* MSVC auto aligns up to 16 bytes in 64 bit builds */
#define EIGEN_DEFINE_STL_VECTOR_SPECIALIZATION(...)

View File

@@ -1,3 +1,10 @@
// This file is part of Eigen, a lightweight C++ template library
// for linear algebra.
//
// This Source Code Form is subject to the terms of the Mozilla
// Public License v. 2.0. If a copy of the MPL was not distributed
// with this file, You can obtain one at http://mozilla.org/MPL/2.0/.
#ifndef EIGEN_SUPERLUSUPPORT_MODULE_H
#define EIGEN_SUPERLUSUPPORT_MODULE_H
@@ -36,6 +43,8 @@ namespace Eigen { struct SluMatrix; }
* - class SuperLU: a supernodal sequential LU factorization.
* - class SuperILU: a supernodal sequential incomplete LU factorization (to be used as a preconditioner for iterative methods).
*
* \warning This wrapper requires at least versions 4.0 of SuperLU. The 3.x versions are not supported.
*
* \warning When including this module, you have to use SUPERLU_EMPTY instead of EMPTY which is no longer defined because it is too polluting.
*
* \code
@@ -48,12 +57,8 @@ namespace Eigen { struct SluMatrix; }
*
*/
#include "src/misc/Solve.h"
#include "src/misc/SparseSolve.h"
#include "src/SuperLUSupport/SuperLUSupport.h"
#include "src/Core/util/ReenableStupidWarnings.h"
#endif // EIGEN_SUPERLUSUPPORT_MODULE_H

View File

@@ -1,3 +1,10 @@
// This file is part of Eigen, a lightweight C++ template library
// for linear algebra.
//
// This Source Code Form is subject to the terms of the Mozilla
// Public License v. 2.0. If a copy of the MPL was not distributed
// with this file, You can obtain one at http://mozilla.org/MPL/2.0/.
#ifndef EIGEN_UMFPACKSUPPORT_MODULE_H
#define EIGEN_UMFPACKSUPPORT_MODULE_H
@@ -12,7 +19,7 @@ extern "C" {
/** \ingroup Support_modules
* \defgroup UmfPackSupport_Module UmfPackSupport module
*
* This module provides an interface to the UmfPack library which is part of the <a href="http://www.cise.ufl.edu/research/sparse/SuiteSparse/">suitesparse</a> package.
* This module provides an interface to the UmfPack library which is part of the <a href="http://www.suitesparse.com">suitesparse</a> package.
* It provides the following factorization class:
* - class UmfPackLU: a multifrontal sequential LU factorization.
*
@@ -26,9 +33,6 @@ extern "C" {
*
*/
#include "src/misc/Solve.h"
#include "src/misc/SparseSolve.h"
#include "src/UmfPackSupport/UmfPackSupport.h"
#include "src/Core/util/ReenableStupidWarnings.h"

View File

@@ -13,7 +13,7 @@
#ifndef EIGEN_LDLT_H
#define EIGEN_LDLT_H
namespace Eigen {
namespace Eigen {
namespace internal {
template<typename MatrixType, int UpLo> struct LDLT_Traits;
@@ -28,8 +28,8 @@ namespace internal {
*
* \brief Robust Cholesky decomposition of a matrix with pivoting
*
* \param MatrixType the type of the matrix of which to compute the LDL^T Cholesky decomposition
* \param UpLo the triangular part that will be used for the decompositon: Lower (default) or Upper.
* \tparam _MatrixType the type of the matrix of which to compute the LDL^T Cholesky decomposition
* \tparam _UpLo the triangular part that will be used for the decompositon: Lower (default) or Upper.
* The other triangular part won't be read.
*
* Perform a robust Cholesky decomposition of a positive semidefinite or negative semidefinite
@@ -43,7 +43,9 @@ namespace internal {
* Remember that Cholesky decompositions are not rank-revealing. Also, do not use a Cholesky
* decomposition to determine whether a system of equations has a solution.
*
* \sa MatrixBase::ldlt(), class LLT
* This class supports the \link InplaceDecomposition inplace decomposition \endlink mechanism.
*
* \sa MatrixBase::ldlt(), SelfAdjointView::ldlt(), class LLT
*/
template<typename _MatrixType, int _UpLo> class LDLT
{
@@ -52,15 +54,15 @@ template<typename _MatrixType, int _UpLo> class LDLT
enum {
RowsAtCompileTime = MatrixType::RowsAtCompileTime,
ColsAtCompileTime = MatrixType::ColsAtCompileTime,
Options = MatrixType::Options & ~RowMajorBit, // these are the options for the TmpMatrixType, we need a ColMajor matrix here!
MaxRowsAtCompileTime = MatrixType::MaxRowsAtCompileTime,
MaxColsAtCompileTime = MatrixType::MaxColsAtCompileTime,
UpLo = _UpLo
};
typedef typename MatrixType::Scalar Scalar;
typedef typename NumTraits<typename MatrixType::Scalar>::Real RealScalar;
typedef typename MatrixType::Index Index;
typedef Matrix<Scalar, RowsAtCompileTime, 1, Options, MaxRowsAtCompileTime, 1> TmpMatrixType;
typedef Eigen::Index Index; ///< \deprecated since Eigen 3.3
typedef typename MatrixType::StorageIndex StorageIndex;
typedef Matrix<Scalar, RowsAtCompileTime, 1, 0, MaxRowsAtCompileTime, 1> TmpMatrixType;
typedef Transpositions<RowsAtCompileTime, MaxRowsAtCompileTime> TranspositionType;
typedef PermutationMatrix<RowsAtCompileTime, MaxRowsAtCompileTime> PermutationType;
@@ -72,11 +74,11 @@ template<typename _MatrixType, int _UpLo> class LDLT
* The default constructor is useful in cases in which the user intends to
* perform decompositions via LDLT::compute(const MatrixType&).
*/
LDLT()
: m_matrix(),
m_transpositions(),
LDLT()
: m_matrix(),
m_transpositions(),
m_sign(internal::ZeroSign),
m_isInitialized(false)
m_isInitialized(false)
{}
/** \brief Default Constructor with memory preallocation
@@ -85,7 +87,7 @@ template<typename _MatrixType, int _UpLo> class LDLT
* according to the specified problem \a size.
* \sa LDLT()
*/
LDLT(Index size)
explicit LDLT(Index size)
: m_matrix(size, size),
m_transpositions(size),
m_temporary(size),
@@ -96,16 +98,35 @@ template<typename _MatrixType, int _UpLo> class LDLT
/** \brief Constructor with decomposition
*
* This calculates the decomposition for the input \a matrix.
*
* \sa LDLT(Index size)
*/
LDLT(const MatrixType& matrix)
template<typename InputType>
explicit LDLT(const EigenBase<InputType>& matrix)
: m_matrix(matrix.rows(), matrix.cols()),
m_transpositions(matrix.rows()),
m_temporary(matrix.rows()),
m_sign(internal::ZeroSign),
m_isInitialized(false)
{
compute(matrix);
compute(matrix.derived());
}
/** \brief Constructs a LDLT factorization from a given matrix
*
* This overloaded constructor is provided for \link InplaceDecomposition inplace decomposition \endlink when \c MatrixType is a Eigen::Ref.
*
* \sa LDLT(const EigenBase&)
*/
template<typename InputType>
explicit LDLT(EigenBase<InputType>& matrix)
: m_matrix(matrix.derived()),
m_transpositions(matrix.rows()),
m_temporary(matrix.rows()),
m_sign(internal::ZeroSign),
m_isInitialized(false)
{
compute(matrix.derived());
}
/** Clear any existing decomposition
@@ -151,13 +172,6 @@ template<typename _MatrixType, int _UpLo> class LDLT
eigen_assert(m_isInitialized && "LDLT is not initialized.");
return m_sign == internal::PositiveSemiDef || m_sign == internal::ZeroSign;
}
#ifdef EIGEN2_SUPPORT
inline bool isPositiveDefinite() const
{
return isPositive();
}
#endif
/** \returns true if the matrix is negative (semidefinite) */
inline bool isNegative(void) const
@@ -173,37 +187,38 @@ template<typename _MatrixType, int _UpLo> class LDLT
* \note_about_checking_solutions
*
* More precisely, this method solves \f$ A x = b \f$ using the decomposition \f$ A = P^T L D L^* P \f$
* by solving the systems \f$ P^T y_1 = b \f$, \f$ L y_2 = y_1 \f$, \f$ D y_3 = y_2 \f$,
* by solving the systems \f$ P^T y_1 = b \f$, \f$ L y_2 = y_1 \f$, \f$ D y_3 = y_2 \f$,
* \f$ L^* y_4 = y_3 \f$ and \f$ P x = y_4 \f$ in succession. If the matrix \f$ A \f$ is singular, then
* \f$ D \f$ will also be singular (all the other matrices are invertible). In that case, the
* least-square solution of \f$ D y_3 = y_2 \f$ is computed. This does not mean that this function
* computes the least-square solution of \f$ A x = b \f$ is \f$ A \f$ is singular.
*
* \sa MatrixBase::ldlt()
* \sa MatrixBase::ldlt(), SelfAdjointView::ldlt()
*/
template<typename Rhs>
inline const internal::solve_retval<LDLT, Rhs>
inline const Solve<LDLT, Rhs>
solve(const MatrixBase<Rhs>& b) const
{
eigen_assert(m_isInitialized && "LDLT is not initialized.");
eigen_assert(m_matrix.rows()==b.rows()
&& "LDLT::solve(): invalid number of rows of the right hand side matrix b");
return internal::solve_retval<LDLT, Rhs>(*this, b.derived());
return Solve<LDLT, Rhs>(*this, b.derived());
}
#ifdef EIGEN2_SUPPORT
template<typename OtherDerived, typename ResultType>
bool solve(const MatrixBase<OtherDerived>& b, ResultType *result) const
{
*result = this->solve(b);
return true;
}
#endif
template<typename Derived>
bool solveInPlace(MatrixBase<Derived> &bAndX) const;
LDLT& compute(const MatrixType& matrix);
template<typename InputType>
LDLT& compute(const EigenBase<InputType>& matrix);
/** \returns an estimate of the reciprocal condition number of the matrix of
* which \c *this is the LDLT decomposition.
*/
RealScalar rcond() const
{
eigen_assert(m_isInitialized && "LDLT is not initialized.");
return internal::rcond_estimate_helper(m_l1_norm, *this);
}
template <typename Derived>
LDLT& rankUpdate(const MatrixBase<Derived>& w, const RealScalar& alpha=1);
@@ -220,22 +235,35 @@ template<typename _MatrixType, int _UpLo> class LDLT
MatrixType reconstructedMatrix() const;
/** \returns the adjoint of \c *this, that is, a const reference to the decomposition itself as the underlying matrix is self-adjoint.
*
* This method is provided for compatibility with other matrix decompositions, thus enabling generic code such as:
* \code x = decomposition.adjoint().solve(b) \endcode
*/
const LDLT& adjoint() const { return *this; };
inline Index rows() const { return m_matrix.rows(); }
inline Index cols() const { return m_matrix.cols(); }
/** \brief Reports whether previous computation was successful.
*
* \returns \c Success if computation was succesful,
* \c NumericalIssue if the matrix.appears to be negative.
* \c NumericalIssue if the factorization failed because of a zero pivot.
*/
ComputationInfo info() const
{
eigen_assert(m_isInitialized && "LDLT is not initialized.");
return Success;
return m_info;
}
#ifndef EIGEN_PARSED_BY_DOXYGEN
template<typename RhsType, typename DstType>
EIGEN_DEVICE_FUNC
void _solve_impl(const RhsType &rhs, DstType &dst) const;
#endif
protected:
static void check_template_parameters()
{
EIGEN_STATIC_ASSERT_NON_INTEGER(Scalar);
@@ -248,10 +276,12 @@ template<typename _MatrixType, int _UpLo> class LDLT
* is not stored), and the diagonal entries correspond to D.
*/
MatrixType m_matrix;
RealScalar m_l1_norm;
TranspositionType m_transpositions;
TmpMatrixType m_temporary;
internal::SignMatrix m_sign;
bool m_isInitialized;
ComputationInfo m_info;
};
namespace internal {
@@ -266,15 +296,18 @@ template<> struct ldlt_inplace<Lower>
using std::abs;
typedef typename MatrixType::Scalar Scalar;
typedef typename MatrixType::RealScalar RealScalar;
typedef typename MatrixType::Index Index;
typedef typename TranspositionType::StorageIndex IndexType;
eigen_assert(mat.rows()==mat.cols());
const Index size = mat.rows();
bool found_zero_pivot = false;
bool ret = true;
if (size <= 1)
{
transpositions.setIdentity();
if (numext::real(mat.coeff(0,0)) > 0) sign = PositiveSemiDef;
else if (numext::real(mat.coeff(0,0)) < 0) sign = NegativeSemiDef;
if(size==0) sign = ZeroSign;
else if (numext::real(mat.coeff(0,0)) > static_cast<RealScalar>(0) ) sign = PositiveSemiDef;
else if (numext::real(mat.coeff(0,0)) < static_cast<RealScalar>(0)) sign = NegativeSemiDef;
else sign = ZeroSign;
return true;
}
@@ -286,7 +319,7 @@ template<> struct ldlt_inplace<Lower>
mat.diagonal().tail(size-k).cwiseAbs().maxCoeff(&index_of_biggest_in_corner);
index_of_biggest_in_corner += k;
transpositions.coeffRef(k) = index_of_biggest_in_corner;
transpositions.coeffRef(k) = IndexType(index_of_biggest_in_corner);
if(k != index_of_biggest_in_corner)
{
// apply the transposition while taking care to consider only
@@ -295,7 +328,7 @@ template<> struct ldlt_inplace<Lower>
mat.row(k).head(k).swap(mat.row(index_of_biggest_in_corner).head(k));
mat.col(k).tail(s).swap(mat.col(index_of_biggest_in_corner).tail(s));
std::swap(mat.coeffRef(k,k),mat.coeffRef(index_of_biggest_in_corner,index_of_biggest_in_corner));
for(int i=k+1;i<index_of_biggest_in_corner;++i)
for(Index i=k+1;i<index_of_biggest_in_corner;++i)
{
Scalar tmp = mat.coeffRef(i,k);
mat.coeffRef(i,k) = numext::conj(mat.coeffRef(index_of_biggest_in_corner,i));
@@ -321,26 +354,46 @@ template<> struct ldlt_inplace<Lower>
if(rs>0)
A21.noalias() -= A20 * temp.head(k);
}
// In some previous versions of Eigen (e.g., 3.2.1), the scaling was omitted if the pivot
// was smaller than the cutoff value. However, soince LDLT is not rank-revealing
// we should only make sure we do not introduce INF or NaN values.
// LAPACK also uses 0 as the cutoff value.
// was smaller than the cutoff value. However, since LDLT is not rank-revealing
// we should only make sure that we do not introduce INF or NaN values.
// Remark that LAPACK also uses 0 as the cutoff value.
RealScalar realAkk = numext::real(mat.coeffRef(k,k));
if((rs>0) && (abs(realAkk) > RealScalar(0)))
bool pivot_is_valid = (abs(realAkk) > RealScalar(0));
if(k==0 && !pivot_is_valid)
{
// The entire diagonal is zero, there is nothing more to do
// except filling the transpositions, and checking whether the matrix is zero.
sign = ZeroSign;
for(Index j = 0; j<size; ++j)
{
transpositions.coeffRef(j) = IndexType(j);
ret = ret && (mat.col(j).tail(size-j-1).array()==Scalar(0)).all();
}
return ret;
}
if((rs>0) && pivot_is_valid)
A21 /= realAkk;
else if(rs>0)
ret = ret && (A21.array()==Scalar(0)).all();
if(found_zero_pivot && pivot_is_valid) ret = false; // factorization failed
else if(!pivot_is_valid) found_zero_pivot = true;
if (sign == PositiveSemiDef) {
if (realAkk < 0) sign = Indefinite;
if (realAkk < static_cast<RealScalar>(0)) sign = Indefinite;
} else if (sign == NegativeSemiDef) {
if (realAkk > 0) sign = Indefinite;
if (realAkk > static_cast<RealScalar>(0)) sign = Indefinite;
} else if (sign == ZeroSign) {
if (realAkk > 0) sign = PositiveSemiDef;
else if (realAkk < 0) sign = NegativeSemiDef;
if (realAkk > static_cast<RealScalar>(0)) sign = PositiveSemiDef;
else if (realAkk < static_cast<RealScalar>(0)) sign = NegativeSemiDef;
}
}
return true;
return ret;
}
// Reference for the algorithm: Davis and Hager, "Multiple Rank
@@ -356,7 +409,6 @@ template<> struct ldlt_inplace<Lower>
using numext::isfinite;
typedef typename MatrixType::Scalar Scalar;
typedef typename MatrixType::RealScalar RealScalar;
typedef typename MatrixType::Index Index;
const Index size = mat.rows();
eigen_assert(mat.cols() == size && w.size()==size);
@@ -420,16 +472,16 @@ template<typename MatrixType> struct LDLT_Traits<MatrixType,Lower>
{
typedef const TriangularView<const MatrixType, UnitLower> MatrixL;
typedef const TriangularView<const typename MatrixType::AdjointReturnType, UnitUpper> MatrixU;
static inline MatrixL getL(const MatrixType& m) { return m; }
static inline MatrixU getU(const MatrixType& m) { return m.adjoint(); }
static inline MatrixL getL(const MatrixType& m) { return MatrixL(m); }
static inline MatrixU getU(const MatrixType& m) { return MatrixU(m.adjoint()); }
};
template<typename MatrixType> struct LDLT_Traits<MatrixType,Upper>
{
typedef const TriangularView<const typename MatrixType::AdjointReturnType, UnitLower> MatrixL;
typedef const TriangularView<const MatrixType, UnitUpper> MatrixU;
static inline MatrixL getL(const MatrixType& m) { return m.adjoint(); }
static inline MatrixU getU(const MatrixType& m) { return m; }
static inline MatrixL getL(const MatrixType& m) { return MatrixL(m.adjoint()); }
static inline MatrixU getU(const MatrixType& m) { return MatrixU(m); }
};
} // end namespace internal
@@ -437,21 +489,35 @@ template<typename MatrixType> struct LDLT_Traits<MatrixType,Upper>
/** Compute / recompute the LDLT decomposition A = L D L^* = U^* D U of \a matrix
*/
template<typename MatrixType, int _UpLo>
LDLT<MatrixType,_UpLo>& LDLT<MatrixType,_UpLo>::compute(const MatrixType& a)
template<typename InputType>
LDLT<MatrixType,_UpLo>& LDLT<MatrixType,_UpLo>::compute(const EigenBase<InputType>& a)
{
check_template_parameters();
eigen_assert(a.rows()==a.cols());
const Index size = a.rows();
m_matrix = a;
m_matrix = a.derived();
// Compute matrix L1 norm = max abs column sum.
m_l1_norm = RealScalar(0);
// TODO move this code to SelfAdjointView
for (Index col = 0; col < size; ++col) {
RealScalar abs_col_sum;
if (_UpLo == Lower)
abs_col_sum = m_matrix.col(col).tail(size - col).template lpNorm<1>() + m_matrix.row(col).head(col).template lpNorm<1>();
else
abs_col_sum = m_matrix.col(col).head(col).template lpNorm<1>() + m_matrix.row(col).tail(size - col).template lpNorm<1>();
if (abs_col_sum > m_l1_norm)
m_l1_norm = abs_col_sum;
}
m_transpositions.resize(size);
m_isInitialized = false;
m_temporary.resize(size);
m_sign = internal::ZeroSign;
internal::ldlt_inplace<UpLo>::unblocked(m_matrix, m_transpositions, m_temporary, m_sign);
m_info = internal::ldlt_inplace<UpLo>::unblocked(m_matrix, m_transpositions, m_temporary, m_sign) ? Success : NumericalIssue;
m_isInitialized = true;
return *this;
@@ -466,18 +532,19 @@ template<typename MatrixType, int _UpLo>
template<typename Derived>
LDLT<MatrixType,_UpLo>& LDLT<MatrixType,_UpLo>::rankUpdate(const MatrixBase<Derived>& w, const typename LDLT<MatrixType,_UpLo>::RealScalar& sigma)
{
typedef typename TranspositionType::StorageIndex IndexType;
const Index size = w.rows();
if (m_isInitialized)
{
eigen_assert(m_matrix.rows()==size);
}
else
{
{
m_matrix.resize(size,size);
m_matrix.setZero();
m_transpositions.resize(size);
for (Index i = 0; i < size; i++)
m_transpositions.coeffRef(i) = i;
m_transpositions.coeffRef(i) = IndexType(i);
m_temporary.resize(size);
m_sign = sigma>=0 ? internal::PositiveSemiDef : internal::NegativeSemiDef;
m_isInitialized = true;
@@ -488,53 +555,46 @@ LDLT<MatrixType,_UpLo>& LDLT<MatrixType,_UpLo>::rankUpdate(const MatrixBase<Deri
return *this;
}
namespace internal {
template<typename _MatrixType, int _UpLo, typename Rhs>
struct solve_retval<LDLT<_MatrixType,_UpLo>, Rhs>
: solve_retval_base<LDLT<_MatrixType,_UpLo>, Rhs>
#ifndef EIGEN_PARSED_BY_DOXYGEN
template<typename _MatrixType, int _UpLo>
template<typename RhsType, typename DstType>
void LDLT<_MatrixType,_UpLo>::_solve_impl(const RhsType &rhs, DstType &dst) const
{
typedef LDLT<_MatrixType,_UpLo> LDLTType;
EIGEN_MAKE_SOLVE_HELPERS(LDLTType,Rhs)
eigen_assert(rhs.rows() == rows());
// dst = P b
dst = m_transpositions * rhs;
template<typename Dest> void evalTo(Dest& dst) const
// dst = L^-1 (P b)
matrixL().solveInPlace(dst);
// dst = D^-1 (L^-1 P b)
// more precisely, use pseudo-inverse of D (see bug 241)
using std::abs;
const typename Diagonal<const MatrixType>::RealReturnType vecD(vectorD());
// In some previous versions, tolerance was set to the max of 1/highest (or rather numeric_limits::min())
// and the maximal diagonal entry * epsilon as motivated by LAPACK's xGELSS:
// RealScalar tolerance = numext::maxi(vecD.array().abs().maxCoeff() * NumTraits<RealScalar>::epsilon(),RealScalar(1) / NumTraits<RealScalar>::highest());
// However, LDLT is not rank revealing, and so adjusting the tolerance wrt to the highest
// diagonal element is not well justified and leads to numerical issues in some cases.
// Moreover, Lapack's xSYTRS routines use 0 for the tolerance.
// Using numeric_limits::min() gives us more robustness to denormals.
RealScalar tolerance = (std::numeric_limits<RealScalar>::min)();
for (Index i = 0; i < vecD.size(); ++i)
{
eigen_assert(rhs().rows() == dec().matrixLDLT().rows());
// dst = P b
dst = dec().transpositionsP() * rhs();
// dst = L^-1 (P b)
dec().matrixL().solveInPlace(dst);
// dst = D^-1 (L^-1 P b)
// more precisely, use pseudo-inverse of D (see bug 241)
using std::abs;
using std::max;
typedef typename LDLTType::MatrixType MatrixType;
typedef typename LDLTType::RealScalar RealScalar;
const typename Diagonal<const MatrixType>::RealReturnType vectorD(dec().vectorD());
// In some previous versions, tolerance was set to the max of 1/highest and the maximal diagonal entry * epsilon
// as motivated by LAPACK's xGELSS:
// RealScalar tolerance = (max)(vectorD.array().abs().maxCoeff() *NumTraits<RealScalar>::epsilon(),RealScalar(1) / NumTraits<RealScalar>::highest());
// However, LDLT is not rank revealing, and so adjusting the tolerance wrt to the highest
// diagonal element is not well justified and to numerical issues in some cases.
// Moreover, Lapack's xSYTRS routines use 0 for the tolerance.
RealScalar tolerance = RealScalar(1) / NumTraits<RealScalar>::highest();
for (Index i = 0; i < vectorD.size(); ++i) {
if(abs(vectorD(i)) > tolerance)
dst.row(i) /= vectorD(i);
else
dst.row(i).setZero();
}
// dst = L^-T (D^-1 L^-1 P b)
dec().matrixU().solveInPlace(dst);
// dst = P^-1 (L^-T D^-1 L^-1 P b) = A^-1 b
dst = dec().transpositionsP().transpose() * dst;
if(abs(vecD(i)) > tolerance)
dst.row(i) /= vecD(i);
else
dst.row(i).setZero();
}
};
// dst = L^-T (D^-1 L^-1 P b)
matrixU().solveInPlace(dst);
// dst = P^-1 (L^-T D^-1 L^-1 P b) = A^-1 b
dst = m_transpositions.transpose() * dst;
}
#endif
/** \internal use x = ldlt_object.solve(x);
*
@@ -588,6 +648,7 @@ MatrixType LDLT<MatrixType,_UpLo>::reconstructedMatrix() const
/** \cholesky_module
* \returns the Cholesky decomposition with full pivoting without square root of \c *this
* \sa MatrixBase::ldlt()
*/
template<typename MatrixType, unsigned int UpLo>
inline const LDLT<typename SelfAdjointView<MatrixType, UpLo>::PlainObject, UpLo>
@@ -598,6 +659,7 @@ SelfAdjointView<MatrixType, UpLo>::ldlt() const
/** \cholesky_module
* \returns the Cholesky decomposition with full pivoting without square root of \c *this
* \sa SelfAdjointView::ldlt()
*/
template<typename Derived>
inline const LDLT<typename MatrixBase<Derived>::PlainObject>

View File

@@ -10,7 +10,7 @@
#ifndef EIGEN_LLT_H
#define EIGEN_LLT_H
namespace Eigen {
namespace Eigen {
namespace internal{
template<typename MatrixType, int UpLo> struct LLT_Traits;
@@ -22,9 +22,9 @@ template<typename MatrixType, int UpLo> struct LLT_Traits;
*
* \brief Standard Cholesky decomposition (LL^T) of a matrix and associated features
*
* \param MatrixType the type of the matrix of which we are computing the LL^T Cholesky decomposition
* \param UpLo the triangular part that will be used for the decompositon: Lower (default) or Upper.
* The other triangular part won't be read.
* \tparam _MatrixType the type of the matrix of which we are computing the LL^T Cholesky decomposition
* \tparam _UpLo the triangular part that will be used for the decompositon: Lower (default) or Upper.
* The other triangular part won't be read.
*
* This class performs a LL^T Cholesky decomposition of a symmetric, positive definite
* matrix A such that A = LL^* = U^*U, where L is lower triangular.
@@ -40,12 +40,18 @@ template<typename MatrixType, int UpLo> struct LLT_Traits;
*
* Example: \include LLT_example.cpp
* Output: \verbinclude LLT_example.out
*
* \sa MatrixBase::llt(), class LDLT
*/
/* HEY THIS DOX IS DISABLED BECAUSE THERE's A BUG EITHER HERE OR IN LDLT ABOUT THAT (OR BOTH)
* Note that during the decomposition, only the upper triangular part of A is considered. Therefore,
* the strict lower part does not have to store correct values.
*
* \b Performance: for best performance, it is recommended to use a column-major storage format
* with the Lower triangular part (the default), or, equivalently, a row-major storage format
* with the Upper triangular part. Otherwise, you might get a 20% slowdown for the full factorization
* step, and rank-updates can be up to 3 times slower.
*
* This class supports the \link InplaceDecomposition inplace decomposition \endlink mechanism.
*
* Note that during the decomposition, only the lower (or upper, as defined by _UpLo) triangular part of A is considered.
* Therefore, the strict lower part does not have to store correct values.
*
* \sa MatrixBase::llt(), SelfAdjointView::llt(), class LDLT
*/
template<typename _MatrixType, int _UpLo> class LLT
{
@@ -54,12 +60,12 @@ template<typename _MatrixType, int _UpLo> class LLT
enum {
RowsAtCompileTime = MatrixType::RowsAtCompileTime,
ColsAtCompileTime = MatrixType::ColsAtCompileTime,
Options = MatrixType::Options,
MaxColsAtCompileTime = MatrixType::MaxColsAtCompileTime
};
typedef typename MatrixType::Scalar Scalar;
typedef typename NumTraits<typename MatrixType::Scalar>::Real RealScalar;
typedef typename MatrixType::Index Index;
typedef Eigen::Index Index; ///< \deprecated since Eigen 3.3
typedef typename MatrixType::StorageIndex StorageIndex;
enum {
PacketSize = internal::packet_traits<Scalar>::size,
@@ -83,14 +89,30 @@ template<typename _MatrixType, int _UpLo> class LLT
* according to the specified problem \a size.
* \sa LLT()
*/
LLT(Index size) : m_matrix(size, size),
explicit LLT(Index size) : m_matrix(size, size),
m_isInitialized(false) {}
LLT(const MatrixType& matrix)
template<typename InputType>
explicit LLT(const EigenBase<InputType>& matrix)
: m_matrix(matrix.rows(), matrix.cols()),
m_isInitialized(false)
{
compute(matrix);
compute(matrix.derived());
}
/** \brief Constructs a LDLT factorization from a given matrix
*
* This overloaded constructor is provided for \link InplaceDecomposition inplace decomposition \endlink when
* \c MatrixType is a Eigen::Ref.
*
* \sa LLT(const EigenBase&)
*/
template<typename InputType>
explicit LLT(EigenBase<InputType>& matrix)
: m_matrix(matrix.derived()),
m_isInitialized(false)
{
compute(matrix.derived());
}
/** \returns a view of the upper triangular matrix U */
@@ -115,33 +137,33 @@ template<typename _MatrixType, int _UpLo> class LLT
* Example: \include LLT_solve.cpp
* Output: \verbinclude LLT_solve.out
*
* \sa solveInPlace(), MatrixBase::llt()
* \sa solveInPlace(), MatrixBase::llt(), SelfAdjointView::llt()
*/
template<typename Rhs>
inline const internal::solve_retval<LLT, Rhs>
inline const Solve<LLT, Rhs>
solve(const MatrixBase<Rhs>& b) const
{
eigen_assert(m_isInitialized && "LLT is not initialized.");
eigen_assert(m_matrix.rows()==b.rows()
&& "LLT::solve(): invalid number of rows of the right hand side matrix b");
return internal::solve_retval<LLT, Rhs>(*this, b.derived());
return Solve<LLT, Rhs>(*this, b.derived());
}
#ifdef EIGEN2_SUPPORT
template<typename OtherDerived, typename ResultType>
bool solve(const MatrixBase<OtherDerived>& b, ResultType *result) const
{
*result = this->solve(b);
return true;
}
bool isPositiveDefinite() const { return true; }
#endif
template<typename Derived>
void solveInPlace(MatrixBase<Derived> &bAndX) const;
void solveInPlace(const MatrixBase<Derived> &bAndX) const;
LLT& compute(const MatrixType& matrix);
template<typename InputType>
LLT& compute(const EigenBase<InputType>& matrix);
/** \returns an estimate of the reciprocal condition number of the matrix of
* which \c *this is the Cholesky decomposition.
*/
RealScalar rcond() const
{
eigen_assert(m_isInitialized && "LLT is not initialized.");
eigen_assert(m_info == Success && "LLT failed because matrix appears to be negative");
return internal::rcond_estimate_helper(m_l1_norm, *this);
}
/** \returns the LLT decomposition matrix
*
@@ -159,7 +181,7 @@ template<typename _MatrixType, int _UpLo> class LLT
/** \brief Reports whether previous computation was successful.
*
* \returns \c Success if computation was succesful,
* \c NumericalIssue if the matrix.appears to be negative.
* \c NumericalIssue if the matrix.appears not to be positive definite.
*/
ComputationInfo info() const
{
@@ -167,24 +189,38 @@ template<typename _MatrixType, int _UpLo> class LLT
return m_info;
}
/** \returns the adjoint of \c *this, that is, a const reference to the decomposition itself as the underlying matrix is self-adjoint.
*
* This method is provided for compatibility with other matrix decompositions, thus enabling generic code such as:
* \code x = decomposition.adjoint().solve(b) \endcode
*/
const LLT& adjoint() const { return *this; };
inline Index rows() const { return m_matrix.rows(); }
inline Index cols() const { return m_matrix.cols(); }
template<typename VectorType>
LLT rankUpdate(const VectorType& vec, const RealScalar& sigma = 1);
#ifndef EIGEN_PARSED_BY_DOXYGEN
template<typename RhsType, typename DstType>
EIGEN_DEVICE_FUNC
void _solve_impl(const RhsType &rhs, DstType &dst) const;
#endif
protected:
static void check_template_parameters()
{
EIGEN_STATIC_ASSERT_NON_INTEGER(Scalar);
}
/** \internal
* Used to compute and store L
* The strict upper part is not used and even not initialized.
*/
MatrixType m_matrix;
RealScalar m_l1_norm;
bool m_isInitialized;
ComputationInfo m_info;
};
@@ -194,12 +230,11 @@ namespace internal {
template<typename Scalar, int UpLo> struct llt_inplace;
template<typename MatrixType, typename VectorType>
static typename MatrixType::Index llt_rank_update_lower(MatrixType& mat, const VectorType& vec, const typename MatrixType::RealScalar& sigma)
static Index llt_rank_update_lower(MatrixType& mat, const VectorType& vec, const typename MatrixType::RealScalar& sigma)
{
using std::sqrt;
typedef typename MatrixType::Scalar Scalar;
typedef typename MatrixType::RealScalar RealScalar;
typedef typename MatrixType::Index Index;
typedef typename MatrixType::ColXpr ColXpr;
typedef typename internal::remove_all<ColXpr>::type ColXprCleaned;
typedef typename ColXprCleaned::SegmentReturnType ColXprSegment;
@@ -268,11 +303,10 @@ template<typename Scalar> struct llt_inplace<Scalar, Lower>
{
typedef typename NumTraits<Scalar>::Real RealScalar;
template<typename MatrixType>
static typename MatrixType::Index unblocked(MatrixType& mat)
static Index unblocked(MatrixType& mat)
{
using std::sqrt;
typedef typename MatrixType::Index Index;
eigen_assert(mat.rows()==mat.cols());
const Index size = mat.rows();
for(Index k = 0; k < size; ++k)
@@ -295,9 +329,8 @@ template<typename Scalar> struct llt_inplace<Scalar, Lower>
}
template<typename MatrixType>
static typename MatrixType::Index blocked(MatrixType& m)
static Index blocked(MatrixType& m)
{
typedef typename MatrixType::Index Index;
eigen_assert(m.rows()==m.cols());
Index size = m.rows();
if(size<32)
@@ -322,36 +355,36 @@ template<typename Scalar> struct llt_inplace<Scalar, Lower>
Index ret;
if((ret=unblocked(A11))>=0) return k+ret;
if(rs>0) A11.adjoint().template triangularView<Upper>().template solveInPlace<OnTheRight>(A21);
if(rs>0) A22.template selfadjointView<Lower>().rankUpdate(A21,-1); // bottleneck
if(rs>0) A22.template selfadjointView<Lower>().rankUpdate(A21,typename NumTraits<RealScalar>::Literal(-1)); // bottleneck
}
return -1;
}
template<typename MatrixType, typename VectorType>
static typename MatrixType::Index rankUpdate(MatrixType& mat, const VectorType& vec, const RealScalar& sigma)
static Index rankUpdate(MatrixType& mat, const VectorType& vec, const RealScalar& sigma)
{
return Eigen::internal::llt_rank_update_lower(mat, vec, sigma);
}
};
template<typename Scalar> struct llt_inplace<Scalar, Upper>
{
typedef typename NumTraits<Scalar>::Real RealScalar;
template<typename MatrixType>
static EIGEN_STRONG_INLINE typename MatrixType::Index unblocked(MatrixType& mat)
static EIGEN_STRONG_INLINE Index unblocked(MatrixType& mat)
{
Transpose<MatrixType> matt(mat);
return llt_inplace<Scalar, Lower>::unblocked(matt);
}
template<typename MatrixType>
static EIGEN_STRONG_INLINE typename MatrixType::Index blocked(MatrixType& mat)
static EIGEN_STRONG_INLINE Index blocked(MatrixType& mat)
{
Transpose<MatrixType> matt(mat);
return llt_inplace<Scalar, Lower>::blocked(matt);
}
template<typename MatrixType, typename VectorType>
static typename MatrixType::Index rankUpdate(MatrixType& mat, const VectorType& vec, const RealScalar& sigma)
static Index rankUpdate(MatrixType& mat, const VectorType& vec, const RealScalar& sigma)
{
Transpose<MatrixType> matt(mat);
return llt_inplace<Scalar, Lower>::rankUpdate(matt, vec.conjugate(), sigma);
@@ -362,8 +395,8 @@ template<typename MatrixType> struct LLT_Traits<MatrixType,Lower>
{
typedef const TriangularView<const MatrixType, Lower> MatrixL;
typedef const TriangularView<const typename MatrixType::AdjointReturnType, Upper> MatrixU;
static inline MatrixL getL(const MatrixType& m) { return m; }
static inline MatrixU getU(const MatrixType& m) { return m.adjoint(); }
static inline MatrixL getL(const MatrixType& m) { return MatrixL(m); }
static inline MatrixU getU(const MatrixType& m) { return MatrixU(m.adjoint()); }
static bool inplace_decomposition(MatrixType& m)
{ return llt_inplace<typename MatrixType::Scalar, Lower>::blocked(m)==-1; }
};
@@ -372,8 +405,8 @@ template<typename MatrixType> struct LLT_Traits<MatrixType,Upper>
{
typedef const TriangularView<const typename MatrixType::AdjointReturnType, Lower> MatrixL;
typedef const TriangularView<const MatrixType, Upper> MatrixU;
static inline MatrixL getL(const MatrixType& m) { return m.adjoint(); }
static inline MatrixU getU(const MatrixType& m) { return m; }
static inline MatrixL getL(const MatrixType& m) { return MatrixL(m.adjoint()); }
static inline MatrixU getU(const MatrixType& m) { return MatrixU(m); }
static bool inplace_decomposition(MatrixType& m)
{ return llt_inplace<typename MatrixType::Scalar, Upper>::blocked(m)==-1; }
};
@@ -388,14 +421,29 @@ template<typename MatrixType> struct LLT_Traits<MatrixType,Upper>
* Output: \verbinclude TutorialLinAlgComputeTwice.out
*/
template<typename MatrixType, int _UpLo>
LLT<MatrixType,_UpLo>& LLT<MatrixType,_UpLo>::compute(const MatrixType& a)
template<typename InputType>
LLT<MatrixType,_UpLo>& LLT<MatrixType,_UpLo>::compute(const EigenBase<InputType>& a)
{
check_template_parameters();
eigen_assert(a.rows()==a.cols());
const Index size = a.rows();
m_matrix.resize(size, size);
m_matrix = a;
if (!internal::is_same_dense(m_matrix, a.derived()))
m_matrix = a.derived();
// Compute matrix L1 norm = max abs column sum.
m_l1_norm = RealScalar(0);
// TODO move this code to SelfAdjointView
for (Index col = 0; col < size; ++col) {
RealScalar abs_col_sum;
if (_UpLo == Lower)
abs_col_sum = m_matrix.col(col).tail(size - col).template lpNorm<1>() + m_matrix.row(col).head(col).template lpNorm<1>();
else
abs_col_sum = m_matrix.col(col).head(col).template lpNorm<1>() + m_matrix.row(col).tail(size - col).template lpNorm<1>();
if (abs_col_sum > m_l1_norm)
m_l1_norm = abs_col_sum;
}
m_isInitialized = true;
bool ok = Traits::inplace_decomposition(m_matrix);
@@ -423,39 +471,33 @@ LLT<_MatrixType,_UpLo> LLT<_MatrixType,_UpLo>::rankUpdate(const VectorType& v, c
return *this;
}
namespace internal {
template<typename _MatrixType, int UpLo, typename Rhs>
struct solve_retval<LLT<_MatrixType, UpLo>, Rhs>
: solve_retval_base<LLT<_MatrixType, UpLo>, Rhs>
{
typedef LLT<_MatrixType,UpLo> LLTType;
EIGEN_MAKE_SOLVE_HELPERS(LLTType,Rhs)
template<typename Dest> void evalTo(Dest& dst) const
{
dst = rhs();
dec().solveInPlace(dst);
}
};
#ifndef EIGEN_PARSED_BY_DOXYGEN
template<typename _MatrixType,int _UpLo>
template<typename RhsType, typename DstType>
void LLT<_MatrixType,_UpLo>::_solve_impl(const RhsType &rhs, DstType &dst) const
{
dst = rhs;
solveInPlace(dst);
}
#endif
/** \internal use x = llt_object.solve(x);
*
*
* This is the \em in-place version of solve().
*
* \param bAndX represents both the right-hand side matrix b and result x.
*
* \returns true always! If you need to check for existence of solutions, use another decomposition like LU, QR, or SVD.
* This version avoids a copy when the right hand side matrix b is not needed anymore.
*
* This version avoids a copy when the right hand side matrix b is not
* needed anymore.
* \warning The parameter is only marked 'const' to make the C++ compiler accept a temporary expression here.
* This function will const_cast it, so constness isn't honored here.
*
* \sa LLT::solve(), MatrixBase::llt()
*/
template<typename MatrixType, int _UpLo>
template<typename Derived>
void LLT<MatrixType,_UpLo>::solveInPlace(MatrixBase<Derived> &bAndX) const
void LLT<MatrixType,_UpLo>::solveInPlace(const MatrixBase<Derived> &bAndX) const
{
eigen_assert(m_isInitialized && "LLT is not initialized.");
eigen_assert(m_matrix.rows()==bAndX.rows());
@@ -475,6 +517,7 @@ MatrixType LLT<MatrixType,_UpLo>::reconstructedMatrix() const
/** \cholesky_module
* \returns the LLT decomposition of \c *this
* \sa SelfAdjointView::llt()
*/
template<typename Derived>
inline const LLT<typename MatrixBase<Derived>::PlainObject>
@@ -485,6 +528,7 @@ MatrixBase<Derived>::llt() const
/** \cholesky_module
* \returns the LLT decomposition of \c *this
* \sa SelfAdjointView::llt()
*/
template<typename MatrixType, unsigned int UpLo>
inline const LLT<typename SelfAdjointView<MatrixType, UpLo>::PlainObject, UpLo>

View File

@@ -0,0 +1,99 @@
/*
Copyright (c) 2011, Intel Corporation. All rights reserved.
Redistribution and use in source and binary forms, with or without modification,
are permitted provided that the following conditions are met:
* Redistributions of source code must retain the above copyright notice, this
list of conditions and the following disclaimer.
* Redistributions in binary form must reproduce the above copyright notice,
this list of conditions and the following disclaimer in the documentation
and/or other materials provided with the distribution.
* Neither the name of Intel Corporation nor the names of its contributors may
be used to endorse or promote products derived from this software without
specific prior written permission.
THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND
ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE FOR
ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
(INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON
ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
(INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
********************************************************************************
* Content : Eigen bindings to LAPACKe
* LLt decomposition based on LAPACKE_?potrf function.
********************************************************************************
*/
#ifndef EIGEN_LLT_LAPACKE_H
#define EIGEN_LLT_LAPACKE_H
namespace Eigen {
namespace internal {
template<typename Scalar> struct lapacke_llt;
#define EIGEN_LAPACKE_LLT(EIGTYPE, BLASTYPE, LAPACKE_PREFIX) \
template<> struct lapacke_llt<EIGTYPE> \
{ \
template<typename MatrixType> \
static inline Index potrf(MatrixType& m, char uplo) \
{ \
lapack_int matrix_order; \
lapack_int size, lda, info, StorageOrder; \
EIGTYPE* a; \
eigen_assert(m.rows()==m.cols()); \
/* Set up parameters for ?potrf */ \
size = convert_index<lapack_int>(m.rows()); \
StorageOrder = MatrixType::Flags&RowMajorBit?RowMajor:ColMajor; \
matrix_order = StorageOrder==RowMajor ? LAPACK_ROW_MAJOR : LAPACK_COL_MAJOR; \
a = &(m.coeffRef(0,0)); \
lda = convert_index<lapack_int>(m.outerStride()); \
\
info = LAPACKE_##LAPACKE_PREFIX##potrf( matrix_order, uplo, size, (BLASTYPE*)a, lda ); \
info = (info==0) ? -1 : info>0 ? info-1 : size; \
return info; \
} \
}; \
template<> struct llt_inplace<EIGTYPE, Lower> \
{ \
template<typename MatrixType> \
static Index blocked(MatrixType& m) \
{ \
return lapacke_llt<EIGTYPE>::potrf(m, 'L'); \
} \
template<typename MatrixType, typename VectorType> \
static Index rankUpdate(MatrixType& mat, const VectorType& vec, const typename MatrixType::RealScalar& sigma) \
{ return Eigen::internal::llt_rank_update_lower(mat, vec, sigma); } \
}; \
template<> struct llt_inplace<EIGTYPE, Upper> \
{ \
template<typename MatrixType> \
static Index blocked(MatrixType& m) \
{ \
return lapacke_llt<EIGTYPE>::potrf(m, 'U'); \
} \
template<typename MatrixType, typename VectorType> \
static Index rankUpdate(MatrixType& mat, const VectorType& vec, const typename MatrixType::RealScalar& sigma) \
{ \
Transpose<MatrixType> matt(mat); \
return llt_inplace<EIGTYPE, Lower>::rankUpdate(matt, vec.conjugate(), sigma); \
} \
};
EIGEN_LAPACKE_LLT(double, double, d)
EIGEN_LAPACKE_LLT(float, float, s)
EIGEN_LAPACKE_LLT(dcomplex, lapack_complex_double, z)
EIGEN_LAPACKE_LLT(scomplex, lapack_complex_float, c)
} // end namespace internal
} // end namespace Eigen
#endif // EIGEN_LLT_LAPACKE_H

View File

@@ -14,46 +14,52 @@ namespace Eigen {
namespace internal {
template<typename Scalar, typename CholmodType>
void cholmod_configure_matrix(CholmodType& mat)
{
if (internal::is_same<Scalar,float>::value)
{
mat.xtype = CHOLMOD_REAL;
mat.dtype = CHOLMOD_SINGLE;
}
else if (internal::is_same<Scalar,double>::value)
{
template<typename Scalar> struct cholmod_configure_matrix;
template<> struct cholmod_configure_matrix<double> {
template<typename CholmodType>
static void run(CholmodType& mat) {
mat.xtype = CHOLMOD_REAL;
mat.dtype = CHOLMOD_DOUBLE;
}
else if (internal::is_same<Scalar,std::complex<float> >::value)
{
mat.xtype = CHOLMOD_COMPLEX;
mat.dtype = CHOLMOD_SINGLE;
}
else if (internal::is_same<Scalar,std::complex<double> >::value)
{
};
template<> struct cholmod_configure_matrix<std::complex<double> > {
template<typename CholmodType>
static void run(CholmodType& mat) {
mat.xtype = CHOLMOD_COMPLEX;
mat.dtype = CHOLMOD_DOUBLE;
}
else
{
eigen_assert(false && "Scalar type not supported by CHOLMOD");
}
}
};
// Other scalar types are not yet suppotred by Cholmod
// template<> struct cholmod_configure_matrix<float> {
// template<typename CholmodType>
// static void run(CholmodType& mat) {
// mat.xtype = CHOLMOD_REAL;
// mat.dtype = CHOLMOD_SINGLE;
// }
// };
//
// template<> struct cholmod_configure_matrix<std::complex<float> > {
// template<typename CholmodType>
// static void run(CholmodType& mat) {
// mat.xtype = CHOLMOD_COMPLEX;
// mat.dtype = CHOLMOD_SINGLE;
// }
// };
} // namespace internal
/** Wraps the Eigen sparse matrix \a mat into a Cholmod sparse matrix object.
* Note that the data are shared.
*/
template<typename _Scalar, int _Options, typename _Index>
cholmod_sparse viewAsCholmod(SparseMatrix<_Scalar,_Options,_Index>& mat)
template<typename _Scalar, int _Options, typename _StorageIndex>
cholmod_sparse viewAsCholmod(Ref<SparseMatrix<_Scalar,_Options,_StorageIndex> > mat)
{
cholmod_sparse res;
res.nzmax = mat.nonZeros();
res.nrow = mat.rows();;
res.nrow = mat.rows();
res.ncol = mat.cols();
res.p = mat.outerIndexPtr();
res.i = mat.innerIndexPtr();
@@ -74,11 +80,11 @@ cholmod_sparse viewAsCholmod(SparseMatrix<_Scalar,_Options,_Index>& mat)
res.dtype = 0;
res.stype = -1;
if (internal::is_same<_Index,int>::value)
if (internal::is_same<_StorageIndex,int>::value)
{
res.itype = CHOLMOD_INT;
}
else if (internal::is_same<_Index,SuiteSparse_long>::value)
else if (internal::is_same<_StorageIndex,long>::value)
{
res.itype = CHOLMOD_LONG;
}
@@ -88,7 +94,7 @@ cholmod_sparse viewAsCholmod(SparseMatrix<_Scalar,_Options,_Index>& mat)
}
// setup res.xtype
internal::cholmod_configure_matrix<_Scalar>(res);
internal::cholmod_configure_matrix<_Scalar>::run(res);
res.stype = 0;
@@ -98,16 +104,23 @@ cholmod_sparse viewAsCholmod(SparseMatrix<_Scalar,_Options,_Index>& mat)
template<typename _Scalar, int _Options, typename _Index>
const cholmod_sparse viewAsCholmod(const SparseMatrix<_Scalar,_Options,_Index>& mat)
{
cholmod_sparse res = viewAsCholmod(mat.const_cast_derived());
cholmod_sparse res = viewAsCholmod(Ref<SparseMatrix<_Scalar,_Options,_Index> >(mat.const_cast_derived()));
return res;
}
template<typename _Scalar, int _Options, typename _Index>
const cholmod_sparse viewAsCholmod(const SparseVector<_Scalar,_Options,_Index>& mat)
{
cholmod_sparse res = viewAsCholmod(Ref<SparseMatrix<_Scalar,_Options,_Index> >(mat.const_cast_derived()));
return res;
}
/** Returns a view of the Eigen sparse matrix \a mat as Cholmod sparse matrix.
* The data are not copied but shared. */
template<typename _Scalar, int _Options, typename _Index, unsigned int UpLo>
cholmod_sparse viewAsCholmod(const SparseSelfAdjointView<SparseMatrix<_Scalar,_Options,_Index>, UpLo>& mat)
cholmod_sparse viewAsCholmod(const SparseSelfAdjointView<const SparseMatrix<_Scalar,_Options,_Index>, UpLo>& mat)
{
cholmod_sparse res = viewAsCholmod(mat.matrix().const_cast_derived());
cholmod_sparse res = viewAsCholmod(Ref<SparseMatrix<_Scalar,_Options,_Index> >(mat.matrix().const_cast_derived()));
if(UpLo==Upper) res.stype = 1;
if(UpLo==Lower) res.stype = -1;
@@ -131,19 +144,19 @@ cholmod_dense viewAsCholmod(MatrixBase<Derived>& mat)
res.x = (void*)(mat.derived().data());
res.z = 0;
internal::cholmod_configure_matrix<Scalar>(res);
internal::cholmod_configure_matrix<Scalar>::run(res);
return res;
}
/** Returns a view of the Cholmod sparse matrix \a cm as an Eigen sparse matrix.
* The data are not copied but shared. */
template<typename Scalar, int Flags, typename Index>
MappedSparseMatrix<Scalar,Flags,Index> viewAsEigen(cholmod_sparse& cm)
template<typename Scalar, int Flags, typename StorageIndex>
MappedSparseMatrix<Scalar,Flags,StorageIndex> viewAsEigen(cholmod_sparse& cm)
{
return MappedSparseMatrix<Scalar,Flags,Index>
(cm.nrow, cm.ncol, static_cast<Index*>(cm.p)[cm.ncol],
static_cast<Index*>(cm.p), static_cast<Index*>(cm.i),static_cast<Scalar*>(cm.x) );
return MappedSparseMatrix<Scalar,Flags,StorageIndex>
(cm.nrow, cm.ncol, static_cast<StorageIndex*>(cm.p)[cm.ncol],
static_cast<StorageIndex*>(cm.p), static_cast<StorageIndex*>(cm.i),static_cast<Scalar*>(cm.x) );
}
enum CholmodMode {
@@ -157,29 +170,39 @@ enum CholmodMode {
* \sa class CholmodSupernodalLLT, class CholmodSimplicialLDLT, class CholmodSimplicialLLT
*/
template<typename _MatrixType, int _UpLo, typename Derived>
class CholmodBase : internal::noncopyable
class CholmodBase : public SparseSolverBase<Derived>
{
protected:
typedef SparseSolverBase<Derived> Base;
using Base::derived;
using Base::m_isInitialized;
public:
typedef _MatrixType MatrixType;
enum { UpLo = _UpLo };
typedef typename MatrixType::Scalar Scalar;
typedef typename MatrixType::RealScalar RealScalar;
typedef MatrixType CholMatrixType;
typedef typename MatrixType::Index Index;
typedef typename MatrixType::StorageIndex StorageIndex;
enum {
ColsAtCompileTime = MatrixType::ColsAtCompileTime,
MaxColsAtCompileTime = MatrixType::MaxColsAtCompileTime
};
public:
CholmodBase()
: m_cholmodFactor(0), m_info(Success), m_isInitialized(false)
: m_cholmodFactor(0), m_info(Success), m_factorizationIsOk(false), m_analysisIsOk(false)
{
m_shiftOffset[0] = m_shiftOffset[1] = RealScalar(0.0);
EIGEN_STATIC_ASSERT((internal::is_same<double,RealScalar>::value), CHOLMOD_SUPPORTS_DOUBLE_PRECISION_ONLY);
m_shiftOffset[0] = m_shiftOffset[1] = 0.0;
cholmod_start(&m_cholmod);
}
CholmodBase(const MatrixType& matrix)
: m_cholmodFactor(0), m_info(Success), m_isInitialized(false)
explicit CholmodBase(const MatrixType& matrix)
: m_cholmodFactor(0), m_info(Success), m_factorizationIsOk(false), m_analysisIsOk(false)
{
m_shiftOffset[0] = m_shiftOffset[1] = RealScalar(0.0);
EIGEN_STATIC_ASSERT((internal::is_same<double,RealScalar>::value), CHOLMOD_SUPPORTS_DOUBLE_PRECISION_ONLY);
m_shiftOffset[0] = m_shiftOffset[1] = 0.0;
cholmod_start(&m_cholmod);
compute(matrix);
}
@@ -191,11 +214,8 @@ class CholmodBase : internal::noncopyable
cholmod_finish(&m_cholmod);
}
inline Index cols() const { return m_cholmodFactor->n; }
inline Index rows() const { return m_cholmodFactor->n; }
Derived& derived() { return *static_cast<Derived*>(this); }
const Derived& derived() const { return *static_cast<const Derived*>(this); }
inline StorageIndex cols() const { return internal::convert_index<StorageIndex, Index>(m_cholmodFactor->n); }
inline StorageIndex rows() const { return internal::convert_index<StorageIndex, Index>(m_cholmodFactor->n); }
/** \brief Reports whether previous computation was successful.
*
@@ -216,34 +236,6 @@ class CholmodBase : internal::noncopyable
return derived();
}
/** \returns the solution x of \f$ A x = b \f$ using the current decomposition of A.
*
* \sa compute()
*/
template<typename Rhs>
inline const internal::solve_retval<CholmodBase, Rhs>
solve(const MatrixBase<Rhs>& b) const
{
eigen_assert(m_isInitialized && "LLT is not initialized.");
eigen_assert(rows()==b.rows()
&& "CholmodDecomposition::solve(): invalid number of rows of the right hand side matrix b");
return internal::solve_retval<CholmodBase, Rhs>(*this, b.derived());
}
/** \returns the solution x of \f$ A x = b \f$ using the current decomposition of A.
*
* \sa compute()
*/
template<typename Rhs>
inline const internal::sparse_solve_retval<CholmodBase, Rhs>
solve(const SparseMatrixBase<Rhs>& b) const
{
eigen_assert(m_isInitialized && "LLT is not initialized.");
eigen_assert(rows()==b.rows()
&& "CholmodDecomposition::solve(): invalid number of rows of the right hand side matrix b");
return internal::sparse_solve_retval<CholmodBase, Rhs>(*this, b.derived());
}
/** Performs a symbolic decomposition on the sparsity pattern of \a matrix.
*
* This function is particularly useful when solving for several problems having the same structure.
@@ -277,7 +269,7 @@ class CholmodBase : internal::noncopyable
eigen_assert(m_analysisIsOk && "You must first call analyzePattern()");
cholmod_sparse A = viewAsCholmod(matrix.template selfadjointView<UpLo>());
cholmod_factorize_p(&A, m_shiftOffset, 0, 0, m_cholmodFactor, &m_cholmod);
// If the factorization failed, minor is the column at which it did. On success minor == n.
this->m_info = (m_cholmodFactor->minor == m_cholmodFactor->n ? Success : NumericalIssue);
m_factorizationIsOk = true;
@@ -290,20 +282,22 @@ class CholmodBase : internal::noncopyable
#ifndef EIGEN_PARSED_BY_DOXYGEN
/** \internal */
template<typename Rhs,typename Dest>
void _solve(const MatrixBase<Rhs> &b, MatrixBase<Dest> &dest) const
void _solve_impl(const MatrixBase<Rhs> &b, MatrixBase<Dest> &dest) const
{
eigen_assert(m_factorizationIsOk && "The decomposition is not in a valid state for solving, you must first call either compute() or symbolic()/numeric()");
const Index size = m_cholmodFactor->n;
EIGEN_UNUSED_VARIABLE(size);
eigen_assert(size==b.rows());
// Cholmod needs column-major stoarge without inner-stride, which corresponds to the default behavior of Ref.
Ref<const Matrix<typename Rhs::Scalar,Dynamic,Dynamic,ColMajor> > b_ref(b.derived());
// note: cd stands for Cholmod Dense
Rhs& b_ref(b.const_cast_derived());
cholmod_dense b_cd = viewAsCholmod(b_ref);
cholmod_dense* x_cd = cholmod_solve(CHOLMOD_A, m_cholmodFactor, &b_cd, &m_cholmod);
if(!x_cd)
{
this->m_info = NumericalIssue;
return;
}
// TODO optimize this copy by swapping when possible (be careful with alignment, etc.)
dest = Matrix<Scalar,Dest::RowsAtCompileTime,Dest::ColsAtCompileTime>::Map(reinterpret_cast<Scalar*>(x_cd->x),b.rows(),b.cols());
@@ -311,8 +305,8 @@ class CholmodBase : internal::noncopyable
}
/** \internal */
template<typename RhsScalar, int RhsOptions, typename RhsIndex, typename DestScalar, int DestOptions, typename DestIndex>
void _solve(const SparseMatrix<RhsScalar,RhsOptions,RhsIndex> &b, SparseMatrix<DestScalar,DestOptions,DestIndex> &dest) const
template<typename RhsDerived, typename DestDerived>
void _solve_impl(const SparseMatrixBase<RhsDerived> &b, SparseMatrixBase<DestDerived> &dest) const
{
eigen_assert(m_factorizationIsOk && "The decomposition is not in a valid state for solving, you must first call either compute() or symbolic()/numeric()");
const Index size = m_cholmodFactor->n;
@@ -320,14 +314,16 @@ class CholmodBase : internal::noncopyable
eigen_assert(size==b.rows());
// note: cs stands for Cholmod Sparse
cholmod_sparse b_cs = viewAsCholmod(b);
Ref<SparseMatrix<typename RhsDerived::Scalar,ColMajor,typename RhsDerived::StorageIndex> > b_ref(b.const_cast_derived());
cholmod_sparse b_cs = viewAsCholmod(b_ref);
cholmod_sparse* x_cs = cholmod_spsolve(CHOLMOD_A, m_cholmodFactor, &b_cs, &m_cholmod);
if(!x_cs)
{
this->m_info = NumericalIssue;
return;
}
// TODO optimize this copy by swapping when possible (be careful with alignment, etc.)
dest = viewAsEigen<DestScalar,DestOptions,DestIndex>(*x_cs);
dest.derived() = viewAsEigen<typename DestDerived::Scalar,ColMajor,typename DestDerived::StorageIndex>(*x_cs);
cholmod_free_sparse(&x_cs, &m_cholmod);
}
#endif // EIGEN_PARSED_BY_DOXYGEN
@@ -344,10 +340,61 @@ class CholmodBase : internal::noncopyable
*/
Derived& setShift(const RealScalar& offset)
{
m_shiftOffset[0] = offset;
m_shiftOffset[0] = double(offset);
return derived();
}
/** \returns the determinant of the underlying matrix from the current factorization */
Scalar determinant() const
{
using std::exp;
return exp(logDeterminant());
}
/** \returns the log determinant of the underlying matrix from the current factorization */
Scalar logDeterminant() const
{
using std::log;
using numext::real;
eigen_assert(m_factorizationIsOk && "The decomposition is not in a valid state for solving, you must first call either compute() or symbolic()/numeric()");
RealScalar logDet = 0;
Scalar *x = static_cast<Scalar*>(m_cholmodFactor->x);
if (m_cholmodFactor->is_super)
{
// Supernodal factorization stored as a packed list of dense column-major blocs,
// as described by the following structure:
// super[k] == index of the first column of the j-th super node
StorageIndex *super = static_cast<StorageIndex*>(m_cholmodFactor->super);
// pi[k] == offset to the description of row indices
StorageIndex *pi = static_cast<StorageIndex*>(m_cholmodFactor->pi);
// px[k] == offset to the respective dense block
StorageIndex *px = static_cast<StorageIndex*>(m_cholmodFactor->px);
Index nb_super_nodes = m_cholmodFactor->nsuper;
for (Index k=0; k < nb_super_nodes; ++k)
{
StorageIndex ncols = super[k + 1] - super[k];
StorageIndex nrows = pi[k + 1] - pi[k];
Map<const Array<Scalar,1,Dynamic>, 0, InnerStride<> > sk(x + px[k], ncols, InnerStride<>(nrows+1));
logDet += sk.real().log().sum();
}
}
else
{
// Simplicial factorization stored as standard CSC matrix.
StorageIndex *p = static_cast<StorageIndex*>(m_cholmodFactor->p);
Index size = m_cholmodFactor->n;
for (Index k=0; k<size; ++k)
logDet += log(real( x[p[k]] ));
}
if (m_cholmodFactor->is_ll)
logDet *= 2.0;
return logDet;
};
template<typename Stream>
void dumpMemory(Stream& /*s*/)
{}
@@ -355,9 +402,8 @@ class CholmodBase : internal::noncopyable
protected:
mutable cholmod_common m_cholmod;
cholmod_factor* m_cholmodFactor;
RealScalar m_shiftOffset[2];
double m_shiftOffset[2];
mutable ComputationInfo m_info;
bool m_isInitialized;
int m_factorizationIsOk;
int m_analysisIsOk;
};
@@ -376,9 +422,13 @@ class CholmodBase : internal::noncopyable
* \tparam _UpLo the triangular part that will be used for the computations. It can be Lower
* or Upper. Default is Lower.
*
* \implsparsesolverconcept
*
* This class supports all kind of SparseMatrix<>: row or column major; upper, lower, or both; compressed or non compressed.
*
* \sa \ref TutorialSparseDirectSolvers, class CholmodSupernodalLLT, class SimplicialLLT
* \warning Only double precision real and complex scalar types are supported by Cholmod.
*
* \sa \ref TutorialSparseSolverConcept, class CholmodSupernodalLLT, class SimplicialLLT
*/
template<typename _MatrixType, int _UpLo = Lower>
class CholmodSimplicialLLT : public CholmodBase<_MatrixType, _UpLo, CholmodSimplicialLLT<_MatrixType, _UpLo> >
@@ -395,7 +445,7 @@ class CholmodSimplicialLLT : public CholmodBase<_MatrixType, _UpLo, CholmodSimpl
CholmodSimplicialLLT(const MatrixType& matrix) : Base()
{
init();
Base::compute(matrix);
this->compute(matrix);
}
~CholmodSimplicialLLT() {}
@@ -423,9 +473,13 @@ class CholmodSimplicialLLT : public CholmodBase<_MatrixType, _UpLo, CholmodSimpl
* \tparam _UpLo the triangular part that will be used for the computations. It can be Lower
* or Upper. Default is Lower.
*
* \implsparsesolverconcept
*
* This class supports all kind of SparseMatrix<>: row or column major; upper, lower, or both; compressed or non compressed.
*
* \sa \ref TutorialSparseDirectSolvers, class CholmodSupernodalLLT, class SimplicialLDLT
* \warning Only double precision real and complex scalar types are supported by Cholmod.
*
* \sa \ref TutorialSparseSolverConcept, class CholmodSupernodalLLT, class SimplicialLDLT
*/
template<typename _MatrixType, int _UpLo = Lower>
class CholmodSimplicialLDLT : public CholmodBase<_MatrixType, _UpLo, CholmodSimplicialLDLT<_MatrixType, _UpLo> >
@@ -442,7 +496,7 @@ class CholmodSimplicialLDLT : public CholmodBase<_MatrixType, _UpLo, CholmodSimp
CholmodSimplicialLDLT(const MatrixType& matrix) : Base()
{
init();
Base::compute(matrix);
this->compute(matrix);
}
~CholmodSimplicialLDLT() {}
@@ -468,9 +522,13 @@ class CholmodSimplicialLDLT : public CholmodBase<_MatrixType, _UpLo, CholmodSimp
* \tparam _UpLo the triangular part that will be used for the computations. It can be Lower
* or Upper. Default is Lower.
*
* \implsparsesolverconcept
*
* This class supports all kind of SparseMatrix<>: row or column major; upper, lower, or both; compressed or non compressed.
*
* \sa \ref TutorialSparseDirectSolvers
* \warning Only double precision real and complex scalar types are supported by Cholmod.
*
* \sa \ref TutorialSparseSolverConcept
*/
template<typename _MatrixType, int _UpLo = Lower>
class CholmodSupernodalLLT : public CholmodBase<_MatrixType, _UpLo, CholmodSupernodalLLT<_MatrixType, _UpLo> >
@@ -487,7 +545,7 @@ class CholmodSupernodalLLT : public CholmodBase<_MatrixType, _UpLo, CholmodSuper
CholmodSupernodalLLT(const MatrixType& matrix) : Base()
{
init();
Base::compute(matrix);
this->compute(matrix);
}
~CholmodSupernodalLLT() {}
@@ -515,9 +573,13 @@ class CholmodSupernodalLLT : public CholmodBase<_MatrixType, _UpLo, CholmodSuper
* \tparam _UpLo the triangular part that will be used for the computations. It can be Lower
* or Upper. Default is Lower.
*
* \implsparsesolverconcept
*
* This class supports all kind of SparseMatrix<>: row or column major; upper, lower, or both; compressed or non compressed.
*
* \sa \ref TutorialSparseDirectSolvers
* \warning Only double precision real and complex scalar types are supported by Cholmod.
*
* \sa \ref TutorialSparseSolverConcept
*/
template<typename _MatrixType, int _UpLo = Lower>
class CholmodDecomposition : public CholmodBase<_MatrixType, _UpLo, CholmodDecomposition<_MatrixType, _UpLo> >
@@ -534,7 +596,7 @@ class CholmodDecomposition : public CholmodBase<_MatrixType, _UpLo, CholmodDecom
CholmodDecomposition(const MatrixType& matrix) : Base()
{
init();
Base::compute(matrix);
this->compute(matrix);
}
~CholmodDecomposition() {}
@@ -572,36 +634,6 @@ class CholmodDecomposition : public CholmodBase<_MatrixType, _UpLo, CholmodDecom
}
};
namespace internal {
template<typename _MatrixType, int _UpLo, typename Derived, typename Rhs>
struct solve_retval<CholmodBase<_MatrixType,_UpLo,Derived>, Rhs>
: solve_retval_base<CholmodBase<_MatrixType,_UpLo,Derived>, Rhs>
{
typedef CholmodBase<_MatrixType,_UpLo,Derived> Dec;
EIGEN_MAKE_SOLVE_HELPERS(Dec,Rhs)
template<typename Dest> void evalTo(Dest& dst) const
{
dec()._solve(rhs(),dst);
}
};
template<typename _MatrixType, int _UpLo, typename Derived, typename Rhs>
struct sparse_solve_retval<CholmodBase<_MatrixType,_UpLo,Derived>, Rhs>
: sparse_solve_retval_base<CholmodBase<_MatrixType,_UpLo,Derived>, Rhs>
{
typedef CholmodBase<_MatrixType,_UpLo,Derived> Dec;
EIGEN_MAKE_SPARSE_SOLVE_HELPERS(Dec,Rhs)
template<typename Dest> void evalTo(Dest& dst) const
{
dec()._solve(rhs(),dst);
}
};
} // end namespace internal
} // end namespace Eigen
#endif // EIGEN_CHOLMODSUPPORT_H

View File

@@ -12,7 +12,16 @@
namespace Eigen {
/** \class Array
namespace internal {
template<typename _Scalar, int _Rows, int _Cols, int _Options, int _MaxRows, int _MaxCols>
struct traits<Array<_Scalar, _Rows, _Cols, _Options, _MaxRows, _MaxCols> > : traits<Matrix<_Scalar, _Rows, _Cols, _Options, _MaxRows, _MaxCols> >
{
typedef ArrayXpr XprKind;
typedef ArrayBase<Array<_Scalar, _Rows, _Cols, _Options, _MaxRows, _MaxCols> > XprBase;
};
}
/** \class Array
* \ingroup Core_Module
*
* \brief General-purpose arrays with easy API for coefficient-wise operations
@@ -24,20 +33,14 @@ namespace Eigen {
* API for the %Matrix class provides easy access to linear-algebra
* operations.
*
* This class can be extended with the help of the plugin mechanism described on the page
* \ref TopicCustomizingEigen by defining the preprocessor symbol \c EIGEN_ARRAY_PLUGIN.
* See documentation of class Matrix for detailed information on the template parameters
* storage layout.
*
* \sa \ref TutorialArrayClass, \ref TopicClassHierarchy
* This class can be extended with the help of the plugin mechanism described on the page
* \ref TopicCustomizing_Plugins by defining the preprocessor symbol \c EIGEN_ARRAY_PLUGIN.
*
* \sa \blank \ref TutorialArrayClass, \ref TopicClassHierarchy
*/
namespace internal {
template<typename _Scalar, int _Rows, int _Cols, int _Options, int _MaxRows, int _MaxCols>
struct traits<Array<_Scalar, _Rows, _Cols, _Options, _MaxRows, _MaxCols> > : traits<Matrix<_Scalar, _Rows, _Cols, _Options, _MaxRows, _MaxCols> >
{
typedef ArrayXpr XprKind;
typedef ArrayBase<Array<_Scalar, _Rows, _Cols, _Options, _MaxRows, _MaxCols> > XprBase;
};
}
template<typename _Scalar, int _Rows, int _Cols, int _Options, int _MaxRows, int _MaxCols>
class Array
: public PlainObjectBase<Array<_Scalar, _Rows, _Cols, _Options, _MaxRows, _MaxCols> >
@@ -69,11 +72,27 @@ class Array
* the usage of 'using'. This should be done only for operator=.
*/
template<typename OtherDerived>
EIGEN_DEVICE_FUNC
EIGEN_STRONG_INLINE Array& operator=(const EigenBase<OtherDerived> &other)
{
return Base::operator=(other);
}
/** Set all the entries to \a value.
* \sa DenseBase::setConstant(), DenseBase::fill()
*/
/* This overload is needed because the usage of
* using Base::operator=;
* fails on MSVC. Since the code below is working with GCC and MSVC, we skipped
* the usage of 'using'. This should be done only for operator=.
*/
EIGEN_DEVICE_FUNC
EIGEN_STRONG_INLINE Array& operator=(const Scalar &value)
{
Base::setConstant(value);
return *this;
}
/** Copies the value of the expression \a other into \c *this with automatic resizing.
*
* *this might be resized to match the dimensions of \a other. If *this was a null matrix (not already initialized),
@@ -84,7 +103,8 @@ class Array
* remain row-vectors and vectors remain vectors.
*/
template<typename OtherDerived>
EIGEN_STRONG_INLINE Array& operator=(const ArrayBase<OtherDerived>& other)
EIGEN_DEVICE_FUNC
EIGEN_STRONG_INLINE Array& operator=(const DenseBase<OtherDerived>& other)
{
return Base::_set(other);
}
@@ -92,11 +112,12 @@ class Array
/** This is a special case of the templated operator=. Its purpose is to
* prevent a default operator= from hiding the templated operator=.
*/
EIGEN_DEVICE_FUNC
EIGEN_STRONG_INLINE Array& operator=(const Array& other)
{
return Base::_set(other);
}
/** Default constructor.
*
* For fixed-size matrices, does nothing.
@@ -107,6 +128,7 @@ class Array
*
* \sa resize(Index,Index)
*/
EIGEN_DEVICE_FUNC
EIGEN_STRONG_INLINE Array() : Base()
{
Base::_check_template_params();
@@ -116,6 +138,7 @@ class Array
#ifndef EIGEN_PARSED_BY_DOXYGEN
// FIXME is it still needed ??
/** \internal */
EIGEN_DEVICE_FUNC
Array(internal::constructor_without_unaligned_array_assert)
: Base(internal::constructor_without_unaligned_array_assert())
{
@@ -124,56 +147,62 @@ class Array
}
#endif
#ifdef EIGEN_HAVE_RVALUE_REFERENCES
Array(Array&& other)
#if EIGEN_HAS_RVALUE_REFERENCES
EIGEN_DEVICE_FUNC
Array(Array&& other) EIGEN_NOEXCEPT_IF(std::is_nothrow_move_constructible<Scalar>::value)
: Base(std::move(other))
{
Base::_check_template_params();
if (RowsAtCompileTime!=Dynamic && ColsAtCompileTime!=Dynamic)
Base::_set_noalias(other);
}
Array& operator=(Array&& other)
EIGEN_DEVICE_FUNC
Array& operator=(Array&& other) EIGEN_NOEXCEPT_IF(std::is_nothrow_move_assignable<Scalar>::value)
{
other.swap(*this);
return *this;
}
#endif
/** Constructs a vector or row-vector with given dimension. \only_for_vectors
*
* Note that this is only useful for dynamic-size vectors. For fixed-size vectors,
* it is redundant to pass the dimension here, so it makes more sense to use the default
* constructor Matrix() instead.
*/
EIGEN_STRONG_INLINE explicit Array(Index dim)
: Base(dim, RowsAtCompileTime == 1 ? 1 : dim, ColsAtCompileTime == 1 ? 1 : dim)
#ifndef EIGEN_PARSED_BY_DOXYGEN
template<typename T>
EIGEN_DEVICE_FUNC
EIGEN_STRONG_INLINE explicit Array(const T& x)
{
Base::_check_template_params();
EIGEN_STATIC_ASSERT_VECTOR_ONLY(Array)
eigen_assert(dim >= 0);
eigen_assert(SizeAtCompileTime == Dynamic || SizeAtCompileTime == dim);
EIGEN_INITIALIZE_COEFFS_IF_THAT_OPTION_IS_ENABLED
Base::template _init1<T>(x);
}
#ifndef EIGEN_PARSED_BY_DOXYGEN
template<typename T0, typename T1>
EIGEN_DEVICE_FUNC
EIGEN_STRONG_INLINE Array(const T0& val0, const T1& val1)
{
Base::_check_template_params();
this->template _init2<T0,T1>(val0, val1);
}
#else
/** constructs an uninitialized matrix with \a rows rows and \a cols columns.
/** \brief Constructs a fixed-sized array initialized with coefficients starting at \a data */
EIGEN_DEVICE_FUNC explicit Array(const Scalar *data);
/** Constructs a vector or row-vector with given dimension. \only_for_vectors
*
* This is useful for dynamic-size matrices. For fixed-size matrices,
* Note that this is only useful for dynamic-size vectors. For fixed-size vectors,
* it is redundant to pass the dimension here, so it makes more sense to use the default
* constructor Array() instead.
*/
EIGEN_DEVICE_FUNC
EIGEN_STRONG_INLINE explicit Array(Index dim);
/** constructs an initialized 1x1 Array with the given coefficient */
Array(const Scalar& value);
/** constructs an uninitialized array with \a rows rows and \a cols columns.
*
* This is useful for dynamic-size arrays. For fixed-size arrays,
* it is redundant to pass these parameters, so one should use the default constructor
* Matrix() instead. */
* Array() instead. */
Array(Index rows, Index cols);
/** constructs an initialized 2D vector with given coefficients */
Array(const Scalar& val0, const Scalar& val1);
#endif
/** constructs an initialized 3D vector with given coefficients */
EIGEN_DEVICE_FUNC
EIGEN_STRONG_INLINE Array(const Scalar& val0, const Scalar& val1, const Scalar& val2)
{
Base::_check_template_params();
@@ -183,6 +212,7 @@ class Array
m_storage.data()[2] = val2;
}
/** constructs an initialized 4D vector with given coefficients */
EIGEN_DEVICE_FUNC
EIGEN_STRONG_INLINE Array(const Scalar& val0, const Scalar& val1, const Scalar& val2, const Scalar& val3)
{
Base::_check_template_params();
@@ -193,51 +223,27 @@ class Array
m_storage.data()[3] = val3;
}
explicit Array(const Scalar *data);
/** Constructor copying the value of the expression \a other */
template<typename OtherDerived>
EIGEN_STRONG_INLINE Array(const ArrayBase<OtherDerived>& other)
: Base(other.rows() * other.cols(), other.rows(), other.cols())
{
Base::_check_template_params();
Base::_set_noalias(other);
}
/** Copy constructor */
EIGEN_DEVICE_FUNC
EIGEN_STRONG_INLINE Array(const Array& other)
: Base(other.rows() * other.cols(), other.rows(), other.cols())
{
Base::_check_template_params();
Base::_set_noalias(other);
}
/** Copy constructor with in-place evaluation */
template<typename OtherDerived>
EIGEN_STRONG_INLINE Array(const ReturnByValue<OtherDerived>& other)
{
Base::_check_template_params();
Base::resize(other.rows(), other.cols());
other.evalTo(*this);
}
: Base(other)
{ }
private:
struct PrivateType {};
public:
/** \sa MatrixBase::operator=(const EigenBase<OtherDerived>&) */
template<typename OtherDerived>
EIGEN_STRONG_INLINE Array(const EigenBase<OtherDerived> &other)
: Base(other.derived().rows() * other.derived().cols(), other.derived().rows(), other.derived().cols())
{
Base::_check_template_params();
Base::_resize_to_match(other);
*this = other;
}
EIGEN_DEVICE_FUNC
EIGEN_STRONG_INLINE Array(const EigenBase<OtherDerived> &other,
typename internal::enable_if<internal::is_convertible<typename OtherDerived::Scalar,Scalar>::value,
PrivateType>::type = PrivateType())
: Base(other.derived())
{ }
/** Override MatrixBase::swap() since for dynamic-sized matrices of same type it is enough to swap the
* data pointers.
*/
template<typename OtherDerived>
void swap(ArrayBase<OtherDerived> const & other)
{ this->_swap(other.derived()); }
inline Index innerStride() const { return 1; }
inline Index outerStride() const { return this->innerSize(); }
EIGEN_DEVICE_FUNC inline Index innerStride() const { return 1; }
EIGEN_DEVICE_FUNC inline Index outerStride() const { return this->innerSize(); }
#ifdef EIGEN_ARRAY_PLUGIN
#include EIGEN_ARRAY_PLUGIN

View File

@@ -32,7 +32,7 @@ template<typename ExpressionType> class MatrixWrapper;
* \tparam Derived is the derived type, e.g., an array or an expression type.
*
* This class can be extended with the help of the plugin mechanism described on the page
* \ref TopicCustomizingEigen by defining the preprocessor symbol \c EIGEN_ARRAYBASE_PLUGIN.
* \ref TopicCustomizing_Plugins by defining the preprocessor symbol \c EIGEN_ARRAYBASE_PLUGIN.
*
* \sa class MatrixBase, \ref TopicClassHierarchy
*/
@@ -47,13 +47,11 @@ template<typename Derived> class ArrayBase
typedef ArrayBase Eigen_BaseClassForSpecializationOfGlobalMathFuncImpl;
typedef typename internal::traits<Derived>::StorageKind StorageKind;
typedef typename internal::traits<Derived>::Index Index;
typedef typename internal::traits<Derived>::Scalar Scalar;
typedef typename internal::packet_traits<Scalar>::type PacketScalar;
typedef typename NumTraits<Scalar>::Real RealScalar;
typedef DenseBase<Derived> Base;
using Base::operator*;
using Base::RowsAtCompileTime;
using Base::ColsAtCompileTime;
using Base::SizeAtCompileTime;
@@ -62,8 +60,7 @@ template<typename Derived> class ArrayBase
using Base::MaxSizeAtCompileTime;
using Base::IsVectorAtCompileTime;
using Base::Flags;
using Base::CoeffReadCost;
using Base::derived;
using Base::const_cast_derived;
using Base::rows;
@@ -83,25 +80,14 @@ template<typename Derived> class ArrayBase
#endif // not EIGEN_PARSED_BY_DOXYGEN
#ifndef EIGEN_PARSED_BY_DOXYGEN
/** \internal the plain matrix type corresponding to this expression. Note that is not necessarily
* exactly the return type of eval(): in the case of plain matrices, the return type of eval() is a const
* reference to a matrix, not a matrix! It is however guaranteed that the return type of eval() is either
* PlainObject or const PlainObject&.
*/
typedef Array<typename internal::traits<Derived>::Scalar,
internal::traits<Derived>::RowsAtCompileTime,
internal::traits<Derived>::ColsAtCompileTime,
AutoAlign | (internal::traits<Derived>::Flags&RowMajorBit ? RowMajor : ColMajor),
internal::traits<Derived>::MaxRowsAtCompileTime,
internal::traits<Derived>::MaxColsAtCompileTime
> PlainObject;
typedef typename Base::PlainObject PlainObject;
/** \internal Represents a matrix with all coefficients equal to one another*/
typedef CwiseNullaryOp<internal::scalar_constant_op<Scalar>,Derived> ConstantReturnType;
typedef CwiseNullaryOp<internal::scalar_constant_op<Scalar>,PlainObject> ConstantReturnType;
#endif // not EIGEN_PARSED_BY_DOXYGEN
#define EIGEN_CURRENT_STORAGE_BASE_CLASS Eigen::ArrayBase
#define EIGEN_DOC_UNARY_ADDONS(X,Y)
# include "../plugins/CommonCwiseUnaryOps.h"
# include "../plugins/MatrixCwiseUnaryOps.h"
# include "../plugins/ArrayCwiseUnaryOps.h"
@@ -112,44 +98,62 @@ template<typename Derived> class ArrayBase
# include EIGEN_ARRAYBASE_PLUGIN
# endif
#undef EIGEN_CURRENT_STORAGE_BASE_CLASS
#undef EIGEN_DOC_UNARY_ADDONS
/** Special case of the template operator=, in order to prevent the compiler
* from generating a default operator= (issue hit with g++ 4.1)
*/
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE
Derived& operator=(const ArrayBase& other)
{
return internal::assign_selector<Derived,Derived>::run(derived(), other.derived());
internal::call_assignment(derived(), other.derived());
return derived();
}
/** Set all the entries to \a value.
* \sa DenseBase::setConstant(), DenseBase::fill() */
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE
Derived& operator=(const Scalar &value)
{ Base::setConstant(value); return derived(); }
Derived& operator+=(const Scalar& scalar)
{ return *this = derived() + scalar; }
Derived& operator-=(const Scalar& scalar)
{ return *this = derived() - scalar; }
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE
Derived& operator+=(const Scalar& scalar);
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE
Derived& operator-=(const Scalar& scalar);
template<typename OtherDerived>
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE
Derived& operator+=(const ArrayBase<OtherDerived>& other);
template<typename OtherDerived>
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE
Derived& operator-=(const ArrayBase<OtherDerived>& other);
template<typename OtherDerived>
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE
Derived& operator*=(const ArrayBase<OtherDerived>& other);
template<typename OtherDerived>
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE
Derived& operator/=(const ArrayBase<OtherDerived>& other);
public:
EIGEN_DEVICE_FUNC
ArrayBase<Derived>& array() { return *this; }
EIGEN_DEVICE_FUNC
const ArrayBase<Derived>& array() const { return *this; }
/** \returns an \link Eigen::MatrixBase Matrix \endlink expression of this array
* \sa MatrixBase::array() */
MatrixWrapper<Derived> matrix() { return derived(); }
const MatrixWrapper<const Derived> matrix() const { return derived(); }
EIGEN_DEVICE_FUNC
MatrixWrapper<Derived> matrix() { return MatrixWrapper<Derived>(derived()); }
EIGEN_DEVICE_FUNC
const MatrixWrapper<const Derived> matrix() const { return MatrixWrapper<const Derived>(derived()); }
// template<typename Dest>
// inline void evalTo(Dest& dst) const { dst = matrix(); }
protected:
EIGEN_DEVICE_FUNC
ArrayBase() : Base() {}
private:
@@ -171,11 +175,10 @@ template<typename Derived> class ArrayBase
*/
template<typename Derived>
template<typename OtherDerived>
EIGEN_STRONG_INLINE Derived &
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE Derived &
ArrayBase<Derived>::operator-=(const ArrayBase<OtherDerived> &other)
{
SelfCwiseBinaryOp<internal::scalar_difference_op<Scalar>, Derived, OtherDerived> tmp(derived());
tmp = other.derived();
call_assignment(derived(), other.derived(), internal::sub_assign_op<Scalar,typename OtherDerived::Scalar>());
return derived();
}
@@ -185,11 +188,10 @@ ArrayBase<Derived>::operator-=(const ArrayBase<OtherDerived> &other)
*/
template<typename Derived>
template<typename OtherDerived>
EIGEN_STRONG_INLINE Derived &
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE Derived &
ArrayBase<Derived>::operator+=(const ArrayBase<OtherDerived>& other)
{
SelfCwiseBinaryOp<internal::scalar_sum_op<Scalar>, Derived, OtherDerived> tmp(derived());
tmp = other.derived();
call_assignment(derived(), other.derived(), internal::add_assign_op<Scalar,typename OtherDerived::Scalar>());
return derived();
}
@@ -199,11 +201,10 @@ ArrayBase<Derived>::operator+=(const ArrayBase<OtherDerived>& other)
*/
template<typename Derived>
template<typename OtherDerived>
EIGEN_STRONG_INLINE Derived &
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE Derived &
ArrayBase<Derived>::operator*=(const ArrayBase<OtherDerived>& other)
{
SelfCwiseBinaryOp<internal::scalar_product_op<Scalar>, Derived, OtherDerived> tmp(derived());
tmp = other.derived();
call_assignment(derived(), other.derived(), internal::mul_assign_op<Scalar,typename OtherDerived::Scalar>());
return derived();
}
@@ -213,11 +214,10 @@ ArrayBase<Derived>::operator*=(const ArrayBase<OtherDerived>& other)
*/
template<typename Derived>
template<typename OtherDerived>
EIGEN_STRONG_INLINE Derived &
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE Derived &
ArrayBase<Derived>::operator/=(const ArrayBase<OtherDerived>& other)
{
SelfCwiseBinaryOp<internal::scalar_quotient_op<Scalar>, Derived, OtherDerived> tmp(derived());
tmp = other.derived();
call_assignment(derived(), other.derived(), internal::div_assign_op<Scalar,typename OtherDerived::Scalar>());
return derived();
}

View File

@@ -32,7 +32,8 @@ struct traits<ArrayWrapper<ExpressionType> >
// Let's remove NestByRefBit
enum {
Flags0 = traits<typename remove_all<typename ExpressionType::Nested>::type >::Flags,
Flags = Flags0 & ~NestByRefBit
LvalueBitFlag = is_lvalue<ExpressionType>::value ? LvalueBit : 0,
Flags = (Flags0 & ~(NestByRefBit | LvalueBit)) | LvalueBitFlag
};
};
}
@@ -44,6 +45,7 @@ class ArrayWrapper : public ArrayBase<ArrayWrapper<ExpressionType> >
typedef ArrayBase<ArrayWrapper> Base;
EIGEN_DENSE_PUBLIC_INTERFACE(ArrayWrapper)
EIGEN_INHERIT_ASSIGNMENT_OPERATORS(ArrayWrapper)
typedef typename internal::remove_all<ExpressionType>::type NestedExpression;
typedef typename internal::conditional<
internal::is_lvalue<ExpressionType>::value,
@@ -51,76 +53,45 @@ class ArrayWrapper : public ArrayBase<ArrayWrapper<ExpressionType> >
const Scalar
>::type ScalarWithConstIfNotLvalue;
typedef typename internal::nested<ExpressionType>::type NestedExpressionType;
typedef typename internal::ref_selector<ExpressionType>::non_const_type NestedExpressionType;
inline ArrayWrapper(ExpressionType& matrix) : m_expression(matrix) {}
using Base::coeffRef;
EIGEN_DEVICE_FUNC
explicit EIGEN_STRONG_INLINE ArrayWrapper(ExpressionType& matrix) : m_expression(matrix) {}
EIGEN_DEVICE_FUNC
inline Index rows() const { return m_expression.rows(); }
EIGEN_DEVICE_FUNC
inline Index cols() const { return m_expression.cols(); }
EIGEN_DEVICE_FUNC
inline Index outerStride() const { return m_expression.outerStride(); }
EIGEN_DEVICE_FUNC
inline Index innerStride() const { return m_expression.innerStride(); }
inline ScalarWithConstIfNotLvalue* data() { return m_expression.const_cast_derived().data(); }
EIGEN_DEVICE_FUNC
inline ScalarWithConstIfNotLvalue* data() { return m_expression.data(); }
EIGEN_DEVICE_FUNC
inline const Scalar* data() const { return m_expression.data(); }
inline CoeffReturnType coeff(Index rowId, Index colId) const
{
return m_expression.coeff(rowId, colId);
}
inline Scalar& coeffRef(Index rowId, Index colId)
{
return m_expression.const_cast_derived().coeffRef(rowId, colId);
}
EIGEN_DEVICE_FUNC
inline const Scalar& coeffRef(Index rowId, Index colId) const
{
return m_expression.const_cast_derived().coeffRef(rowId, colId);
}
inline CoeffReturnType coeff(Index index) const
{
return m_expression.coeff(index);
}
inline Scalar& coeffRef(Index index)
{
return m_expression.const_cast_derived().coeffRef(index);
return m_expression.coeffRef(rowId, colId);
}
EIGEN_DEVICE_FUNC
inline const Scalar& coeffRef(Index index) const
{
return m_expression.const_cast_derived().coeffRef(index);
}
template<int LoadMode>
inline const PacketScalar packet(Index rowId, Index colId) const
{
return m_expression.template packet<LoadMode>(rowId, colId);
}
template<int LoadMode>
inline void writePacket(Index rowId, Index colId, const PacketScalar& val)
{
m_expression.const_cast_derived().template writePacket<LoadMode>(rowId, colId, val);
}
template<int LoadMode>
inline const PacketScalar packet(Index index) const
{
return m_expression.template packet<LoadMode>(index);
}
template<int LoadMode>
inline void writePacket(Index index, const PacketScalar& val)
{
m_expression.const_cast_derived().template writePacket<LoadMode>(index, val);
return m_expression.coeffRef(index);
}
template<typename Dest>
EIGEN_DEVICE_FUNC
inline void evalTo(Dest& dst) const { dst = m_expression; }
const typename internal::remove_all<NestedExpressionType>::type&
EIGEN_DEVICE_FUNC
nestedExpression() const
{
return m_expression;
@@ -128,10 +99,12 @@ class ArrayWrapper : public ArrayBase<ArrayWrapper<ExpressionType> >
/** Forwards the resizing request to the nested expression
* \sa DenseBase::resize(Index) */
void resize(Index newSize) { m_expression.const_cast_derived().resize(newSize); }
EIGEN_DEVICE_FUNC
void resize(Index newSize) { m_expression.resize(newSize); }
/** Forwards the resizing request to the nested expression
* \sa DenseBase::resize(Index,Index)*/
void resize(Index nbRows, Index nbCols) { m_expression.const_cast_derived().resize(nbRows,nbCols); }
EIGEN_DEVICE_FUNC
void resize(Index rows, Index cols) { m_expression.resize(rows,cols); }
protected:
NestedExpressionType m_expression;
@@ -157,7 +130,8 @@ struct traits<MatrixWrapper<ExpressionType> >
// Let's remove NestByRefBit
enum {
Flags0 = traits<typename remove_all<typename ExpressionType::Nested>::type >::Flags,
Flags = Flags0 & ~NestByRefBit
LvalueBitFlag = is_lvalue<ExpressionType>::value ? LvalueBit : 0,
Flags = (Flags0 & ~(NestByRefBit | LvalueBit)) | LvalueBitFlag
};
};
}
@@ -169,6 +143,7 @@ class MatrixWrapper : public MatrixBase<MatrixWrapper<ExpressionType> >
typedef MatrixBase<MatrixWrapper<ExpressionType> > Base;
EIGEN_DENSE_PUBLIC_INTERFACE(MatrixWrapper)
EIGEN_INHERIT_ASSIGNMENT_OPERATORS(MatrixWrapper)
typedef typename internal::remove_all<ExpressionType>::type NestedExpression;
typedef typename internal::conditional<
internal::is_lvalue<ExpressionType>::value,
@@ -176,72 +151,40 @@ class MatrixWrapper : public MatrixBase<MatrixWrapper<ExpressionType> >
const Scalar
>::type ScalarWithConstIfNotLvalue;
typedef typename internal::nested<ExpressionType>::type NestedExpressionType;
typedef typename internal::ref_selector<ExpressionType>::non_const_type NestedExpressionType;
inline MatrixWrapper(ExpressionType& a_matrix) : m_expression(a_matrix) {}
using Base::coeffRef;
EIGEN_DEVICE_FUNC
explicit inline MatrixWrapper(ExpressionType& matrix) : m_expression(matrix) {}
EIGEN_DEVICE_FUNC
inline Index rows() const { return m_expression.rows(); }
EIGEN_DEVICE_FUNC
inline Index cols() const { return m_expression.cols(); }
EIGEN_DEVICE_FUNC
inline Index outerStride() const { return m_expression.outerStride(); }
EIGEN_DEVICE_FUNC
inline Index innerStride() const { return m_expression.innerStride(); }
inline ScalarWithConstIfNotLvalue* data() { return m_expression.const_cast_derived().data(); }
EIGEN_DEVICE_FUNC
inline ScalarWithConstIfNotLvalue* data() { return m_expression.data(); }
EIGEN_DEVICE_FUNC
inline const Scalar* data() const { return m_expression.data(); }
inline CoeffReturnType coeff(Index rowId, Index colId) const
{
return m_expression.coeff(rowId, colId);
}
inline Scalar& coeffRef(Index rowId, Index colId)
{
return m_expression.const_cast_derived().coeffRef(rowId, colId);
}
EIGEN_DEVICE_FUNC
inline const Scalar& coeffRef(Index rowId, Index colId) const
{
return m_expression.derived().coeffRef(rowId, colId);
}
inline CoeffReturnType coeff(Index index) const
{
return m_expression.coeff(index);
}
inline Scalar& coeffRef(Index index)
{
return m_expression.const_cast_derived().coeffRef(index);
}
EIGEN_DEVICE_FUNC
inline const Scalar& coeffRef(Index index) const
{
return m_expression.const_cast_derived().coeffRef(index);
}
template<int LoadMode>
inline const PacketScalar packet(Index rowId, Index colId) const
{
return m_expression.template packet<LoadMode>(rowId, colId);
}
template<int LoadMode>
inline void writePacket(Index rowId, Index colId, const PacketScalar& val)
{
m_expression.const_cast_derived().template writePacket<LoadMode>(rowId, colId, val);
}
template<int LoadMode>
inline const PacketScalar packet(Index index) const
{
return m_expression.template packet<LoadMode>(index);
}
template<int LoadMode>
inline void writePacket(Index index, const PacketScalar& val)
{
m_expression.const_cast_derived().template writePacket<LoadMode>(index, val);
return m_expression.coeffRef(index);
}
EIGEN_DEVICE_FUNC
const typename internal::remove_all<NestedExpressionType>::type&
nestedExpression() const
{
@@ -250,10 +193,12 @@ class MatrixWrapper : public MatrixBase<MatrixWrapper<ExpressionType> >
/** Forwards the resizing request to the nested expression
* \sa DenseBase::resize(Index) */
void resize(Index newSize) { m_expression.const_cast_derived().resize(newSize); }
EIGEN_DEVICE_FUNC
void resize(Index newSize) { m_expression.resize(newSize); }
/** Forwards the resizing request to the nested expression
* \sa DenseBase::resize(Index,Index)*/
void resize(Index nbRows, Index nbCols) { m_expression.const_cast_derived().resize(nbRows,nbCols); }
EIGEN_DEVICE_FUNC
void resize(Index rows, Index cols) { m_expression.resize(rows,cols); }
protected:
NestedExpressionType m_expression;

View File

@@ -14,478 +14,6 @@
namespace Eigen {
namespace internal {
/***************************************************************************
* Part 1 : the logic deciding a strategy for traversal and unrolling *
***************************************************************************/
template <typename Derived, typename OtherDerived>
struct assign_traits
{
public:
enum {
DstIsAligned = Derived::Flags & AlignedBit,
DstHasDirectAccess = Derived::Flags & DirectAccessBit,
SrcIsAligned = OtherDerived::Flags & AlignedBit,
JointAlignment = bool(DstIsAligned) && bool(SrcIsAligned) ? Aligned : Unaligned
};
private:
enum {
InnerSize = int(Derived::IsVectorAtCompileTime) ? int(Derived::SizeAtCompileTime)
: int(Derived::Flags)&RowMajorBit ? int(Derived::ColsAtCompileTime)
: int(Derived::RowsAtCompileTime),
InnerMaxSize = int(Derived::IsVectorAtCompileTime) ? int(Derived::MaxSizeAtCompileTime)
: int(Derived::Flags)&RowMajorBit ? int(Derived::MaxColsAtCompileTime)
: int(Derived::MaxRowsAtCompileTime),
MaxSizeAtCompileTime = Derived::SizeAtCompileTime,
PacketSize = packet_traits<typename Derived::Scalar>::size
};
enum {
StorageOrdersAgree = (int(Derived::IsRowMajor) == int(OtherDerived::IsRowMajor)),
MightVectorize = StorageOrdersAgree
&& (int(Derived::Flags) & int(OtherDerived::Flags) & ActualPacketAccessBit),
MayInnerVectorize = MightVectorize && int(InnerSize)!=Dynamic && int(InnerSize)%int(PacketSize)==0
&& int(DstIsAligned) && int(SrcIsAligned),
MayLinearize = StorageOrdersAgree && (int(Derived::Flags) & int(OtherDerived::Flags) & LinearAccessBit),
MayLinearVectorize = MightVectorize && MayLinearize && DstHasDirectAccess
&& (DstIsAligned || MaxSizeAtCompileTime == Dynamic),
/* If the destination isn't aligned, we have to do runtime checks and we don't unroll,
so it's only good for large enough sizes. */
MaySliceVectorize = MightVectorize && DstHasDirectAccess
&& (int(InnerMaxSize)==Dynamic || int(InnerMaxSize)>=3*PacketSize)
/* slice vectorization can be slow, so we only want it if the slices are big, which is
indicated by InnerMaxSize rather than InnerSize, think of the case of a dynamic block
in a fixed-size matrix */
};
public:
enum {
Traversal = int(MayInnerVectorize) ? int(InnerVectorizedTraversal)
: int(MayLinearVectorize) ? int(LinearVectorizedTraversal)
: int(MaySliceVectorize) ? int(SliceVectorizedTraversal)
: int(MayLinearize) ? int(LinearTraversal)
: int(DefaultTraversal),
Vectorized = int(Traversal) == InnerVectorizedTraversal
|| int(Traversal) == LinearVectorizedTraversal
|| int(Traversal) == SliceVectorizedTraversal
};
private:
enum {
UnrollingLimit = EIGEN_UNROLLING_LIMIT * (Vectorized ? int(PacketSize) : 1),
MayUnrollCompletely = int(Derived::SizeAtCompileTime) != Dynamic
&& int(OtherDerived::CoeffReadCost) != Dynamic
&& int(Derived::SizeAtCompileTime) * int(OtherDerived::CoeffReadCost) <= int(UnrollingLimit),
MayUnrollInner = int(InnerSize) != Dynamic
&& int(OtherDerived::CoeffReadCost) != Dynamic
&& int(InnerSize) * int(OtherDerived::CoeffReadCost) <= int(UnrollingLimit)
};
public:
enum {
Unrolling = (int(Traversal) == int(InnerVectorizedTraversal) || int(Traversal) == int(DefaultTraversal))
? (
int(MayUnrollCompletely) ? int(CompleteUnrolling)
: int(MayUnrollInner) ? int(InnerUnrolling)
: int(NoUnrolling)
)
: int(Traversal) == int(LinearVectorizedTraversal)
? ( bool(MayUnrollCompletely) && bool(DstIsAligned) ? int(CompleteUnrolling) : int(NoUnrolling) )
: int(Traversal) == int(LinearTraversal)
? ( bool(MayUnrollCompletely) ? int(CompleteUnrolling) : int(NoUnrolling) )
: int(NoUnrolling)
};
#ifdef EIGEN_DEBUG_ASSIGN
static void debug()
{
EIGEN_DEBUG_VAR(DstIsAligned)
EIGEN_DEBUG_VAR(SrcIsAligned)
EIGEN_DEBUG_VAR(JointAlignment)
EIGEN_DEBUG_VAR(InnerSize)
EIGEN_DEBUG_VAR(InnerMaxSize)
EIGEN_DEBUG_VAR(PacketSize)
EIGEN_DEBUG_VAR(StorageOrdersAgree)
EIGEN_DEBUG_VAR(MightVectorize)
EIGEN_DEBUG_VAR(MayLinearize)
EIGEN_DEBUG_VAR(MayInnerVectorize)
EIGEN_DEBUG_VAR(MayLinearVectorize)
EIGEN_DEBUG_VAR(MaySliceVectorize)
EIGEN_DEBUG_VAR(Traversal)
EIGEN_DEBUG_VAR(UnrollingLimit)
EIGEN_DEBUG_VAR(MayUnrollCompletely)
EIGEN_DEBUG_VAR(MayUnrollInner)
EIGEN_DEBUG_VAR(Unrolling)
}
#endif
};
/***************************************************************************
* Part 2 : meta-unrollers
***************************************************************************/
/************************
*** Default traversal ***
************************/
template<typename Derived1, typename Derived2, int Index, int Stop>
struct assign_DefaultTraversal_CompleteUnrolling
{
enum {
outer = Index / Derived1::InnerSizeAtCompileTime,
inner = Index % Derived1::InnerSizeAtCompileTime
};
static EIGEN_STRONG_INLINE void run(Derived1 &dst, const Derived2 &src)
{
dst.copyCoeffByOuterInner(outer, inner, src);
assign_DefaultTraversal_CompleteUnrolling<Derived1, Derived2, Index+1, Stop>::run(dst, src);
}
};
template<typename Derived1, typename Derived2, int Stop>
struct assign_DefaultTraversal_CompleteUnrolling<Derived1, Derived2, Stop, Stop>
{
static EIGEN_STRONG_INLINE void run(Derived1 &, const Derived2 &) {}
};
template<typename Derived1, typename Derived2, int Index, int Stop>
struct assign_DefaultTraversal_InnerUnrolling
{
static EIGEN_STRONG_INLINE void run(Derived1 &dst, const Derived2 &src, typename Derived1::Index outer)
{
dst.copyCoeffByOuterInner(outer, Index, src);
assign_DefaultTraversal_InnerUnrolling<Derived1, Derived2, Index+1, Stop>::run(dst, src, outer);
}
};
template<typename Derived1, typename Derived2, int Stop>
struct assign_DefaultTraversal_InnerUnrolling<Derived1, Derived2, Stop, Stop>
{
static EIGEN_STRONG_INLINE void run(Derived1 &, const Derived2 &, typename Derived1::Index) {}
};
/***********************
*** Linear traversal ***
***********************/
template<typename Derived1, typename Derived2, int Index, int Stop>
struct assign_LinearTraversal_CompleteUnrolling
{
static EIGEN_STRONG_INLINE void run(Derived1 &dst, const Derived2 &src)
{
dst.copyCoeff(Index, src);
assign_LinearTraversal_CompleteUnrolling<Derived1, Derived2, Index+1, Stop>::run(dst, src);
}
};
template<typename Derived1, typename Derived2, int Stop>
struct assign_LinearTraversal_CompleteUnrolling<Derived1, Derived2, Stop, Stop>
{
static EIGEN_STRONG_INLINE void run(Derived1 &, const Derived2 &) {}
};
/**************************
*** Inner vectorization ***
**************************/
template<typename Derived1, typename Derived2, int Index, int Stop>
struct assign_innervec_CompleteUnrolling
{
enum {
outer = Index / Derived1::InnerSizeAtCompileTime,
inner = Index % Derived1::InnerSizeAtCompileTime,
JointAlignment = assign_traits<Derived1,Derived2>::JointAlignment
};
static EIGEN_STRONG_INLINE void run(Derived1 &dst, const Derived2 &src)
{
dst.template copyPacketByOuterInner<Derived2, Aligned, JointAlignment>(outer, inner, src);
assign_innervec_CompleteUnrolling<Derived1, Derived2,
Index+packet_traits<typename Derived1::Scalar>::size, Stop>::run(dst, src);
}
};
template<typename Derived1, typename Derived2, int Stop>
struct assign_innervec_CompleteUnrolling<Derived1, Derived2, Stop, Stop>
{
static EIGEN_STRONG_INLINE void run(Derived1 &, const Derived2 &) {}
};
template<typename Derived1, typename Derived2, int Index, int Stop>
struct assign_innervec_InnerUnrolling
{
static EIGEN_STRONG_INLINE void run(Derived1 &dst, const Derived2 &src, typename Derived1::Index outer)
{
dst.template copyPacketByOuterInner<Derived2, Aligned, Aligned>(outer, Index, src);
assign_innervec_InnerUnrolling<Derived1, Derived2,
Index+packet_traits<typename Derived1::Scalar>::size, Stop>::run(dst, src, outer);
}
};
template<typename Derived1, typename Derived2, int Stop>
struct assign_innervec_InnerUnrolling<Derived1, Derived2, Stop, Stop>
{
static EIGEN_STRONG_INLINE void run(Derived1 &, const Derived2 &, typename Derived1::Index) {}
};
/***************************************************************************
* Part 3 : implementation of all cases
***************************************************************************/
template<typename Derived1, typename Derived2,
int Traversal = assign_traits<Derived1, Derived2>::Traversal,
int Unrolling = assign_traits<Derived1, Derived2>::Unrolling,
int Version = Specialized>
struct assign_impl;
/************************
*** Default traversal ***
************************/
template<typename Derived1, typename Derived2, int Unrolling, int Version>
struct assign_impl<Derived1, Derived2, InvalidTraversal, Unrolling, Version>
{
static inline void run(Derived1 &, const Derived2 &) { }
};
template<typename Derived1, typename Derived2, int Version>
struct assign_impl<Derived1, Derived2, DefaultTraversal, NoUnrolling, Version>
{
typedef typename Derived1::Index Index;
static inline void run(Derived1 &dst, const Derived2 &src)
{
const Index innerSize = dst.innerSize();
const Index outerSize = dst.outerSize();
for(Index outer = 0; outer < outerSize; ++outer)
for(Index inner = 0; inner < innerSize; ++inner)
dst.copyCoeffByOuterInner(outer, inner, src);
}
};
template<typename Derived1, typename Derived2, int Version>
struct assign_impl<Derived1, Derived2, DefaultTraversal, CompleteUnrolling, Version>
{
static EIGEN_STRONG_INLINE void run(Derived1 &dst, const Derived2 &src)
{
assign_DefaultTraversal_CompleteUnrolling<Derived1, Derived2, 0, Derived1::SizeAtCompileTime>
::run(dst, src);
}
};
template<typename Derived1, typename Derived2, int Version>
struct assign_impl<Derived1, Derived2, DefaultTraversal, InnerUnrolling, Version>
{
typedef typename Derived1::Index Index;
static EIGEN_STRONG_INLINE void run(Derived1 &dst, const Derived2 &src)
{
const Index outerSize = dst.outerSize();
for(Index outer = 0; outer < outerSize; ++outer)
assign_DefaultTraversal_InnerUnrolling<Derived1, Derived2, 0, Derived1::InnerSizeAtCompileTime>
::run(dst, src, outer);
}
};
/***********************
*** Linear traversal ***
***********************/
template<typename Derived1, typename Derived2, int Version>
struct assign_impl<Derived1, Derived2, LinearTraversal, NoUnrolling, Version>
{
typedef typename Derived1::Index Index;
static inline void run(Derived1 &dst, const Derived2 &src)
{
const Index size = dst.size();
for(Index i = 0; i < size; ++i)
dst.copyCoeff(i, src);
}
};
template<typename Derived1, typename Derived2, int Version>
struct assign_impl<Derived1, Derived2, LinearTraversal, CompleteUnrolling, Version>
{
static EIGEN_STRONG_INLINE void run(Derived1 &dst, const Derived2 &src)
{
assign_LinearTraversal_CompleteUnrolling<Derived1, Derived2, 0, Derived1::SizeAtCompileTime>
::run(dst, src);
}
};
/**************************
*** Inner vectorization ***
**************************/
template<typename Derived1, typename Derived2, int Version>
struct assign_impl<Derived1, Derived2, InnerVectorizedTraversal, NoUnrolling, Version>
{
typedef typename Derived1::Index Index;
static inline void run(Derived1 &dst, const Derived2 &src)
{
const Index innerSize = dst.innerSize();
const Index outerSize = dst.outerSize();
const Index packetSize = packet_traits<typename Derived1::Scalar>::size;
for(Index outer = 0; outer < outerSize; ++outer)
for(Index inner = 0; inner < innerSize; inner+=packetSize)
dst.template copyPacketByOuterInner<Derived2, Aligned, Aligned>(outer, inner, src);
}
};
template<typename Derived1, typename Derived2, int Version>
struct assign_impl<Derived1, Derived2, InnerVectorizedTraversal, CompleteUnrolling, Version>
{
static EIGEN_STRONG_INLINE void run(Derived1 &dst, const Derived2 &src)
{
assign_innervec_CompleteUnrolling<Derived1, Derived2, 0, Derived1::SizeAtCompileTime>
::run(dst, src);
}
};
template<typename Derived1, typename Derived2, int Version>
struct assign_impl<Derived1, Derived2, InnerVectorizedTraversal, InnerUnrolling, Version>
{
typedef typename Derived1::Index Index;
static EIGEN_STRONG_INLINE void run(Derived1 &dst, const Derived2 &src)
{
const Index outerSize = dst.outerSize();
for(Index outer = 0; outer < outerSize; ++outer)
assign_innervec_InnerUnrolling<Derived1, Derived2, 0, Derived1::InnerSizeAtCompileTime>
::run(dst, src, outer);
}
};
/***************************
*** Linear vectorization ***
***************************/
template <bool IsAligned = false>
struct unaligned_assign_impl
{
template <typename Derived, typename OtherDerived>
static EIGEN_STRONG_INLINE void run(const Derived&, OtherDerived&, typename Derived::Index, typename Derived::Index) {}
};
template <>
struct unaligned_assign_impl<false>
{
// MSVC must not inline this functions. If it does, it fails to optimize the
// packet access path.
#ifdef _MSC_VER
template <typename Derived, typename OtherDerived>
static EIGEN_DONT_INLINE void run(const Derived& src, OtherDerived& dst, typename Derived::Index start, typename Derived::Index end)
#else
template <typename Derived, typename OtherDerived>
static EIGEN_STRONG_INLINE void run(const Derived& src, OtherDerived& dst, typename Derived::Index start, typename Derived::Index end)
#endif
{
for (typename Derived::Index index = start; index < end; ++index)
dst.copyCoeff(index, src);
}
};
template<typename Derived1, typename Derived2, int Version>
struct assign_impl<Derived1, Derived2, LinearVectorizedTraversal, NoUnrolling, Version>
{
typedef typename Derived1::Index Index;
static EIGEN_STRONG_INLINE void run(Derived1 &dst, const Derived2 &src)
{
const Index size = dst.size();
typedef packet_traits<typename Derived1::Scalar> PacketTraits;
enum {
packetSize = PacketTraits::size,
dstAlignment = PacketTraits::AlignedOnScalar ? Aligned : int(assign_traits<Derived1,Derived2>::DstIsAligned) ,
srcAlignment = assign_traits<Derived1,Derived2>::JointAlignment
};
const Index alignedStart = assign_traits<Derived1,Derived2>::DstIsAligned ? 0
: internal::first_aligned(&dst.coeffRef(0), size);
const Index alignedEnd = alignedStart + ((size-alignedStart)/packetSize)*packetSize;
unaligned_assign_impl<assign_traits<Derived1,Derived2>::DstIsAligned!=0>::run(src,dst,0,alignedStart);
for(Index index = alignedStart; index < alignedEnd; index += packetSize)
{
dst.template copyPacket<Derived2, dstAlignment, srcAlignment>(index, src);
}
unaligned_assign_impl<>::run(src,dst,alignedEnd,size);
}
};
template<typename Derived1, typename Derived2, int Version>
struct assign_impl<Derived1, Derived2, LinearVectorizedTraversal, CompleteUnrolling, Version>
{
typedef typename Derived1::Index Index;
static EIGEN_STRONG_INLINE void run(Derived1 &dst, const Derived2 &src)
{
enum { size = Derived1::SizeAtCompileTime,
packetSize = packet_traits<typename Derived1::Scalar>::size,
alignedSize = (size/packetSize)*packetSize };
assign_innervec_CompleteUnrolling<Derived1, Derived2, 0, alignedSize>::run(dst, src);
assign_DefaultTraversal_CompleteUnrolling<Derived1, Derived2, alignedSize, size>::run(dst, src);
}
};
/**************************
*** Slice vectorization ***
***************************/
template<typename Derived1, typename Derived2, int Version>
struct assign_impl<Derived1, Derived2, SliceVectorizedTraversal, NoUnrolling, Version>
{
typedef typename Derived1::Index Index;
static inline void run(Derived1 &dst, const Derived2 &src)
{
typedef typename Derived1::Scalar Scalar;
typedef packet_traits<Scalar> PacketTraits;
enum {
packetSize = PacketTraits::size,
alignable = PacketTraits::AlignedOnScalar,
dstIsAligned = assign_traits<Derived1,Derived2>::DstIsAligned,
dstAlignment = alignable ? Aligned : int(dstIsAligned),
srcAlignment = assign_traits<Derived1,Derived2>::JointAlignment
};
const Scalar *dst_ptr = &dst.coeffRef(0,0);
if((!bool(dstIsAligned)) && (size_t(dst_ptr) % sizeof(Scalar))>0)
{
// the pointer is not aligend-on scalar, so alignment is not possible
return assign_impl<Derived1,Derived2,DefaultTraversal,NoUnrolling>::run(dst, src);
}
const Index packetAlignedMask = packetSize - 1;
const Index innerSize = dst.innerSize();
const Index outerSize = dst.outerSize();
const Index alignedStep = alignable ? (packetSize - dst.outerStride() % packetSize) & packetAlignedMask : 0;
Index alignedStart = ((!alignable) || bool(dstIsAligned)) ? 0 : internal::first_aligned(dst_ptr, innerSize);
for(Index outer = 0; outer < outerSize; ++outer)
{
const Index alignedEnd = alignedStart + ((innerSize-alignedStart) & ~packetAlignedMask);
// do the non-vectorizable part of the assignment
for(Index inner = 0; inner<alignedStart ; ++inner)
dst.copyCoeffByOuterInner(outer, inner, src);
// do the vectorizable part of the assignment
for(Index inner = alignedStart; inner<alignedEnd; inner+=packetSize)
dst.template copyPacketByOuterInner<Derived2, dstAlignment, Unaligned>(outer, inner, src);
// do the non-vectorizable part of the assignment
for(Index inner = alignedEnd; inner<innerSize ; ++inner)
dst.copyCoeffByOuterInner(outer, inner, src);
alignedStart = std::min<Index>((alignedStart+alignedStep)%packetSize, innerSize);
}
}
};
} // end namespace internal
/***************************************************************************
* Part 4 : implementation of DenseBase methods
***************************************************************************/
template<typename Derived>
template<typename OtherDerived>
EIGEN_STRONG_INLINE Derived& DenseBase<Derived>
@@ -499,90 +27,62 @@ EIGEN_STRONG_INLINE Derived& DenseBase<Derived>
EIGEN_STATIC_ASSERT_SAME_MATRIX_SIZE(Derived,OtherDerived)
EIGEN_STATIC_ASSERT(SameType,YOU_MIXED_DIFFERENT_NUMERIC_TYPES__YOU_NEED_TO_USE_THE_CAST_METHOD_OF_MATRIXBASE_TO_CAST_NUMERIC_TYPES_EXPLICITLY)
#ifdef EIGEN_DEBUG_ASSIGN
internal::assign_traits<Derived, OtherDerived>::debug();
#endif
eigen_assert(rows() == other.rows() && cols() == other.cols());
internal::assign_impl<Derived, OtherDerived, int(SameType) ? int(internal::assign_traits<Derived, OtherDerived>::Traversal)
: int(InvalidTraversal)>::run(derived(),other.derived());
#ifndef EIGEN_NO_DEBUG
checkTransposeAliasing(other.derived());
#endif
internal::call_assignment_no_alias(derived(),other.derived());
return derived();
}
namespace internal {
template<typename Derived, typename OtherDerived,
bool EvalBeforeAssigning = (int(internal::traits<OtherDerived>::Flags) & EvalBeforeAssigningBit) != 0,
bool NeedToTranspose = ((int(Derived::RowsAtCompileTime) == 1 && int(OtherDerived::ColsAtCompileTime) == 1)
| // FIXME | instead of || to please GCC 4.4.0 stupid warning "suggest parentheses around &&".
// revert to || as soon as not needed anymore.
(int(Derived::ColsAtCompileTime) == 1 && int(OtherDerived::RowsAtCompileTime) == 1))
&& int(Derived::SizeAtCompileTime) != 1>
struct assign_selector;
template<typename Derived, typename OtherDerived>
struct assign_selector<Derived,OtherDerived,false,false> {
static EIGEN_STRONG_INLINE Derived& run(Derived& dst, const OtherDerived& other) { return dst.lazyAssign(other.derived()); }
template<typename ActualDerived, typename ActualOtherDerived>
static EIGEN_STRONG_INLINE Derived& evalTo(ActualDerived& dst, const ActualOtherDerived& other) { other.evalTo(dst); return dst; }
};
template<typename Derived, typename OtherDerived>
struct assign_selector<Derived,OtherDerived,true,false> {
static EIGEN_STRONG_INLINE Derived& run(Derived& dst, const OtherDerived& other) { return dst.lazyAssign(other.eval()); }
};
template<typename Derived, typename OtherDerived>
struct assign_selector<Derived,OtherDerived,false,true> {
static EIGEN_STRONG_INLINE Derived& run(Derived& dst, const OtherDerived& other) { return dst.lazyAssign(other.transpose()); }
template<typename ActualDerived, typename ActualOtherDerived>
static EIGEN_STRONG_INLINE Derived& evalTo(ActualDerived& dst, const ActualOtherDerived& other) { Transpose<ActualDerived> dstTrans(dst); other.evalTo(dstTrans); return dst; }
};
template<typename Derived, typename OtherDerived>
struct assign_selector<Derived,OtherDerived,true,true> {
static EIGEN_STRONG_INLINE Derived& run(Derived& dst, const OtherDerived& other) { return dst.lazyAssign(other.transpose().eval()); }
};
} // end namespace internal
template<typename Derived>
template<typename OtherDerived>
EIGEN_DEVICE_FUNC
EIGEN_STRONG_INLINE Derived& DenseBase<Derived>::operator=(const DenseBase<OtherDerived>& other)
{
return internal::assign_selector<Derived,OtherDerived>::run(derived(), other.derived());
internal::call_assignment(derived(), other.derived());
return derived();
}
template<typename Derived>
EIGEN_DEVICE_FUNC
EIGEN_STRONG_INLINE Derived& DenseBase<Derived>::operator=(const DenseBase& other)
{
return internal::assign_selector<Derived,Derived>::run(derived(), other.derived());
internal::call_assignment(derived(), other.derived());
return derived();
}
template<typename Derived>
EIGEN_DEVICE_FUNC
EIGEN_STRONG_INLINE Derived& MatrixBase<Derived>::operator=(const MatrixBase& other)
{
return internal::assign_selector<Derived,Derived>::run(derived(), other.derived());
internal::call_assignment(derived(), other.derived());
return derived();
}
template<typename Derived>
template <typename OtherDerived>
EIGEN_DEVICE_FUNC
EIGEN_STRONG_INLINE Derived& MatrixBase<Derived>::operator=(const DenseBase<OtherDerived>& other)
{
return internal::assign_selector<Derived,OtherDerived>::run(derived(), other.derived());
internal::call_assignment(derived(), other.derived());
return derived();
}
template<typename Derived>
template <typename OtherDerived>
EIGEN_DEVICE_FUNC
EIGEN_STRONG_INLINE Derived& MatrixBase<Derived>::operator=(const EigenBase<OtherDerived>& other)
{
return internal::assign_selector<Derived,OtherDerived,false>::evalTo(derived(), other.derived());
internal::call_assignment(derived(), other.derived());
return derived();
}
template<typename Derived>
template<typename OtherDerived>
EIGEN_DEVICE_FUNC
EIGEN_STRONG_INLINE Derived& MatrixBase<Derived>::operator=(const ReturnByValue<OtherDerived>& other)
{
return internal::assign_selector<Derived,OtherDerived,false>::evalTo(derived(), other.derived());
other.derived().evalTo(derived());
return derived();
}
} // end namespace Eigen

View File

@@ -0,0 +1,935 @@
// This file is part of Eigen, a lightweight C++ template library
// for linear algebra.
//
// Copyright (C) 2011 Benoit Jacob <jacob.benoit.1@gmail.com>
// Copyright (C) 2011-2014 Gael Guennebaud <gael.guennebaud@inria.fr>
// Copyright (C) 2011-2012 Jitse Niesen <jitse@maths.leeds.ac.uk>
//
// This Source Code Form is subject to the terms of the Mozilla
// Public License v. 2.0. If a copy of the MPL was not distributed
// with this file, You can obtain one at http://mozilla.org/MPL/2.0/.
#ifndef EIGEN_ASSIGN_EVALUATOR_H
#define EIGEN_ASSIGN_EVALUATOR_H
namespace Eigen {
// This implementation is based on Assign.h
namespace internal {
/***************************************************************************
* Part 1 : the logic deciding a strategy for traversal and unrolling *
***************************************************************************/
// copy_using_evaluator_traits is based on assign_traits
template <typename DstEvaluator, typename SrcEvaluator, typename AssignFunc>
struct copy_using_evaluator_traits
{
typedef typename DstEvaluator::XprType Dst;
typedef typename Dst::Scalar DstScalar;
enum {
DstFlags = DstEvaluator::Flags,
SrcFlags = SrcEvaluator::Flags
};
public:
enum {
DstAlignment = DstEvaluator::Alignment,
SrcAlignment = SrcEvaluator::Alignment,
DstHasDirectAccess = (DstFlags & DirectAccessBit) == DirectAccessBit,
JointAlignment = EIGEN_PLAIN_ENUM_MIN(DstAlignment,SrcAlignment)
};
private:
enum {
InnerSize = int(Dst::IsVectorAtCompileTime) ? int(Dst::SizeAtCompileTime)
: int(DstFlags)&RowMajorBit ? int(Dst::ColsAtCompileTime)
: int(Dst::RowsAtCompileTime),
InnerMaxSize = int(Dst::IsVectorAtCompileTime) ? int(Dst::MaxSizeAtCompileTime)
: int(DstFlags)&RowMajorBit ? int(Dst::MaxColsAtCompileTime)
: int(Dst::MaxRowsAtCompileTime),
OuterStride = int(outer_stride_at_compile_time<Dst>::ret),
MaxSizeAtCompileTime = Dst::SizeAtCompileTime
};
// TODO distinguish between linear traversal and inner-traversals
typedef typename find_best_packet<DstScalar,Dst::SizeAtCompileTime>::type LinearPacketType;
typedef typename find_best_packet<DstScalar,InnerSize>::type InnerPacketType;
enum {
LinearPacketSize = unpacket_traits<LinearPacketType>::size,
InnerPacketSize = unpacket_traits<InnerPacketType>::size
};
public:
enum {
LinearRequiredAlignment = unpacket_traits<LinearPacketType>::alignment,
InnerRequiredAlignment = unpacket_traits<InnerPacketType>::alignment
};
private:
enum {
DstIsRowMajor = DstFlags&RowMajorBit,
SrcIsRowMajor = SrcFlags&RowMajorBit,
StorageOrdersAgree = (int(DstIsRowMajor) == int(SrcIsRowMajor)),
MightVectorize = bool(StorageOrdersAgree)
&& (int(DstFlags) & int(SrcFlags) & ActualPacketAccessBit)
&& bool(functor_traits<AssignFunc>::PacketAccess),
MayInnerVectorize = MightVectorize
&& int(InnerSize)!=Dynamic && int(InnerSize)%int(InnerPacketSize)==0
&& int(OuterStride)!=Dynamic && int(OuterStride)%int(InnerPacketSize)==0
&& (EIGEN_UNALIGNED_VECTORIZE || int(JointAlignment)>=int(InnerRequiredAlignment)),
MayLinearize = bool(StorageOrdersAgree) && (int(DstFlags) & int(SrcFlags) & LinearAccessBit),
MayLinearVectorize = bool(MightVectorize) && bool(MayLinearize) && bool(DstHasDirectAccess)
&& (EIGEN_UNALIGNED_VECTORIZE || (int(DstAlignment)>=int(LinearRequiredAlignment)) || MaxSizeAtCompileTime == Dynamic),
/* If the destination isn't aligned, we have to do runtime checks and we don't unroll,
so it's only good for large enough sizes. */
MaySliceVectorize = bool(MightVectorize) && bool(DstHasDirectAccess)
&& (int(InnerMaxSize)==Dynamic || int(InnerMaxSize)>=(EIGEN_UNALIGNED_VECTORIZE?InnerPacketSize:(3*InnerPacketSize)))
/* slice vectorization can be slow, so we only want it if the slices are big, which is
indicated by InnerMaxSize rather than InnerSize, think of the case of a dynamic block
in a fixed-size matrix
However, with EIGEN_UNALIGNED_VECTORIZE and unrolling, slice vectorization is still worth it */
};
public:
enum {
Traversal = int(MayLinearVectorize) && (LinearPacketSize>InnerPacketSize) ? int(LinearVectorizedTraversal)
: int(MayInnerVectorize) ? int(InnerVectorizedTraversal)
: int(MayLinearVectorize) ? int(LinearVectorizedTraversal)
: int(MaySliceVectorize) ? int(SliceVectorizedTraversal)
: int(MayLinearize) ? int(LinearTraversal)
: int(DefaultTraversal),
Vectorized = int(Traversal) == InnerVectorizedTraversal
|| int(Traversal) == LinearVectorizedTraversal
|| int(Traversal) == SliceVectorizedTraversal
};
typedef typename conditional<int(Traversal)==LinearVectorizedTraversal, LinearPacketType, InnerPacketType>::type PacketType;
private:
enum {
ActualPacketSize = int(Traversal)==LinearVectorizedTraversal ? LinearPacketSize
: Vectorized ? InnerPacketSize
: 1,
UnrollingLimit = EIGEN_UNROLLING_LIMIT * ActualPacketSize,
MayUnrollCompletely = int(Dst::SizeAtCompileTime) != Dynamic
&& int(Dst::SizeAtCompileTime) * (int(DstEvaluator::CoeffReadCost)+int(SrcEvaluator::CoeffReadCost)) <= int(UnrollingLimit),
MayUnrollInner = int(InnerSize) != Dynamic
&& int(InnerSize) * (int(DstEvaluator::CoeffReadCost)+int(SrcEvaluator::CoeffReadCost)) <= int(UnrollingLimit)
};
public:
enum {
Unrolling = (int(Traversal) == int(InnerVectorizedTraversal) || int(Traversal) == int(DefaultTraversal))
? (
int(MayUnrollCompletely) ? int(CompleteUnrolling)
: int(MayUnrollInner) ? int(InnerUnrolling)
: int(NoUnrolling)
)
: int(Traversal) == int(LinearVectorizedTraversal)
? ( bool(MayUnrollCompletely) && ( EIGEN_UNALIGNED_VECTORIZE || (int(DstAlignment)>=int(LinearRequiredAlignment)))
? int(CompleteUnrolling)
: int(NoUnrolling) )
: int(Traversal) == int(LinearTraversal)
? ( bool(MayUnrollCompletely) ? int(CompleteUnrolling)
: int(NoUnrolling) )
#if EIGEN_UNALIGNED_VECTORIZE
: int(Traversal) == int(SliceVectorizedTraversal)
? ( bool(MayUnrollInner) ? int(InnerUnrolling)
: int(NoUnrolling) )
#endif
: int(NoUnrolling)
};
#ifdef EIGEN_DEBUG_ASSIGN
static void debug()
{
std::cerr << "DstXpr: " << typeid(typename DstEvaluator::XprType).name() << std::endl;
std::cerr << "SrcXpr: " << typeid(typename SrcEvaluator::XprType).name() << std::endl;
std::cerr.setf(std::ios::hex, std::ios::basefield);
std::cerr << "DstFlags" << " = " << DstFlags << " (" << demangle_flags(DstFlags) << " )" << std::endl;
std::cerr << "SrcFlags" << " = " << SrcFlags << " (" << demangle_flags(SrcFlags) << " )" << std::endl;
std::cerr.unsetf(std::ios::hex);
EIGEN_DEBUG_VAR(DstAlignment)
EIGEN_DEBUG_VAR(SrcAlignment)
EIGEN_DEBUG_VAR(LinearRequiredAlignment)
EIGEN_DEBUG_VAR(InnerRequiredAlignment)
EIGEN_DEBUG_VAR(JointAlignment)
EIGEN_DEBUG_VAR(InnerSize)
EIGEN_DEBUG_VAR(InnerMaxSize)
EIGEN_DEBUG_VAR(LinearPacketSize)
EIGEN_DEBUG_VAR(InnerPacketSize)
EIGEN_DEBUG_VAR(ActualPacketSize)
EIGEN_DEBUG_VAR(StorageOrdersAgree)
EIGEN_DEBUG_VAR(MightVectorize)
EIGEN_DEBUG_VAR(MayLinearize)
EIGEN_DEBUG_VAR(MayInnerVectorize)
EIGEN_DEBUG_VAR(MayLinearVectorize)
EIGEN_DEBUG_VAR(MaySliceVectorize)
std::cerr << "Traversal" << " = " << Traversal << " (" << demangle_traversal(Traversal) << ")" << std::endl;
EIGEN_DEBUG_VAR(SrcEvaluator::CoeffReadCost)
EIGEN_DEBUG_VAR(UnrollingLimit)
EIGEN_DEBUG_VAR(MayUnrollCompletely)
EIGEN_DEBUG_VAR(MayUnrollInner)
std::cerr << "Unrolling" << " = " << Unrolling << " (" << demangle_unrolling(Unrolling) << ")" << std::endl;
std::cerr << std::endl;
}
#endif
};
/***************************************************************************
* Part 2 : meta-unrollers
***************************************************************************/
/************************
*** Default traversal ***
************************/
template<typename Kernel, int Index, int Stop>
struct copy_using_evaluator_DefaultTraversal_CompleteUnrolling
{
// FIXME: this is not very clean, perhaps this information should be provided by the kernel?
typedef typename Kernel::DstEvaluatorType DstEvaluatorType;
typedef typename DstEvaluatorType::XprType DstXprType;
enum {
outer = Index / DstXprType::InnerSizeAtCompileTime,
inner = Index % DstXprType::InnerSizeAtCompileTime
};
EIGEN_DEVICE_FUNC static EIGEN_STRONG_INLINE void run(Kernel &kernel)
{
kernel.assignCoeffByOuterInner(outer, inner);
copy_using_evaluator_DefaultTraversal_CompleteUnrolling<Kernel, Index+1, Stop>::run(kernel);
}
};
template<typename Kernel, int Stop>
struct copy_using_evaluator_DefaultTraversal_CompleteUnrolling<Kernel, Stop, Stop>
{
EIGEN_DEVICE_FUNC static EIGEN_STRONG_INLINE void run(Kernel&) { }
};
template<typename Kernel, int Index_, int Stop>
struct copy_using_evaluator_DefaultTraversal_InnerUnrolling
{
EIGEN_DEVICE_FUNC static EIGEN_STRONG_INLINE void run(Kernel &kernel, Index outer)
{
kernel.assignCoeffByOuterInner(outer, Index_);
copy_using_evaluator_DefaultTraversal_InnerUnrolling<Kernel, Index_+1, Stop>::run(kernel, outer);
}
};
template<typename Kernel, int Stop>
struct copy_using_evaluator_DefaultTraversal_InnerUnrolling<Kernel, Stop, Stop>
{
EIGEN_DEVICE_FUNC static EIGEN_STRONG_INLINE void run(Kernel&, Index) { }
};
/***********************
*** Linear traversal ***
***********************/
template<typename Kernel, int Index, int Stop>
struct copy_using_evaluator_LinearTraversal_CompleteUnrolling
{
EIGEN_DEVICE_FUNC static EIGEN_STRONG_INLINE void run(Kernel& kernel)
{
kernel.assignCoeff(Index);
copy_using_evaluator_LinearTraversal_CompleteUnrolling<Kernel, Index+1, Stop>::run(kernel);
}
};
template<typename Kernel, int Stop>
struct copy_using_evaluator_LinearTraversal_CompleteUnrolling<Kernel, Stop, Stop>
{
EIGEN_DEVICE_FUNC static EIGEN_STRONG_INLINE void run(Kernel&) { }
};
/**************************
*** Inner vectorization ***
**************************/
template<typename Kernel, int Index, int Stop>
struct copy_using_evaluator_innervec_CompleteUnrolling
{
// FIXME: this is not very clean, perhaps this information should be provided by the kernel?
typedef typename Kernel::DstEvaluatorType DstEvaluatorType;
typedef typename DstEvaluatorType::XprType DstXprType;
typedef typename Kernel::PacketType PacketType;
enum {
outer = Index / DstXprType::InnerSizeAtCompileTime,
inner = Index % DstXprType::InnerSizeAtCompileTime,
SrcAlignment = Kernel::AssignmentTraits::SrcAlignment,
DstAlignment = Kernel::AssignmentTraits::DstAlignment
};
EIGEN_DEVICE_FUNC static EIGEN_STRONG_INLINE void run(Kernel &kernel)
{
kernel.template assignPacketByOuterInner<DstAlignment, SrcAlignment, PacketType>(outer, inner);
enum { NextIndex = Index + unpacket_traits<PacketType>::size };
copy_using_evaluator_innervec_CompleteUnrolling<Kernel, NextIndex, Stop>::run(kernel);
}
};
template<typename Kernel, int Stop>
struct copy_using_evaluator_innervec_CompleteUnrolling<Kernel, Stop, Stop>
{
EIGEN_DEVICE_FUNC static EIGEN_STRONG_INLINE void run(Kernel&) { }
};
template<typename Kernel, int Index_, int Stop, int SrcAlignment, int DstAlignment>
struct copy_using_evaluator_innervec_InnerUnrolling
{
typedef typename Kernel::PacketType PacketType;
EIGEN_DEVICE_FUNC static EIGEN_STRONG_INLINE void run(Kernel &kernel, Index outer)
{
kernel.template assignPacketByOuterInner<DstAlignment, SrcAlignment, PacketType>(outer, Index_);
enum { NextIndex = Index_ + unpacket_traits<PacketType>::size };
copy_using_evaluator_innervec_InnerUnrolling<Kernel, NextIndex, Stop, SrcAlignment, DstAlignment>::run(kernel, outer);
}
};
template<typename Kernel, int Stop, int SrcAlignment, int DstAlignment>
struct copy_using_evaluator_innervec_InnerUnrolling<Kernel, Stop, Stop, SrcAlignment, DstAlignment>
{
EIGEN_DEVICE_FUNC static EIGEN_STRONG_INLINE void run(Kernel &, Index) { }
};
/***************************************************************************
* Part 3 : implementation of all cases
***************************************************************************/
// dense_assignment_loop is based on assign_impl
template<typename Kernel,
int Traversal = Kernel::AssignmentTraits::Traversal,
int Unrolling = Kernel::AssignmentTraits::Unrolling>
struct dense_assignment_loop;
/************************
*** Default traversal ***
************************/
template<typename Kernel>
struct dense_assignment_loop<Kernel, DefaultTraversal, NoUnrolling>
{
EIGEN_DEVICE_FUNC static void EIGEN_STRONG_INLINE run(Kernel &kernel)
{
for(Index outer = 0; outer < kernel.outerSize(); ++outer) {
for(Index inner = 0; inner < kernel.innerSize(); ++inner) {
kernel.assignCoeffByOuterInner(outer, inner);
}
}
}
};
template<typename Kernel>
struct dense_assignment_loop<Kernel, DefaultTraversal, CompleteUnrolling>
{
EIGEN_DEVICE_FUNC static EIGEN_STRONG_INLINE void run(Kernel &kernel)
{
typedef typename Kernel::DstEvaluatorType::XprType DstXprType;
copy_using_evaluator_DefaultTraversal_CompleteUnrolling<Kernel, 0, DstXprType::SizeAtCompileTime>::run(kernel);
}
};
template<typename Kernel>
struct dense_assignment_loop<Kernel, DefaultTraversal, InnerUnrolling>
{
EIGEN_DEVICE_FUNC static EIGEN_STRONG_INLINE void run(Kernel &kernel)
{
typedef typename Kernel::DstEvaluatorType::XprType DstXprType;
const Index outerSize = kernel.outerSize();
for(Index outer = 0; outer < outerSize; ++outer)
copy_using_evaluator_DefaultTraversal_InnerUnrolling<Kernel, 0, DstXprType::InnerSizeAtCompileTime>::run(kernel, outer);
}
};
/***************************
*** Linear vectorization ***
***************************/
// The goal of unaligned_dense_assignment_loop is simply to factorize the handling
// of the non vectorizable beginning and ending parts
template <bool IsAligned = false>
struct unaligned_dense_assignment_loop
{
// if IsAligned = true, then do nothing
template <typename Kernel>
EIGEN_DEVICE_FUNC static EIGEN_STRONG_INLINE void run(Kernel&, Index, Index) {}
};
template <>
struct unaligned_dense_assignment_loop<false>
{
// MSVC must not inline this functions. If it does, it fails to optimize the
// packet access path.
// FIXME check which version exhibits this issue
#if EIGEN_COMP_MSVC
template <typename Kernel>
static EIGEN_DONT_INLINE void run(Kernel &kernel,
Index start,
Index end)
#else
template <typename Kernel>
EIGEN_DEVICE_FUNC static EIGEN_STRONG_INLINE void run(Kernel &kernel,
Index start,
Index end)
#endif
{
for (Index index = start; index < end; ++index)
kernel.assignCoeff(index);
}
};
template<typename Kernel>
struct dense_assignment_loop<Kernel, LinearVectorizedTraversal, NoUnrolling>
{
EIGEN_DEVICE_FUNC static EIGEN_STRONG_INLINE void run(Kernel &kernel)
{
const Index size = kernel.size();
typedef typename Kernel::Scalar Scalar;
typedef typename Kernel::PacketType PacketType;
enum {
requestedAlignment = Kernel::AssignmentTraits::LinearRequiredAlignment,
packetSize = unpacket_traits<PacketType>::size,
dstIsAligned = int(Kernel::AssignmentTraits::DstAlignment)>=int(requestedAlignment),
dstAlignment = packet_traits<Scalar>::AlignedOnScalar ? int(requestedAlignment)
: int(Kernel::AssignmentTraits::DstAlignment),
srcAlignment = Kernel::AssignmentTraits::JointAlignment
};
const Index alignedStart = dstIsAligned ? 0 : internal::first_aligned<requestedAlignment>(kernel.dstDataPtr(), size);
const Index alignedEnd = alignedStart + ((size-alignedStart)/packetSize)*packetSize;
unaligned_dense_assignment_loop<dstIsAligned!=0>::run(kernel, 0, alignedStart);
for(Index index = alignedStart; index < alignedEnd; index += packetSize)
kernel.template assignPacket<dstAlignment, srcAlignment, PacketType>(index);
unaligned_dense_assignment_loop<>::run(kernel, alignedEnd, size);
}
};
template<typename Kernel>
struct dense_assignment_loop<Kernel, LinearVectorizedTraversal, CompleteUnrolling>
{
EIGEN_DEVICE_FUNC static EIGEN_STRONG_INLINE void run(Kernel &kernel)
{
typedef typename Kernel::DstEvaluatorType::XprType DstXprType;
typedef typename Kernel::PacketType PacketType;
enum { size = DstXprType::SizeAtCompileTime,
packetSize =unpacket_traits<PacketType>::size,
alignedSize = (size/packetSize)*packetSize };
copy_using_evaluator_innervec_CompleteUnrolling<Kernel, 0, alignedSize>::run(kernel);
copy_using_evaluator_DefaultTraversal_CompleteUnrolling<Kernel, alignedSize, size>::run(kernel);
}
};
/**************************
*** Inner vectorization ***
**************************/
template<typename Kernel>
struct dense_assignment_loop<Kernel, InnerVectorizedTraversal, NoUnrolling>
{
typedef typename Kernel::PacketType PacketType;
enum {
SrcAlignment = Kernel::AssignmentTraits::SrcAlignment,
DstAlignment = Kernel::AssignmentTraits::DstAlignment
};
EIGEN_DEVICE_FUNC static EIGEN_STRONG_INLINE void run(Kernel &kernel)
{
const Index innerSize = kernel.innerSize();
const Index outerSize = kernel.outerSize();
const Index packetSize = unpacket_traits<PacketType>::size;
for(Index outer = 0; outer < outerSize; ++outer)
for(Index inner = 0; inner < innerSize; inner+=packetSize)
kernel.template assignPacketByOuterInner<DstAlignment, SrcAlignment, PacketType>(outer, inner);
}
};
template<typename Kernel>
struct dense_assignment_loop<Kernel, InnerVectorizedTraversal, CompleteUnrolling>
{
EIGEN_DEVICE_FUNC static EIGEN_STRONG_INLINE void run(Kernel &kernel)
{
typedef typename Kernel::DstEvaluatorType::XprType DstXprType;
copy_using_evaluator_innervec_CompleteUnrolling<Kernel, 0, DstXprType::SizeAtCompileTime>::run(kernel);
}
};
template<typename Kernel>
struct dense_assignment_loop<Kernel, InnerVectorizedTraversal, InnerUnrolling>
{
EIGEN_DEVICE_FUNC static EIGEN_STRONG_INLINE void run(Kernel &kernel)
{
typedef typename Kernel::DstEvaluatorType::XprType DstXprType;
typedef typename Kernel::AssignmentTraits Traits;
const Index outerSize = kernel.outerSize();
for(Index outer = 0; outer < outerSize; ++outer)
copy_using_evaluator_innervec_InnerUnrolling<Kernel, 0, DstXprType::InnerSizeAtCompileTime,
Traits::SrcAlignment, Traits::DstAlignment>::run(kernel, outer);
}
};
/***********************
*** Linear traversal ***
***********************/
template<typename Kernel>
struct dense_assignment_loop<Kernel, LinearTraversal, NoUnrolling>
{
EIGEN_DEVICE_FUNC static EIGEN_STRONG_INLINE void run(Kernel &kernel)
{
const Index size = kernel.size();
for(Index i = 0; i < size; ++i)
kernel.assignCoeff(i);
}
};
template<typename Kernel>
struct dense_assignment_loop<Kernel, LinearTraversal, CompleteUnrolling>
{
EIGEN_DEVICE_FUNC static EIGEN_STRONG_INLINE void run(Kernel &kernel)
{
typedef typename Kernel::DstEvaluatorType::XprType DstXprType;
copy_using_evaluator_LinearTraversal_CompleteUnrolling<Kernel, 0, DstXprType::SizeAtCompileTime>::run(kernel);
}
};
/**************************
*** Slice vectorization ***
***************************/
template<typename Kernel>
struct dense_assignment_loop<Kernel, SliceVectorizedTraversal, NoUnrolling>
{
EIGEN_DEVICE_FUNC static EIGEN_STRONG_INLINE void run(Kernel &kernel)
{
typedef typename Kernel::Scalar Scalar;
typedef typename Kernel::PacketType PacketType;
enum {
packetSize = unpacket_traits<PacketType>::size,
requestedAlignment = int(Kernel::AssignmentTraits::InnerRequiredAlignment),
alignable = packet_traits<Scalar>::AlignedOnScalar || int(Kernel::AssignmentTraits::DstAlignment)>=sizeof(Scalar),
dstIsAligned = int(Kernel::AssignmentTraits::DstAlignment)>=int(requestedAlignment),
dstAlignment = alignable ? int(requestedAlignment)
: int(Kernel::AssignmentTraits::DstAlignment)
};
const Scalar *dst_ptr = kernel.dstDataPtr();
if((!bool(dstIsAligned)) && (UIntPtr(dst_ptr) % sizeof(Scalar))>0)
{
// the pointer is not aligend-on scalar, so alignment is not possible
return dense_assignment_loop<Kernel,DefaultTraversal,NoUnrolling>::run(kernel);
}
const Index packetAlignedMask = packetSize - 1;
const Index innerSize = kernel.innerSize();
const Index outerSize = kernel.outerSize();
const Index alignedStep = alignable ? (packetSize - kernel.outerStride() % packetSize) & packetAlignedMask : 0;
Index alignedStart = ((!alignable) || bool(dstIsAligned)) ? 0 : internal::first_aligned<requestedAlignment>(dst_ptr, innerSize);
for(Index outer = 0; outer < outerSize; ++outer)
{
const Index alignedEnd = alignedStart + ((innerSize-alignedStart) & ~packetAlignedMask);
// do the non-vectorizable part of the assignment
for(Index inner = 0; inner<alignedStart ; ++inner)
kernel.assignCoeffByOuterInner(outer, inner);
// do the vectorizable part of the assignment
for(Index inner = alignedStart; inner<alignedEnd; inner+=packetSize)
kernel.template assignPacketByOuterInner<dstAlignment, Unaligned, PacketType>(outer, inner);
// do the non-vectorizable part of the assignment
for(Index inner = alignedEnd; inner<innerSize ; ++inner)
kernel.assignCoeffByOuterInner(outer, inner);
alignedStart = numext::mini((alignedStart+alignedStep)%packetSize, innerSize);
}
}
};
#if EIGEN_UNALIGNED_VECTORIZE
template<typename Kernel>
struct dense_assignment_loop<Kernel, SliceVectorizedTraversal, InnerUnrolling>
{
EIGEN_DEVICE_FUNC static EIGEN_STRONG_INLINE void run(Kernel &kernel)
{
typedef typename Kernel::DstEvaluatorType::XprType DstXprType;
typedef typename Kernel::PacketType PacketType;
enum { size = DstXprType::InnerSizeAtCompileTime,
packetSize =unpacket_traits<PacketType>::size,
vectorizableSize = (size/packetSize)*packetSize };
for(Index outer = 0; outer < kernel.outerSize(); ++outer)
{
copy_using_evaluator_innervec_InnerUnrolling<Kernel, 0, vectorizableSize, 0, 0>::run(kernel, outer);
copy_using_evaluator_DefaultTraversal_InnerUnrolling<Kernel, vectorizableSize, size>::run(kernel, outer);
}
}
};
#endif
/***************************************************************************
* Part 4 : Generic dense assignment kernel
***************************************************************************/
// This class generalize the assignment of a coefficient (or packet) from one dense evaluator
// to another dense writable evaluator.
// It is parametrized by the two evaluators, and the actual assignment functor.
// This abstraction level permits to keep the evaluation loops as simple and as generic as possible.
// One can customize the assignment using this generic dense_assignment_kernel with different
// functors, or by completely overloading it, by-passing a functor.
template<typename DstEvaluatorTypeT, typename SrcEvaluatorTypeT, typename Functor, int Version = Specialized>
class generic_dense_assignment_kernel
{
protected:
typedef typename DstEvaluatorTypeT::XprType DstXprType;
typedef typename SrcEvaluatorTypeT::XprType SrcXprType;
public:
typedef DstEvaluatorTypeT DstEvaluatorType;
typedef SrcEvaluatorTypeT SrcEvaluatorType;
typedef typename DstEvaluatorType::Scalar Scalar;
typedef copy_using_evaluator_traits<DstEvaluatorTypeT, SrcEvaluatorTypeT, Functor> AssignmentTraits;
typedef typename AssignmentTraits::PacketType PacketType;
EIGEN_DEVICE_FUNC generic_dense_assignment_kernel(DstEvaluatorType &dst, const SrcEvaluatorType &src, const Functor &func, DstXprType& dstExpr)
: m_dst(dst), m_src(src), m_functor(func), m_dstExpr(dstExpr)
{
#ifdef EIGEN_DEBUG_ASSIGN
AssignmentTraits::debug();
#endif
}
EIGEN_DEVICE_FUNC Index size() const { return m_dstExpr.size(); }
EIGEN_DEVICE_FUNC Index innerSize() const { return m_dstExpr.innerSize(); }
EIGEN_DEVICE_FUNC Index outerSize() const { return m_dstExpr.outerSize(); }
EIGEN_DEVICE_FUNC Index rows() const { return m_dstExpr.rows(); }
EIGEN_DEVICE_FUNC Index cols() const { return m_dstExpr.cols(); }
EIGEN_DEVICE_FUNC Index outerStride() const { return m_dstExpr.outerStride(); }
EIGEN_DEVICE_FUNC DstEvaluatorType& dstEvaluator() { return m_dst; }
EIGEN_DEVICE_FUNC const SrcEvaluatorType& srcEvaluator() const { return m_src; }
/// Assign src(row,col) to dst(row,col) through the assignment functor.
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE void assignCoeff(Index row, Index col)
{
m_functor.assignCoeff(m_dst.coeffRef(row,col), m_src.coeff(row,col));
}
/// \sa assignCoeff(Index,Index)
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE void assignCoeff(Index index)
{
m_functor.assignCoeff(m_dst.coeffRef(index), m_src.coeff(index));
}
/// \sa assignCoeff(Index,Index)
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE void assignCoeffByOuterInner(Index outer, Index inner)
{
Index row = rowIndexByOuterInner(outer, inner);
Index col = colIndexByOuterInner(outer, inner);
assignCoeff(row, col);
}
template<int StoreMode, int LoadMode, typename PacketType>
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE void assignPacket(Index row, Index col)
{
m_functor.template assignPacket<StoreMode>(&m_dst.coeffRef(row,col), m_src.template packet<LoadMode,PacketType>(row,col));
}
template<int StoreMode, int LoadMode, typename PacketType>
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE void assignPacket(Index index)
{
m_functor.template assignPacket<StoreMode>(&m_dst.coeffRef(index), m_src.template packet<LoadMode,PacketType>(index));
}
template<int StoreMode, int LoadMode, typename PacketType>
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE void assignPacketByOuterInner(Index outer, Index inner)
{
Index row = rowIndexByOuterInner(outer, inner);
Index col = colIndexByOuterInner(outer, inner);
assignPacket<StoreMode,LoadMode,PacketType>(row, col);
}
EIGEN_DEVICE_FUNC static EIGEN_STRONG_INLINE Index rowIndexByOuterInner(Index outer, Index inner)
{
typedef typename DstEvaluatorType::ExpressionTraits Traits;
return int(Traits::RowsAtCompileTime) == 1 ? 0
: int(Traits::ColsAtCompileTime) == 1 ? inner
: int(DstEvaluatorType::Flags)&RowMajorBit ? outer
: inner;
}
EIGEN_DEVICE_FUNC static EIGEN_STRONG_INLINE Index colIndexByOuterInner(Index outer, Index inner)
{
typedef typename DstEvaluatorType::ExpressionTraits Traits;
return int(Traits::ColsAtCompileTime) == 1 ? 0
: int(Traits::RowsAtCompileTime) == 1 ? inner
: int(DstEvaluatorType::Flags)&RowMajorBit ? inner
: outer;
}
EIGEN_DEVICE_FUNC const Scalar* dstDataPtr() const
{
return m_dstExpr.data();
}
protected:
DstEvaluatorType& m_dst;
const SrcEvaluatorType& m_src;
const Functor &m_functor;
// TODO find a way to avoid the needs of the original expression
DstXprType& m_dstExpr;
};
/***************************************************************************
* Part 5 : Entry point for dense rectangular assignment
***************************************************************************/
template<typename DstXprType,typename SrcXprType, typename Functor>
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE
void resize_if_allowed(DstXprType &dst, const SrcXprType& src, const Functor &/*func*/)
{
EIGEN_ONLY_USED_FOR_DEBUG(dst);
EIGEN_ONLY_USED_FOR_DEBUG(src);
eigen_assert(dst.rows() == src.rows() && dst.cols() == src.cols());
}
template<typename DstXprType,typename SrcXprType, typename T1, typename T2>
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE
void resize_if_allowed(DstXprType &dst, const SrcXprType& src, const internal::assign_op<T1,T2> &/*func*/)
{
Index dstRows = src.rows();
Index dstCols = src.cols();
if(((dst.rows()!=dstRows) || (dst.cols()!=dstCols)))
dst.resize(dstRows, dstCols);
eigen_assert(dst.rows() == dstRows && dst.cols() == dstCols);
}
template<typename DstXprType, typename SrcXprType, typename Functor>
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE void call_dense_assignment_loop(DstXprType& dst, const SrcXprType& src, const Functor &func)
{
typedef evaluator<DstXprType> DstEvaluatorType;
typedef evaluator<SrcXprType> SrcEvaluatorType;
SrcEvaluatorType srcEvaluator(src);
// NOTE To properly handle A = (A*A.transpose())/s with A rectangular,
// we need to resize the destination after the source evaluator has been created.
resize_if_allowed(dst, src, func);
DstEvaluatorType dstEvaluator(dst);
typedef generic_dense_assignment_kernel<DstEvaluatorType,SrcEvaluatorType,Functor> Kernel;
Kernel kernel(dstEvaluator, srcEvaluator, func, dst.const_cast_derived());
dense_assignment_loop<Kernel>::run(kernel);
}
template<typename DstXprType, typename SrcXprType>
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE void call_dense_assignment_loop(DstXprType& dst, const SrcXprType& src)
{
call_dense_assignment_loop(dst, src, internal::assign_op<typename DstXprType::Scalar,typename SrcXprType::Scalar>());
}
/***************************************************************************
* Part 6 : Generic assignment
***************************************************************************/
// Based on the respective shapes of the destination and source,
// the class AssignmentKind determine the kind of assignment mechanism.
// AssignmentKind must define a Kind typedef.
template<typename DstShape, typename SrcShape> struct AssignmentKind;
// Assignement kind defined in this file:
struct Dense2Dense {};
struct EigenBase2EigenBase {};
template<typename,typename> struct AssignmentKind { typedef EigenBase2EigenBase Kind; };
template<> struct AssignmentKind<DenseShape,DenseShape> { typedef Dense2Dense Kind; };
// This is the main assignment class
template< typename DstXprType, typename SrcXprType, typename Functor,
typename Kind = typename AssignmentKind< typename evaluator_traits<DstXprType>::Shape , typename evaluator_traits<SrcXprType>::Shape >::Kind,
typename EnableIf = void>
struct Assignment;
// The only purpose of this call_assignment() function is to deal with noalias() / "assume-aliasing" and automatic transposition.
// Indeed, I (Gael) think that this concept of "assume-aliasing" was a mistake, and it makes thing quite complicated.
// So this intermediate function removes everything related to "assume-aliasing" such that Assignment
// does not has to bother about these annoying details.
template<typename Dst, typename Src>
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE
void call_assignment(Dst& dst, const Src& src)
{
call_assignment(dst, src, internal::assign_op<typename Dst::Scalar,typename Src::Scalar>());
}
template<typename Dst, typename Src>
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE
void call_assignment(const Dst& dst, const Src& src)
{
call_assignment(dst, src, internal::assign_op<typename Dst::Scalar,typename Src::Scalar>());
}
// Deal with "assume-aliasing"
template<typename Dst, typename Src, typename Func>
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE
void call_assignment(Dst& dst, const Src& src, const Func& func, typename enable_if< evaluator_assume_aliasing<Src>::value, void*>::type = 0)
{
typename plain_matrix_type<Src>::type tmp(src);
call_assignment_no_alias(dst, tmp, func);
}
template<typename Dst, typename Src, typename Func>
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE
void call_assignment(Dst& dst, const Src& src, const Func& func, typename enable_if<!evaluator_assume_aliasing<Src>::value, void*>::type = 0)
{
call_assignment_no_alias(dst, src, func);
}
// by-pass "assume-aliasing"
// When there is no aliasing, we require that 'dst' has been properly resized
template<typename Dst, template <typename> class StorageBase, typename Src, typename Func>
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE
void call_assignment(NoAlias<Dst,StorageBase>& dst, const Src& src, const Func& func)
{
call_assignment_no_alias(dst.expression(), src, func);
}
template<typename Dst, typename Src, typename Func>
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE
void call_assignment_no_alias(Dst& dst, const Src& src, const Func& func)
{
enum {
NeedToTranspose = ( (int(Dst::RowsAtCompileTime) == 1 && int(Src::ColsAtCompileTime) == 1)
|| (int(Dst::ColsAtCompileTime) == 1 && int(Src::RowsAtCompileTime) == 1)
) && int(Dst::SizeAtCompileTime) != 1
};
typedef typename internal::conditional<NeedToTranspose, Transpose<Dst>, Dst>::type ActualDstTypeCleaned;
typedef typename internal::conditional<NeedToTranspose, Transpose<Dst>, Dst&>::type ActualDstType;
ActualDstType actualDst(dst);
// TODO check whether this is the right place to perform these checks:
EIGEN_STATIC_ASSERT_LVALUE(Dst)
EIGEN_STATIC_ASSERT_SAME_MATRIX_SIZE(ActualDstTypeCleaned,Src)
EIGEN_CHECK_BINARY_COMPATIBILIY(Func,typename ActualDstTypeCleaned::Scalar,typename Src::Scalar);
Assignment<ActualDstTypeCleaned,Src,Func>::run(actualDst, src, func);
}
template<typename Dst, typename Src>
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE
void call_assignment_no_alias(Dst& dst, const Src& src)
{
call_assignment_no_alias(dst, src, internal::assign_op<typename Dst::Scalar,typename Src::Scalar>());
}
template<typename Dst, typename Src, typename Func>
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE
void call_assignment_no_alias_no_transpose(Dst& dst, const Src& src, const Func& func)
{
// TODO check whether this is the right place to perform these checks:
EIGEN_STATIC_ASSERT_LVALUE(Dst)
EIGEN_STATIC_ASSERT_SAME_MATRIX_SIZE(Dst,Src)
EIGEN_CHECK_BINARY_COMPATIBILIY(Func,typename Dst::Scalar,typename Src::Scalar);
Assignment<Dst,Src,Func>::run(dst, src, func);
}
template<typename Dst, typename Src>
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE
void call_assignment_no_alias_no_transpose(Dst& dst, const Src& src)
{
call_assignment_no_alias_no_transpose(dst, src, internal::assign_op<typename Dst::Scalar,typename Src::Scalar>());
}
// forward declaration
template<typename Dst, typename Src> void check_for_aliasing(const Dst &dst, const Src &src);
// Generic Dense to Dense assignment
// Note that the last template argument "Weak" is needed to make it possible to perform
// both partial specialization+SFINAE without ambiguous specialization
template< typename DstXprType, typename SrcXprType, typename Functor, typename Weak>
struct Assignment<DstXprType, SrcXprType, Functor, Dense2Dense, Weak>
{
EIGEN_DEVICE_FUNC
static EIGEN_STRONG_INLINE void run(DstXprType &dst, const SrcXprType &src, const Functor &func)
{
#ifndef EIGEN_NO_DEBUG
internal::check_for_aliasing(dst, src);
#endif
call_dense_assignment_loop(dst, src, func);
}
};
// Generic assignment through evalTo.
// TODO: not sure we have to keep that one, but it helps porting current code to new evaluator mechanism.
// Note that the last template argument "Weak" is needed to make it possible to perform
// both partial specialization+SFINAE without ambiguous specialization
template< typename DstXprType, typename SrcXprType, typename Functor, typename Weak>
struct Assignment<DstXprType, SrcXprType, Functor, EigenBase2EigenBase, Weak>
{
EIGEN_DEVICE_FUNC
static EIGEN_STRONG_INLINE void run(DstXprType &dst, const SrcXprType &src, const internal::assign_op<typename DstXprType::Scalar,typename SrcXprType::Scalar> &/*func*/)
{
Index dstRows = src.rows();
Index dstCols = src.cols();
if((dst.rows()!=dstRows) || (dst.cols()!=dstCols))
dst.resize(dstRows, dstCols);
eigen_assert(dst.rows() == src.rows() && dst.cols() == src.cols());
src.evalTo(dst);
}
// NOTE The following two functions are templated to avoid their instanciation if not needed
// This is needed because some expressions supports evalTo only and/or have 'void' as scalar type.
template<typename SrcScalarType>
EIGEN_DEVICE_FUNC
static EIGEN_STRONG_INLINE void run(DstXprType &dst, const SrcXprType &src, const internal::add_assign_op<typename DstXprType::Scalar,SrcScalarType> &/*func*/)
{
Index dstRows = src.rows();
Index dstCols = src.cols();
if((dst.rows()!=dstRows) || (dst.cols()!=dstCols))
dst.resize(dstRows, dstCols);
eigen_assert(dst.rows() == src.rows() && dst.cols() == src.cols());
src.addTo(dst);
}
template<typename SrcScalarType>
EIGEN_DEVICE_FUNC
static EIGEN_STRONG_INLINE void run(DstXprType &dst, const SrcXprType &src, const internal::sub_assign_op<typename DstXprType::Scalar,SrcScalarType> &/*func*/)
{
Index dstRows = src.rows();
Index dstCols = src.cols();
if((dst.rows()!=dstRows) || (dst.cols()!=dstCols))
dst.resize(dstRows, dstCols);
eigen_assert(dst.rows() == src.rows() && dst.cols() == src.cols());
src.subTo(dst);
}
};
} // namespace internal
} // end namespace Eigen
#endif // EIGEN_ASSIGN_EVALUATOR_H

View File

@@ -1,6 +1,7 @@
/*
Copyright (c) 2011, Intel Corporation. All rights reserved.
Copyright (C) 2015 Gael Guennebaud <gael.guennebaud@inria.fr>
Redistribution and use in source and binary forms, with or without modification,
are permitted provided that the following conditions are met:
@@ -37,17 +38,13 @@ namespace Eigen {
namespace internal {
template<typename Op> struct vml_call
{ enum { IsSupported = 0 }; };
template<typename Dst, typename Src, typename UnaryOp>
template<typename Dst, typename Src>
class vml_assign_traits
{
private:
enum {
DstHasDirectAccess = Dst::Flags & DirectAccessBit,
SrcHasDirectAccess = Src::Flags & DirectAccessBit,
StorageOrdersAgree = (int(Dst::IsRowMajor) == int(Src::IsRowMajor)),
InnerSize = int(Dst::IsVectorAtCompileTime) ? int(Dst::SizeAtCompileTime)
: int(Dst::Flags)&RowMajorBit ? int(Dst::ColsAtCompileTime)
@@ -57,165 +54,122 @@ class vml_assign_traits
: int(Dst::MaxRowsAtCompileTime),
MaxSizeAtCompileTime = Dst::SizeAtCompileTime,
MightEnableVml = vml_call<UnaryOp>::IsSupported && StorageOrdersAgree && DstHasDirectAccess && SrcHasDirectAccess
&& Src::InnerStrideAtCompileTime==1 && Dst::InnerStrideAtCompileTime==1,
MightEnableVml = StorageOrdersAgree && DstHasDirectAccess && SrcHasDirectAccess && Src::InnerStrideAtCompileTime==1 && Dst::InnerStrideAtCompileTime==1,
MightLinearize = MightEnableVml && (int(Dst::Flags) & int(Src::Flags) & LinearAccessBit),
VmlSize = MightLinearize ? MaxSizeAtCompileTime : InnerMaxSize,
LargeEnough = VmlSize==Dynamic || VmlSize>=EIGEN_MKL_VML_THRESHOLD,
MayEnableVml = MightEnableVml && LargeEnough,
MayLinearize = MayEnableVml && MightLinearize
LargeEnough = VmlSize==Dynamic || VmlSize>=EIGEN_MKL_VML_THRESHOLD
};
public:
enum {
Traversal = MayLinearize ? LinearVectorizedTraversal
: MayEnableVml ? InnerVectorizedTraversal
: DefaultTraversal
EnableVml = MightEnableVml && LargeEnough,
Traversal = MightLinearize ? LinearTraversal : DefaultTraversal
};
};
template<typename Derived1, typename Derived2, typename UnaryOp, int Traversal, int Unrolling,
int VmlTraversal = vml_assign_traits<Derived1, Derived2, UnaryOp>::Traversal >
struct vml_assign_impl
: assign_impl<Derived1, Eigen::CwiseUnaryOp<UnaryOp, Derived2>,Traversal,Unrolling,BuiltIn>
{
};
template<typename Derived1, typename Derived2, typename UnaryOp, int Traversal, int Unrolling>
struct vml_assign_impl<Derived1, Derived2, UnaryOp, Traversal, Unrolling, InnerVectorizedTraversal>
{
typedef typename Derived1::Scalar Scalar;
typedef typename Derived1::Index Index;
static inline void run(Derived1& dst, const CwiseUnaryOp<UnaryOp, Derived2>& src)
{
// in case we want to (or have to) skip VML at runtime we can call:
// assign_impl<Derived1,Eigen::CwiseUnaryOp<UnaryOp, Derived2>,Traversal,Unrolling,BuiltIn>::run(dst,src);
const Index innerSize = dst.innerSize();
const Index outerSize = dst.outerSize();
for(Index outer = 0; outer < outerSize; ++outer) {
const Scalar *src_ptr = src.IsRowMajor ? &(src.nestedExpression().coeffRef(outer,0)) :
&(src.nestedExpression().coeffRef(0, outer));
Scalar *dst_ptr = dst.IsRowMajor ? &(dst.coeffRef(outer,0)) : &(dst.coeffRef(0, outer));
vml_call<UnaryOp>::run(src.functor(), innerSize, src_ptr, dst_ptr );
}
}
};
template<typename Derived1, typename Derived2, typename UnaryOp, int Traversal, int Unrolling>
struct vml_assign_impl<Derived1, Derived2, UnaryOp, Traversal, Unrolling, LinearVectorizedTraversal>
{
static inline void run(Derived1& dst, const CwiseUnaryOp<UnaryOp, Derived2>& src)
{
// in case we want to (or have to) skip VML at runtime we can call:
// assign_impl<Derived1,Eigen::CwiseUnaryOp<UnaryOp, Derived2>,Traversal,Unrolling,BuiltIn>::run(dst,src);
vml_call<UnaryOp>::run(src.functor(), dst.size(), src.nestedExpression().data(), dst.data() );
}
};
// Macroses
#define EIGEN_MKL_VML_SPECIALIZE_ASSIGN(TRAVERSAL,UNROLLING) \
template<typename Derived1, typename Derived2, typename UnaryOp> \
struct assign_impl<Derived1, Eigen::CwiseUnaryOp<UnaryOp, Derived2>, TRAVERSAL, UNROLLING, Specialized> { \
static inline void run(Derived1 &dst, const Eigen::CwiseUnaryOp<UnaryOp, Derived2> &src) { \
vml_assign_impl<Derived1,Derived2,UnaryOp,TRAVERSAL,UNROLLING>::run(dst, src); \
} \
};
EIGEN_MKL_VML_SPECIALIZE_ASSIGN(DefaultTraversal,NoUnrolling)
EIGEN_MKL_VML_SPECIALIZE_ASSIGN(DefaultTraversal,CompleteUnrolling)
EIGEN_MKL_VML_SPECIALIZE_ASSIGN(DefaultTraversal,InnerUnrolling)
EIGEN_MKL_VML_SPECIALIZE_ASSIGN(LinearTraversal,NoUnrolling)
EIGEN_MKL_VML_SPECIALIZE_ASSIGN(LinearTraversal,CompleteUnrolling)
EIGEN_MKL_VML_SPECIALIZE_ASSIGN(InnerVectorizedTraversal,NoUnrolling)
EIGEN_MKL_VML_SPECIALIZE_ASSIGN(InnerVectorizedTraversal,CompleteUnrolling)
EIGEN_MKL_VML_SPECIALIZE_ASSIGN(InnerVectorizedTraversal,InnerUnrolling)
EIGEN_MKL_VML_SPECIALIZE_ASSIGN(LinearVectorizedTraversal,CompleteUnrolling)
EIGEN_MKL_VML_SPECIALIZE_ASSIGN(LinearVectorizedTraversal,NoUnrolling)
EIGEN_MKL_VML_SPECIALIZE_ASSIGN(SliceVectorizedTraversal,NoUnrolling)
#define EIGEN_PP_EXPAND(ARG) ARG
#if !defined (EIGEN_FAST_MATH) || (EIGEN_FAST_MATH != 1)
#define EIGEN_MKL_VML_MODE VML_HA
#define EIGEN_VMLMODE_EXPAND_LA , VML_HA
#else
#define EIGEN_MKL_VML_MODE VML_LA
#define EIGEN_VMLMODE_EXPAND_LA , VML_LA
#endif
#define EIGEN_MKL_VML_DECLARE_UNARY_CALL(EIGENOP, VMLOP, EIGENTYPE, VMLTYPE) \
template<> struct vml_call< scalar_##EIGENOP##_op<EIGENTYPE> > { \
enum { IsSupported = 1 }; \
static inline void run( const scalar_##EIGENOP##_op<EIGENTYPE>& /*func*/, \
int size, const EIGENTYPE* src, EIGENTYPE* dst) { \
VMLOP(size, (const VMLTYPE*)src, (VMLTYPE*)dst); \
} \
#define EIGEN_VMLMODE_EXPAND__
#define EIGEN_VMLMODE_PREFIX_LA vm
#define EIGEN_VMLMODE_PREFIX__ v
#define EIGEN_VMLMODE_PREFIX(VMLMODE) EIGEN_CAT(EIGEN_VMLMODE_PREFIX_,VMLMODE)
#define EIGEN_MKL_VML_DECLARE_UNARY_CALL(EIGENOP, VMLOP, EIGENTYPE, VMLTYPE, VMLMODE) \
template< typename DstXprType, typename SrcXprNested> \
struct Assignment<DstXprType, CwiseUnaryOp<scalar_##EIGENOP##_op<EIGENTYPE>, SrcXprNested>, assign_op<EIGENTYPE,EIGENTYPE>, \
Dense2Dense, typename enable_if<vml_assign_traits<DstXprType,SrcXprNested>::EnableVml>::type> { \
typedef CwiseUnaryOp<scalar_##EIGENOP##_op<EIGENTYPE>, SrcXprNested> SrcXprType; \
static void run(DstXprType &dst, const SrcXprType &src, const assign_op<EIGENTYPE,EIGENTYPE> &func) { \
resize_if_allowed(dst, src, func); \
eigen_assert(dst.rows() == src.rows() && dst.cols() == src.cols()); \
if(vml_assign_traits<DstXprType,SrcXprNested>::Traversal==LinearTraversal) { \
VMLOP(dst.size(), (const VMLTYPE*)src.nestedExpression().data(), \
(VMLTYPE*)dst.data() EIGEN_PP_EXPAND(EIGEN_VMLMODE_EXPAND_##VMLMODE) ); \
} else { \
const Index outerSize = dst.outerSize(); \
for(Index outer = 0; outer < outerSize; ++outer) { \
const EIGENTYPE *src_ptr = src.IsRowMajor ? &(src.nestedExpression().coeffRef(outer,0)) : \
&(src.nestedExpression().coeffRef(0, outer)); \
EIGENTYPE *dst_ptr = dst.IsRowMajor ? &(dst.coeffRef(outer,0)) : &(dst.coeffRef(0, outer)); \
VMLOP( dst.innerSize(), (const VMLTYPE*)src_ptr, \
(VMLTYPE*)dst_ptr EIGEN_PP_EXPAND(EIGEN_VMLMODE_EXPAND_##VMLMODE)); \
} \
} \
} \
}; \
#define EIGEN_MKL_VML_DECLARE_UNARY_CALLS_REAL(EIGENOP, VMLOP, VMLMODE) \
EIGEN_MKL_VML_DECLARE_UNARY_CALL(EIGENOP, EIGEN_CAT(EIGEN_VMLMODE_PREFIX(VMLMODE),s##VMLOP), float, float, VMLMODE) \
EIGEN_MKL_VML_DECLARE_UNARY_CALL(EIGENOP, EIGEN_CAT(EIGEN_VMLMODE_PREFIX(VMLMODE),d##VMLOP), double, double, VMLMODE)
#define EIGEN_MKL_VML_DECLARE_UNARY_CALLS_CPLX(EIGENOP, VMLOP, VMLMODE) \
EIGEN_MKL_VML_DECLARE_UNARY_CALL(EIGENOP, EIGEN_CAT(EIGEN_VMLMODE_PREFIX(VMLMODE),c##VMLOP), scomplex, MKL_Complex8, VMLMODE) \
EIGEN_MKL_VML_DECLARE_UNARY_CALL(EIGENOP, EIGEN_CAT(EIGEN_VMLMODE_PREFIX(VMLMODE),z##VMLOP), dcomplex, MKL_Complex16, VMLMODE)
#define EIGEN_MKL_VML_DECLARE_UNARY_CALLS(EIGENOP, VMLOP, VMLMODE) \
EIGEN_MKL_VML_DECLARE_UNARY_CALLS_REAL(EIGENOP, VMLOP, VMLMODE) \
EIGEN_MKL_VML_DECLARE_UNARY_CALLS_CPLX(EIGENOP, VMLOP, VMLMODE)
EIGEN_MKL_VML_DECLARE_UNARY_CALLS(sin, Sin, LA)
EIGEN_MKL_VML_DECLARE_UNARY_CALLS(asin, Asin, LA)
EIGEN_MKL_VML_DECLARE_UNARY_CALLS(sinh, Sinh, LA)
EIGEN_MKL_VML_DECLARE_UNARY_CALLS(cos, Cos, LA)
EIGEN_MKL_VML_DECLARE_UNARY_CALLS(acos, Acos, LA)
EIGEN_MKL_VML_DECLARE_UNARY_CALLS(cosh, Cosh, LA)
EIGEN_MKL_VML_DECLARE_UNARY_CALLS(tan, Tan, LA)
EIGEN_MKL_VML_DECLARE_UNARY_CALLS(atan, Atan, LA)
EIGEN_MKL_VML_DECLARE_UNARY_CALLS(tanh, Tanh, LA)
// EIGEN_MKL_VML_DECLARE_UNARY_CALLS(abs, Abs, _)
EIGEN_MKL_VML_DECLARE_UNARY_CALLS(exp, Exp, LA)
EIGEN_MKL_VML_DECLARE_UNARY_CALLS(log, Ln, LA)
EIGEN_MKL_VML_DECLARE_UNARY_CALLS(log10, Log10, LA)
EIGEN_MKL_VML_DECLARE_UNARY_CALLS(sqrt, Sqrt, _)
EIGEN_MKL_VML_DECLARE_UNARY_CALLS_REAL(square, Sqr, _)
EIGEN_MKL_VML_DECLARE_UNARY_CALLS_CPLX(arg, Arg, _)
EIGEN_MKL_VML_DECLARE_UNARY_CALLS_REAL(round, Round, _)
EIGEN_MKL_VML_DECLARE_UNARY_CALLS_REAL(floor, Floor, _)
EIGEN_MKL_VML_DECLARE_UNARY_CALLS_REAL(ceil, Ceil, _)
#define EIGEN_MKL_VML_DECLARE_POW_CALL(EIGENOP, VMLOP, EIGENTYPE, VMLTYPE, VMLMODE) \
template< typename DstXprType, typename SrcXprNested, typename Plain> \
struct Assignment<DstXprType, CwiseBinaryOp<scalar_##EIGENOP##_op<EIGENTYPE,EIGENTYPE>, SrcXprNested, \
const CwiseNullaryOp<internal::scalar_constant_op<EIGENTYPE>,Plain> >, assign_op<EIGENTYPE,EIGENTYPE>, \
Dense2Dense, typename enable_if<vml_assign_traits<DstXprType,SrcXprNested>::EnableVml>::type> { \
typedef CwiseBinaryOp<scalar_##EIGENOP##_op<EIGENTYPE,EIGENTYPE>, SrcXprNested, \
const CwiseNullaryOp<internal::scalar_constant_op<EIGENTYPE>,Plain> > SrcXprType; \
static void run(DstXprType &dst, const SrcXprType &src, const assign_op<EIGENTYPE,EIGENTYPE> &func) { \
resize_if_allowed(dst, src, func); \
eigen_assert(dst.rows() == src.rows() && dst.cols() == src.cols()); \
VMLTYPE exponent = reinterpret_cast<const VMLTYPE&>(src.rhs().functor().m_other); \
if(vml_assign_traits<DstXprType,SrcXprNested>::Traversal==LinearTraversal) \
{ \
VMLOP( dst.size(), (const VMLTYPE*)src.lhs().data(), exponent, \
(VMLTYPE*)dst.data() EIGEN_PP_EXPAND(EIGEN_VMLMODE_EXPAND_##VMLMODE) ); \
} else { \
const Index outerSize = dst.outerSize(); \
for(Index outer = 0; outer < outerSize; ++outer) { \
const EIGENTYPE *src_ptr = src.IsRowMajor ? &(src.lhs().coeffRef(outer,0)) : \
&(src.lhs().coeffRef(0, outer)); \
EIGENTYPE *dst_ptr = dst.IsRowMajor ? &(dst.coeffRef(outer,0)) : &(dst.coeffRef(0, outer)); \
VMLOP( dst.innerSize(), (const VMLTYPE*)src_ptr, exponent, \
(VMLTYPE*)dst_ptr EIGEN_PP_EXPAND(EIGEN_VMLMODE_EXPAND_##VMLMODE)); \
} \
} \
} \
};
#define EIGEN_MKL_VML_DECLARE_UNARY_CALL_LA(EIGENOP, VMLOP, EIGENTYPE, VMLTYPE) \
template<> struct vml_call< scalar_##EIGENOP##_op<EIGENTYPE> > { \
enum { IsSupported = 1 }; \
static inline void run( const scalar_##EIGENOP##_op<EIGENTYPE>& /*func*/, \
int size, const EIGENTYPE* src, EIGENTYPE* dst) { \
MKL_INT64 vmlMode = EIGEN_MKL_VML_MODE; \
VMLOP(size, (const VMLTYPE*)src, (VMLTYPE*)dst, vmlMode); \
} \
};
#define EIGEN_MKL_VML_DECLARE_POW_CALL(EIGENOP, VMLOP, EIGENTYPE, VMLTYPE) \
template<> struct vml_call< scalar_##EIGENOP##_op<EIGENTYPE> > { \
enum { IsSupported = 1 }; \
static inline void run( const scalar_##EIGENOP##_op<EIGENTYPE>& func, \
int size, const EIGENTYPE* src, EIGENTYPE* dst) { \
EIGENTYPE exponent = func.m_exponent; \
MKL_INT64 vmlMode = EIGEN_MKL_VML_MODE; \
VMLOP(&size, (const VMLTYPE*)src, (const VMLTYPE*)&exponent, \
(VMLTYPE*)dst, &vmlMode); \
} \
};
#define EIGEN_MKL_VML_DECLARE_UNARY_CALLS_REAL(EIGENOP, VMLOP) \
EIGEN_MKL_VML_DECLARE_UNARY_CALL(EIGENOP, vs##VMLOP, float, float) \
EIGEN_MKL_VML_DECLARE_UNARY_CALL(EIGENOP, vd##VMLOP, double, double)
#define EIGEN_MKL_VML_DECLARE_UNARY_CALLS_COMPLEX(EIGENOP, VMLOP) \
EIGEN_MKL_VML_DECLARE_UNARY_CALL(EIGENOP, vc##VMLOP, scomplex, MKL_Complex8) \
EIGEN_MKL_VML_DECLARE_UNARY_CALL(EIGENOP, vz##VMLOP, dcomplex, MKL_Complex16)
#define EIGEN_MKL_VML_DECLARE_UNARY_CALLS(EIGENOP, VMLOP) \
EIGEN_MKL_VML_DECLARE_UNARY_CALLS_REAL(EIGENOP, VMLOP) \
EIGEN_MKL_VML_DECLARE_UNARY_CALLS_COMPLEX(EIGENOP, VMLOP)
#define EIGEN_MKL_VML_DECLARE_UNARY_CALLS_REAL_LA(EIGENOP, VMLOP) \
EIGEN_MKL_VML_DECLARE_UNARY_CALL_LA(EIGENOP, vms##VMLOP, float, float) \
EIGEN_MKL_VML_DECLARE_UNARY_CALL_LA(EIGENOP, vmd##VMLOP, double, double)
#define EIGEN_MKL_VML_DECLARE_UNARY_CALLS_COMPLEX_LA(EIGENOP, VMLOP) \
EIGEN_MKL_VML_DECLARE_UNARY_CALL_LA(EIGENOP, vmc##VMLOP, scomplex, MKL_Complex8) \
EIGEN_MKL_VML_DECLARE_UNARY_CALL_LA(EIGENOP, vmz##VMLOP, dcomplex, MKL_Complex16)
#define EIGEN_MKL_VML_DECLARE_UNARY_CALLS_LA(EIGENOP, VMLOP) \
EIGEN_MKL_VML_DECLARE_UNARY_CALLS_REAL_LA(EIGENOP, VMLOP) \
EIGEN_MKL_VML_DECLARE_UNARY_CALLS_COMPLEX_LA(EIGENOP, VMLOP)
EIGEN_MKL_VML_DECLARE_UNARY_CALLS_LA(sin, Sin)
EIGEN_MKL_VML_DECLARE_UNARY_CALLS_LA(asin, Asin)
EIGEN_MKL_VML_DECLARE_UNARY_CALLS_LA(cos, Cos)
EIGEN_MKL_VML_DECLARE_UNARY_CALLS_LA(acos, Acos)
EIGEN_MKL_VML_DECLARE_UNARY_CALLS_LA(tan, Tan)
//EIGEN_MKL_VML_DECLARE_UNARY_CALLS(abs, Abs)
EIGEN_MKL_VML_DECLARE_UNARY_CALLS_LA(exp, Exp)
EIGEN_MKL_VML_DECLARE_UNARY_CALLS_LA(log, Ln)
EIGEN_MKL_VML_DECLARE_UNARY_CALLS_LA(sqrt, Sqrt)
EIGEN_MKL_VML_DECLARE_UNARY_CALLS_REAL(square, Sqr)
// The vm*powx functions are not avaibale in the windows version of MKL.
#ifndef _WIN32
EIGEN_MKL_VML_DECLARE_POW_CALL(pow, vmspowx_, float, float)
EIGEN_MKL_VML_DECLARE_POW_CALL(pow, vmdpowx_, double, double)
EIGEN_MKL_VML_DECLARE_POW_CALL(pow, vmcpowx_, scomplex, MKL_Complex8)
EIGEN_MKL_VML_DECLARE_POW_CALL(pow, vmzpowx_, dcomplex, MKL_Complex16)
#endif
EIGEN_MKL_VML_DECLARE_POW_CALL(pow, vmsPowx, float, float, LA)
EIGEN_MKL_VML_DECLARE_POW_CALL(pow, vmdPowx, double, double, LA)
EIGEN_MKL_VML_DECLARE_POW_CALL(pow, vmcPowx, scomplex, MKL_Complex8, LA)
EIGEN_MKL_VML_DECLARE_POW_CALL(pow, vmzPowx, dcomplex, MKL_Complex16, LA)
} // end namespace internal

View File

@@ -32,7 +32,7 @@ class BandMatrixBase : public EigenBase<Derived>
};
typedef typename internal::traits<Derived>::Scalar Scalar;
typedef Matrix<Scalar,RowsAtCompileTime,ColsAtCompileTime> DenseMatrixType;
typedef typename DenseMatrixType::Index Index;
typedef typename DenseMatrixType::StorageIndex StorageIndex;
typedef typename internal::traits<Derived>::CoefficientsType CoefficientsType;
typedef EigenBase<Derived> Base;
@@ -161,15 +161,15 @@ class BandMatrixBase : public EigenBase<Derived>
*
* \brief Represents a rectangular matrix with a banded storage
*
* \param _Scalar Numeric type, i.e. float, double, int
* \param Rows Number of rows, or \b Dynamic
* \param Cols Number of columns, or \b Dynamic
* \param Supers Number of super diagonal
* \param Subs Number of sub diagonal
* \param _Options A combination of either \b #RowMajor or \b #ColMajor, and of \b #SelfAdjoint
* The former controls \ref TopicStorageOrders "storage order", and defaults to
* column-major. The latter controls whether the matrix represents a selfadjoint
* matrix in which case either Supers of Subs have to be null.
* \tparam _Scalar Numeric type, i.e. float, double, int
* \tparam _Rows Number of rows, or \b Dynamic
* \tparam _Cols Number of columns, or \b Dynamic
* \tparam _Supers Number of super diagonal
* \tparam _Subs Number of sub diagonal
* \tparam _Options A combination of either \b #RowMajor or \b #ColMajor, and of \b #SelfAdjoint
* The former controls \ref TopicStorageOrders "storage order", and defaults to
* column-major. The latter controls whether the matrix represents a selfadjoint
* matrix in which case either Supers of Subs have to be null.
*
* \sa class TridiagonalMatrix
*/
@@ -179,7 +179,7 @@ struct traits<BandMatrix<_Scalar,_Rows,_Cols,_Supers,_Subs,_Options> >
{
typedef _Scalar Scalar;
typedef Dense StorageKind;
typedef DenseIndex Index;
typedef Eigen::Index StorageIndex;
enum {
CoeffReadCost = NumTraits<Scalar>::ReadCost,
RowsAtCompileTime = _Rows,
@@ -201,10 +201,10 @@ class BandMatrix : public BandMatrixBase<BandMatrix<_Scalar,Rows,Cols,Supers,Sub
public:
typedef typename internal::traits<BandMatrix>::Scalar Scalar;
typedef typename internal::traits<BandMatrix>::Index Index;
typedef typename internal::traits<BandMatrix>::StorageIndex StorageIndex;
typedef typename internal::traits<BandMatrix>::CoefficientsType CoefficientsType;
inline BandMatrix(Index rows=Rows, Index cols=Cols, Index supers=Supers, Index subs=Subs)
explicit inline BandMatrix(Index rows=Rows, Index cols=Cols, Index supers=Supers, Index subs=Subs)
: m_coeffs(1+supers+subs,cols),
m_rows(rows), m_supers(supers), m_subs(subs)
{
@@ -241,7 +241,7 @@ struct traits<BandMatrixWrapper<_CoefficientsType,_Rows,_Cols,_Supers,_Subs,_Opt
{
typedef typename _CoefficientsType::Scalar Scalar;
typedef typename _CoefficientsType::StorageKind StorageKind;
typedef typename _CoefficientsType::Index Index;
typedef typename _CoefficientsType::StorageIndex StorageIndex;
enum {
CoeffReadCost = internal::traits<_CoefficientsType>::CoeffReadCost,
RowsAtCompileTime = _Rows,
@@ -264,9 +264,9 @@ class BandMatrixWrapper : public BandMatrixBase<BandMatrixWrapper<_CoefficientsT
typedef typename internal::traits<BandMatrixWrapper>::Scalar Scalar;
typedef typename internal::traits<BandMatrixWrapper>::CoefficientsType CoefficientsType;
typedef typename internal::traits<BandMatrixWrapper>::Index Index;
typedef typename internal::traits<BandMatrixWrapper>::StorageIndex StorageIndex;
inline BandMatrixWrapper(const CoefficientsType& coeffs, Index rows=_Rows, Index cols=_Cols, Index supers=_Supers, Index subs=_Subs)
explicit inline BandMatrixWrapper(const CoefficientsType& coeffs, Index rows=_Rows, Index cols=_Cols, Index supers=_Supers, Index subs=_Subs)
: m_coeffs(coeffs),
m_rows(rows), m_supers(supers), m_subs(subs)
{
@@ -302,9 +302,9 @@ class BandMatrixWrapper : public BandMatrixBase<BandMatrixWrapper<_CoefficientsT
*
* \brief Represents a tridiagonal matrix with a compact banded storage
*
* \param _Scalar Numeric type, i.e. float, double, int
* \param Size Number of rows and cols, or \b Dynamic
* \param _Options Can be 0 or \b SelfAdjoint
* \tparam Scalar Numeric type, i.e. float, double, int
* \tparam Size Number of rows and cols, or \b Dynamic
* \tparam Options Can be 0 or \b SelfAdjoint
*
* \sa class BandMatrix
*/
@@ -312,9 +312,9 @@ template<typename Scalar, int Size, int Options>
class TridiagonalMatrix : public BandMatrix<Scalar,Size,Size,Options&SelfAdjoint?0:1,1,Options|RowMajor>
{
typedef BandMatrix<Scalar,Size,Size,Options&SelfAdjoint?0:1,1,Options|RowMajor> Base;
typedef typename Base::Index Index;
typedef typename Base::StorageIndex StorageIndex;
public:
TridiagonalMatrix(Index size = Size) : Base(size,size,Options&SelfAdjoint?0:1,1) {}
explicit TridiagonalMatrix(Index size = Size) : Base(size,size,Options&SelfAdjoint?0:1,1) {}
inline typename Base::template DiagonalIntReturnType<1>::Type super()
{ return Base::template diagonal<1>(); }
@@ -327,6 +327,25 @@ class TridiagonalMatrix : public BandMatrix<Scalar,Size,Size,Options&SelfAdjoint
protected:
};
struct BandShape {};
template<typename _Scalar, int _Rows, int _Cols, int _Supers, int _Subs, int _Options>
struct evaluator_traits<BandMatrix<_Scalar,_Rows,_Cols,_Supers,_Subs,_Options> >
: public evaluator_traits_base<BandMatrix<_Scalar,_Rows,_Cols,_Supers,_Subs,_Options> >
{
typedef BandShape Shape;
};
template<typename _CoefficientsType,int _Rows, int _Cols, int _Supers, int _Subs,int _Options>
struct evaluator_traits<BandMatrixWrapper<_CoefficientsType,_Rows,_Cols,_Supers,_Subs,_Options> >
: public evaluator_traits_base<BandMatrixWrapper<_CoefficientsType,_Rows,_Cols,_Supers,_Subs,_Options> >
{
typedef BandShape Shape;
};
template<> struct AssignmentKind<DenseShape,BandShape> { typedef EigenBase2EigenBase Kind; };
} // end namespace internal
} // end namespace Eigen

View File

@@ -13,14 +13,70 @@
namespace Eigen {
namespace internal {
template<typename XprType, int BlockRows, int BlockCols, bool InnerPanel>
struct traits<Block<XprType, BlockRows, BlockCols, InnerPanel> > : traits<XprType>
{
typedef typename traits<XprType>::Scalar Scalar;
typedef typename traits<XprType>::StorageKind StorageKind;
typedef typename traits<XprType>::XprKind XprKind;
typedef typename ref_selector<XprType>::type XprTypeNested;
typedef typename remove_reference<XprTypeNested>::type _XprTypeNested;
enum{
MatrixRows = traits<XprType>::RowsAtCompileTime,
MatrixCols = traits<XprType>::ColsAtCompileTime,
RowsAtCompileTime = MatrixRows == 0 ? 0 : BlockRows,
ColsAtCompileTime = MatrixCols == 0 ? 0 : BlockCols,
MaxRowsAtCompileTime = BlockRows==0 ? 0
: RowsAtCompileTime != Dynamic ? int(RowsAtCompileTime)
: int(traits<XprType>::MaxRowsAtCompileTime),
MaxColsAtCompileTime = BlockCols==0 ? 0
: ColsAtCompileTime != Dynamic ? int(ColsAtCompileTime)
: int(traits<XprType>::MaxColsAtCompileTime),
XprTypeIsRowMajor = (int(traits<XprType>::Flags)&RowMajorBit) != 0,
IsRowMajor = (MaxRowsAtCompileTime==1&&MaxColsAtCompileTime!=1) ? 1
: (MaxColsAtCompileTime==1&&MaxRowsAtCompileTime!=1) ? 0
: XprTypeIsRowMajor,
HasSameStorageOrderAsXprType = (IsRowMajor == XprTypeIsRowMajor),
InnerSize = IsRowMajor ? int(ColsAtCompileTime) : int(RowsAtCompileTime),
InnerStrideAtCompileTime = HasSameStorageOrderAsXprType
? int(inner_stride_at_compile_time<XprType>::ret)
: int(outer_stride_at_compile_time<XprType>::ret),
OuterStrideAtCompileTime = HasSameStorageOrderAsXprType
? int(outer_stride_at_compile_time<XprType>::ret)
: int(inner_stride_at_compile_time<XprType>::ret),
// FIXME, this traits is rather specialized for dense object and it needs to be cleaned further
FlagsLvalueBit = is_lvalue<XprType>::value ? LvalueBit : 0,
FlagsRowMajorBit = IsRowMajor ? RowMajorBit : 0,
Flags = (traits<XprType>::Flags & (DirectAccessBit | (InnerPanel?CompressedAccessBit:0))) | FlagsLvalueBit | FlagsRowMajorBit,
// FIXME DirectAccessBit should not be handled by expressions
//
// Alignment is needed by MapBase's assertions
// We can sefely set it to false here. Internal alignment errors will be detected by an eigen_internal_assert in the respective evaluator
Alignment = 0
};
};
template<typename XprType, int BlockRows=Dynamic, int BlockCols=Dynamic, bool InnerPanel = false,
bool HasDirectAccess = internal::has_direct_access<XprType>::ret> class BlockImpl_dense;
} // end namespace internal
template<typename XprType, int BlockRows, int BlockCols, bool InnerPanel, typename StorageKind> class BlockImpl;
/** \class Block
* \ingroup Core_Module
*
* \brief Expression of a fixed-size or dynamic-size block
*
* \param XprType the type of the expression in which we are taking a block
* \param BlockRows the number of rows of the block we are taking at compile time (optional)
* \param BlockCols the number of columns of the block we are taking at compile time (optional)
* \tparam XprType the type of the expression in which we are taking a block
* \tparam BlockRows the number of rows of the block we are taking at compile time (optional)
* \tparam BlockCols the number of columns of the block we are taking at compile time (optional)
* \tparam InnerPanel is true, if the block maps to a set of rows of a row major matrix or
* to set of columns of a column major matrix (optional). The parameter allows to determine
* at compile time whether aligned access is possible on the block expression.
*
* This class represents an expression of either a fixed-size or dynamic-size block. It is the return
* type of DenseBase::block(Index,Index,Index,Index) and DenseBase::block<int,int>(Index,Index) and
@@ -44,62 +100,6 @@ namespace Eigen {
*
* \sa DenseBase::block(Index,Index,Index,Index), DenseBase::block(Index,Index), class VectorBlock
*/
namespace internal {
template<typename XprType, int BlockRows, int BlockCols, bool InnerPanel>
struct traits<Block<XprType, BlockRows, BlockCols, InnerPanel> > : traits<XprType>
{
typedef typename traits<XprType>::Scalar Scalar;
typedef typename traits<XprType>::StorageKind StorageKind;
typedef typename traits<XprType>::XprKind XprKind;
typedef typename nested<XprType>::type XprTypeNested;
typedef typename remove_reference<XprTypeNested>::type _XprTypeNested;
enum{
MatrixRows = traits<XprType>::RowsAtCompileTime,
MatrixCols = traits<XprType>::ColsAtCompileTime,
RowsAtCompileTime = MatrixRows == 0 ? 0 : BlockRows,
ColsAtCompileTime = MatrixCols == 0 ? 0 : BlockCols,
MaxRowsAtCompileTime = BlockRows==0 ? 0
: RowsAtCompileTime != Dynamic ? int(RowsAtCompileTime)
: int(traits<XprType>::MaxRowsAtCompileTime),
MaxColsAtCompileTime = BlockCols==0 ? 0
: ColsAtCompileTime != Dynamic ? int(ColsAtCompileTime)
: int(traits<XprType>::MaxColsAtCompileTime),
XprTypeIsRowMajor = (int(traits<XprType>::Flags)&RowMajorBit) != 0,
IsDense = is_same<StorageKind,Dense>::value,
IsRowMajor = (IsDense&&MaxRowsAtCompileTime==1&&MaxColsAtCompileTime!=1) ? 1
: (IsDense&&MaxColsAtCompileTime==1&&MaxRowsAtCompileTime!=1) ? 0
: XprTypeIsRowMajor,
HasSameStorageOrderAsXprType = (IsRowMajor == XprTypeIsRowMajor),
InnerSize = IsRowMajor ? int(ColsAtCompileTime) : int(RowsAtCompileTime),
InnerStrideAtCompileTime = HasSameStorageOrderAsXprType
? int(inner_stride_at_compile_time<XprType>::ret)
: int(outer_stride_at_compile_time<XprType>::ret),
OuterStrideAtCompileTime = HasSameStorageOrderAsXprType
? int(outer_stride_at_compile_time<XprType>::ret)
: int(inner_stride_at_compile_time<XprType>::ret),
MaskPacketAccessBit = (InnerSize == Dynamic || (InnerSize % packet_traits<Scalar>::size) == 0)
&& (InnerStrideAtCompileTime == 1)
? PacketAccessBit : 0,
MaskAlignedBit = (InnerPanel && (OuterStrideAtCompileTime!=Dynamic) && (((OuterStrideAtCompileTime * int(sizeof(Scalar))) % 16) == 0)) ? AlignedBit : 0,
FlagsLinearAccessBit = (RowsAtCompileTime == 1 || ColsAtCompileTime == 1 || (InnerPanel && (traits<XprType>::Flags&LinearAccessBit))) ? LinearAccessBit : 0,
FlagsLvalueBit = is_lvalue<XprType>::value ? LvalueBit : 0,
FlagsRowMajorBit = IsRowMajor ? RowMajorBit : 0,
Flags0 = traits<XprType>::Flags & ( (HereditaryBits & ~RowMajorBit) |
DirectAccessBit |
MaskPacketAccessBit |
MaskAlignedBit),
Flags = Flags0 | FlagsLinearAccessBit | FlagsLvalueBit | FlagsRowMajorBit
};
};
template<typename XprType, int BlockRows=Dynamic, int BlockCols=Dynamic, bool InnerPanel = false,
bool HasDirectAccess = internal::has_direct_access<XprType>::ret> class BlockImpl_dense;
} // end namespace internal
template<typename XprType, int BlockRows, int BlockCols, bool InnerPanel, typename StorageKind> class BlockImpl;
template<typename XprType, int BlockRows, int BlockCols, bool InnerPanel> class Block
: public BlockImpl<XprType, BlockRows, BlockCols, InnerPanel, typename internal::traits<XprType>::StorageKind>
{
@@ -109,9 +109,12 @@ template<typename XprType, int BlockRows, int BlockCols, bool InnerPanel> class
typedef Impl Base;
EIGEN_GENERIC_PUBLIC_INTERFACE(Block)
EIGEN_INHERIT_ASSIGNMENT_OPERATORS(Block)
typedef typename internal::remove_all<XprType>::type NestedExpression;
/** Column or Row constructor
*/
EIGEN_DEVICE_FUNC
inline Block(XprType& xpr, Index i) : Impl(xpr,i)
{
eigen_assert( (i>=0) && (
@@ -121,25 +124,27 @@ template<typename XprType, int BlockRows, int BlockCols, bool InnerPanel> class
/** Fixed-size constructor
*/
inline Block(XprType& xpr, Index a_startRow, Index a_startCol)
: Impl(xpr, a_startRow, a_startCol)
EIGEN_DEVICE_FUNC
inline Block(XprType& xpr, Index startRow, Index startCol)
: Impl(xpr, startRow, startCol)
{
EIGEN_STATIC_ASSERT(RowsAtCompileTime!=Dynamic && ColsAtCompileTime!=Dynamic,THIS_METHOD_IS_ONLY_FOR_FIXED_SIZE)
eigen_assert(a_startRow >= 0 && BlockRows >= 1 && a_startRow + BlockRows <= xpr.rows()
&& a_startCol >= 0 && BlockCols >= 1 && a_startCol + BlockCols <= xpr.cols());
eigen_assert(startRow >= 0 && BlockRows >= 0 && startRow + BlockRows <= xpr.rows()
&& startCol >= 0 && BlockCols >= 0 && startCol + BlockCols <= xpr.cols());
}
/** Dynamic-size constructor
*/
EIGEN_DEVICE_FUNC
inline Block(XprType& xpr,
Index a_startRow, Index a_startCol,
Index startRow, Index startCol,
Index blockRows, Index blockCols)
: Impl(xpr, a_startRow, a_startCol, blockRows, blockCols)
: Impl(xpr, startRow, startCol, blockRows, blockCols)
{
eigen_assert((RowsAtCompileTime==Dynamic || RowsAtCompileTime==blockRows)
&& (ColsAtCompileTime==Dynamic || ColsAtCompileTime==blockCols));
eigen_assert(a_startRow >= 0 && blockRows >= 0 && a_startRow <= xpr.rows() - blockRows
&& a_startCol >= 0 && blockCols >= 0 && a_startCol <= xpr.cols() - blockCols);
eigen_assert(startRow >= 0 && blockRows >= 0 && startRow <= xpr.rows() - blockRows
&& startCol >= 0 && blockCols >= 0 && startCol <= xpr.cols() - blockCols);
}
};
@@ -150,14 +155,15 @@ class BlockImpl<XprType, BlockRows, BlockCols, InnerPanel, Dense>
: public internal::BlockImpl_dense<XprType, BlockRows, BlockCols, InnerPanel>
{
typedef internal::BlockImpl_dense<XprType, BlockRows, BlockCols, InnerPanel> Impl;
typedef typename XprType::Index Index;
typedef typename XprType::StorageIndex StorageIndex;
public:
typedef Impl Base;
EIGEN_INHERIT_ASSIGNMENT_OPERATORS(BlockImpl)
inline BlockImpl(XprType& xpr, Index i) : Impl(xpr,i) {}
inline BlockImpl(XprType& xpr, Index a_startRow, Index a_startCol) : Impl(xpr, a_startRow, a_startCol) {}
inline BlockImpl(XprType& xpr, Index a_startRow, Index a_startCol, Index blockRows, Index blockCols)
: Impl(xpr, a_startRow, a_startCol, blockRows, blockCols) {}
EIGEN_DEVICE_FUNC inline BlockImpl(XprType& xpr, Index i) : Impl(xpr,i) {}
EIGEN_DEVICE_FUNC inline BlockImpl(XprType& xpr, Index startRow, Index startCol) : Impl(xpr, startRow, startCol) {}
EIGEN_DEVICE_FUNC
inline BlockImpl(XprType& xpr, Index startRow, Index startCol, Index blockRows, Index blockCols)
: Impl(xpr, startRow, startCol, blockRows, blockCols) {}
};
namespace internal {
@@ -167,16 +173,18 @@ template<typename XprType, int BlockRows, int BlockCols, bool InnerPanel, bool H
: public internal::dense_xpr_base<Block<XprType, BlockRows, BlockCols, InnerPanel> >::type
{
typedef Block<XprType, BlockRows, BlockCols, InnerPanel> BlockType;
typedef typename internal::ref_selector<XprType>::non_const_type XprTypeNested;
public:
typedef typename internal::dense_xpr_base<BlockType>::type Base;
EIGEN_DENSE_PUBLIC_INTERFACE(BlockType)
EIGEN_INHERIT_ASSIGNMENT_OPERATORS(BlockImpl_dense)
class InnerIterator;
// class InnerIterator; // FIXME apparently never used
/** Column or Row constructor
*/
EIGEN_DEVICE_FUNC
inline BlockImpl_dense(XprType& xpr, Index i)
: m_xpr(xpr),
// It is a row if and only if BlockRows==1 and BlockCols==XprType::ColsAtCompileTime,
@@ -191,75 +199,76 @@ template<typename XprType, int BlockRows, int BlockCols, bool InnerPanel, bool H
/** Fixed-size constructor
*/
inline BlockImpl_dense(XprType& xpr, Index a_startRow, Index a_startCol)
: m_xpr(xpr), m_startRow(a_startRow), m_startCol(a_startCol),
EIGEN_DEVICE_FUNC
inline BlockImpl_dense(XprType& xpr, Index startRow, Index startCol)
: m_xpr(xpr), m_startRow(startRow), m_startCol(startCol),
m_blockRows(BlockRows), m_blockCols(BlockCols)
{}
/** Dynamic-size constructor
*/
EIGEN_DEVICE_FUNC
inline BlockImpl_dense(XprType& xpr,
Index a_startRow, Index a_startCol,
Index startRow, Index startCol,
Index blockRows, Index blockCols)
: m_xpr(xpr), m_startRow(a_startRow), m_startCol(a_startCol),
: m_xpr(xpr), m_startRow(startRow), m_startCol(startCol),
m_blockRows(blockRows), m_blockCols(blockCols)
{}
inline Index rows() const { return m_blockRows.value(); }
inline Index cols() const { return m_blockCols.value(); }
EIGEN_DEVICE_FUNC inline Index rows() const { return m_blockRows.value(); }
EIGEN_DEVICE_FUNC inline Index cols() const { return m_blockCols.value(); }
EIGEN_DEVICE_FUNC
inline Scalar& coeffRef(Index rowId, Index colId)
{
EIGEN_STATIC_ASSERT_LVALUE(XprType)
return m_xpr.const_cast_derived()
.coeffRef(rowId + m_startRow.value(), colId + m_startCol.value());
return m_xpr.coeffRef(rowId + m_startRow.value(), colId + m_startCol.value());
}
EIGEN_DEVICE_FUNC
inline const Scalar& coeffRef(Index rowId, Index colId) const
{
return m_xpr.derived()
.coeffRef(rowId + m_startRow.value(), colId + m_startCol.value());
return m_xpr.derived().coeffRef(rowId + m_startRow.value(), colId + m_startCol.value());
}
EIGEN_DEVICE_FUNC
EIGEN_STRONG_INLINE const CoeffReturnType coeff(Index rowId, Index colId) const
{
return m_xpr.coeff(rowId + m_startRow.value(), colId + m_startCol.value());
}
EIGEN_DEVICE_FUNC
inline Scalar& coeffRef(Index index)
{
EIGEN_STATIC_ASSERT_LVALUE(XprType)
return m_xpr.const_cast_derived()
.coeffRef(m_startRow.value() + (RowsAtCompileTime == 1 ? 0 : index),
m_startCol.value() + (RowsAtCompileTime == 1 ? index : 0));
return m_xpr.coeffRef(m_startRow.value() + (RowsAtCompileTime == 1 ? 0 : index),
m_startCol.value() + (RowsAtCompileTime == 1 ? index : 0));
}
EIGEN_DEVICE_FUNC
inline const Scalar& coeffRef(Index index) const
{
return m_xpr.const_cast_derived()
.coeffRef(m_startRow.value() + (RowsAtCompileTime == 1 ? 0 : index),
m_startCol.value() + (RowsAtCompileTime == 1 ? index : 0));
return m_xpr.coeffRef(m_startRow.value() + (RowsAtCompileTime == 1 ? 0 : index),
m_startCol.value() + (RowsAtCompileTime == 1 ? index : 0));
}
EIGEN_DEVICE_FUNC
inline const CoeffReturnType coeff(Index index) const
{
return m_xpr
.coeff(m_startRow.value() + (RowsAtCompileTime == 1 ? 0 : index),
m_startCol.value() + (RowsAtCompileTime == 1 ? index : 0));
return m_xpr.coeff(m_startRow.value() + (RowsAtCompileTime == 1 ? 0 : index),
m_startCol.value() + (RowsAtCompileTime == 1 ? index : 0));
}
template<int LoadMode>
inline PacketScalar packet(Index rowId, Index colId) const
{
return m_xpr.template packet<Unaligned>
(rowId + m_startRow.value(), colId + m_startCol.value());
return m_xpr.template packet<Unaligned>(rowId + m_startRow.value(), colId + m_startCol.value());
}
template<int LoadMode>
inline void writePacket(Index rowId, Index colId, const PacketScalar& val)
{
m_xpr.const_cast_derived().template writePacket<Unaligned>
(rowId + m_startRow.value(), colId + m_startCol.value(), val);
m_xpr.template writePacket<Unaligned>(rowId + m_startRow.value(), colId + m_startCol.value(), val);
}
template<int LoadMode>
@@ -273,40 +282,46 @@ template<typename XprType, int BlockRows, int BlockCols, bool InnerPanel, bool H
template<int LoadMode>
inline void writePacket(Index index, const PacketScalar& val)
{
m_xpr.const_cast_derived().template writePacket<Unaligned>
m_xpr.template writePacket<Unaligned>
(m_startRow.value() + (RowsAtCompileTime == 1 ? 0 : index),
m_startCol.value() + (RowsAtCompileTime == 1 ? index : 0), val);
}
#ifdef EIGEN_PARSED_BY_DOXYGEN
/** \sa MapBase::data() */
inline const Scalar* data() const;
inline Index innerStride() const;
inline Index outerStride() const;
EIGEN_DEVICE_FUNC inline const Scalar* data() const;
EIGEN_DEVICE_FUNC inline Index innerStride() const;
EIGEN_DEVICE_FUNC inline Index outerStride() const;
#endif
const typename internal::remove_all<typename XprType::Nested>::type& nestedExpression() const
EIGEN_DEVICE_FUNC
const typename internal::remove_all<XprTypeNested>::type& nestedExpression() const
{
return m_xpr;
}
EIGEN_DEVICE_FUNC
XprType& nestedExpression() { return m_xpr; }
Index startRow() const
EIGEN_DEVICE_FUNC
StorageIndex startRow() const
{
return m_startRow.value();
}
Index startCol() const
EIGEN_DEVICE_FUNC
StorageIndex startCol() const
{
return m_startCol.value();
}
protected:
const typename XprType::Nested m_xpr;
const internal::variable_if_dynamic<Index, XprType::RowsAtCompileTime == 1 ? 0 : Dynamic> m_startRow;
const internal::variable_if_dynamic<Index, XprType::ColsAtCompileTime == 1 ? 0 : Dynamic> m_startCol;
const internal::variable_if_dynamic<Index, RowsAtCompileTime> m_blockRows;
const internal::variable_if_dynamic<Index, ColsAtCompileTime> m_blockCols;
XprTypeNested m_xpr;
const internal::variable_if_dynamic<StorageIndex, (XprType::RowsAtCompileTime == 1 && BlockRows==1) ? 0 : Dynamic> m_startRow;
const internal::variable_if_dynamic<StorageIndex, (XprType::ColsAtCompileTime == 1 && BlockCols==1) ? 0 : Dynamic> m_startCol;
const internal::variable_if_dynamic<StorageIndex, RowsAtCompileTime> m_blockRows;
const internal::variable_if_dynamic<StorageIndex, ColsAtCompileTime> m_blockCols;
};
/** \internal Internal implementation of dense Blocks in the direct access case.*/
@@ -315,6 +330,10 @@ class BlockImpl_dense<XprType,BlockRows,BlockCols, InnerPanel,true>
: public MapBase<Block<XprType, BlockRows, BlockCols, InnerPanel> >
{
typedef Block<XprType, BlockRows, BlockCols, InnerPanel> BlockType;
typedef typename internal::ref_selector<XprType>::non_const_type XprTypeNested;
enum {
XprTypeIsRowMajor = (int(traits<XprType>::Flags)&RowMajorBit) != 0
};
public:
typedef MapBase<BlockType> Base;
@@ -323,42 +342,52 @@ class BlockImpl_dense<XprType,BlockRows,BlockCols, InnerPanel,true>
/** Column or Row constructor
*/
EIGEN_DEVICE_FUNC
inline BlockImpl_dense(XprType& xpr, Index i)
: Base(internal::const_cast_ptr(&xpr.coeffRef(
(BlockRows==1) && (BlockCols==XprType::ColsAtCompileTime) ? i : 0,
(BlockRows==XprType::RowsAtCompileTime) && (BlockCols==1) ? i : 0)),
: Base(xpr.data() + i * ( ((BlockRows==1) && (BlockCols==XprType::ColsAtCompileTime) && (!XprTypeIsRowMajor))
|| ((BlockRows==XprType::RowsAtCompileTime) && (BlockCols==1) && ( XprTypeIsRowMajor)) ? xpr.innerStride() : xpr.outerStride()),
BlockRows==1 ? 1 : xpr.rows(),
BlockCols==1 ? 1 : xpr.cols()),
m_xpr(xpr)
m_xpr(xpr),
m_startRow( (BlockRows==1) && (BlockCols==XprType::ColsAtCompileTime) ? i : 0),
m_startCol( (BlockRows==XprType::RowsAtCompileTime) && (BlockCols==1) ? i : 0)
{
init();
}
/** Fixed-size constructor
*/
EIGEN_DEVICE_FUNC
inline BlockImpl_dense(XprType& xpr, Index startRow, Index startCol)
: Base(internal::const_cast_ptr(&xpr.coeffRef(startRow,startCol))), m_xpr(xpr)
: Base(xpr.data()+xpr.innerStride()*(XprTypeIsRowMajor?startCol:startRow) + xpr.outerStride()*(XprTypeIsRowMajor?startRow:startCol)),
m_xpr(xpr), m_startRow(startRow), m_startCol(startCol)
{
init();
}
/** Dynamic-size constructor
*/
EIGEN_DEVICE_FUNC
inline BlockImpl_dense(XprType& xpr,
Index startRow, Index startCol,
Index blockRows, Index blockCols)
: Base(internal::const_cast_ptr(&xpr.coeffRef(startRow,startCol)), blockRows, blockCols),
m_xpr(xpr)
: Base(xpr.data()+xpr.innerStride()*(XprTypeIsRowMajor?startCol:startRow) + xpr.outerStride()*(XprTypeIsRowMajor?startRow:startCol), blockRows, blockCols),
m_xpr(xpr), m_startRow(startRow), m_startCol(startCol)
{
init();
}
const typename internal::remove_all<typename XprType::Nested>::type& nestedExpression() const
EIGEN_DEVICE_FUNC
const typename internal::remove_all<XprTypeNested>::type& nestedExpression() const
{
return m_xpr;
}
EIGEN_DEVICE_FUNC
XprType& nestedExpression() { return m_xpr; }
/** \sa MapBase::innerStride() */
EIGEN_DEVICE_FUNC
inline Index innerStride() const
{
return internal::traits<BlockType>::HasSameStorageOrderAsXprType
@@ -367,11 +396,24 @@ class BlockImpl_dense<XprType,BlockRows,BlockCols, InnerPanel,true>
}
/** \sa MapBase::outerStride() */
EIGEN_DEVICE_FUNC
inline Index outerStride() const
{
return m_outerStride;
}
EIGEN_DEVICE_FUNC
StorageIndex startRow() const
{
return m_startRow.value();
}
EIGEN_DEVICE_FUNC
StorageIndex startCol() const
{
return m_startCol.value();
}
#ifndef __SUNPRO_CC
// FIXME sunstudio is not friendly with the above friend...
// META-FIXME there is no 'friend' keyword around here. Is this obsolete?
@@ -380,6 +422,7 @@ class BlockImpl_dense<XprType,BlockRows,BlockCols, InnerPanel,true>
#ifndef EIGEN_PARSED_BY_DOXYGEN
/** \internal used by allowAligned() */
EIGEN_DEVICE_FUNC
inline BlockImpl_dense(XprType& xpr, const Scalar* data, Index blockRows, Index blockCols)
: Base(data, blockRows, blockCols), m_xpr(xpr)
{
@@ -388,6 +431,7 @@ class BlockImpl_dense<XprType,BlockRows,BlockCols, InnerPanel,true>
#endif
protected:
EIGEN_DEVICE_FUNC
void init()
{
m_outerStride = internal::traits<BlockType>::HasSameStorageOrderAsXprType
@@ -395,7 +439,9 @@ class BlockImpl_dense<XprType,BlockRows,BlockCols, InnerPanel,true>
: m_xpr.innerStride();
}
typename XprType::Nested m_xpr;
XprTypeNested m_xpr;
const internal::variable_if_dynamic<StorageIndex, (XprType::RowsAtCompileTime == 1 && BlockRows==1) ? 0 : Dynamic> m_startRow;
const internal::variable_if_dynamic<StorageIndex, (XprType::ColsAtCompileTime == 1 && BlockCols==1) ? 0 : Dynamic> m_startCol;
Index m_outerStride;
};

View File

@@ -17,9 +17,10 @@ namespace internal {
template<typename Derived, int UnrollCount>
struct all_unroller
{
typedef typename Derived::ExpressionTraits Traits;
enum {
col = (UnrollCount-1) / Derived::RowsAtCompileTime,
row = (UnrollCount-1) % Derived::RowsAtCompileTime
col = (UnrollCount-1) / Traits::RowsAtCompileTime,
row = (UnrollCount-1) % Traits::RowsAtCompileTime
};
static inline bool run(const Derived &mat)
@@ -43,11 +44,12 @@ struct all_unroller<Derived, Dynamic>
template<typename Derived, int UnrollCount>
struct any_unroller
{
typedef typename Derived::ExpressionTraits Traits;
enum {
col = (UnrollCount-1) / Derived::RowsAtCompileTime,
row = (UnrollCount-1) % Derived::RowsAtCompileTime
col = (UnrollCount-1) / Traits::RowsAtCompileTime,
row = (UnrollCount-1) % Traits::RowsAtCompileTime
};
static inline bool run(const Derived &mat)
{
return any_unroller<Derived, UnrollCount-1>::run(mat) || mat.coeff(row, col);
@@ -78,19 +80,19 @@ struct any_unroller<Derived, Dynamic>
template<typename Derived>
inline bool DenseBase<Derived>::all() const
{
typedef internal::evaluator<Derived> Evaluator;
enum {
unroll = SizeAtCompileTime != Dynamic
&& CoeffReadCost != Dynamic
&& NumTraits<Scalar>::AddCost != Dynamic
&& SizeAtCompileTime * (CoeffReadCost + NumTraits<Scalar>::AddCost) <= EIGEN_UNROLLING_LIMIT
&& SizeAtCompileTime * (Evaluator::CoeffReadCost + NumTraits<Scalar>::AddCost) <= EIGEN_UNROLLING_LIMIT
};
Evaluator evaluator(derived());
if(unroll)
return internal::all_unroller<Derived, unroll ? int(SizeAtCompileTime) : Dynamic>::run(derived());
return internal::all_unroller<Evaluator, unroll ? int(SizeAtCompileTime) : Dynamic>::run(evaluator);
else
{
for(Index j = 0; j < cols(); ++j)
for(Index i = 0; i < rows(); ++i)
if (!coeff(i, j)) return false;
if (!evaluator.coeff(i, j)) return false;
return true;
}
}
@@ -102,19 +104,19 @@ inline bool DenseBase<Derived>::all() const
template<typename Derived>
inline bool DenseBase<Derived>::any() const
{
typedef internal::evaluator<Derived> Evaluator;
enum {
unroll = SizeAtCompileTime != Dynamic
&& CoeffReadCost != Dynamic
&& NumTraits<Scalar>::AddCost != Dynamic
&& SizeAtCompileTime * (CoeffReadCost + NumTraits<Scalar>::AddCost) <= EIGEN_UNROLLING_LIMIT
&& SizeAtCompileTime * (Evaluator::CoeffReadCost + NumTraits<Scalar>::AddCost) <= EIGEN_UNROLLING_LIMIT
};
Evaluator evaluator(derived());
if(unroll)
return internal::any_unroller<Derived, unroll ? int(SizeAtCompileTime) : Dynamic>::run(derived());
return internal::any_unroller<Evaluator, unroll ? int(SizeAtCompileTime) : Dynamic>::run(evaluator);
else
{
for(Index j = 0; j < cols(); ++j)
for(Index i = 0; i < rows(); ++i)
if (coeff(i, j)) return true;
if (evaluator.coeff(i, j)) return true;
return false;
}
}
@@ -124,7 +126,7 @@ inline bool DenseBase<Derived>::any() const
* \sa all(), any()
*/
template<typename Derived>
inline typename DenseBase<Derived>::Index DenseBase<Derived>::count() const
inline Eigen::Index DenseBase<Derived>::count() const
{
return derived().template cast<bool>().template cast<Index>().sum();
}
@@ -136,7 +138,11 @@ inline typename DenseBase<Derived>::Index DenseBase<Derived>::count() const
template<typename Derived>
inline bool DenseBase<Derived>::hasNaN() const
{
#if EIGEN_COMP_MSVC || (defined __FAST_MATH__)
return derived().array().isNaN().any();
#else
return !((derived().array()==derived().array()).all());
#endif
}
/** \returns true if \c *this contains only finite numbers, i.e., no NaN and no +/-INF values.
@@ -146,7 +152,11 @@ inline bool DenseBase<Derived>::hasNaN() const
template<typename Derived>
inline bool DenseBase<Derived>::allFinite() const
{
#if EIGEN_COMP_MSVC || (defined __FAST_MATH__)
return derived().array().isFinite().all();
#else
return !((derived()-derived()).hasNaN());
#endif
}
} // end namespace Eigen

View File

@@ -22,14 +22,14 @@ namespace Eigen {
* the return type of MatrixBase::operator<<, and most of the time this is the only
* way it is used.
*
* \sa \ref MatrixBaseCommaInitRef "MatrixBase::operator<<", CommaInitializer::finished()
* \sa \blank \ref MatrixBaseCommaInitRef "MatrixBase::operator<<", CommaInitializer::finished()
*/
template<typename XprType>
struct CommaInitializer
{
typedef typename XprType::Scalar Scalar;
typedef typename XprType::Index Index;
EIGEN_DEVICE_FUNC
inline CommaInitializer(XprType& xpr, const Scalar& s)
: m_xpr(xpr), m_row(0), m_col(1), m_currentBlockRows(1)
{
@@ -37,6 +37,7 @@ struct CommaInitializer
}
template<typename OtherDerived>
EIGEN_DEVICE_FUNC
inline CommaInitializer(XprType& xpr, const DenseBase<OtherDerived>& other)
: m_xpr(xpr), m_row(0), m_col(other.cols()), m_currentBlockRows(other.rows())
{
@@ -46,6 +47,7 @@ struct CommaInitializer
/* Copy/Move constructor which transfers ownership. This is crucial in
* absence of return value optimization to avoid assertions during destruction. */
// FIXME in C++11 mode this could be replaced by a proper RValue constructor
EIGEN_DEVICE_FUNC
inline CommaInitializer(const CommaInitializer& o)
: m_xpr(o.m_xpr), m_row(o.m_row), m_col(o.m_col), m_currentBlockRows(o.m_currentBlockRows) {
// Mark original object as finished. In absence of R-value references we need to const_cast:
@@ -55,6 +57,7 @@ struct CommaInitializer
}
/* inserts a scalar value in the target matrix */
EIGEN_DEVICE_FUNC
CommaInitializer& operator,(const Scalar& s)
{
if (m_col==m_xpr.cols())
@@ -74,11 +77,10 @@ struct CommaInitializer
/* inserts a matrix expression in the target matrix */
template<typename OtherDerived>
EIGEN_DEVICE_FUNC
CommaInitializer& operator,(const DenseBase<OtherDerived>& other)
{
if(other.cols()==0 || other.rows()==0)
return *this;
if (m_col==m_xpr.cols())
if (m_col==m_xpr.cols() && (other.cols()!=0 || other.rows()!=m_currentBlockRows))
{
m_row+=m_currentBlockRows;
m_col = 0;
@@ -86,24 +88,22 @@ struct CommaInitializer
eigen_assert(m_row+m_currentBlockRows<=m_xpr.rows()
&& "Too many rows passed to comma initializer (operator<<)");
}
eigen_assert(m_col<m_xpr.cols()
eigen_assert((m_col + other.cols() <= m_xpr.cols())
&& "Too many coefficients passed to comma initializer (operator<<)");
eigen_assert(m_currentBlockRows==other.rows());
if (OtherDerived::SizeAtCompileTime != Dynamic)
m_xpr.template block<OtherDerived::RowsAtCompileTime != Dynamic ? OtherDerived::RowsAtCompileTime : 1,
OtherDerived::ColsAtCompileTime != Dynamic ? OtherDerived::ColsAtCompileTime : 1>
(m_row, m_col) = other;
else
m_xpr.block(m_row, m_col, other.rows(), other.cols()) = other;
m_xpr.template block<OtherDerived::RowsAtCompileTime, OtherDerived::ColsAtCompileTime>
(m_row, m_col, other.rows(), other.cols()) = other;
m_col += other.cols();
return *this;
}
EIGEN_DEVICE_FUNC
inline ~CommaInitializer()
#if defined VERIFY_RAISES_ASSERT && (!defined EIGEN_NO_ASSERTION_CHECKING) && defined EIGEN_EXCEPTIONS
EIGEN_EXCEPTION_SPEC(Eigen::eigen_assert_exception)
#endif
{
eigen_assert((m_row+m_currentBlockRows) == m_xpr.rows()
&& m_col == m_xpr.cols()
&& "Too few coefficients passed to comma initializer (operator<<)");
finished();
}
/** \returns the built matrix once all its coefficients have been set.
@@ -113,9 +113,15 @@ struct CommaInitializer
* quaternion.fromRotationMatrix((Matrix3f() << axis0, axis1, axis2).finished());
* \endcode
*/
inline XprType& finished() { return m_xpr; }
EIGEN_DEVICE_FUNC
inline XprType& finished() {
eigen_assert(((m_row+m_currentBlockRows) == m_xpr.rows() || m_xpr.cols() == 0)
&& m_col == m_xpr.cols()
&& "Too few coefficients passed to comma initializer (operator<<)");
return m_xpr;
}
XprType& m_xpr; // target expression
XprType& m_xpr; // target expression
Index m_row; // current row id
Index m_col; // current col id
Index m_currentBlockRows; // current block height

View File

@@ -0,0 +1,175 @@
// This file is part of Eigen, a lightweight C++ template library
// for linear algebra.
//
// Copyright (C) 2016 Rasmus Munk Larsen (rmlarsen@google.com)
//
// This Source Code Form is subject to the terms of the Mozilla
// Public License v. 2.0. If a copy of the MPL was not distributed
// with this file, You can obtain one at http://mozilla.org/MPL/2.0/.
#ifndef EIGEN_CONDITIONESTIMATOR_H
#define EIGEN_CONDITIONESTIMATOR_H
namespace Eigen {
namespace internal {
template <typename Vector, typename RealVector, bool IsComplex>
struct rcond_compute_sign {
static inline Vector run(const Vector& v) {
const RealVector v_abs = v.cwiseAbs();
return (v_abs.array() == static_cast<typename Vector::RealScalar>(0))
.select(Vector::Ones(v.size()), v.cwiseQuotient(v_abs));
}
};
// Partial specialization to avoid elementwise division for real vectors.
template <typename Vector>
struct rcond_compute_sign<Vector, Vector, false> {
static inline Vector run(const Vector& v) {
return (v.array() < static_cast<typename Vector::RealScalar>(0))
.select(-Vector::Ones(v.size()), Vector::Ones(v.size()));
}
};
/**
* \returns an estimate of ||inv(matrix)||_1 given a decomposition of
* \a matrix that implements .solve() and .adjoint().solve() methods.
*
* This function implements Algorithms 4.1 and 5.1 from
* http://www.maths.manchester.ac.uk/~higham/narep/narep135.pdf
* which also forms the basis for the condition number estimators in
* LAPACK. Since at most 10 calls to the solve method of dec are
* performed, the total cost is O(dims^2), as opposed to O(dims^3)
* needed to compute the inverse matrix explicitly.
*
* The most common usage is in estimating the condition number
* ||matrix||_1 * ||inv(matrix)||_1. The first term ||matrix||_1 can be
* computed directly in O(n^2) operations.
*
* Supports the following decompositions: FullPivLU, PartialPivLU, LDLT, and
* LLT.
*
* \sa FullPivLU, PartialPivLU, LDLT, LLT.
*/
template <typename Decomposition>
typename Decomposition::RealScalar rcond_invmatrix_L1_norm_estimate(const Decomposition& dec)
{
typedef typename Decomposition::MatrixType MatrixType;
typedef typename Decomposition::Scalar Scalar;
typedef typename Decomposition::RealScalar RealScalar;
typedef typename internal::plain_col_type<MatrixType>::type Vector;
typedef typename internal::plain_col_type<MatrixType, RealScalar>::type RealVector;
const bool is_complex = (NumTraits<Scalar>::IsComplex != 0);
eigen_assert(dec.rows() == dec.cols());
const Index n = dec.rows();
if (n == 0)
return 0;
// Disable Index to float conversion warning
#ifdef __INTEL_COMPILER
#pragma warning push
#pragma warning ( disable : 2259 )
#endif
Vector v = dec.solve(Vector::Ones(n) / Scalar(n));
#ifdef __INTEL_COMPILER
#pragma warning pop
#endif
// lower_bound is a lower bound on
// ||inv(matrix)||_1 = sup_v ||inv(matrix) v||_1 / ||v||_1
// and is the objective maximized by the ("super-") gradient ascent
// algorithm below.
RealScalar lower_bound = v.template lpNorm<1>();
if (n == 1)
return lower_bound;
// Gradient ascent algorithm follows: We know that the optimum is achieved at
// one of the simplices v = e_i, so in each iteration we follow a
// super-gradient to move towards the optimal one.
RealScalar old_lower_bound = lower_bound;
Vector sign_vector(n);
Vector old_sign_vector;
Index v_max_abs_index = -1;
Index old_v_max_abs_index = v_max_abs_index;
for (int k = 0; k < 4; ++k)
{
sign_vector = internal::rcond_compute_sign<Vector, RealVector, is_complex>::run(v);
if (k > 0 && !is_complex && sign_vector == old_sign_vector) {
// Break if the solution stagnated.
break;
}
// v_max_abs_index = argmax |real( inv(matrix)^T * sign_vector )|
v = dec.adjoint().solve(sign_vector);
v.real().cwiseAbs().maxCoeff(&v_max_abs_index);
if (v_max_abs_index == old_v_max_abs_index) {
// Break if the solution stagnated.
break;
}
// Move to the new simplex e_j, where j = v_max_abs_index.
v = dec.solve(Vector::Unit(n, v_max_abs_index)); // v = inv(matrix) * e_j.
lower_bound = v.template lpNorm<1>();
if (lower_bound <= old_lower_bound) {
// Break if the gradient step did not increase the lower_bound.
break;
}
if (!is_complex) {
old_sign_vector = sign_vector;
}
old_v_max_abs_index = v_max_abs_index;
old_lower_bound = lower_bound;
}
// The following calculates an independent estimate of ||matrix||_1 by
// multiplying matrix by a vector with entries of slowly increasing
// magnitude and alternating sign:
// v_i = (-1)^{i} (1 + (i / (dim-1))), i = 0,...,dim-1.
// This improvement to Hager's algorithm above is due to Higham. It was
// added to make the algorithm more robust in certain corner cases where
// large elements in the matrix might otherwise escape detection due to
// exact cancellation (especially when op and op_adjoint correspond to a
// sequence of backsubstitutions and permutations), which could cause
// Hager's algorithm to vastly underestimate ||matrix||_1.
Scalar alternating_sign(RealScalar(1));
for (Index i = 0; i < n; ++i) {
// The static_cast is needed when Scalar is a complex and RealScalar implements expression templates
v[i] = alternating_sign * static_cast<RealScalar>(RealScalar(1) + (RealScalar(i) / (RealScalar(n - 1))));
alternating_sign = -alternating_sign;
}
v = dec.solve(v);
const RealScalar alternate_lower_bound = (2 * v.template lpNorm<1>()) / (3 * RealScalar(n));
return numext::maxi(lower_bound, alternate_lower_bound);
}
/** \brief Reciprocal condition number estimator.
*
* Computing a decomposition of a dense matrix takes O(n^3) operations, while
* this method estimates the condition number quickly and reliably in O(n^2)
* operations.
*
* \returns an estimate of the reciprocal condition number
* (1 / (||matrix||_1 * ||inv(matrix)||_1)) of matrix, given ||matrix||_1 and
* its decomposition. Supports the following decompositions: FullPivLU,
* PartialPivLU, LDLT, and LLT.
*
* \sa FullPivLU, PartialPivLU, LDLT, LLT.
*/
template <typename Decomposition>
typename Decomposition::RealScalar
rcond_estimate_helper(typename Decomposition::RealScalar matrix_norm, const Decomposition& dec)
{
typedef typename Decomposition::RealScalar RealScalar;
eigen_assert(dec.rows() == dec.cols());
if (dec.rows() == 0) return NumTraits<RealScalar>::infinity();
if (matrix_norm == RealScalar(0)) return RealScalar(0);
if (dec.rows() == 1) return RealScalar(1);
const RealScalar inverse_matrix_norm = rcond_invmatrix_L1_norm_estimate(dec);
return (inverse_matrix_norm == RealScalar(0) ? RealScalar(0)
: (RealScalar(1) / inverse_matrix_norm) / matrix_norm);
}
} // namespace internal
} // namespace Eigen
#endif

File diff suppressed because it is too large Load Diff

View File

@@ -1,7 +1,7 @@
// This file is part of Eigen, a lightweight C++ template library
// for linear algebra.
//
// Copyright (C) 2008-2010 Gael Guennebaud <gael.guennebaud@inria.fr>
// Copyright (C) 2008-2014 Gael Guennebaud <gael.guennebaud@inria.fr>
//
// This Source Code Form is subject to the terms of the Mozilla
// Public License v. 2.0. If a copy of the MPL was not distributed
@@ -15,47 +15,113 @@ namespace Eigen {
/* This file contains the respective InnerIterator definition of the expressions defined in Eigen/Core
*/
/** \ingroup SparseCore_Module
* \class InnerIterator
* \brief An InnerIterator allows to loop over the element of a sparse (or dense) matrix or expression
*
* todo
namespace internal {
template<typename XprType, typename EvaluatorKind>
class inner_iterator_selector;
}
/** \class InnerIterator
* \brief An InnerIterator allows to loop over the element of any matrix expression.
*
* \warning To be used with care because an evaluator is constructed every time an InnerIterator iterator is constructed.
*
* TODO: add a usage example
*/
// generic version for dense matrix and expressions
template<typename Derived> class DenseBase<Derived>::InnerIterator
template<typename XprType>
class InnerIterator
{
protected:
typedef typename Derived::Scalar Scalar;
typedef typename Derived::Index Index;
enum { IsRowMajor = (Derived::Flags&RowMajorBit)==RowMajorBit };
public:
EIGEN_STRONG_INLINE InnerIterator(const Derived& expr, Index outer)
: m_expression(expr), m_inner(0), m_outer(outer), m_end(expr.innerSize())
{}
EIGEN_STRONG_INLINE Scalar value() const
{
return (IsRowMajor) ? m_expression.coeff(m_outer, m_inner)
: m_expression.coeff(m_inner, m_outer);
}
EIGEN_STRONG_INLINE InnerIterator& operator++() { m_inner++; return *this; }
EIGEN_STRONG_INLINE Index index() const { return m_inner; }
inline Index row() const { return IsRowMajor ? m_outer : index(); }
inline Index col() const { return IsRowMajor ? index() : m_outer; }
EIGEN_STRONG_INLINE operator bool() const { return m_inner < m_end && m_inner>=0; }
protected:
const Derived& m_expression;
Index m_inner;
const Index m_outer;
const Index m_end;
protected:
typedef internal::inner_iterator_selector<XprType, typename internal::evaluator_traits<XprType>::Kind> IteratorType;
typedef internal::evaluator<XprType> EvaluatorType;
typedef typename internal::traits<XprType>::Scalar Scalar;
public:
/** Construct an iterator over the \a outerId -th row or column of \a xpr */
InnerIterator(const XprType &xpr, const Index &outerId)
: m_eval(xpr), m_iter(m_eval, outerId, xpr.innerSize())
{}
/// \returns the value of the current coefficient.
EIGEN_STRONG_INLINE Scalar value() const { return m_iter.value(); }
/** Increment the iterator \c *this to the next non-zero coefficient.
* Explicit zeros are not skipped over. To skip explicit zeros, see class SparseView
*/
EIGEN_STRONG_INLINE InnerIterator& operator++() { m_iter.operator++(); return *this; }
/// \returns the column or row index of the current coefficient.
EIGEN_STRONG_INLINE Index index() const { return m_iter.index(); }
/// \returns the row index of the current coefficient.
EIGEN_STRONG_INLINE Index row() const { return m_iter.row(); }
/// \returns the column index of the current coefficient.
EIGEN_STRONG_INLINE Index col() const { return m_iter.col(); }
/// \returns \c true if the iterator \c *this still references a valid coefficient.
EIGEN_STRONG_INLINE operator bool() const { return m_iter; }
protected:
EvaluatorType m_eval;
IteratorType m_iter;
private:
// If you get here, then you're not using the right InnerIterator type, e.g.:
// SparseMatrix<double,RowMajor> A;
// SparseMatrix<double>::InnerIterator it(A,0);
template<typename T> InnerIterator(const EigenBase<T>&,Index outer);
};
namespace internal {
// Generic inner iterator implementation for dense objects
template<typename XprType>
class inner_iterator_selector<XprType, IndexBased>
{
protected:
typedef evaluator<XprType> EvaluatorType;
typedef typename traits<XprType>::Scalar Scalar;
enum { IsRowMajor = (XprType::Flags&RowMajorBit)==RowMajorBit };
public:
EIGEN_STRONG_INLINE inner_iterator_selector(const EvaluatorType &eval, const Index &outerId, const Index &innerSize)
: m_eval(eval), m_inner(0), m_outer(outerId), m_end(innerSize)
{}
EIGEN_STRONG_INLINE Scalar value() const
{
return (IsRowMajor) ? m_eval.coeff(m_outer, m_inner)
: m_eval.coeff(m_inner, m_outer);
}
EIGEN_STRONG_INLINE inner_iterator_selector& operator++() { m_inner++; return *this; }
EIGEN_STRONG_INLINE Index index() const { return m_inner; }
inline Index row() const { return IsRowMajor ? m_outer : index(); }
inline Index col() const { return IsRowMajor ? index() : m_outer; }
EIGEN_STRONG_INLINE operator bool() const { return m_inner < m_end && m_inner>=0; }
protected:
const EvaluatorType& m_eval;
Index m_inner;
const Index m_outer;
const Index m_end;
};
// For iterator-based evaluator, inner-iterator is already implemented as
// evaluator<>::InnerIterator
template<typename XprType>
class inner_iterator_selector<XprType, IteratorBased>
: public evaluator<XprType>::InnerIterator
{
protected:
typedef typename evaluator<XprType>::InnerIterator Base;
typedef evaluator<XprType> EvaluatorType;
public:
EIGEN_STRONG_INLINE inner_iterator_selector(const EvaluatorType &eval, const Index &outerId, const Index &/*innerSize*/)
: Base(eval, outerId)
{}
};
} // end namespace internal
} // end namespace Eigen
#endif // EIGEN_COREITERATORS_H

View File

@@ -1,7 +1,7 @@
// This file is part of Eigen, a lightweight C++ template library
// for linear algebra.
//
// Copyright (C) 2008-2009 Gael Guennebaud <gael.guennebaud@inria.fr>
// Copyright (C) 2008-2014 Gael Guennebaud <gael.guennebaud@inria.fr>
// Copyright (C) 2006-2008 Benoit Jacob <jacob.benoit.1@gmail.com>
//
// This Source Code Form is subject to the terms of the Mozilla
@@ -13,26 +13,6 @@
namespace Eigen {
/** \class CwiseBinaryOp
* \ingroup Core_Module
*
* \brief Generic expression where a coefficient-wise binary operator is applied to two expressions
*
* \param BinaryOp template functor implementing the operator
* \param Lhs the type of the left-hand side
* \param Rhs the type of the right-hand side
*
* This class represents an expression where a coefficient-wise binary operator is applied to two expressions.
* It is the return type of binary operators, by which we mean only those binary operators where
* both the left-hand side and the right-hand side are Eigen expressions.
* For example, the return type of matrix1+matrix2 is a CwiseBinaryOp.
*
* Most of the time, this is the only way that it is used, so you typically don't have to name
* CwiseBinaryOp types explicitly.
*
* \sa MatrixBase::binaryExpr(const MatrixBase<OtherDerived> &,const CustomBinaryOp &) const, class CwiseUnaryOp, class CwiseNullaryOp
*/
namespace internal {
template<typename BinaryOp, typename Lhs, typename Rhs>
struct traits<CwiseBinaryOp<BinaryOp, Lhs, Rhs> >
@@ -52,77 +32,75 @@ struct traits<CwiseBinaryOp<BinaryOp, Lhs, Rhs> >
// we still want to handle the case when the result type is different.
typedef typename result_of<
BinaryOp(
typename Lhs::Scalar,
typename Rhs::Scalar
const typename Lhs::Scalar&,
const typename Rhs::Scalar&
)
>::type Scalar;
typedef typename promote_storage_type<typename traits<Lhs>::StorageKind,
typename traits<Rhs>::StorageKind>::ret StorageKind;
typedef typename promote_index_type<typename traits<Lhs>::Index,
typename traits<Rhs>::Index>::type Index;
typedef typename cwise_promote_storage_type<typename traits<Lhs>::StorageKind,
typename traits<Rhs>::StorageKind,
BinaryOp>::ret StorageKind;
typedef typename promote_index_type<typename traits<Lhs>::StorageIndex,
typename traits<Rhs>::StorageIndex>::type StorageIndex;
typedef typename Lhs::Nested LhsNested;
typedef typename Rhs::Nested RhsNested;
typedef typename remove_reference<LhsNested>::type _LhsNested;
typedef typename remove_reference<RhsNested>::type _RhsNested;
enum {
LhsCoeffReadCost = _LhsNested::CoeffReadCost,
RhsCoeffReadCost = _RhsNested::CoeffReadCost,
LhsFlags = _LhsNested::Flags,
RhsFlags = _RhsNested::Flags,
SameType = is_same<typename _LhsNested::Scalar,typename _RhsNested::Scalar>::value,
StorageOrdersAgree = (int(Lhs::Flags)&RowMajorBit)==(int(Rhs::Flags)&RowMajorBit),
Flags0 = (int(LhsFlags) | int(RhsFlags)) & (
HereditaryBits
| (int(LhsFlags) & int(RhsFlags) &
( AlignedBit
| (StorageOrdersAgree ? LinearAccessBit : 0)
| (functor_traits<BinaryOp>::PacketAccess && StorageOrdersAgree && SameType ? PacketAccessBit : 0)
)
)
),
Flags = (Flags0 & ~RowMajorBit) | (LhsFlags & RowMajorBit),
Cost0 = EIGEN_ADD_COST(LhsCoeffReadCost,RhsCoeffReadCost),
CoeffReadCost = EIGEN_ADD_COST(Cost0,functor_traits<BinaryOp>::Cost)
Flags = cwise_promote_storage_order<typename traits<Lhs>::StorageKind,typename traits<Rhs>::StorageKind,_LhsNested::Flags & RowMajorBit,_RhsNested::Flags & RowMajorBit>::value
};
};
} // end namespace internal
// we require Lhs and Rhs to have the same scalar type. Currently there is no example of a binary functor
// that would take two operands of different types. If there were such an example, then this check should be
// moved to the BinaryOp functors, on a per-case basis. This would however require a change in the BinaryOp functors, as
// currently they take only one typename Scalar template parameter.
// It is tempting to always allow mixing different types but remember that this is often impossible in the vectorized paths.
// So allowing mixing different types gives very unexpected errors when enabling vectorization, when the user tries to
// add together a float matrix and a double matrix.
#define EIGEN_CHECK_BINARY_COMPATIBILIY(BINOP,LHS,RHS) \
EIGEN_STATIC_ASSERT((internal::functor_is_product_like<BINOP>::ret \
? int(internal::scalar_product_traits<LHS, RHS>::Defined) \
: int(internal::is_same<LHS, RHS>::value)), \
YOU_MIXED_DIFFERENT_NUMERIC_TYPES__YOU_NEED_TO_USE_THE_CAST_METHOD_OF_MATRIXBASE_TO_CAST_NUMERIC_TYPES_EXPLICITLY)
template<typename BinaryOp, typename Lhs, typename Rhs, typename StorageKind>
class CwiseBinaryOpImpl;
template<typename BinaryOp, typename Lhs, typename Rhs>
class CwiseBinaryOp : internal::no_assignment_operator,
/** \class CwiseBinaryOp
* \ingroup Core_Module
*
* \brief Generic expression where a coefficient-wise binary operator is applied to two expressions
*
* \tparam BinaryOp template functor implementing the operator
* \tparam LhsType the type of the left-hand side
* \tparam RhsType the type of the right-hand side
*
* This class represents an expression where a coefficient-wise binary operator is applied to two expressions.
* It is the return type of binary operators, by which we mean only those binary operators where
* both the left-hand side and the right-hand side are Eigen expressions.
* For example, the return type of matrix1+matrix2 is a CwiseBinaryOp.
*
* Most of the time, this is the only way that it is used, so you typically don't have to name
* CwiseBinaryOp types explicitly.
*
* \sa MatrixBase::binaryExpr(const MatrixBase<OtherDerived> &,const CustomBinaryOp &) const, class CwiseUnaryOp, class CwiseNullaryOp
*/
template<typename BinaryOp, typename LhsType, typename RhsType>
class CwiseBinaryOp :
public CwiseBinaryOpImpl<
BinaryOp, Lhs, Rhs,
typename internal::promote_storage_type<typename internal::traits<Lhs>::StorageKind,
typename internal::traits<Rhs>::StorageKind>::ret>
BinaryOp, LhsType, RhsType,
typename internal::cwise_promote_storage_type<typename internal::traits<LhsType>::StorageKind,
typename internal::traits<RhsType>::StorageKind,
BinaryOp>::ret>,
internal::no_assignment_operator
{
public:
typedef typename internal::remove_all<BinaryOp>::type Functor;
typedef typename internal::remove_all<LhsType>::type Lhs;
typedef typename internal::remove_all<RhsType>::type Rhs;
typedef typename CwiseBinaryOpImpl<
BinaryOp, Lhs, Rhs,
typename internal::promote_storage_type<typename internal::traits<Lhs>::StorageKind,
typename internal::traits<Rhs>::StorageKind>::ret>::Base Base;
BinaryOp, LhsType, RhsType,
typename internal::cwise_promote_storage_type<typename internal::traits<LhsType>::StorageKind,
typename internal::traits<Rhs>::StorageKind,
BinaryOp>::ret>::Base Base;
EIGEN_GENERIC_PUBLIC_INTERFACE(CwiseBinaryOp)
typedef typename internal::nested<Lhs>::type LhsNested;
typedef typename internal::nested<Rhs>::type RhsNested;
typedef typename internal::ref_selector<LhsType>::type LhsNested;
typedef typename internal::ref_selector<RhsType>::type RhsNested;
typedef typename internal::remove_reference<LhsNested>::type _LhsNested;
typedef typename internal::remove_reference<RhsNested>::type _RhsNested;
EIGEN_DEVICE_FUNC
EIGEN_STRONG_INLINE CwiseBinaryOp(const Lhs& aLhs, const Rhs& aRhs, const BinaryOp& func = BinaryOp())
: m_lhs(aLhs), m_rhs(aRhs), m_functor(func)
{
@@ -132,6 +110,7 @@ class CwiseBinaryOp : internal::no_assignment_operator,
eigen_assert(aLhs.rows() == aRhs.rows() && aLhs.cols() == aRhs.cols());
}
EIGEN_DEVICE_FUNC
EIGEN_STRONG_INLINE Index rows() const {
// return the fixed size type if available to enable compile time optimizations
if (internal::traits<typename internal::remove_all<LhsNested>::type>::RowsAtCompileTime==Dynamic)
@@ -139,6 +118,7 @@ class CwiseBinaryOp : internal::no_assignment_operator,
else
return m_lhs.rows();
}
EIGEN_DEVICE_FUNC
EIGEN_STRONG_INLINE Index cols() const {
// return the fixed size type if available to enable compile time optimizations
if (internal::traits<typename internal::remove_all<LhsNested>::type>::ColsAtCompileTime==Dynamic)
@@ -148,10 +128,13 @@ class CwiseBinaryOp : internal::no_assignment_operator,
}
/** \returns the left hand side nested expression */
EIGEN_DEVICE_FUNC
const _LhsNested& lhs() const { return m_lhs; }
/** \returns the right hand side nested expression */
EIGEN_DEVICE_FUNC
const _RhsNested& rhs() const { return m_rhs; }
/** \returns the functor representing the binary operation */
EIGEN_DEVICE_FUNC
const BinaryOp& functor() const { return m_functor; }
protected:
@@ -160,41 +143,13 @@ class CwiseBinaryOp : internal::no_assignment_operator,
const BinaryOp m_functor;
};
template<typename BinaryOp, typename Lhs, typename Rhs>
class CwiseBinaryOpImpl<BinaryOp, Lhs, Rhs, Dense>
: public internal::dense_xpr_base<CwiseBinaryOp<BinaryOp, Lhs, Rhs> >::type
// Generic API dispatcher
template<typename BinaryOp, typename Lhs, typename Rhs, typename StorageKind>
class CwiseBinaryOpImpl
: public internal::generic_xpr_base<CwiseBinaryOp<BinaryOp, Lhs, Rhs> >::type
{
typedef CwiseBinaryOp<BinaryOp, Lhs, Rhs> Derived;
public:
typedef typename internal::dense_xpr_base<CwiseBinaryOp<BinaryOp, Lhs, Rhs> >::type Base;
EIGEN_DENSE_PUBLIC_INTERFACE( Derived )
EIGEN_STRONG_INLINE const Scalar coeff(Index rowId, Index colId) const
{
return derived().functor()(derived().lhs().coeff(rowId, colId),
derived().rhs().coeff(rowId, colId));
}
template<int LoadMode>
EIGEN_STRONG_INLINE PacketScalar packet(Index rowId, Index colId) const
{
return derived().functor().packetOp(derived().lhs().template packet<LoadMode>(rowId, colId),
derived().rhs().template packet<LoadMode>(rowId, colId));
}
EIGEN_STRONG_INLINE const Scalar coeff(Index index) const
{
return derived().functor()(derived().lhs().coeff(index),
derived().rhs().coeff(index));
}
template<int LoadMode>
EIGEN_STRONG_INLINE PacketScalar packet(Index index) const
{
return derived().functor().packetOp(derived().lhs().template packet<LoadMode>(index),
derived().rhs().template packet<LoadMode>(index));
}
public:
typedef typename internal::generic_xpr_base<CwiseBinaryOp<BinaryOp, Lhs, Rhs> >::type Base;
};
/** replaces \c *this by \c *this - \a other.
@@ -206,8 +161,7 @@ template<typename OtherDerived>
EIGEN_STRONG_INLINE Derived &
MatrixBase<Derived>::operator-=(const MatrixBase<OtherDerived> &other)
{
SelfCwiseBinaryOp<internal::scalar_difference_op<Scalar>, Derived, OtherDerived> tmp(derived());
tmp = other.derived();
call_assignment(derived(), other.derived(), internal::sub_assign_op<Scalar,typename OtherDerived::Scalar>());
return derived();
}
@@ -220,11 +174,11 @@ template<typename OtherDerived>
EIGEN_STRONG_INLINE Derived &
MatrixBase<Derived>::operator+=(const MatrixBase<OtherDerived>& other)
{
SelfCwiseBinaryOp<internal::scalar_sum_op<Scalar>, Derived, OtherDerived> tmp(derived());
tmp = other.derived();
call_assignment(derived(), other.derived(), internal::add_assign_op<Scalar,typename OtherDerived::Scalar>());
return derived();
}
} // end namespace Eigen
#endif // EIGEN_CWISE_BINARY_OP_H

View File

@@ -12,13 +12,24 @@
namespace Eigen {
namespace internal {
template<typename NullaryOp, typename PlainObjectType>
struct traits<CwiseNullaryOp<NullaryOp, PlainObjectType> > : traits<PlainObjectType>
{
enum {
Flags = traits<PlainObjectType>::Flags & RowMajorBit
};
};
} // namespace internal
/** \class CwiseNullaryOp
* \ingroup Core_Module
*
* \brief Generic expression of a matrix where all coefficients are defined by a functor
*
* \param NullaryOp template functor implementing the operator
* \param PlainObjectType the underlying plain matrix/array type
* \tparam NullaryOp template functor implementing the operator
* \tparam PlainObjectType the underlying plain matrix/array type
*
* This class represents an expression of a generic nullary operator.
* It is the return type of the Ones(), Zero(), Constant(), Identity() and Random() methods,
@@ -27,68 +38,49 @@ namespace Eigen {
* However, if you want to write a function returning such an expression, you
* will need to use this class.
*
* \sa class CwiseUnaryOp, class CwiseBinaryOp, DenseBase::NullaryExpr()
* The functor NullaryOp must expose one of the following method:
<table class="manual">
<tr ><td>\c operator()() </td><td>if the procedural generation does not depend on the coefficient entries (e.g., random numbers)</td></tr>
<tr class="alt"><td>\c operator()(Index i)</td><td>if the procedural generation makes sense for vectors only and that it depends on the coefficient index \c i (e.g., linspace) </td></tr>
<tr ><td>\c operator()(Index i,Index j)</td><td>if the procedural generation depends on the matrix coordinates \c i, \c j (e.g., to generate a checkerboard with 0 and 1)</td></tr>
</table>
* It is also possible to expose the last two operators if the generation makes sense for matrices but can be optimized for vectors.
*
* See DenseBase::NullaryExpr(Index,const CustomNullaryOp&) for an example binding
* C++11 random number generators.
*
* A nullary expression can also be used to implement custom sophisticated matrix manipulations
* that cannot be covered by the existing set of natively supported matrix manipulations.
* See this \ref TopicCustomizing_NullaryExpr "page" for some examples and additional explanations
* on the behavior of CwiseNullaryOp.
*
* \sa class CwiseUnaryOp, class CwiseBinaryOp, DenseBase::NullaryExpr
*/
namespace internal {
template<typename NullaryOp, typename PlainObjectType>
struct traits<CwiseNullaryOp<NullaryOp, PlainObjectType> > : traits<PlainObjectType>
{
enum {
Flags = (traits<PlainObjectType>::Flags
& ( HereditaryBits
| (functor_has_linear_access<NullaryOp>::ret ? LinearAccessBit : 0)
| (functor_traits<NullaryOp>::PacketAccess ? PacketAccessBit : 0)))
| (functor_traits<NullaryOp>::IsRepeatable ? 0 : EvalBeforeNestingBit),
CoeffReadCost = functor_traits<NullaryOp>::Cost
};
};
}
template<typename NullaryOp, typename PlainObjectType>
class CwiseNullaryOp : internal::no_assignment_operator,
public internal::dense_xpr_base< CwiseNullaryOp<NullaryOp, PlainObjectType> >::type
class CwiseNullaryOp : public internal::dense_xpr_base< CwiseNullaryOp<NullaryOp, PlainObjectType> >::type, internal::no_assignment_operator
{
public:
typedef typename internal::dense_xpr_base<CwiseNullaryOp>::type Base;
EIGEN_DENSE_PUBLIC_INTERFACE(CwiseNullaryOp)
CwiseNullaryOp(Index nbRows, Index nbCols, const NullaryOp& func = NullaryOp())
: m_rows(nbRows), m_cols(nbCols), m_functor(func)
EIGEN_DEVICE_FUNC
CwiseNullaryOp(Index rows, Index cols, const NullaryOp& func = NullaryOp())
: m_rows(rows), m_cols(cols), m_functor(func)
{
eigen_assert(nbRows >= 0
&& (RowsAtCompileTime == Dynamic || RowsAtCompileTime == nbRows)
&& nbCols >= 0
&& (ColsAtCompileTime == Dynamic || ColsAtCompileTime == nbCols));
eigen_assert(rows >= 0
&& (RowsAtCompileTime == Dynamic || RowsAtCompileTime == rows)
&& cols >= 0
&& (ColsAtCompileTime == Dynamic || ColsAtCompileTime == cols));
}
EIGEN_DEVICE_FUNC
EIGEN_STRONG_INLINE Index rows() const { return m_rows.value(); }
EIGEN_DEVICE_FUNC
EIGEN_STRONG_INLINE Index cols() const { return m_cols.value(); }
EIGEN_STRONG_INLINE const Scalar coeff(Index rowId, Index colId) const
{
return m_functor(rowId, colId);
}
template<int LoadMode>
EIGEN_STRONG_INLINE PacketScalar packet(Index rowId, Index colId) const
{
return m_functor.packetOp(rowId, colId);
}
EIGEN_STRONG_INLINE const Scalar coeff(Index index) const
{
return m_functor(index);
}
template<int LoadMode>
EIGEN_STRONG_INLINE PacketScalar packet(Index index) const
{
return m_functor.packetOp(index);
}
/** \returns the functor representing the nullary operation */
EIGEN_DEVICE_FUNC
const NullaryOp& functor() const { return m_functor; }
protected:
@@ -113,10 +105,10 @@ class CwiseNullaryOp : internal::no_assignment_operator,
*/
template<typename Derived>
template<typename CustomNullaryOp>
EIGEN_STRONG_INLINE const CwiseNullaryOp<CustomNullaryOp, Derived>
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const CwiseNullaryOp<CustomNullaryOp, typename DenseBase<Derived>::PlainObject>
DenseBase<Derived>::NullaryExpr(Index rows, Index cols, const CustomNullaryOp& func)
{
return CwiseNullaryOp<CustomNullaryOp, Derived>(rows, cols, func);
return CwiseNullaryOp<CustomNullaryOp, PlainObject>(rows, cols, func);
}
/** \returns an expression of a matrix defined by a custom functor \a func
@@ -132,16 +124,19 @@ DenseBase<Derived>::NullaryExpr(Index rows, Index cols, const CustomNullaryOp& f
*
* The template parameter \a CustomNullaryOp is the type of the functor.
*
* Here is an example with C++11 random generators: \include random_cpp11.cpp
* Output: \verbinclude random_cpp11.out
*
* \sa class CwiseNullaryOp
*/
template<typename Derived>
template<typename CustomNullaryOp>
EIGEN_STRONG_INLINE const CwiseNullaryOp<CustomNullaryOp, Derived>
EIGEN_STRONG_INLINE const CwiseNullaryOp<CustomNullaryOp, typename DenseBase<Derived>::PlainObject>
DenseBase<Derived>::NullaryExpr(Index size, const CustomNullaryOp& func)
{
EIGEN_STATIC_ASSERT_VECTOR_ONLY(Derived)
if(RowsAtCompileTime == 1) return CwiseNullaryOp<CustomNullaryOp, Derived>(1, size, func);
else return CwiseNullaryOp<CustomNullaryOp, Derived>(size, 1, func);
if(RowsAtCompileTime == 1) return CwiseNullaryOp<CustomNullaryOp, PlainObject>(1, size, func);
else return CwiseNullaryOp<CustomNullaryOp, PlainObject>(size, 1, func);
}
/** \returns an expression of a matrix defined by a custom functor \a func
@@ -155,19 +150,19 @@ DenseBase<Derived>::NullaryExpr(Index size, const CustomNullaryOp& func)
*/
template<typename Derived>
template<typename CustomNullaryOp>
EIGEN_STRONG_INLINE const CwiseNullaryOp<CustomNullaryOp, Derived>
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const CwiseNullaryOp<CustomNullaryOp, typename DenseBase<Derived>::PlainObject>
DenseBase<Derived>::NullaryExpr(const CustomNullaryOp& func)
{
return CwiseNullaryOp<CustomNullaryOp, Derived>(RowsAtCompileTime, ColsAtCompileTime, func);
return CwiseNullaryOp<CustomNullaryOp, PlainObject>(RowsAtCompileTime, ColsAtCompileTime, func);
}
/** \returns an expression of a constant matrix of value \a value
*
* The parameters \a nbRows and \a nbCols are the number of rows and of columns of
* The parameters \a rows and \a cols are the number of rows and of columns of
* the returned matrix. Must be compatible with this DenseBase type.
*
* This variant is meant to be used for dynamic-size matrix types. For fixed-size types,
* it is redundant to pass \a nbRows and \a nbCols as arguments, so Zero() should be used
* it is redundant to pass \a rows and \a cols as arguments, so Zero() should be used
* instead.
*
* The template parameter \a CustomNullaryOp is the type of the functor.
@@ -176,9 +171,9 @@ DenseBase<Derived>::NullaryExpr(const CustomNullaryOp& func)
*/
template<typename Derived>
EIGEN_STRONG_INLINE const typename DenseBase<Derived>::ConstantReturnType
DenseBase<Derived>::Constant(Index nbRows, Index nbCols, const Scalar& value)
DenseBase<Derived>::Constant(Index rows, Index cols, const Scalar& value)
{
return DenseBase<Derived>::NullaryExpr(nbRows, nbCols, internal::scalar_constant_op<Scalar>(value));
return DenseBase<Derived>::NullaryExpr(rows, cols, internal::scalar_constant_op<Scalar>(value));
}
/** \returns an expression of a constant matrix of value \a value
@@ -197,7 +192,7 @@ DenseBase<Derived>::Constant(Index nbRows, Index nbCols, const Scalar& value)
* \sa class CwiseNullaryOp
*/
template<typename Derived>
EIGEN_STRONG_INLINE const typename DenseBase<Derived>::ConstantReturnType
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const typename DenseBase<Derived>::ConstantReturnType
DenseBase<Derived>::Constant(Index size, const Scalar& value)
{
return DenseBase<Derived>::NullaryExpr(size, internal::scalar_constant_op<Scalar>(value));
@@ -213,53 +208,40 @@ DenseBase<Derived>::Constant(Index size, const Scalar& value)
* \sa class CwiseNullaryOp
*/
template<typename Derived>
EIGEN_STRONG_INLINE const typename DenseBase<Derived>::ConstantReturnType
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const typename DenseBase<Derived>::ConstantReturnType
DenseBase<Derived>::Constant(const Scalar& value)
{
EIGEN_STATIC_ASSERT_FIXED_SIZE(Derived)
return DenseBase<Derived>::NullaryExpr(RowsAtCompileTime, ColsAtCompileTime, internal::scalar_constant_op<Scalar>(value));
}
/**
* \brief Sets a linearly space vector.
/** \deprecated because of accuracy loss. In Eigen 3.3, it is an alias for LinSpaced(Index,const Scalar&,const Scalar&)
*
* The function generates 'size' equally spaced values in the closed interval [low,high].
* This particular version of LinSpaced() uses sequential access, i.e. vector access is
* assumed to be a(0), a(1), ..., a(size). This assumption allows for better vectorization
* and yields faster code than the random access version.
*
* When size is set to 1, a vector of length 1 containing 'high' is returned.
*
* \only_for_vectors
*
* Example: \include DenseBase_LinSpaced_seq.cpp
* Output: \verbinclude DenseBase_LinSpaced_seq.out
*
* \sa setLinSpaced(Index,const Scalar&,const Scalar&), LinSpaced(Index,Scalar,Scalar), CwiseNullaryOp
* \sa LinSpaced(Index,Scalar,Scalar), setLinSpaced(Index,const Scalar&,const Scalar&)
*/
template<typename Derived>
EIGEN_STRONG_INLINE const typename DenseBase<Derived>::SequentialLinSpacedReturnType
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const typename DenseBase<Derived>::RandomAccessLinSpacedReturnType
DenseBase<Derived>::LinSpaced(Sequential_t, Index size, const Scalar& low, const Scalar& high)
{
EIGEN_STATIC_ASSERT_VECTOR_ONLY(Derived)
return DenseBase<Derived>::NullaryExpr(size, internal::linspaced_op<Scalar,false>(low,high,size));
return DenseBase<Derived>::NullaryExpr(size, internal::linspaced_op<Scalar,PacketScalar>(low,high,size));
}
/**
* \copydoc DenseBase::LinSpaced(Sequential_t, Index, const Scalar&, const Scalar&)
* Special version for fixed size types which does not require the size parameter.
/** \deprecated because of accuracy loss. In Eigen 3.3, it is an alias for LinSpaced(const Scalar&,const Scalar&)
*
* \sa LinSpaced(Scalar,Scalar)
*/
template<typename Derived>
EIGEN_STRONG_INLINE const typename DenseBase<Derived>::SequentialLinSpacedReturnType
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const typename DenseBase<Derived>::RandomAccessLinSpacedReturnType
DenseBase<Derived>::LinSpaced(Sequential_t, const Scalar& low, const Scalar& high)
{
EIGEN_STATIC_ASSERT_VECTOR_ONLY(Derived)
EIGEN_STATIC_ASSERT_FIXED_SIZE(Derived)
return DenseBase<Derived>::NullaryExpr(Derived::SizeAtCompileTime, internal::linspaced_op<Scalar,false>(low,high,Derived::SizeAtCompileTime));
return DenseBase<Derived>::NullaryExpr(Derived::SizeAtCompileTime, internal::linspaced_op<Scalar,PacketScalar>(low,high,Derived::SizeAtCompileTime));
}
/**
* \brief Sets a linearly space vector.
* \brief Sets a linearly spaced vector.
*
* The function generates 'size' equally spaced values in the closed interval [low,high].
* When size is set to 1, a vector of length 1 containing 'high' is returned.
@@ -269,14 +251,24 @@ DenseBase<Derived>::LinSpaced(Sequential_t, const Scalar& low, const Scalar& hig
* Example: \include DenseBase_LinSpaced.cpp
* Output: \verbinclude DenseBase_LinSpaced.out
*
* \sa setLinSpaced(Index,const Scalar&,const Scalar&), LinSpaced(Sequential_t,Index,const Scalar&,const Scalar&,Index), CwiseNullaryOp
* For integer scalar types, an even spacing is possible if and only if the length of the range,
* i.e., \c high-low is a scalar multiple of \c size-1, or if \c size is a scalar multiple of the
* number of values \c high-low+1 (meaning each value can be repeated the same number of time).
* If one of these two considions is not satisfied, then \c high is lowered to the largest value
* satisfying one of this constraint.
* Here are some examples:
*
* Example: \include DenseBase_LinSpacedInt.cpp
* Output: \verbinclude DenseBase_LinSpacedInt.out
*
* \sa setLinSpaced(Index,const Scalar&,const Scalar&), CwiseNullaryOp
*/
template<typename Derived>
EIGEN_STRONG_INLINE const typename DenseBase<Derived>::RandomAccessLinSpacedReturnType
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const typename DenseBase<Derived>::RandomAccessLinSpacedReturnType
DenseBase<Derived>::LinSpaced(Index size, const Scalar& low, const Scalar& high)
{
EIGEN_STATIC_ASSERT_VECTOR_ONLY(Derived)
return DenseBase<Derived>::NullaryExpr(size, internal::linspaced_op<Scalar,true>(low,high,size));
return DenseBase<Derived>::NullaryExpr(size, internal::linspaced_op<Scalar,PacketScalar>(low,high,size));
}
/**
@@ -284,22 +276,23 @@ DenseBase<Derived>::LinSpaced(Index size, const Scalar& low, const Scalar& high)
* Special version for fixed size types which does not require the size parameter.
*/
template<typename Derived>
EIGEN_STRONG_INLINE const typename DenseBase<Derived>::RandomAccessLinSpacedReturnType
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const typename DenseBase<Derived>::RandomAccessLinSpacedReturnType
DenseBase<Derived>::LinSpaced(const Scalar& low, const Scalar& high)
{
EIGEN_STATIC_ASSERT_VECTOR_ONLY(Derived)
EIGEN_STATIC_ASSERT_FIXED_SIZE(Derived)
return DenseBase<Derived>::NullaryExpr(Derived::SizeAtCompileTime, internal::linspaced_op<Scalar,true>(low,high,Derived::SizeAtCompileTime));
return DenseBase<Derived>::NullaryExpr(Derived::SizeAtCompileTime, internal::linspaced_op<Scalar,PacketScalar>(low,high,Derived::SizeAtCompileTime));
}
/** \returns true if all coefficients in this matrix are approximately equal to \a val, to within precision \a prec */
template<typename Derived>
bool DenseBase<Derived>::isApproxToConstant
EIGEN_DEVICE_FUNC bool DenseBase<Derived>::isApproxToConstant
(const Scalar& val, const RealScalar& prec) const
{
typename internal::nested_eval<Derived,1>::type self(derived());
for(Index j = 0; j < cols(); ++j)
for(Index i = 0; i < rows(); ++i)
if(!internal::isApprox(this->coeff(i, j), val, prec))
if(!internal::isApprox(self.coeff(i, j), val, prec))
return false;
return true;
}
@@ -308,7 +301,7 @@ bool DenseBase<Derived>::isApproxToConstant
*
* \returns true if all coefficients in this matrix are approximately equal to \a value, to within precision \a prec */
template<typename Derived>
bool DenseBase<Derived>::isConstant
EIGEN_DEVICE_FUNC bool DenseBase<Derived>::isConstant
(const Scalar& val, const RealScalar& prec) const
{
return isApproxToConstant(val, prec);
@@ -319,22 +312,22 @@ bool DenseBase<Derived>::isConstant
* \sa setConstant(), Constant(), class CwiseNullaryOp
*/
template<typename Derived>
EIGEN_STRONG_INLINE void DenseBase<Derived>::fill(const Scalar& val)
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE void DenseBase<Derived>::fill(const Scalar& val)
{
setConstant(val);
}
/** Sets all coefficients in this expression to \a value.
/** Sets all coefficients in this expression to value \a val.
*
* \sa fill(), setConstant(Index,const Scalar&), setConstant(Index,Index,const Scalar&), setZero(), setOnes(), Constant(), class CwiseNullaryOp, setZero(), setOnes()
*/
template<typename Derived>
EIGEN_STRONG_INLINE Derived& DenseBase<Derived>::setConstant(const Scalar& val)
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE Derived& DenseBase<Derived>::setConstant(const Scalar& val)
{
return derived() = Constant(rows(), cols(), val);
}
/** Resizes to the given \a size, and sets all coefficients in this expression to the given \a value.
/** Resizes to the given \a size, and sets all coefficients in this expression to the given value \a val.
*
* \only_for_vectors
*
@@ -344,17 +337,17 @@ EIGEN_STRONG_INLINE Derived& DenseBase<Derived>::setConstant(const Scalar& val)
* \sa MatrixBase::setConstant(const Scalar&), setConstant(Index,Index,const Scalar&), class CwiseNullaryOp, MatrixBase::Constant(const Scalar&)
*/
template<typename Derived>
EIGEN_STRONG_INLINE Derived&
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE Derived&
PlainObjectBase<Derived>::setConstant(Index size, const Scalar& val)
{
resize(size);
return setConstant(val);
}
/** Resizes to the given size, and sets all coefficients in this expression to the given \a value.
/** Resizes to the given size, and sets all coefficients in this expression to the given value \a val.
*
* \param nbRows the new number of rows
* \param nbCols the new number of columns
* \param rows the new number of rows
* \param cols the new number of columns
* \param val the value to which all coefficients are set
*
* Example: \include Matrix_setConstant_int_int.cpp
@@ -363,15 +356,15 @@ PlainObjectBase<Derived>::setConstant(Index size, const Scalar& val)
* \sa MatrixBase::setConstant(const Scalar&), setConstant(Index,const Scalar&), class CwiseNullaryOp, MatrixBase::Constant(const Scalar&)
*/
template<typename Derived>
EIGEN_STRONG_INLINE Derived&
PlainObjectBase<Derived>::setConstant(Index nbRows, Index nbCols, const Scalar& val)
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE Derived&
PlainObjectBase<Derived>::setConstant(Index rows, Index cols, const Scalar& val)
{
resize(nbRows, nbCols);
resize(rows, cols);
return setConstant(val);
}
/**
* \brief Sets a linearly space vector.
* \brief Sets a linearly spaced vector.
*
* The function generates 'size' equally spaced values in the closed interval [low,high].
* When size is set to 1, a vector of length 1 containing 'high' is returned.
@@ -381,27 +374,33 @@ PlainObjectBase<Derived>::setConstant(Index nbRows, Index nbCols, const Scalar&
* Example: \include DenseBase_setLinSpaced.cpp
* Output: \verbinclude DenseBase_setLinSpaced.out
*
* \sa CwiseNullaryOp
* For integer scalar types, do not miss the explanations on the definition
* of \link LinSpaced(Index,const Scalar&,const Scalar&) even spacing \endlink.
*
* \sa LinSpaced(Index,const Scalar&,const Scalar&), CwiseNullaryOp
*/
template<typename Derived>
EIGEN_STRONG_INLINE Derived& DenseBase<Derived>::setLinSpaced(Index newSize, const Scalar& low, const Scalar& high)
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE Derived& DenseBase<Derived>::setLinSpaced(Index newSize, const Scalar& low, const Scalar& high)
{
EIGEN_STATIC_ASSERT_VECTOR_ONLY(Derived)
return derived() = Derived::NullaryExpr(newSize, internal::linspaced_op<Scalar,false>(low,high,newSize));
return derived() = Derived::NullaryExpr(newSize, internal::linspaced_op<Scalar,PacketScalar>(low,high,newSize));
}
/**
* \brief Sets a linearly space vector.
* \brief Sets a linearly spaced vector.
*
* The function fill *this with equally spaced values in the closed interval [low,high].
* The function fills \c *this with equally spaced values in the closed interval [low,high].
* When size is set to 1, a vector of length 1 containing 'high' is returned.
*
* \only_for_vectors
*
* \sa setLinSpaced(Index, const Scalar&, const Scalar&), CwiseNullaryOp
* For integer scalar types, do not miss the explanations on the definition
* of \link LinSpaced(Index,const Scalar&,const Scalar&) even spacing \endlink.
*
* \sa LinSpaced(Index,const Scalar&,const Scalar&), setLinSpaced(Index, const Scalar&, const Scalar&), CwiseNullaryOp
*/
template<typename Derived>
EIGEN_STRONG_INLINE Derived& DenseBase<Derived>::setLinSpaced(const Scalar& low, const Scalar& high)
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE Derived& DenseBase<Derived>::setLinSpaced(const Scalar& low, const Scalar& high)
{
EIGEN_STATIC_ASSERT_VECTOR_ONLY(Derived)
return setLinSpaced(size(), low, high);
@@ -424,10 +423,10 @@ EIGEN_STRONG_INLINE Derived& DenseBase<Derived>::setLinSpaced(const Scalar& low,
* \sa Zero(), Zero(Index)
*/
template<typename Derived>
EIGEN_STRONG_INLINE const typename DenseBase<Derived>::ConstantReturnType
DenseBase<Derived>::Zero(Index nbRows, Index nbCols)
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const typename DenseBase<Derived>::ConstantReturnType
DenseBase<Derived>::Zero(Index rows, Index cols)
{
return Constant(nbRows, nbCols, Scalar(0));
return Constant(rows, cols, Scalar(0));
}
/** \returns an expression of a zero vector.
@@ -447,7 +446,7 @@ DenseBase<Derived>::Zero(Index nbRows, Index nbCols)
* \sa Zero(), Zero(Index,Index)
*/
template<typename Derived>
EIGEN_STRONG_INLINE const typename DenseBase<Derived>::ConstantReturnType
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const typename DenseBase<Derived>::ConstantReturnType
DenseBase<Derived>::Zero(Index size)
{
return Constant(size, Scalar(0));
@@ -464,7 +463,7 @@ DenseBase<Derived>::Zero(Index size)
* \sa Zero(Index), Zero(Index,Index)
*/
template<typename Derived>
EIGEN_STRONG_INLINE const typename DenseBase<Derived>::ConstantReturnType
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const typename DenseBase<Derived>::ConstantReturnType
DenseBase<Derived>::Zero()
{
return Constant(Scalar(0));
@@ -479,11 +478,12 @@ DenseBase<Derived>::Zero()
* \sa class CwiseNullaryOp, Zero()
*/
template<typename Derived>
bool DenseBase<Derived>::isZero(const RealScalar& prec) const
EIGEN_DEVICE_FUNC bool DenseBase<Derived>::isZero(const RealScalar& prec) const
{
typename internal::nested_eval<Derived,1>::type self(derived());
for(Index j = 0; j < cols(); ++j)
for(Index i = 0; i < rows(); ++i)
if(!internal::isMuchSmallerThan(this->coeff(i, j), static_cast<Scalar>(1), prec))
if(!internal::isMuchSmallerThan(self.coeff(i, j), static_cast<Scalar>(1), prec))
return false;
return true;
}
@@ -496,7 +496,7 @@ bool DenseBase<Derived>::isZero(const RealScalar& prec) const
* \sa class CwiseNullaryOp, Zero()
*/
template<typename Derived>
EIGEN_STRONG_INLINE Derived& DenseBase<Derived>::setZero()
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE Derived& DenseBase<Derived>::setZero()
{
return setConstant(Scalar(0));
}
@@ -511,7 +511,7 @@ EIGEN_STRONG_INLINE Derived& DenseBase<Derived>::setZero()
* \sa DenseBase::setZero(), setZero(Index,Index), class CwiseNullaryOp, DenseBase::Zero()
*/
template<typename Derived>
EIGEN_STRONG_INLINE Derived&
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE Derived&
PlainObjectBase<Derived>::setZero(Index newSize)
{
resize(newSize);
@@ -520,8 +520,8 @@ PlainObjectBase<Derived>::setZero(Index newSize)
/** Resizes to the given size, and sets all coefficients in this expression to zero.
*
* \param nbRows the new number of rows
* \param nbCols the new number of columns
* \param rows the new number of rows
* \param cols the new number of columns
*
* Example: \include Matrix_setZero_int_int.cpp
* Output: \verbinclude Matrix_setZero_int_int.out
@@ -529,10 +529,10 @@ PlainObjectBase<Derived>::setZero(Index newSize)
* \sa DenseBase::setZero(), setZero(Index), class CwiseNullaryOp, DenseBase::Zero()
*/
template<typename Derived>
EIGEN_STRONG_INLINE Derived&
PlainObjectBase<Derived>::setZero(Index nbRows, Index nbCols)
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE Derived&
PlainObjectBase<Derived>::setZero(Index rows, Index cols)
{
resize(nbRows, nbCols);
resize(rows, cols);
return setConstant(Scalar(0));
}
@@ -540,7 +540,7 @@ PlainObjectBase<Derived>::setZero(Index nbRows, Index nbCols)
/** \returns an expression of a matrix where all coefficients equal one.
*
* The parameters \a nbRows and \a nbCols are the number of rows and of columns of
* The parameters \a rows and \a cols are the number of rows and of columns of
* the returned matrix. Must be compatible with this MatrixBase type.
*
* This variant is meant to be used for dynamic-size matrix types. For fixed-size types,
@@ -553,10 +553,10 @@ PlainObjectBase<Derived>::setZero(Index nbRows, Index nbCols)
* \sa Ones(), Ones(Index), isOnes(), class Ones
*/
template<typename Derived>
EIGEN_STRONG_INLINE const typename DenseBase<Derived>::ConstantReturnType
DenseBase<Derived>::Ones(Index nbRows, Index nbCols)
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const typename DenseBase<Derived>::ConstantReturnType
DenseBase<Derived>::Ones(Index rows, Index cols)
{
return Constant(nbRows, nbCols, Scalar(1));
return Constant(rows, cols, Scalar(1));
}
/** \returns an expression of a vector where all coefficients equal one.
@@ -576,7 +576,7 @@ DenseBase<Derived>::Ones(Index nbRows, Index nbCols)
* \sa Ones(), Ones(Index,Index), isOnes(), class Ones
*/
template<typename Derived>
EIGEN_STRONG_INLINE const typename DenseBase<Derived>::ConstantReturnType
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const typename DenseBase<Derived>::ConstantReturnType
DenseBase<Derived>::Ones(Index newSize)
{
return Constant(newSize, Scalar(1));
@@ -593,7 +593,7 @@ DenseBase<Derived>::Ones(Index newSize)
* \sa Ones(Index), Ones(Index,Index), isOnes(), class Ones
*/
template<typename Derived>
EIGEN_STRONG_INLINE const typename DenseBase<Derived>::ConstantReturnType
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const typename DenseBase<Derived>::ConstantReturnType
DenseBase<Derived>::Ones()
{
return Constant(Scalar(1));
@@ -608,7 +608,7 @@ DenseBase<Derived>::Ones()
* \sa class CwiseNullaryOp, Ones()
*/
template<typename Derived>
bool DenseBase<Derived>::isOnes
EIGEN_DEVICE_FUNC bool DenseBase<Derived>::isOnes
(const RealScalar& prec) const
{
return isApproxToConstant(Scalar(1), prec);
@@ -622,7 +622,7 @@ bool DenseBase<Derived>::isOnes
* \sa class CwiseNullaryOp, Ones()
*/
template<typename Derived>
EIGEN_STRONG_INLINE Derived& DenseBase<Derived>::setOnes()
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE Derived& DenseBase<Derived>::setOnes()
{
return setConstant(Scalar(1));
}
@@ -637,7 +637,7 @@ EIGEN_STRONG_INLINE Derived& DenseBase<Derived>::setOnes()
* \sa MatrixBase::setOnes(), setOnes(Index,Index), class CwiseNullaryOp, MatrixBase::Ones()
*/
template<typename Derived>
EIGEN_STRONG_INLINE Derived&
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE Derived&
PlainObjectBase<Derived>::setOnes(Index newSize)
{
resize(newSize);
@@ -646,8 +646,8 @@ PlainObjectBase<Derived>::setOnes(Index newSize)
/** Resizes to the given size, and sets all coefficients in this expression to one.
*
* \param nbRows the new number of rows
* \param nbCols the new number of columns
* \param rows the new number of rows
* \param cols the new number of columns
*
* Example: \include Matrix_setOnes_int_int.cpp
* Output: \verbinclude Matrix_setOnes_int_int.out
@@ -655,10 +655,10 @@ PlainObjectBase<Derived>::setOnes(Index newSize)
* \sa MatrixBase::setOnes(), setOnes(Index), class CwiseNullaryOp, MatrixBase::Ones()
*/
template<typename Derived>
EIGEN_STRONG_INLINE Derived&
PlainObjectBase<Derived>::setOnes(Index nbRows, Index nbCols)
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE Derived&
PlainObjectBase<Derived>::setOnes(Index rows, Index cols)
{
resize(nbRows, nbCols);
resize(rows, cols);
return setConstant(Scalar(1));
}
@@ -666,7 +666,7 @@ PlainObjectBase<Derived>::setOnes(Index nbRows, Index nbCols)
/** \returns an expression of the identity matrix (not necessarily square).
*
* The parameters \a nbRows and \a nbCols are the number of rows and of columns of
* The parameters \a rows and \a cols are the number of rows and of columns of
* the returned matrix. Must be compatible with this MatrixBase type.
*
* This variant is meant to be used for dynamic-size matrix types. For fixed-size types,
@@ -679,10 +679,10 @@ PlainObjectBase<Derived>::setOnes(Index nbRows, Index nbCols)
* \sa Identity(), setIdentity(), isIdentity()
*/
template<typename Derived>
EIGEN_STRONG_INLINE const typename MatrixBase<Derived>::IdentityReturnType
MatrixBase<Derived>::Identity(Index nbRows, Index nbCols)
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const typename MatrixBase<Derived>::IdentityReturnType
MatrixBase<Derived>::Identity(Index rows, Index cols)
{
return DenseBase<Derived>::NullaryExpr(nbRows, nbCols, internal::scalar_identity_op<Scalar>());
return DenseBase<Derived>::NullaryExpr(rows, cols, internal::scalar_identity_op<Scalar>());
}
/** \returns an expression of the identity matrix (not necessarily square).
@@ -696,7 +696,7 @@ MatrixBase<Derived>::Identity(Index nbRows, Index nbCols)
* \sa Identity(Index,Index), setIdentity(), isIdentity()
*/
template<typename Derived>
EIGEN_STRONG_INLINE const typename MatrixBase<Derived>::IdentityReturnType
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const typename MatrixBase<Derived>::IdentityReturnType
MatrixBase<Derived>::Identity()
{
EIGEN_STATIC_ASSERT_FIXED_SIZE(Derived)
@@ -716,18 +716,19 @@ template<typename Derived>
bool MatrixBase<Derived>::isIdentity
(const RealScalar& prec) const
{
typename internal::nested_eval<Derived,1>::type self(derived());
for(Index j = 0; j < cols(); ++j)
{
for(Index i = 0; i < rows(); ++i)
{
if(i == j)
{
if(!internal::isApprox(this->coeff(i, j), static_cast<Scalar>(1), prec))
if(!internal::isApprox(self.coeff(i, j), static_cast<Scalar>(1), prec))
return false;
}
else
{
if(!internal::isMuchSmallerThan(this->coeff(i, j), static_cast<RealScalar>(1), prec))
if(!internal::isMuchSmallerThan(self.coeff(i, j), static_cast<RealScalar>(1), prec))
return false;
}
}
@@ -740,6 +741,7 @@ namespace internal {
template<typename Derived, bool Big = (Derived::SizeAtCompileTime>=16)>
struct setIdentity_impl
{
EIGEN_DEVICE_FUNC
static EIGEN_STRONG_INLINE Derived& run(Derived& m)
{
return m = Derived::Identity(m.rows(), m.cols());
@@ -749,11 +751,11 @@ struct setIdentity_impl
template<typename Derived>
struct setIdentity_impl<Derived, true>
{
typedef typename Derived::Index Index;
EIGEN_DEVICE_FUNC
static EIGEN_STRONG_INLINE Derived& run(Derived& m)
{
m.setZero();
const Index size = (std::min)(m.rows(), m.cols());
const Index size = numext::mini(m.rows(), m.cols());
for(Index i = 0; i < size; ++i) m.coeffRef(i,i) = typename Derived::Scalar(1);
return m;
}
@@ -769,15 +771,15 @@ struct setIdentity_impl<Derived, true>
* \sa class CwiseNullaryOp, Identity(), Identity(Index,Index), isIdentity()
*/
template<typename Derived>
EIGEN_STRONG_INLINE Derived& MatrixBase<Derived>::setIdentity()
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE Derived& MatrixBase<Derived>::setIdentity()
{
return internal::setIdentity_impl<Derived>::run(derived());
}
/** \brief Resizes to the given size, and writes the identity expression (not necessarily square) into *this.
*
* \param nbRows the new number of rows
* \param nbCols the new number of columns
* \param rows the new number of rows
* \param cols the new number of columns
*
* Example: \include Matrix_setIdentity_int_int.cpp
* Output: \verbinclude Matrix_setIdentity_int_int.out
@@ -785,9 +787,9 @@ EIGEN_STRONG_INLINE Derived& MatrixBase<Derived>::setIdentity()
* \sa MatrixBase::setIdentity(), class CwiseNullaryOp, MatrixBase::Identity()
*/
template<typename Derived>
EIGEN_STRONG_INLINE Derived& MatrixBase<Derived>::setIdentity(Index nbRows, Index nbCols)
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE Derived& MatrixBase<Derived>::setIdentity(Index rows, Index cols)
{
derived().resize(nbRows, nbCols);
derived().resize(rows, cols);
return setIdentity();
}
@@ -798,7 +800,7 @@ EIGEN_STRONG_INLINE Derived& MatrixBase<Derived>::setIdentity(Index nbRows, Inde
* \sa MatrixBase::Unit(Index), MatrixBase::UnitX(), MatrixBase::UnitY(), MatrixBase::UnitZ(), MatrixBase::UnitW()
*/
template<typename Derived>
EIGEN_STRONG_INLINE const typename MatrixBase<Derived>::BasisReturnType MatrixBase<Derived>::Unit(Index newSize, Index i)
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const typename MatrixBase<Derived>::BasisReturnType MatrixBase<Derived>::Unit(Index newSize, Index i)
{
EIGEN_STATIC_ASSERT_VECTOR_ONLY(Derived)
return BasisReturnType(SquareMatrixType::Identity(newSize,newSize), i);
@@ -813,7 +815,7 @@ EIGEN_STRONG_INLINE const typename MatrixBase<Derived>::BasisReturnType MatrixBa
* \sa MatrixBase::Unit(Index,Index), MatrixBase::UnitX(), MatrixBase::UnitY(), MatrixBase::UnitZ(), MatrixBase::UnitW()
*/
template<typename Derived>
EIGEN_STRONG_INLINE const typename MatrixBase<Derived>::BasisReturnType MatrixBase<Derived>::Unit(Index i)
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const typename MatrixBase<Derived>::BasisReturnType MatrixBase<Derived>::Unit(Index i)
{
EIGEN_STATIC_ASSERT_VECTOR_ONLY(Derived)
return BasisReturnType(SquareMatrixType::Identity(),i);
@@ -826,7 +828,7 @@ EIGEN_STRONG_INLINE const typename MatrixBase<Derived>::BasisReturnType MatrixBa
* \sa MatrixBase::Unit(Index,Index), MatrixBase::Unit(Index), MatrixBase::UnitY(), MatrixBase::UnitZ(), MatrixBase::UnitW()
*/
template<typename Derived>
EIGEN_STRONG_INLINE const typename MatrixBase<Derived>::BasisReturnType MatrixBase<Derived>::UnitX()
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const typename MatrixBase<Derived>::BasisReturnType MatrixBase<Derived>::UnitX()
{ return Derived::Unit(0); }
/** \returns an expression of the Y axis unit vector (0,1{,0}^*)
@@ -836,7 +838,7 @@ EIGEN_STRONG_INLINE const typename MatrixBase<Derived>::BasisReturnType MatrixBa
* \sa MatrixBase::Unit(Index,Index), MatrixBase::Unit(Index), MatrixBase::UnitY(), MatrixBase::UnitZ(), MatrixBase::UnitW()
*/
template<typename Derived>
EIGEN_STRONG_INLINE const typename MatrixBase<Derived>::BasisReturnType MatrixBase<Derived>::UnitY()
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const typename MatrixBase<Derived>::BasisReturnType MatrixBase<Derived>::UnitY()
{ return Derived::Unit(1); }
/** \returns an expression of the Z axis unit vector (0,0,1{,0}^*)
@@ -846,7 +848,7 @@ EIGEN_STRONG_INLINE const typename MatrixBase<Derived>::BasisReturnType MatrixBa
* \sa MatrixBase::Unit(Index,Index), MatrixBase::Unit(Index), MatrixBase::UnitY(), MatrixBase::UnitZ(), MatrixBase::UnitW()
*/
template<typename Derived>
EIGEN_STRONG_INLINE const typename MatrixBase<Derived>::BasisReturnType MatrixBase<Derived>::UnitZ()
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const typename MatrixBase<Derived>::BasisReturnType MatrixBase<Derived>::UnitZ()
{ return Derived::Unit(2); }
/** \returns an expression of the W axis unit vector (0,0,0,1)
@@ -856,7 +858,7 @@ EIGEN_STRONG_INLINE const typename MatrixBase<Derived>::BasisReturnType MatrixBa
* \sa MatrixBase::Unit(Index,Index), MatrixBase::Unit(Index), MatrixBase::UnitY(), MatrixBase::UnitZ(), MatrixBase::UnitW()
*/
template<typename Derived>
EIGEN_STRONG_INLINE const typename MatrixBase<Derived>::BasisReturnType MatrixBase<Derived>::UnitW()
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const typename MatrixBase<Derived>::BasisReturnType MatrixBase<Derived>::UnitW()
{ return Derived::Unit(3); }
} // end namespace Eigen

View File

@@ -0,0 +1,197 @@
// This file is part of Eigen, a lightweight C++ template library
// for linear algebra.
//
// Copyright (C) 2008-2014 Gael Guennebaud <gael.guennebaud@inria.fr>
// Copyright (C) 2006-2008 Benoit Jacob <jacob.benoit.1@gmail.com>
// Copyright (C) 2016 Eugene Brevdo <ebrevdo@gmail.com>
//
// This Source Code Form is subject to the terms of the Mozilla
// Public License v. 2.0. If a copy of the MPL was not distributed
// with this file, You can obtain one at http://mozilla.org/MPL/2.0/.
#ifndef EIGEN_CWISE_TERNARY_OP_H
#define EIGEN_CWISE_TERNARY_OP_H
namespace Eigen {
namespace internal {
template <typename TernaryOp, typename Arg1, typename Arg2, typename Arg3>
struct traits<CwiseTernaryOp<TernaryOp, Arg1, Arg2, Arg3> > {
// we must not inherit from traits<Arg1> since it has
// the potential to cause problems with MSVC
typedef typename remove_all<Arg1>::type Ancestor;
typedef typename traits<Ancestor>::XprKind XprKind;
enum {
RowsAtCompileTime = traits<Ancestor>::RowsAtCompileTime,
ColsAtCompileTime = traits<Ancestor>::ColsAtCompileTime,
MaxRowsAtCompileTime = traits<Ancestor>::MaxRowsAtCompileTime,
MaxColsAtCompileTime = traits<Ancestor>::MaxColsAtCompileTime
};
// even though we require Arg1, Arg2, and Arg3 to have the same scalar type
// (see CwiseTernaryOp constructor),
// we still want to handle the case when the result type is different.
typedef typename result_of<TernaryOp(
const typename Arg1::Scalar&, const typename Arg2::Scalar&,
const typename Arg3::Scalar&)>::type Scalar;
typedef typename internal::traits<Arg1>::StorageKind StorageKind;
typedef typename internal::traits<Arg1>::StorageIndex StorageIndex;
typedef typename Arg1::Nested Arg1Nested;
typedef typename Arg2::Nested Arg2Nested;
typedef typename Arg3::Nested Arg3Nested;
typedef typename remove_reference<Arg1Nested>::type _Arg1Nested;
typedef typename remove_reference<Arg2Nested>::type _Arg2Nested;
typedef typename remove_reference<Arg3Nested>::type _Arg3Nested;
enum { Flags = _Arg1Nested::Flags & RowMajorBit };
};
} // end namespace internal
template <typename TernaryOp, typename Arg1, typename Arg2, typename Arg3,
typename StorageKind>
class CwiseTernaryOpImpl;
/** \class CwiseTernaryOp
* \ingroup Core_Module
*
* \brief Generic expression where a coefficient-wise ternary operator is
* applied to two expressions
*
* \tparam TernaryOp template functor implementing the operator
* \tparam Arg1Type the type of the first argument
* \tparam Arg2Type the type of the second argument
* \tparam Arg3Type the type of the third argument
*
* This class represents an expression where a coefficient-wise ternary
* operator is applied to three expressions.
* It is the return type of ternary operators, by which we mean only those
* ternary operators where
* all three arguments are Eigen expressions.
* For example, the return type of betainc(matrix1, matrix2, matrix3) is a
* CwiseTernaryOp.
*
* Most of the time, this is the only way that it is used, so you typically
* don't have to name
* CwiseTernaryOp types explicitly.
*
* \sa MatrixBase::ternaryExpr(const MatrixBase<Argument2> &, const
* MatrixBase<Argument3> &, const CustomTernaryOp &) const, class CwiseBinaryOp,
* class CwiseUnaryOp, class CwiseNullaryOp
*/
template <typename TernaryOp, typename Arg1Type, typename Arg2Type,
typename Arg3Type>
class CwiseTernaryOp : public CwiseTernaryOpImpl<
TernaryOp, Arg1Type, Arg2Type, Arg3Type,
typename internal::traits<Arg1Type>::StorageKind>,
internal::no_assignment_operator
{
public:
typedef typename internal::remove_all<Arg1Type>::type Arg1;
typedef typename internal::remove_all<Arg2Type>::type Arg2;
typedef typename internal::remove_all<Arg3Type>::type Arg3;
typedef typename CwiseTernaryOpImpl<
TernaryOp, Arg1Type, Arg2Type, Arg3Type,
typename internal::traits<Arg1Type>::StorageKind>::Base Base;
EIGEN_GENERIC_PUBLIC_INTERFACE(CwiseTernaryOp)
typedef typename internal::ref_selector<Arg1Type>::type Arg1Nested;
typedef typename internal::ref_selector<Arg2Type>::type Arg2Nested;
typedef typename internal::ref_selector<Arg3Type>::type Arg3Nested;
typedef typename internal::remove_reference<Arg1Nested>::type _Arg1Nested;
typedef typename internal::remove_reference<Arg2Nested>::type _Arg2Nested;
typedef typename internal::remove_reference<Arg3Nested>::type _Arg3Nested;
EIGEN_DEVICE_FUNC
EIGEN_STRONG_INLINE CwiseTernaryOp(const Arg1& a1, const Arg2& a2,
const Arg3& a3,
const TernaryOp& func = TernaryOp())
: m_arg1(a1), m_arg2(a2), m_arg3(a3), m_functor(func) {
// require the sizes to match
EIGEN_STATIC_ASSERT_SAME_MATRIX_SIZE(Arg1, Arg2)
EIGEN_STATIC_ASSERT_SAME_MATRIX_SIZE(Arg1, Arg3)
// The index types should match
EIGEN_STATIC_ASSERT((internal::is_same<
typename internal::traits<Arg1Type>::StorageKind,
typename internal::traits<Arg2Type>::StorageKind>::value),
STORAGE_KIND_MUST_MATCH)
EIGEN_STATIC_ASSERT((internal::is_same<
typename internal::traits<Arg1Type>::StorageKind,
typename internal::traits<Arg3Type>::StorageKind>::value),
STORAGE_KIND_MUST_MATCH)
eigen_assert(a1.rows() == a2.rows() && a1.cols() == a2.cols() &&
a1.rows() == a3.rows() && a1.cols() == a3.cols());
}
EIGEN_DEVICE_FUNC
EIGEN_STRONG_INLINE Index rows() const {
// return the fixed size type if available to enable compile time
// optimizations
if (internal::traits<typename internal::remove_all<Arg1Nested>::type>::
RowsAtCompileTime == Dynamic &&
internal::traits<typename internal::remove_all<Arg2Nested>::type>::
RowsAtCompileTime == Dynamic)
return m_arg3.rows();
else if (internal::traits<typename internal::remove_all<Arg1Nested>::type>::
RowsAtCompileTime == Dynamic &&
internal::traits<typename internal::remove_all<Arg3Nested>::type>::
RowsAtCompileTime == Dynamic)
return m_arg2.rows();
else
return m_arg1.rows();
}
EIGEN_DEVICE_FUNC
EIGEN_STRONG_INLINE Index cols() const {
// return the fixed size type if available to enable compile time
// optimizations
if (internal::traits<typename internal::remove_all<Arg1Nested>::type>::
ColsAtCompileTime == Dynamic &&
internal::traits<typename internal::remove_all<Arg2Nested>::type>::
ColsAtCompileTime == Dynamic)
return m_arg3.cols();
else if (internal::traits<typename internal::remove_all<Arg1Nested>::type>::
ColsAtCompileTime == Dynamic &&
internal::traits<typename internal::remove_all<Arg3Nested>::type>::
ColsAtCompileTime == Dynamic)
return m_arg2.cols();
else
return m_arg1.cols();
}
/** \returns the first argument nested expression */
EIGEN_DEVICE_FUNC
const _Arg1Nested& arg1() const { return m_arg1; }
/** \returns the first argument nested expression */
EIGEN_DEVICE_FUNC
const _Arg2Nested& arg2() const { return m_arg2; }
/** \returns the third argument nested expression */
EIGEN_DEVICE_FUNC
const _Arg3Nested& arg3() const { return m_arg3; }
/** \returns the functor representing the ternary operation */
EIGEN_DEVICE_FUNC
const TernaryOp& functor() const { return m_functor; }
protected:
Arg1Nested m_arg1;
Arg2Nested m_arg2;
Arg3Nested m_arg3;
const TernaryOp m_functor;
};
// Generic API dispatcher
template <typename TernaryOp, typename Arg1, typename Arg2, typename Arg3,
typename StorageKind>
class CwiseTernaryOpImpl
: public internal::generic_xpr_base<
CwiseTernaryOp<TernaryOp, Arg1, Arg2, Arg3> >::type {
public:
typedef typename internal::generic_xpr_base<
CwiseTernaryOp<TernaryOp, Arg1, Arg2, Arg3> >::type Base;
};
} // end namespace Eigen
#endif // EIGEN_CWISE_TERNARY_OP_H

View File

@@ -1,7 +1,7 @@
// This file is part of Eigen, a lightweight C++ template library
// for linear algebra.
//
// Copyright (C) 2008-2010 Gael Guennebaud <gael.guennebaud@inria.fr>
// Copyright (C) 2008-2014 Gael Guennebaud <gael.guennebaud@inria.fr>
// Copyright (C) 2006-2008 Benoit Jacob <jacob.benoit.1@gmail.com>
//
// This Source Code Form is subject to the terms of the Mozilla
@@ -13,13 +13,32 @@
namespace Eigen {
namespace internal {
template<typename UnaryOp, typename XprType>
struct traits<CwiseUnaryOp<UnaryOp, XprType> >
: traits<XprType>
{
typedef typename result_of<
UnaryOp(const typename XprType::Scalar&)
>::type Scalar;
typedef typename XprType::Nested XprTypeNested;
typedef typename remove_reference<XprTypeNested>::type _XprTypeNested;
enum {
Flags = _XprTypeNested::Flags & RowMajorBit
};
};
}
template<typename UnaryOp, typename XprType, typename StorageKind>
class CwiseUnaryOpImpl;
/** \class CwiseUnaryOp
* \ingroup Core_Module
*
* \brief Generic expression where a coefficient-wise unary operator is applied to an expression
*
* \param UnaryOp template functor implementing the operator
* \param XprType the type of the expression to which we are applying the unary operator
* \tparam UnaryOp template functor implementing the operator
* \tparam XprType the type of the expression to which we are applying the unary operator
*
* This class represents an expression where a unary operator is applied to an expression.
* It is the return type of all operations taking exactly 1 input expression, regardless of the
@@ -32,93 +51,51 @@ namespace Eigen {
*
* \sa MatrixBase::unaryExpr(const CustomUnaryOp &) const, class CwiseBinaryOp, class CwiseNullaryOp
*/
namespace internal {
template<typename UnaryOp, typename XprType>
struct traits<CwiseUnaryOp<UnaryOp, XprType> >
: traits<XprType>
{
typedef typename result_of<
UnaryOp(typename XprType::Scalar)
>::type Scalar;
typedef typename XprType::Nested XprTypeNested;
typedef typename remove_reference<XprTypeNested>::type _XprTypeNested;
enum {
Flags = _XprTypeNested::Flags & (
HereditaryBits | LinearAccessBit | AlignedBit
| (functor_traits<UnaryOp>::PacketAccess ? PacketAccessBit : 0)),
CoeffReadCost = EIGEN_ADD_COST(_XprTypeNested::CoeffReadCost, functor_traits<UnaryOp>::Cost)
};
};
}
template<typename UnaryOp, typename XprType, typename StorageKind>
class CwiseUnaryOpImpl;
template<typename UnaryOp, typename XprType>
class CwiseUnaryOp : internal::no_assignment_operator,
public CwiseUnaryOpImpl<UnaryOp, XprType, typename internal::traits<XprType>::StorageKind>
class CwiseUnaryOp : public CwiseUnaryOpImpl<UnaryOp, XprType, typename internal::traits<XprType>::StorageKind>, internal::no_assignment_operator
{
public:
typedef typename CwiseUnaryOpImpl<UnaryOp, XprType,typename internal::traits<XprType>::StorageKind>::Base Base;
EIGEN_GENERIC_PUBLIC_INTERFACE(CwiseUnaryOp)
typedef typename internal::ref_selector<XprType>::type XprTypeNested;
typedef typename internal::remove_all<XprType>::type NestedExpression;
inline CwiseUnaryOp(const XprType& xpr, const UnaryOp& func = UnaryOp())
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE
explicit CwiseUnaryOp(const XprType& xpr, const UnaryOp& func = UnaryOp())
: m_xpr(xpr), m_functor(func) {}
EIGEN_STRONG_INLINE Index rows() const { return m_xpr.rows(); }
EIGEN_STRONG_INLINE Index cols() const { return m_xpr.cols(); }
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE
Index rows() const { return m_xpr.rows(); }
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE
Index cols() const { return m_xpr.cols(); }
/** \returns the functor representing the unary operation */
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE
const UnaryOp& functor() const { return m_functor; }
/** \returns the nested expression */
const typename internal::remove_all<typename XprType::Nested>::type&
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE
const typename internal::remove_all<XprTypeNested>::type&
nestedExpression() const { return m_xpr; }
/** \returns the nested expression */
typename internal::remove_all<typename XprType::Nested>::type&
nestedExpression() { return m_xpr.const_cast_derived(); }
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE
typename internal::remove_all<XprTypeNested>::type&
nestedExpression() { return m_xpr; }
protected:
typename XprType::Nested m_xpr;
XprTypeNested m_xpr;
const UnaryOp m_functor;
};
// This is the generic implementation for dense storage.
// It can be used for any expression types implementing the dense concept.
template<typename UnaryOp, typename XprType>
class CwiseUnaryOpImpl<UnaryOp,XprType,Dense>
: public internal::dense_xpr_base<CwiseUnaryOp<UnaryOp, XprType> >::type
// Generic API dispatcher
template<typename UnaryOp, typename XprType, typename StorageKind>
class CwiseUnaryOpImpl
: public internal::generic_xpr_base<CwiseUnaryOp<UnaryOp, XprType> >::type
{
public:
typedef CwiseUnaryOp<UnaryOp, XprType> Derived;
typedef typename internal::dense_xpr_base<CwiseUnaryOp<UnaryOp, XprType> >::type Base;
EIGEN_DENSE_PUBLIC_INTERFACE(Derived)
EIGEN_STRONG_INLINE const Scalar coeff(Index rowId, Index colId) const
{
return derived().functor()(derived().nestedExpression().coeff(rowId, colId));
}
template<int LoadMode>
EIGEN_STRONG_INLINE PacketScalar packet(Index rowId, Index colId) const
{
return derived().functor().packetOp(derived().nestedExpression().template packet<LoadMode>(rowId, colId));
}
EIGEN_STRONG_INLINE const Scalar coeff(Index index) const
{
return derived().functor()(derived().nestedExpression().coeff(index));
}
template<int LoadMode>
EIGEN_STRONG_INLINE PacketScalar packet(Index index) const
{
return derived().functor().packetOp(derived().nestedExpression().template packet<LoadMode>(index));
}
public:
typedef typename internal::generic_xpr_base<CwiseUnaryOp<UnaryOp, XprType> >::type Base;
};
} // end namespace Eigen

View File

@@ -12,33 +12,19 @@
namespace Eigen {
/** \class CwiseUnaryView
* \ingroup Core_Module
*
* \brief Generic lvalue expression of a coefficient-wise unary operator of a matrix or a vector
*
* \param ViewOp template functor implementing the view
* \param MatrixType the type of the matrix we are applying the unary operator
*
* This class represents a lvalue expression of a generic unary view operator of a matrix or a vector.
* It is the return type of real() and imag(), and most of the time this is the only way it is used.
*
* \sa MatrixBase::unaryViewExpr(const CustomUnaryOp &) const, class CwiseUnaryOp
*/
namespace internal {
template<typename ViewOp, typename MatrixType>
struct traits<CwiseUnaryView<ViewOp, MatrixType> >
: traits<MatrixType>
{
typedef typename result_of<
ViewOp(typename traits<MatrixType>::Scalar)
ViewOp(const typename traits<MatrixType>::Scalar&)
>::type Scalar;
typedef typename MatrixType::Nested MatrixTypeNested;
typedef typename remove_all<MatrixTypeNested>::type _MatrixTypeNested;
enum {
Flags = (traits<_MatrixTypeNested>::Flags & (HereditaryBits | LvalueBit | LinearAccessBit | DirectAccessBit)),
CoeffReadCost = traits<_MatrixTypeNested>::CoeffReadCost + functor_traits<ViewOp>::Cost,
FlagsLvalueBit = is_lvalue<MatrixType>::value ? LvalueBit : 0,
Flags = traits<_MatrixTypeNested>::Flags & (RowMajorBit | FlagsLvalueBit | DirectAccessBit), // FIXME DirectAccessBit should not be handled by expressions
MatrixTypeInnerStride = inner_stride_at_compile_time<MatrixType>::ret,
// need to cast the sizeof's from size_t to int explicitly, otherwise:
// "error: no integral type can represent all of the enumerator values
@@ -55,6 +41,19 @@ struct traits<CwiseUnaryView<ViewOp, MatrixType> >
template<typename ViewOp, typename MatrixType, typename StorageKind>
class CwiseUnaryViewImpl;
/** \class CwiseUnaryView
* \ingroup Core_Module
*
* \brief Generic lvalue expression of a coefficient-wise unary operator of a matrix or a vector
*
* \tparam ViewOp template functor implementing the view
* \tparam MatrixType the type of the matrix we are applying the unary operator
*
* This class represents a lvalue expression of a generic unary view operator of a matrix or a vector.
* It is the return type of real() and imag(), and most of the time this is the only way it is used.
*
* \sa MatrixBase::unaryViewExpr(const CustomUnaryOp &) const, class CwiseUnaryOp
*/
template<typename ViewOp, typename MatrixType>
class CwiseUnaryView : public CwiseUnaryViewImpl<ViewOp, MatrixType, typename internal::traits<MatrixType>::StorageKind>
{
@@ -62,8 +61,10 @@ class CwiseUnaryView : public CwiseUnaryViewImpl<ViewOp, MatrixType, typename in
typedef typename CwiseUnaryViewImpl<ViewOp, MatrixType,typename internal::traits<MatrixType>::StorageKind>::Base Base;
EIGEN_GENERIC_PUBLIC_INTERFACE(CwiseUnaryView)
typedef typename internal::ref_selector<MatrixType>::non_const_type MatrixTypeNested;
typedef typename internal::remove_all<MatrixType>::type NestedExpression;
inline CwiseUnaryView(const MatrixType& mat, const ViewOp& func = ViewOp())
explicit inline CwiseUnaryView(MatrixType& mat, const ViewOp& func = ViewOp())
: m_matrix(mat), m_functor(func) {}
EIGEN_INHERIT_ASSIGNMENT_OPERATORS(CwiseUnaryView)
@@ -75,19 +76,27 @@ class CwiseUnaryView : public CwiseUnaryViewImpl<ViewOp, MatrixType, typename in
const ViewOp& functor() const { return m_functor; }
/** \returns the nested expression */
const typename internal::remove_all<typename MatrixType::Nested>::type&
const typename internal::remove_all<MatrixTypeNested>::type&
nestedExpression() const { return m_matrix; }
/** \returns the nested expression */
typename internal::remove_all<typename MatrixType::Nested>::type&
typename internal::remove_reference<MatrixTypeNested>::type&
nestedExpression() { return m_matrix.const_cast_derived(); }
protected:
// FIXME changed from MatrixType::Nested because of a weird compilation error with sun CC
typename internal::nested<MatrixType>::type m_matrix;
MatrixTypeNested m_matrix;
ViewOp m_functor;
};
// Generic API dispatcher
template<typename ViewOp, typename XprType, typename StorageKind>
class CwiseUnaryViewImpl
: public internal::generic_xpr_base<CwiseUnaryView<ViewOp, XprType> >::type
{
public:
typedef typename internal::generic_xpr_base<CwiseUnaryView<ViewOp, XprType> >::type Base;
};
template<typename ViewOp, typename MatrixType>
class CwiseUnaryViewImpl<ViewOp,MatrixType,Dense>
: public internal::dense_xpr_base< CwiseUnaryView<ViewOp, MatrixType> >::type
@@ -100,38 +109,18 @@ class CwiseUnaryViewImpl<ViewOp,MatrixType,Dense>
EIGEN_DENSE_PUBLIC_INTERFACE(Derived)
EIGEN_INHERIT_ASSIGNMENT_OPERATORS(CwiseUnaryViewImpl)
inline Scalar* data() { return &coeffRef(0); }
inline const Scalar* data() const { return &coeff(0); }
EIGEN_DEVICE_FUNC inline Scalar* data() { return &(this->coeffRef(0)); }
EIGEN_DEVICE_FUNC inline const Scalar* data() const { return &(this->coeff(0)); }
inline Index innerStride() const
EIGEN_DEVICE_FUNC inline Index innerStride() const
{
return derived().nestedExpression().innerStride() * sizeof(typename internal::traits<MatrixType>::Scalar) / sizeof(Scalar);
}
inline Index outerStride() const
EIGEN_DEVICE_FUNC inline Index outerStride() const
{
return derived().nestedExpression().outerStride() * sizeof(typename internal::traits<MatrixType>::Scalar) / sizeof(Scalar);
}
EIGEN_STRONG_INLINE CoeffReturnType coeff(Index row, Index col) const
{
return derived().functor()(derived().nestedExpression().coeff(row, col));
}
EIGEN_STRONG_INLINE CoeffReturnType coeff(Index index) const
{
return derived().functor()(derived().nestedExpression().coeff(index));
}
EIGEN_STRONG_INLINE Scalar& coeffRef(Index row, Index col)
{
return derived().functor()(const_cast_derived().nestedExpression().coeffRef(row, col));
}
EIGEN_STRONG_INLINE Scalar& coeffRef(Index index)
{
return derived().functor()(const_cast_derived().nestedExpression().coeffRef(index));
}
};
} // end namespace Eigen

View File

@@ -34,37 +34,45 @@ static inline void check_DenseIndex_is_signed() {
* \tparam Derived is the derived type, e.g., a matrix type or an expression.
*
* This class can be extended with the help of the plugin mechanism described on the page
* \ref TopicCustomizingEigen by defining the preprocessor symbol \c EIGEN_DENSEBASE_PLUGIN.
* \ref TopicCustomizing_Plugins by defining the preprocessor symbol \c EIGEN_DENSEBASE_PLUGIN.
*
* \sa \ref TopicClassHierarchy
* \sa \blank \ref TopicClassHierarchy
*/
template<typename Derived> class DenseBase
#ifndef EIGEN_PARSED_BY_DOXYGEN
: public internal::special_scalar_op_base<Derived, typename internal::traits<Derived>::Scalar,
typename NumTraits<typename internal::traits<Derived>::Scalar>::Real,
DenseCoeffsBase<Derived> >
#else
: public DenseCoeffsBase<Derived>
#else
: public DenseCoeffsBase<Derived,DirectWriteAccessors>
#endif // not EIGEN_PARSED_BY_DOXYGEN
{
public:
class InnerIterator;
/** Inner iterator type to iterate over the coefficients of a row or column.
* \sa class InnerIterator
*/
typedef Eigen::InnerIterator<Derived> InnerIterator;
typedef typename internal::traits<Derived>::StorageKind StorageKind;
/** \brief The type of indices
* \details To change this, \c \#define the preprocessor symbol \c EIGEN_DEFAULT_DENSE_INDEX_TYPE.
* \sa \ref TopicPreprocessorDirectives.
*/
typedef typename internal::traits<Derived>::Index Index;
/**
* \brief The type used to store indices
* \details This typedef is relevant for types that store multiple indices such as
* PermutationMatrix or Transpositions, otherwise it defaults to Eigen::Index
* \sa \blank \ref TopicPreprocessorDirectives, Eigen::Index, SparseMatrixBase.
*/
typedef typename internal::traits<Derived>::StorageIndex StorageIndex;
/** The numeric type of the expression' coefficients, e.g. float, double, int or std::complex<float>, etc. */
typedef typename internal::traits<Derived>::Scalar Scalar;
typedef typename internal::packet_traits<Scalar>::type PacketScalar;
/** The numeric type of the expression' coefficients, e.g. float, double, int or std::complex<float>, etc.
*
* It is an alias for the Scalar type */
typedef Scalar value_type;
typedef typename NumTraits<Scalar>::Real RealScalar;
typedef internal::special_scalar_op_base<Derived,Scalar,RealScalar, DenseCoeffsBase<Derived> > Base;
typedef DenseCoeffsBase<Derived> Base;
using Base::operator*;
using Base::derived;
using Base::const_cast_derived;
using Base::rows;
@@ -74,16 +82,6 @@ template<typename Derived> class DenseBase
using Base::colIndexByOuterInner;
using Base::coeff;
using Base::coeffByOuterInner;
using Base::packet;
using Base::packetByOuterInner;
using Base::writePacket;
using Base::writePacketByOuterInner;
using Base::coeffRef;
using Base::coeffRefByOuterInner;
using Base::copyCoeff;
using Base::copyCoeffByOuterInner;
using Base::copyPacket;
using Base::copyPacketByOuterInner;
using Base::operator();
using Base::operator[];
using Base::x;
@@ -169,19 +167,46 @@ template<typename Derived> class DenseBase
InnerSizeAtCompileTime = int(IsVectorAtCompileTime) ? int(SizeAtCompileTime)
: int(IsRowMajor) ? int(ColsAtCompileTime) : int(RowsAtCompileTime),
CoeffReadCost = internal::traits<Derived>::CoeffReadCost,
/**< This is a rough measure of how expensive it is to read one coefficient from
* this expression.
*/
InnerStrideAtCompileTime = internal::inner_stride_at_compile_time<Derived>::ret,
OuterStrideAtCompileTime = internal::outer_stride_at_compile_time<Derived>::ret
};
typedef typename internal::find_best_packet<Scalar,SizeAtCompileTime>::type PacketScalar;
enum { ThisConstantIsPrivateInPlainObjectBase };
enum { IsPlainObjectBase = 0 };
/** The plain matrix type corresponding to this expression.
* \sa PlainObject */
typedef Matrix<typename internal::traits<Derived>::Scalar,
internal::traits<Derived>::RowsAtCompileTime,
internal::traits<Derived>::ColsAtCompileTime,
AutoAlign | (internal::traits<Derived>::Flags&RowMajorBit ? RowMajor : ColMajor),
internal::traits<Derived>::MaxRowsAtCompileTime,
internal::traits<Derived>::MaxColsAtCompileTime
> PlainMatrix;
/** The plain array type corresponding to this expression.
* \sa PlainObject */
typedef Array<typename internal::traits<Derived>::Scalar,
internal::traits<Derived>::RowsAtCompileTime,
internal::traits<Derived>::ColsAtCompileTime,
AutoAlign | (internal::traits<Derived>::Flags&RowMajorBit ? RowMajor : ColMajor),
internal::traits<Derived>::MaxRowsAtCompileTime,
internal::traits<Derived>::MaxColsAtCompileTime
> PlainArray;
/** \brief The plain matrix or array type corresponding to this expression.
*
* This is not necessarily exactly the return type of eval(). In the case of plain matrices,
* the return type of eval() is a const reference to a matrix, not a matrix! It is however guaranteed
* that the return type of eval() is either PlainObject or const PlainObject&.
*/
typedef typename internal::conditional<internal::is_same<typename internal::traits<Derived>::XprKind,MatrixXpr >::value,
PlainMatrix, PlainArray>::type PlainObject;
/** \returns the number of nonzero coefficients which is in practice the number
* of stored coefficients. */
EIGEN_DEVICE_FUNC
inline Index nonZeros() const { return size(); }
/** \returns the outer size.
@@ -189,6 +214,7 @@ template<typename Derived> class DenseBase
* \note For a vector, this returns just 1. For a matrix (non-vector), this is the major dimension
* with respect to the \ref TopicStorageOrders "storage order", i.e., the number of columns for a
* column-major matrix, and the number of rows for a row-major matrix. */
EIGEN_DEVICE_FUNC
Index outerSize() const
{
return IsVectorAtCompileTime ? 1
@@ -200,6 +226,7 @@ template<typename Derived> class DenseBase
* \note For a vector, this is just the size. For a matrix (non-vector), this is the minor dimension
* with respect to the \ref TopicStorageOrders "storage order", i.e., the number of rows for a
* column-major matrix, and the number of columns for a row-major matrix. */
EIGEN_DEVICE_FUNC
Index innerSize() const
{
return IsVectorAtCompileTime ? this->size()
@@ -210,6 +237,7 @@ template<typename Derived> class DenseBase
* Matrix::resize() and Array::resize(). The present method only asserts that the new size equals the old size, and does
* nothing else.
*/
EIGEN_DEVICE_FUNC
void resize(Index newSize)
{
EIGEN_ONLY_USED_FOR_DEBUG(newSize);
@@ -220,22 +248,22 @@ template<typename Derived> class DenseBase
* Matrix::resize() and Array::resize(). The present method only asserts that the new size equals the old size, and does
* nothing else.
*/
void resize(Index nbRows, Index nbCols)
EIGEN_DEVICE_FUNC
void resize(Index rows, Index cols)
{
EIGEN_ONLY_USED_FOR_DEBUG(nbRows);
EIGEN_ONLY_USED_FOR_DEBUG(nbCols);
eigen_assert(nbRows == this->rows() && nbCols == this->cols()
EIGEN_ONLY_USED_FOR_DEBUG(rows);
EIGEN_ONLY_USED_FOR_DEBUG(cols);
eigen_assert(rows == this->rows() && cols == this->cols()
&& "DenseBase::resize() does not actually allow to resize.");
}
#ifndef EIGEN_PARSED_BY_DOXYGEN
/** \internal Represents a matrix with all coefficients equal to one another*/
typedef CwiseNullaryOp<internal::scalar_constant_op<Scalar>,Derived> ConstantReturnType;
/** \internal Represents a vector with linearly spaced coefficients that allows sequential access only. */
typedef CwiseNullaryOp<internal::linspaced_op<Scalar,false>,Derived> SequentialLinSpacedReturnType;
typedef CwiseNullaryOp<internal::scalar_constant_op<Scalar>,PlainObject> ConstantReturnType;
/** \internal \deprecated Represents a vector with linearly spaced coefficients that allows sequential access only. */
typedef CwiseNullaryOp<internal::linspaced_op<Scalar,PacketScalar>,PlainObject> SequentialLinSpacedReturnType;
/** \internal Represents a vector with linearly spaced coefficients that allows random access. */
typedef CwiseNullaryOp<internal::linspaced_op<Scalar,true>,Derived> RandomAccessLinSpacedReturnType;
typedef CwiseNullaryOp<internal::linspaced_op<Scalar,PacketScalar>,PlainObject> RandomAccessLinSpacedReturnType;
/** \internal the return type of MatrixBase::eigenvalues() */
typedef Matrix<typename NumTraits<typename internal::traits<Derived>::Scalar>::Real, internal::traits<Derived>::ColsAtCompileTime, 1> EigenvaluesReturnType;
@@ -243,120 +271,133 @@ template<typename Derived> class DenseBase
/** Copies \a other into *this. \returns a reference to *this. */
template<typename OtherDerived>
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE
Derived& operator=(const DenseBase<OtherDerived>& other);
/** Special case of the template operator=, in order to prevent the compiler
* from generating a default operator= (issue hit with g++ 4.1)
*/
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE
Derived& operator=(const DenseBase& other);
template<typename OtherDerived>
EIGEN_DEVICE_FUNC
Derived& operator=(const EigenBase<OtherDerived> &other);
template<typename OtherDerived>
EIGEN_DEVICE_FUNC
Derived& operator+=(const EigenBase<OtherDerived> &other);
template<typename OtherDerived>
EIGEN_DEVICE_FUNC
Derived& operator-=(const EigenBase<OtherDerived> &other);
template<typename OtherDerived>
EIGEN_DEVICE_FUNC
Derived& operator=(const ReturnByValue<OtherDerived>& func);
/** \internal Copies \a other into *this without evaluating other. \returns a reference to *this. */
/** \internal
* Copies \a other into *this without evaluating other. \returns a reference to *this.
* \deprecated */
template<typename OtherDerived>
EIGEN_DEVICE_FUNC
Derived& lazyAssign(const DenseBase<OtherDerived>& other);
/** \internal Evaluates \a other into *this. \returns a reference to *this. */
template<typename OtherDerived>
Derived& lazyAssign(const ReturnByValue<OtherDerived>& other);
EIGEN_DEVICE_FUNC
CommaInitializer<Derived> operator<< (const Scalar& s);
/** \deprecated it now returns \c *this */
template<unsigned int Added,unsigned int Removed>
const Flagged<Derived, Added, Removed> flagged() const;
EIGEN_DEPRECATED
const Derived& flagged() const
{ return derived(); }
template<typename OtherDerived>
EIGEN_DEVICE_FUNC
CommaInitializer<Derived> operator<< (const DenseBase<OtherDerived>& other);
Eigen::Transpose<Derived> transpose();
typedef typename internal::add_const<Transpose<const Derived> >::type ConstTransposeReturnType;
typedef Transpose<Derived> TransposeReturnType;
EIGEN_DEVICE_FUNC
TransposeReturnType transpose();
typedef typename internal::add_const<Transpose<const Derived> >::type ConstTransposeReturnType;
EIGEN_DEVICE_FUNC
ConstTransposeReturnType transpose() const;
EIGEN_DEVICE_FUNC
void transposeInPlace();
#ifndef EIGEN_NO_DEBUG
protected:
template<typename OtherDerived>
void checkTransposeAliasing(const OtherDerived& other) const;
public:
#endif
static const ConstantReturnType
EIGEN_DEVICE_FUNC static const ConstantReturnType
Constant(Index rows, Index cols, const Scalar& value);
static const ConstantReturnType
EIGEN_DEVICE_FUNC static const ConstantReturnType
Constant(Index size, const Scalar& value);
static const ConstantReturnType
EIGEN_DEVICE_FUNC static const ConstantReturnType
Constant(const Scalar& value);
static const SequentialLinSpacedReturnType
EIGEN_DEVICE_FUNC static const SequentialLinSpacedReturnType
LinSpaced(Sequential_t, Index size, const Scalar& low, const Scalar& high);
static const RandomAccessLinSpacedReturnType
EIGEN_DEVICE_FUNC static const RandomAccessLinSpacedReturnType
LinSpaced(Index size, const Scalar& low, const Scalar& high);
static const SequentialLinSpacedReturnType
EIGEN_DEVICE_FUNC static const SequentialLinSpacedReturnType
LinSpaced(Sequential_t, const Scalar& low, const Scalar& high);
static const RandomAccessLinSpacedReturnType
EIGEN_DEVICE_FUNC static const RandomAccessLinSpacedReturnType
LinSpaced(const Scalar& low, const Scalar& high);
template<typename CustomNullaryOp>
static const CwiseNullaryOp<CustomNullaryOp, Derived>
template<typename CustomNullaryOp> EIGEN_DEVICE_FUNC
static const CwiseNullaryOp<CustomNullaryOp, PlainObject>
NullaryExpr(Index rows, Index cols, const CustomNullaryOp& func);
template<typename CustomNullaryOp>
static const CwiseNullaryOp<CustomNullaryOp, Derived>
template<typename CustomNullaryOp> EIGEN_DEVICE_FUNC
static const CwiseNullaryOp<CustomNullaryOp, PlainObject>
NullaryExpr(Index size, const CustomNullaryOp& func);
template<typename CustomNullaryOp>
static const CwiseNullaryOp<CustomNullaryOp, Derived>
template<typename CustomNullaryOp> EIGEN_DEVICE_FUNC
static const CwiseNullaryOp<CustomNullaryOp, PlainObject>
NullaryExpr(const CustomNullaryOp& func);
static const ConstantReturnType Zero(Index rows, Index cols);
static const ConstantReturnType Zero(Index size);
static const ConstantReturnType Zero();
static const ConstantReturnType Ones(Index rows, Index cols);
static const ConstantReturnType Ones(Index size);
static const ConstantReturnType Ones();
EIGEN_DEVICE_FUNC static const ConstantReturnType Zero(Index rows, Index cols);
EIGEN_DEVICE_FUNC static const ConstantReturnType Zero(Index size);
EIGEN_DEVICE_FUNC static const ConstantReturnType Zero();
EIGEN_DEVICE_FUNC static const ConstantReturnType Ones(Index rows, Index cols);
EIGEN_DEVICE_FUNC static const ConstantReturnType Ones(Index size);
EIGEN_DEVICE_FUNC static const ConstantReturnType Ones();
void fill(const Scalar& value);
Derived& setConstant(const Scalar& value);
Derived& setLinSpaced(Index size, const Scalar& low, const Scalar& high);
Derived& setLinSpaced(const Scalar& low, const Scalar& high);
Derived& setZero();
Derived& setOnes();
Derived& setRandom();
EIGEN_DEVICE_FUNC void fill(const Scalar& value);
EIGEN_DEVICE_FUNC Derived& setConstant(const Scalar& value);
EIGEN_DEVICE_FUNC Derived& setLinSpaced(Index size, const Scalar& low, const Scalar& high);
EIGEN_DEVICE_FUNC Derived& setLinSpaced(const Scalar& low, const Scalar& high);
EIGEN_DEVICE_FUNC Derived& setZero();
EIGEN_DEVICE_FUNC Derived& setOnes();
EIGEN_DEVICE_FUNC Derived& setRandom();
template<typename OtherDerived>
template<typename OtherDerived> EIGEN_DEVICE_FUNC
bool isApprox(const DenseBase<OtherDerived>& other,
const RealScalar& prec = NumTraits<Scalar>::dummy_precision()) const;
EIGEN_DEVICE_FUNC
bool isMuchSmallerThan(const RealScalar& other,
const RealScalar& prec = NumTraits<Scalar>::dummy_precision()) const;
template<typename OtherDerived>
template<typename OtherDerived> EIGEN_DEVICE_FUNC
bool isMuchSmallerThan(const DenseBase<OtherDerived>& other,
const RealScalar& prec = NumTraits<Scalar>::dummy_precision()) const;
bool isApproxToConstant(const Scalar& value, const RealScalar& prec = NumTraits<Scalar>::dummy_precision()) const;
bool isConstant(const Scalar& value, const RealScalar& prec = NumTraits<Scalar>::dummy_precision()) const;
bool isZero(const RealScalar& prec = NumTraits<Scalar>::dummy_precision()) const;
bool isOnes(const RealScalar& prec = NumTraits<Scalar>::dummy_precision()) const;
EIGEN_DEVICE_FUNC bool isApproxToConstant(const Scalar& value, const RealScalar& prec = NumTraits<Scalar>::dummy_precision()) const;
EIGEN_DEVICE_FUNC bool isConstant(const Scalar& value, const RealScalar& prec = NumTraits<Scalar>::dummy_precision()) const;
EIGEN_DEVICE_FUNC bool isZero(const RealScalar& prec = NumTraits<Scalar>::dummy_precision()) const;
EIGEN_DEVICE_FUNC bool isOnes(const RealScalar& prec = NumTraits<Scalar>::dummy_precision()) const;
inline bool hasNaN() const;
inline bool allFinite() const;
inline Derived& operator*=(const Scalar& other);
inline Derived& operator/=(const Scalar& other);
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE
Derived& operator*=(const Scalar& other);
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE
Derived& operator/=(const Scalar& other);
typedef typename internal::add_const_on_value_type<typename internal::eval<Derived>::type>::type EvalReturnType;
/** \returns the matrix or vector obtained by evaluating this expression.
*
* Notice that in the case of a plain matrix or vector (not an expression) this function just returns
* a const reference, in order to avoid a useless copy.
*
* \warning Be carefull with eval() and the auto C++ keyword, as detailed in this \link TopicPitfalls_auto_keyword page \endlink.
*/
EIGEN_DEVICE_FUNC
EIGEN_STRONG_INLINE EvalReturnType eval() const
{
// Even though MSVC does not honor strong inlining when the return type
@@ -364,61 +405,78 @@ template<typename Derived> class DenseBase
// size types on MSVC.
return typename internal::eval<Derived>::type(derived());
}
/** swaps *this with the expression \a other.
*
*/
template<typename OtherDerived>
void swap(const DenseBase<OtherDerived>& other,
int = OtherDerived::ThisConstantIsPrivateInPlainObjectBase)
EIGEN_DEVICE_FUNC
void swap(const DenseBase<OtherDerived>& other)
{
SwapWrapper<Derived>(derived()).lazyAssign(other.derived());
EIGEN_STATIC_ASSERT(!OtherDerived::IsPlainObjectBase,THIS_EXPRESSION_IS_NOT_A_LVALUE__IT_IS_READ_ONLY);
eigen_assert(rows()==other.rows() && cols()==other.cols());
call_assignment(derived(), other.const_cast_derived(), internal::swap_assign_op<Scalar>());
}
/** swaps *this with the matrix or array \a other.
*
*/
template<typename OtherDerived>
EIGEN_DEVICE_FUNC
void swap(PlainObjectBase<OtherDerived>& other)
{
SwapWrapper<Derived>(derived()).lazyAssign(other.derived());
eigen_assert(rows()==other.rows() && cols()==other.cols());
call_assignment(derived(), other.derived(), internal::swap_assign_op<Scalar>());
}
EIGEN_DEVICE_FUNC inline const NestByValue<Derived> nestByValue() const;
EIGEN_DEVICE_FUNC inline const ForceAlignedAccess<Derived> forceAlignedAccess() const;
EIGEN_DEVICE_FUNC inline ForceAlignedAccess<Derived> forceAlignedAccess();
template<bool Enable> EIGEN_DEVICE_FUNC
inline const typename internal::conditional<Enable,ForceAlignedAccess<Derived>,Derived&>::type forceAlignedAccessIf() const;
template<bool Enable> EIGEN_DEVICE_FUNC
inline typename internal::conditional<Enable,ForceAlignedAccess<Derived>,Derived&>::type forceAlignedAccessIf();
inline const NestByValue<Derived> nestByValue() const;
inline const ForceAlignedAccess<Derived> forceAlignedAccess() const;
inline ForceAlignedAccess<Derived> forceAlignedAccess();
template<bool Enable> inline const typename internal::conditional<Enable,ForceAlignedAccess<Derived>,Derived&>::type forceAlignedAccessIf() const;
template<bool Enable> inline typename internal::conditional<Enable,ForceAlignedAccess<Derived>,Derived&>::type forceAlignedAccessIf();
EIGEN_DEVICE_FUNC Scalar sum() const;
EIGEN_DEVICE_FUNC Scalar mean() const;
EIGEN_DEVICE_FUNC Scalar trace() const;
Scalar sum() const;
Scalar mean() const;
Scalar trace() const;
EIGEN_DEVICE_FUNC Scalar prod() const;
Scalar prod() const;
EIGEN_DEVICE_FUNC typename internal::traits<Derived>::Scalar minCoeff() const;
EIGEN_DEVICE_FUNC typename internal::traits<Derived>::Scalar maxCoeff() const;
typename internal::traits<Derived>::Scalar minCoeff() const;
typename internal::traits<Derived>::Scalar maxCoeff() const;
template<typename IndexType>
template<typename IndexType> EIGEN_DEVICE_FUNC
typename internal::traits<Derived>::Scalar minCoeff(IndexType* row, IndexType* col) const;
template<typename IndexType>
template<typename IndexType> EIGEN_DEVICE_FUNC
typename internal::traits<Derived>::Scalar maxCoeff(IndexType* row, IndexType* col) const;
template<typename IndexType>
template<typename IndexType> EIGEN_DEVICE_FUNC
typename internal::traits<Derived>::Scalar minCoeff(IndexType* index) const;
template<typename IndexType>
template<typename IndexType> EIGEN_DEVICE_FUNC
typename internal::traits<Derived>::Scalar maxCoeff(IndexType* index) const;
template<typename BinaryOp>
typename internal::result_of<BinaryOp(typename internal::traits<Derived>::Scalar)>::type
redux(const BinaryOp& func) const;
EIGEN_DEVICE_FUNC
Scalar redux(const BinaryOp& func) const;
template<typename Visitor>
EIGEN_DEVICE_FUNC
void visit(Visitor& func) const;
inline const WithFormat<Derived> format(const IOFormat& fmt) const;
/** \returns a WithFormat proxy object allowing to print a matrix the with given
* format \a fmt.
*
* See class IOFormat for some examples.
*
* \sa class IOFormat, class WithFormat
*/
inline const WithFormat<Derived> format(const IOFormat& fmt) const
{
return WithFormat<Derived>(derived(), fmt);
}
/** \returns the unique coefficient of a 1x1 expression */
EIGEN_DEVICE_FUNC
CoeffReturnType value() const
{
EIGEN_STATIC_ASSERT_SIZE_1x1(Derived)
@@ -426,23 +484,44 @@ template<typename Derived> class DenseBase
return derived().coeff(0,0);
}
bool all(void) const;
bool any(void) const;
Index count() const;
EIGEN_DEVICE_FUNC bool all() const;
EIGEN_DEVICE_FUNC bool any() const;
EIGEN_DEVICE_FUNC Index count() const;
typedef VectorwiseOp<Derived, Horizontal> RowwiseReturnType;
typedef const VectorwiseOp<const Derived, Horizontal> ConstRowwiseReturnType;
typedef VectorwiseOp<Derived, Vertical> ColwiseReturnType;
typedef const VectorwiseOp<const Derived, Vertical> ConstColwiseReturnType;
ConstRowwiseReturnType rowwise() const;
RowwiseReturnType rowwise();
ConstColwiseReturnType colwise() const;
ColwiseReturnType colwise();
/** \returns a VectorwiseOp wrapper of *this providing additional partial reduction operations
*
* Example: \include MatrixBase_rowwise.cpp
* Output: \verbinclude MatrixBase_rowwise.out
*
* \sa colwise(), class VectorwiseOp, \ref TutorialReductionsVisitorsBroadcasting
*/
//Code moved here due to a CUDA compiler bug
EIGEN_DEVICE_FUNC inline ConstRowwiseReturnType rowwise() const {
return ConstRowwiseReturnType(derived());
}
EIGEN_DEVICE_FUNC RowwiseReturnType rowwise();
static const CwiseNullaryOp<internal::scalar_random_op<Scalar>,Derived> Random(Index rows, Index cols);
static const CwiseNullaryOp<internal::scalar_random_op<Scalar>,Derived> Random(Index size);
static const CwiseNullaryOp<internal::scalar_random_op<Scalar>,Derived> Random();
/** \returns a VectorwiseOp wrapper of *this providing additional partial reduction operations
*
* Example: \include MatrixBase_colwise.cpp
* Output: \verbinclude MatrixBase_colwise.out
*
* \sa rowwise(), class VectorwiseOp, \ref TutorialReductionsVisitorsBroadcasting
*/
EIGEN_DEVICE_FUNC inline ConstColwiseReturnType colwise() const {
return ConstColwiseReturnType(derived());
}
EIGEN_DEVICE_FUNC ColwiseReturnType colwise();
typedef CwiseNullaryOp<internal::scalar_random_op<Scalar>,PlainObject> RandomReturnType;
static const RandomReturnType Random(Index rows, Index cols);
static const RandomReturnType Random(Index size);
static const RandomReturnType Random();
template<typename ThenDerived,typename ElseDerived>
const Select<Derived,ThenDerived,ElseDerived>
@@ -460,45 +539,56 @@ template<typename Derived> class DenseBase
template<int p> RealScalar lpNorm() const;
template<int RowFactor, int ColFactor>
inline const Replicate<Derived,RowFactor,ColFactor> replicate() const;
typedef Replicate<Derived,Dynamic,Dynamic> ReplicateReturnType;
inline const ReplicateReturnType replicate(Index rowFacor,Index colFactor) const;
EIGEN_DEVICE_FUNC
const Replicate<Derived,RowFactor,ColFactor> replicate() const;
/**
* \return an expression of the replication of \c *this
*
* Example: \include MatrixBase_replicate_int_int.cpp
* Output: \verbinclude MatrixBase_replicate_int_int.out
*
* \sa VectorwiseOp::replicate(), DenseBase::replicate<int,int>(), class Replicate
*/
//Code moved here due to a CUDA compiler bug
EIGEN_DEVICE_FUNC
const Replicate<Derived, Dynamic, Dynamic> replicate(Index rowFactor, Index colFactor) const
{
return Replicate<Derived, Dynamic, Dynamic>(derived(), rowFactor, colFactor);
}
typedef Reverse<Derived, BothDirections> ReverseReturnType;
typedef const Reverse<const Derived, BothDirections> ConstReverseReturnType;
ReverseReturnType reverse();
ConstReverseReturnType reverse() const;
void reverseInPlace();
EIGEN_DEVICE_FUNC ReverseReturnType reverse();
/** This is the const version of reverse(). */
//Code moved here due to a CUDA compiler bug
EIGEN_DEVICE_FUNC ConstReverseReturnType reverse() const
{
return ConstReverseReturnType(derived());
}
EIGEN_DEVICE_FUNC void reverseInPlace();
#define EIGEN_CURRENT_STORAGE_BASE_CLASS Eigen::DenseBase
#define EIGEN_DOC_BLOCK_ADDONS_NOT_INNER_PANEL
#define EIGEN_DOC_BLOCK_ADDONS_INNER_PANEL_IF(COND)
# include "../plugins/BlockMethods.h"
# ifdef EIGEN_DENSEBASE_PLUGIN
# include EIGEN_DENSEBASE_PLUGIN
# endif
#undef EIGEN_CURRENT_STORAGE_BASE_CLASS
#ifdef EIGEN2_SUPPORT
Block<Derived> corner(CornerType type, Index cRows, Index cCols);
const Block<Derived> corner(CornerType type, Index cRows, Index cCols) const;
template<int CRows, int CCols>
Block<Derived, CRows, CCols> corner(CornerType type);
template<int CRows, int CCols>
const Block<Derived, CRows, CCols> corner(CornerType type) const;
#endif // EIGEN2_SUPPORT
#undef EIGEN_DOC_BLOCK_ADDONS_NOT_INNER_PANEL
#undef EIGEN_DOC_BLOCK_ADDONS_INNER_PANEL_IF
// disable the use of evalTo for dense objects with a nice compilation error
template<typename Dest> inline void evalTo(Dest& ) const
template<typename Dest>
EIGEN_DEVICE_FUNC
inline void evalTo(Dest& ) const
{
EIGEN_STATIC_ASSERT((internal::is_same<Dest,void>::value),THE_EVAL_EVALTO_FUNCTION_SHOULD_NEVER_BE_CALLED_FOR_DENSE_OBJECTS);
}
protected:
/** Default constructor. Do nothing. */
DenseBase()
EIGEN_DEVICE_FUNC DenseBase()
{
/* Just checks for self-consistency of the flags.
* Only do it when debugging Eigen, as this borders on paranoiac and could slow compilation down
@@ -511,9 +601,9 @@ template<typename Derived> class DenseBase
}
private:
explicit DenseBase(int);
DenseBase(int,int);
template<typename OtherDerived> explicit DenseBase(const DenseBase<OtherDerived>&);
EIGEN_DEVICE_FUNC explicit DenseBase(int);
EIGEN_DEVICE_FUNC DenseBase(int,int);
template<typename OtherDerived> EIGEN_DEVICE_FUNC explicit DenseBase(const DenseBase<OtherDerived>&);
};
} // end namespace Eigen

View File

@@ -35,7 +35,6 @@ class DenseCoeffsBase<Derived,ReadOnlyAccessors> : public EigenBase<Derived>
{
public:
typedef typename internal::traits<Derived>::StorageKind StorageKind;
typedef typename internal::traits<Derived>::Index Index;
typedef typename internal::traits<Derived>::Scalar Scalar;
typedef typename internal::packet_traits<Scalar>::type PacketScalar;
@@ -61,6 +60,7 @@ class DenseCoeffsBase<Derived,ReadOnlyAccessors> : public EigenBase<Derived>
using Base::size;
using Base::derived;
EIGEN_DEVICE_FUNC
EIGEN_STRONG_INLINE Index rowIndexByOuterInner(Index outer, Index inner) const
{
return int(Derived::RowsAtCompileTime) == 1 ? 0
@@ -69,6 +69,7 @@ class DenseCoeffsBase<Derived,ReadOnlyAccessors> : public EigenBase<Derived>
: inner;
}
EIGEN_DEVICE_FUNC
EIGEN_STRONG_INLINE Index colIndexByOuterInner(Index outer, Index inner) const
{
return int(Derived::ColsAtCompileTime) == 1 ? 0
@@ -91,13 +92,15 @@ class DenseCoeffsBase<Derived,ReadOnlyAccessors> : public EigenBase<Derived>
*
* \sa operator()(Index,Index) const, coeffRef(Index,Index), coeff(Index) const
*/
EIGEN_DEVICE_FUNC
EIGEN_STRONG_INLINE CoeffReturnType coeff(Index row, Index col) const
{
eigen_internal_assert(row >= 0 && row < rows()
&& col >= 0 && col < cols());
return derived().coeff(row, col);
&& col >= 0 && col < cols());
return internal::evaluator<Derived>(derived()).coeff(row,col);
}
EIGEN_DEVICE_FUNC
EIGEN_STRONG_INLINE CoeffReturnType coeffByOuterInner(Index outer, Index inner) const
{
return coeff(rowIndexByOuterInner(outer, inner),
@@ -108,11 +111,12 @@ class DenseCoeffsBase<Derived,ReadOnlyAccessors> : public EigenBase<Derived>
*
* \sa operator()(Index,Index), operator[](Index)
*/
EIGEN_DEVICE_FUNC
EIGEN_STRONG_INLINE CoeffReturnType operator()(Index row, Index col) const
{
eigen_assert(row >= 0 && row < rows()
&& col >= 0 && col < cols());
return derived().coeff(row, col);
return coeff(row, col);
}
/** Short version: don't use this function, use
@@ -130,11 +134,14 @@ class DenseCoeffsBase<Derived,ReadOnlyAccessors> : public EigenBase<Derived>
* \sa operator[](Index) const, coeffRef(Index), coeff(Index,Index) const
*/
EIGEN_DEVICE_FUNC
EIGEN_STRONG_INLINE CoeffReturnType
coeff(Index index) const
{
EIGEN_STATIC_ASSERT(internal::evaluator<Derived>::Flags & LinearAccessBit,
THIS_COEFFICIENT_ACCESSOR_TAKING_ONE_ACCESS_IS_ONLY_FOR_EXPRESSIONS_ALLOWING_LINEAR_ACCESS)
eigen_internal_assert(index >= 0 && index < size());
return derived().coeff(index);
return internal::evaluator<Derived>(derived()).coeff(index);
}
@@ -146,15 +153,14 @@ class DenseCoeffsBase<Derived,ReadOnlyAccessors> : public EigenBase<Derived>
* z() const, w() const
*/
EIGEN_DEVICE_FUNC
EIGEN_STRONG_INLINE CoeffReturnType
operator[](Index index) const
{
#ifndef EIGEN2_SUPPORT
EIGEN_STATIC_ASSERT(Derived::IsVectorAtCompileTime,
THE_BRACKET_OPERATOR_IS_ONLY_FOR_VECTORS__USE_THE_PARENTHESIS_OPERATOR_INSTEAD)
#endif
eigen_assert(index >= 0 && index < size());
return derived().coeff(index);
return coeff(index);
}
/** \returns the coefficient at given index.
@@ -167,32 +173,49 @@ class DenseCoeffsBase<Derived,ReadOnlyAccessors> : public EigenBase<Derived>
* z() const, w() const
*/
EIGEN_DEVICE_FUNC
EIGEN_STRONG_INLINE CoeffReturnType
operator()(Index index) const
{
eigen_assert(index >= 0 && index < size());
return derived().coeff(index);
return coeff(index);
}
/** equivalent to operator[](0). */
EIGEN_DEVICE_FUNC
EIGEN_STRONG_INLINE CoeffReturnType
x() const { return (*this)[0]; }
/** equivalent to operator[](1). */
EIGEN_DEVICE_FUNC
EIGEN_STRONG_INLINE CoeffReturnType
y() const { return (*this)[1]; }
y() const
{
EIGEN_STATIC_ASSERT(Derived::SizeAtCompileTime==-1 || Derived::SizeAtCompileTime>=2, OUT_OF_RANGE_ACCESS);
return (*this)[1];
}
/** equivalent to operator[](2). */
EIGEN_DEVICE_FUNC
EIGEN_STRONG_INLINE CoeffReturnType
z() const { return (*this)[2]; }
z() const
{
EIGEN_STATIC_ASSERT(Derived::SizeAtCompileTime==-1 || Derived::SizeAtCompileTime>=3, OUT_OF_RANGE_ACCESS);
return (*this)[2];
}
/** equivalent to operator[](3). */
EIGEN_DEVICE_FUNC
EIGEN_STRONG_INLINE CoeffReturnType
w() const { return (*this)[3]; }
w() const
{
EIGEN_STATIC_ASSERT(Derived::SizeAtCompileTime==-1 || Derived::SizeAtCompileTime>=4, OUT_OF_RANGE_ACCESS);
return (*this)[3];
}
/** \internal
* \returns the packet of coefficients starting at the given row and column. It is your responsibility
@@ -207,9 +230,9 @@ class DenseCoeffsBase<Derived,ReadOnlyAccessors> : public EigenBase<Derived>
template<int LoadMode>
EIGEN_STRONG_INLINE PacketReturnType packet(Index row, Index col) const
{
eigen_internal_assert(row >= 0 && row < rows()
&& col >= 0 && col < cols());
return derived().template packet<LoadMode>(row,col);
typedef typename internal::packet_traits<Scalar>::type DefaultPacketType;
eigen_internal_assert(row >= 0 && row < rows() && col >= 0 && col < cols());
return internal::evaluator<Derived>(derived()).template packet<LoadMode,DefaultPacketType>(row,col);
}
@@ -234,8 +257,11 @@ class DenseCoeffsBase<Derived,ReadOnlyAccessors> : public EigenBase<Derived>
template<int LoadMode>
EIGEN_STRONG_INLINE PacketReturnType packet(Index index) const
{
EIGEN_STATIC_ASSERT(internal::evaluator<Derived>::Flags & LinearAccessBit,
THIS_COEFFICIENT_ACCESSOR_TAKING_ONE_ACCESS_IS_ONLY_FOR_EXPRESSIONS_ALLOWING_LINEAR_ACCESS)
typedef typename internal::packet_traits<Scalar>::type DefaultPacketType;
eigen_internal_assert(index >= 0 && index < size());
return derived().template packet<LoadMode>(index);
return internal::evaluator<Derived>(derived()).template packet<LoadMode,DefaultPacketType>(index);
}
protected:
@@ -278,7 +304,6 @@ class DenseCoeffsBase<Derived, WriteAccessors> : public DenseCoeffsBase<Derived,
typedef DenseCoeffsBase<Derived, ReadOnlyAccessors> Base;
typedef typename internal::traits<Derived>::StorageKind StorageKind;
typedef typename internal::traits<Derived>::Index Index;
typedef typename internal::traits<Derived>::Scalar Scalar;
typedef typename internal::packet_traits<Scalar>::type PacketScalar;
typedef typename NumTraits<Scalar>::Real RealScalar;
@@ -311,13 +336,15 @@ class DenseCoeffsBase<Derived, WriteAccessors> : public DenseCoeffsBase<Derived,
*
* \sa operator()(Index,Index), coeff(Index, Index) const, coeffRef(Index)
*/
EIGEN_DEVICE_FUNC
EIGEN_STRONG_INLINE Scalar& coeffRef(Index row, Index col)
{
eigen_internal_assert(row >= 0 && row < rows()
&& col >= 0 && col < cols());
return derived().coeffRef(row, col);
&& col >= 0 && col < cols());
return internal::evaluator<Derived>(derived()).coeffRef(row,col);
}
EIGEN_DEVICE_FUNC
EIGEN_STRONG_INLINE Scalar&
coeffRefByOuterInner(Index outer, Index inner)
{
@@ -330,12 +357,13 @@ class DenseCoeffsBase<Derived, WriteAccessors> : public DenseCoeffsBase<Derived,
* \sa operator[](Index)
*/
EIGEN_DEVICE_FUNC
EIGEN_STRONG_INLINE Scalar&
operator()(Index row, Index col)
{
eigen_assert(row >= 0 && row < rows()
&& col >= 0 && col < cols());
return derived().coeffRef(row, col);
return coeffRef(row, col);
}
@@ -354,11 +382,14 @@ class DenseCoeffsBase<Derived, WriteAccessors> : public DenseCoeffsBase<Derived,
* \sa operator[](Index), coeff(Index) const, coeffRef(Index,Index)
*/
EIGEN_DEVICE_FUNC
EIGEN_STRONG_INLINE Scalar&
coeffRef(Index index)
{
EIGEN_STATIC_ASSERT(internal::evaluator<Derived>::Flags & LinearAccessBit,
THIS_COEFFICIENT_ACCESSOR_TAKING_ONE_ACCESS_IS_ONLY_FOR_EXPRESSIONS_ALLOWING_LINEAR_ACCESS)
eigen_internal_assert(index >= 0 && index < size());
return derived().coeffRef(index);
return internal::evaluator<Derived>(derived()).coeffRef(index);
}
/** \returns a reference to the coefficient at given index.
@@ -368,15 +399,14 @@ class DenseCoeffsBase<Derived, WriteAccessors> : public DenseCoeffsBase<Derived,
* \sa operator[](Index) const, operator()(Index,Index), x(), y(), z(), w()
*/
EIGEN_DEVICE_FUNC
EIGEN_STRONG_INLINE Scalar&
operator[](Index index)
{
#ifndef EIGEN2_SUPPORT
EIGEN_STATIC_ASSERT(Derived::IsVectorAtCompileTime,
THE_BRACKET_OPERATOR_IS_ONLY_FOR_VECTORS__USE_THE_PARENTHESIS_OPERATOR_INSTEAD)
#endif
eigen_assert(index >= 0 && index < size());
return derived().coeffRef(index);
return coeffRef(index);
}
/** \returns a reference to the coefficient at given index.
@@ -388,167 +418,49 @@ class DenseCoeffsBase<Derived, WriteAccessors> : public DenseCoeffsBase<Derived,
* \sa operator[](Index) const, operator()(Index,Index), x(), y(), z(), w()
*/
EIGEN_DEVICE_FUNC
EIGEN_STRONG_INLINE Scalar&
operator()(Index index)
{
eigen_assert(index >= 0 && index < size());
return derived().coeffRef(index);
return coeffRef(index);
}
/** equivalent to operator[](0). */
EIGEN_DEVICE_FUNC
EIGEN_STRONG_INLINE Scalar&
x() { return (*this)[0]; }
/** equivalent to operator[](1). */
EIGEN_DEVICE_FUNC
EIGEN_STRONG_INLINE Scalar&
y() { return (*this)[1]; }
y()
{
EIGEN_STATIC_ASSERT(Derived::SizeAtCompileTime==-1 || Derived::SizeAtCompileTime>=2, OUT_OF_RANGE_ACCESS);
return (*this)[1];
}
/** equivalent to operator[](2). */
EIGEN_DEVICE_FUNC
EIGEN_STRONG_INLINE Scalar&
z() { return (*this)[2]; }
z()
{
EIGEN_STATIC_ASSERT(Derived::SizeAtCompileTime==-1 || Derived::SizeAtCompileTime>=3, OUT_OF_RANGE_ACCESS);
return (*this)[2];
}
/** equivalent to operator[](3). */
EIGEN_DEVICE_FUNC
EIGEN_STRONG_INLINE Scalar&
w() { return (*this)[3]; }
/** \internal
* Stores the given packet of coefficients, at the given row and column of this expression. It is your responsibility
* to ensure that a packet really starts there. This method is only available on expressions having the
* PacketAccessBit.
*
* The \a LoadMode parameter may have the value \a #Aligned or \a #Unaligned. Its effect is to select
* the appropriate vectorization instruction. Aligned access is faster, but is only possible for packets
* starting at an address which is a multiple of the packet size.
*/
template<int StoreMode>
EIGEN_STRONG_INLINE void writePacket
(Index row, Index col, const typename internal::packet_traits<Scalar>::type& val)
w()
{
eigen_internal_assert(row >= 0 && row < rows()
&& col >= 0 && col < cols());
derived().template writePacket<StoreMode>(row,col,val);
EIGEN_STATIC_ASSERT(Derived::SizeAtCompileTime==-1 || Derived::SizeAtCompileTime>=4, OUT_OF_RANGE_ACCESS);
return (*this)[3];
}
/** \internal */
template<int StoreMode>
EIGEN_STRONG_INLINE void writePacketByOuterInner
(Index outer, Index inner, const typename internal::packet_traits<Scalar>::type& val)
{
writePacket<StoreMode>(rowIndexByOuterInner(outer, inner),
colIndexByOuterInner(outer, inner),
val);
}
/** \internal
* Stores the given packet of coefficients, at the given index in this expression. It is your responsibility
* to ensure that a packet really starts there. This method is only available on expressions having the
* PacketAccessBit and the LinearAccessBit.
*
* The \a LoadMode parameter may have the value \a Aligned or \a Unaligned. Its effect is to select
* the appropriate vectorization instruction. Aligned access is faster, but is only possible for packets
* starting at an address which is a multiple of the packet size.
*/
template<int StoreMode>
EIGEN_STRONG_INLINE void writePacket
(Index index, const typename internal::packet_traits<Scalar>::type& val)
{
eigen_internal_assert(index >= 0 && index < size());
derived().template writePacket<StoreMode>(index,val);
}
#ifndef EIGEN_PARSED_BY_DOXYGEN
/** \internal Copies the coefficient at position (row,col) of other into *this.
*
* This method is overridden in SwapWrapper, allowing swap() assignments to share 99% of their code
* with usual assignments.
*
* Outside of this internal usage, this method has probably no usefulness. It is hidden in the public API dox.
*/
template<typename OtherDerived>
EIGEN_STRONG_INLINE void copyCoeff(Index row, Index col, const DenseBase<OtherDerived>& other)
{
eigen_internal_assert(row >= 0 && row < rows()
&& col >= 0 && col < cols());
derived().coeffRef(row, col) = other.derived().coeff(row, col);
}
/** \internal Copies the coefficient at the given index of other into *this.
*
* This method is overridden in SwapWrapper, allowing swap() assignments to share 99% of their code
* with usual assignments.
*
* Outside of this internal usage, this method has probably no usefulness. It is hidden in the public API dox.
*/
template<typename OtherDerived>
EIGEN_STRONG_INLINE void copyCoeff(Index index, const DenseBase<OtherDerived>& other)
{
eigen_internal_assert(index >= 0 && index < size());
derived().coeffRef(index) = other.derived().coeff(index);
}
template<typename OtherDerived>
EIGEN_STRONG_INLINE void copyCoeffByOuterInner(Index outer, Index inner, const DenseBase<OtherDerived>& other)
{
const Index row = rowIndexByOuterInner(outer,inner);
const Index col = colIndexByOuterInner(outer,inner);
// derived() is important here: copyCoeff() may be reimplemented in Derived!
derived().copyCoeff(row, col, other);
}
/** \internal Copies the packet at position (row,col) of other into *this.
*
* This method is overridden in SwapWrapper, allowing swap() assignments to share 99% of their code
* with usual assignments.
*
* Outside of this internal usage, this method has probably no usefulness. It is hidden in the public API dox.
*/
template<typename OtherDerived, int StoreMode, int LoadMode>
EIGEN_STRONG_INLINE void copyPacket(Index row, Index col, const DenseBase<OtherDerived>& other)
{
eigen_internal_assert(row >= 0 && row < rows()
&& col >= 0 && col < cols());
derived().template writePacket<StoreMode>(row, col,
other.derived().template packet<LoadMode>(row, col));
}
/** \internal Copies the packet at the given index of other into *this.
*
* This method is overridden in SwapWrapper, allowing swap() assignments to share 99% of their code
* with usual assignments.
*
* Outside of this internal usage, this method has probably no usefulness. It is hidden in the public API dox.
*/
template<typename OtherDerived, int StoreMode, int LoadMode>
EIGEN_STRONG_INLINE void copyPacket(Index index, const DenseBase<OtherDerived>& other)
{
eigen_internal_assert(index >= 0 && index < size());
derived().template writePacket<StoreMode>(index,
other.derived().template packet<LoadMode>(index));
}
/** \internal */
template<typename OtherDerived, int StoreMode, int LoadMode>
EIGEN_STRONG_INLINE void copyPacketByOuterInner(Index outer, Index inner, const DenseBase<OtherDerived>& other)
{
const Index row = rowIndexByOuterInner(outer,inner);
const Index col = colIndexByOuterInner(outer,inner);
// derived() is important here: copyCoeff() may be reimplemented in Derived!
derived().template copyPacket< OtherDerived, StoreMode, LoadMode>(row, col, other);
}
#endif
};
/** \brief Base class providing direct read-only coefficient access to matrices and arrays.
@@ -560,7 +472,7 @@ class DenseCoeffsBase<Derived, WriteAccessors> : public DenseCoeffsBase<Derived,
* inherits DenseCoeffsBase<Derived, ReadOnlyAccessors> which defines functions to access entries read-only using
* \c operator() .
*
* \sa \ref TopicClassHierarchy
* \sa \blank \ref TopicClassHierarchy
*/
template<typename Derived>
class DenseCoeffsBase<Derived, DirectAccessors> : public DenseCoeffsBase<Derived, ReadOnlyAccessors>
@@ -568,7 +480,6 @@ class DenseCoeffsBase<Derived, DirectAccessors> : public DenseCoeffsBase<Derived
public:
typedef DenseCoeffsBase<Derived, ReadOnlyAccessors> Base;
typedef typename internal::traits<Derived>::Index Index;
typedef typename internal::traits<Derived>::Scalar Scalar;
typedef typename NumTraits<Scalar>::Real RealScalar;
@@ -581,6 +492,7 @@ class DenseCoeffsBase<Derived, DirectAccessors> : public DenseCoeffsBase<Derived
*
* \sa outerStride(), rowStride(), colStride()
*/
EIGEN_DEVICE_FUNC
inline Index innerStride() const
{
return derived().innerStride();
@@ -591,6 +503,7 @@ class DenseCoeffsBase<Derived, DirectAccessors> : public DenseCoeffsBase<Derived
*
* \sa innerStride(), rowStride(), colStride()
*/
EIGEN_DEVICE_FUNC
inline Index outerStride() const
{
return derived().outerStride();
@@ -606,6 +519,7 @@ class DenseCoeffsBase<Derived, DirectAccessors> : public DenseCoeffsBase<Derived
*
* \sa innerStride(), outerStride(), colStride()
*/
EIGEN_DEVICE_FUNC
inline Index rowStride() const
{
return Derived::IsRowMajor ? outerStride() : innerStride();
@@ -615,6 +529,7 @@ class DenseCoeffsBase<Derived, DirectAccessors> : public DenseCoeffsBase<Derived
*
* \sa innerStride(), outerStride(), rowStride()
*/
EIGEN_DEVICE_FUNC
inline Index colStride() const
{
return Derived::IsRowMajor ? innerStride() : outerStride();
@@ -630,7 +545,7 @@ class DenseCoeffsBase<Derived, DirectAccessors> : public DenseCoeffsBase<Derived
* inherits DenseCoeffsBase<Derived, WriteAccessors> which defines functions to access entries read/write using
* \c operator().
*
* \sa \ref TopicClassHierarchy
* \sa \blank \ref TopicClassHierarchy
*/
template<typename Derived>
class DenseCoeffsBase<Derived, DirectWriteAccessors>
@@ -639,7 +554,6 @@ class DenseCoeffsBase<Derived, DirectWriteAccessors>
public:
typedef DenseCoeffsBase<Derived, WriteAccessors> Base;
typedef typename internal::traits<Derived>::Index Index;
typedef typename internal::traits<Derived>::Scalar Scalar;
typedef typename NumTraits<Scalar>::Real RealScalar;
@@ -652,6 +566,7 @@ class DenseCoeffsBase<Derived, DirectWriteAccessors>
*
* \sa outerStride(), rowStride(), colStride()
*/
EIGEN_DEVICE_FUNC
inline Index innerStride() const
{
return derived().innerStride();
@@ -662,6 +577,7 @@ class DenseCoeffsBase<Derived, DirectWriteAccessors>
*
* \sa innerStride(), rowStride(), colStride()
*/
EIGEN_DEVICE_FUNC
inline Index outerStride() const
{
return derived().outerStride();
@@ -677,6 +593,7 @@ class DenseCoeffsBase<Derived, DirectWriteAccessors>
*
* \sa innerStride(), outerStride(), colStride()
*/
EIGEN_DEVICE_FUNC
inline Index rowStride() const
{
return Derived::IsRowMajor ? outerStride() : innerStride();
@@ -686,6 +603,7 @@ class DenseCoeffsBase<Derived, DirectWriteAccessors>
*
* \sa innerStride(), outerStride(), rowStride()
*/
EIGEN_DEVICE_FUNC
inline Index colStride() const
{
return Derived::IsRowMajor ? innerStride() : outerStride();
@@ -694,33 +612,42 @@ class DenseCoeffsBase<Derived, DirectWriteAccessors>
namespace internal {
template<typename Derived, bool JustReturnZero>
template<int Alignment, typename Derived, bool JustReturnZero>
struct first_aligned_impl
{
static inline typename Derived::Index run(const Derived&)
static inline Index run(const Derived&)
{ return 0; }
};
template<typename Derived>
struct first_aligned_impl<Derived, false>
template<int Alignment, typename Derived>
struct first_aligned_impl<Alignment, Derived, false>
{
static inline typename Derived::Index run(const Derived& m)
static inline Index run(const Derived& m)
{
return internal::first_aligned(&m.const_cast_derived().coeffRef(0,0), m.size());
return internal::first_aligned<Alignment>(m.data(), m.size());
}
};
/** \internal \returns the index of the first element of the array that is well aligned for vectorization.
/** \internal \returns the index of the first element of the array stored by \a m that is properly aligned with respect to \a Alignment for vectorization.
*
* \tparam Alignment requested alignment in Bytes.
*
* There is also the variant first_aligned(const Scalar*, Integer) defined in Memory.h. See it for more
* documentation.
*/
template<typename Derived>
static inline typename Derived::Index first_aligned(const Derived& m)
template<int Alignment, typename Derived>
static inline Index first_aligned(const DenseBase<Derived>& m)
{
return first_aligned_impl
<Derived, (Derived::Flags & AlignedBit) || !(Derived::Flags & DirectAccessBit)>
::run(m);
enum { ReturnZero = (int(evaluator<Derived>::Alignment) >= Alignment) || !(Derived::Flags & DirectAccessBit) };
return first_aligned_impl<Alignment, Derived, ReturnZero>::run(m.derived());
}
template<typename Derived>
static inline Index first_default_aligned(const DenseBase<Derived>& m)
{
typedef typename Derived::Scalar Scalar;
typedef typename packet_traits<Scalar>::type DefaultPacketType;
return internal::first_aligned<int(unpacket_traits<DefaultPacketType>::alignment),Derived>(m);
}
template<typename Derived, bool HasDirectAccess = has_direct_access<Derived>::ret>

View File

@@ -3,7 +3,7 @@
//
// Copyright (C) 2008 Gael Guennebaud <gael.guennebaud@inria.fr>
// Copyright (C) 2006-2009 Benoit Jacob <jacob.benoit.1@gmail.com>
// Copyright (C) 2010 Hauke Heibel <hauke.heibel@gmail.com>
// Copyright (C) 2010-2013 Hauke Heibel <hauke.heibel@gmail.com>
//
// This Source Code Form is subject to the terms of the Mozilla
// Public License v. 2.0. If a copy of the MPL was not distributed
@@ -13,9 +13,9 @@
#define EIGEN_MATRIXSTORAGE_H
#ifdef EIGEN_DENSE_STORAGE_CTOR_PLUGIN
#define EIGEN_INTERNAL_DENSE_STORAGE_CTOR_PLUGIN EIGEN_DENSE_STORAGE_CTOR_PLUGIN;
#define EIGEN_INTERNAL_DENSE_STORAGE_CTOR_PLUGIN(X) X; EIGEN_DENSE_STORAGE_CTOR_PLUGIN;
#else
#define EIGEN_INTERNAL_DENSE_STORAGE_CTOR_PLUGIN
#define EIGEN_INTERNAL_DENSE_STORAGE_CTOR_PLUGIN(X)
#endif
namespace Eigen {
@@ -24,7 +24,9 @@ namespace internal {
struct constructor_without_unaligned_array_assert {};
template<typename T, int Size> void check_static_allocation_size()
template<typename T, int Size>
EIGEN_DEVICE_FUNC
void check_static_allocation_size()
{
// if EIGEN_STACK_ALLOCATION_LIMIT is defined to 0, then no limit
#if EIGEN_STACK_ALLOCATION_LIMIT
@@ -38,18 +40,19 @@ template<typename T, int Size> void check_static_allocation_size()
*/
template <typename T, int Size, int MatrixOrArrayOptions,
int Alignment = (MatrixOrArrayOptions&DontAlign) ? 0
: (((Size*sizeof(T))%16)==0) ? 16
: 0 >
: compute_default_alignment<T,Size>::value >
struct plain_array
{
T array[Size];
plain_array()
EIGEN_DEVICE_FUNC
plain_array()
{
check_static_allocation_size<T,Size>();
}
plain_array(constructor_without_unaligned_array_assert)
EIGEN_DEVICE_FUNC
plain_array(constructor_without_unaligned_array_assert)
{
check_static_allocation_size<T,Size>();
}
@@ -64,29 +67,88 @@ struct plain_array
template<typename PtrType>
EIGEN_ALWAYS_INLINE PtrType eigen_unaligned_array_assert_workaround_gcc47(PtrType array) { return array; }
#define EIGEN_MAKE_UNALIGNED_ARRAY_ASSERT(sizemask) \
eigen_assert((reinterpret_cast<size_t>(eigen_unaligned_array_assert_workaround_gcc47(array)) & sizemask) == 0 \
eigen_assert((internal::UIntPtr(eigen_unaligned_array_assert_workaround_gcc47(array)) & (sizemask)) == 0 \
&& "this assertion is explained here: " \
"http://eigen.tuxfamily.org/dox-devel/group__TopicUnalignedArrayAssert.html" \
" **** READ THIS WEB PAGE !!! ****");
#else
#define EIGEN_MAKE_UNALIGNED_ARRAY_ASSERT(sizemask) \
eigen_assert((reinterpret_cast<size_t>(array) & sizemask) == 0 \
eigen_assert((internal::UIntPtr(array) & (sizemask)) == 0 \
&& "this assertion is explained here: " \
"http://eigen.tuxfamily.org/dox-devel/group__TopicUnalignedArrayAssert.html" \
" **** READ THIS WEB PAGE !!! ****");
#endif
template <typename T, int Size, int MatrixOrArrayOptions>
struct plain_array<T, Size, MatrixOrArrayOptions, 16>
struct plain_array<T, Size, MatrixOrArrayOptions, 8>
{
EIGEN_USER_ALIGN16 T array[Size];
EIGEN_ALIGN_TO_BOUNDARY(8) T array[Size];
EIGEN_DEVICE_FUNC
plain_array()
{
EIGEN_MAKE_UNALIGNED_ARRAY_ASSERT(0xf);
{
EIGEN_MAKE_UNALIGNED_ARRAY_ASSERT(7);
check_static_allocation_size<T,Size>();
}
EIGEN_DEVICE_FUNC
plain_array(constructor_without_unaligned_array_assert)
{
check_static_allocation_size<T,Size>();
}
};
template <typename T, int Size, int MatrixOrArrayOptions>
struct plain_array<T, Size, MatrixOrArrayOptions, 16>
{
EIGEN_ALIGN_TO_BOUNDARY(16) T array[Size];
EIGEN_DEVICE_FUNC
plain_array()
{
EIGEN_MAKE_UNALIGNED_ARRAY_ASSERT(15);
check_static_allocation_size<T,Size>();
}
EIGEN_DEVICE_FUNC
plain_array(constructor_without_unaligned_array_assert)
{
check_static_allocation_size<T,Size>();
}
};
template <typename T, int Size, int MatrixOrArrayOptions>
struct plain_array<T, Size, MatrixOrArrayOptions, 32>
{
EIGEN_ALIGN_TO_BOUNDARY(32) T array[Size];
EIGEN_DEVICE_FUNC
plain_array()
{
EIGEN_MAKE_UNALIGNED_ARRAY_ASSERT(31);
check_static_allocation_size<T,Size>();
}
EIGEN_DEVICE_FUNC
plain_array(constructor_without_unaligned_array_assert)
{
check_static_allocation_size<T,Size>();
}
};
template <typename T, int Size, int MatrixOrArrayOptions>
struct plain_array<T, Size, MatrixOrArrayOptions, 64>
{
EIGEN_ALIGN_TO_BOUNDARY(64) T array[Size];
EIGEN_DEVICE_FUNC
plain_array()
{
EIGEN_MAKE_UNALIGNED_ARRAY_ASSERT(63);
check_static_allocation_size<T,Size>();
}
EIGEN_DEVICE_FUNC
plain_array(constructor_without_unaligned_array_assert)
{
check_static_allocation_size<T,Size>();
@@ -96,9 +158,9 @@ struct plain_array<T, Size, MatrixOrArrayOptions, 16>
template <typename T, int MatrixOrArrayOptions, int Alignment>
struct plain_array<T, 0, MatrixOrArrayOptions, Alignment>
{
EIGEN_USER_ALIGN16 T array[1];
plain_array() {}
plain_array(constructor_without_unaligned_array_assert) {}
T array[1];
EIGEN_DEVICE_FUNC plain_array() {}
EIGEN_DEVICE_FUNC plain_array(constructor_without_unaligned_array_assert) {}
};
} // end namespace internal
@@ -122,41 +184,54 @@ template<typename T, int Size, int _Rows, int _Cols, int _Options> class DenseSt
{
internal::plain_array<T,Size,_Options> m_data;
public:
DenseStorage() {}
DenseStorage(internal::constructor_without_unaligned_array_assert)
: m_data(internal::constructor_without_unaligned_array_assert()) {}
DenseStorage(const DenseStorage& other) : m_data(other.m_data) {}
DenseStorage& operator=(const DenseStorage& other)
{
if (this != &other) m_data = other.m_data;
return *this;
EIGEN_DEVICE_FUNC DenseStorage() {
EIGEN_INTERNAL_DENSE_STORAGE_CTOR_PLUGIN(Index size = Size)
}
DenseStorage(DenseIndex,DenseIndex,DenseIndex) {}
void swap(DenseStorage& other) { std::swap(m_data,other.m_data); }
static DenseIndex rows(void) {return _Rows;}
static DenseIndex cols(void) {return _Cols;}
void conservativeResize(DenseIndex,DenseIndex,DenseIndex) {}
void resize(DenseIndex,DenseIndex,DenseIndex) {}
const T *data() const { return m_data.array; }
T *data() { return m_data.array; }
EIGEN_DEVICE_FUNC
explicit DenseStorage(internal::constructor_without_unaligned_array_assert)
: m_data(internal::constructor_without_unaligned_array_assert()) {}
EIGEN_DEVICE_FUNC
DenseStorage(const DenseStorage& other) : m_data(other.m_data) {
EIGEN_INTERNAL_DENSE_STORAGE_CTOR_PLUGIN(Index size = Size)
}
EIGEN_DEVICE_FUNC
DenseStorage& operator=(const DenseStorage& other)
{
if (this != &other) m_data = other.m_data;
return *this;
}
EIGEN_DEVICE_FUNC DenseStorage(Index size, Index rows, Index cols) {
EIGEN_INTERNAL_DENSE_STORAGE_CTOR_PLUGIN({})
eigen_internal_assert(size==rows*cols && rows==_Rows && cols==_Cols);
EIGEN_UNUSED_VARIABLE(size);
EIGEN_UNUSED_VARIABLE(rows);
EIGEN_UNUSED_VARIABLE(cols);
}
EIGEN_DEVICE_FUNC void swap(DenseStorage& other) { std::swap(m_data,other.m_data); }
EIGEN_DEVICE_FUNC static Index rows(void) {return _Rows;}
EIGEN_DEVICE_FUNC static Index cols(void) {return _Cols;}
EIGEN_DEVICE_FUNC void conservativeResize(Index,Index,Index) {}
EIGEN_DEVICE_FUNC void resize(Index,Index,Index) {}
EIGEN_DEVICE_FUNC const T *data() const { return m_data.array; }
EIGEN_DEVICE_FUNC T *data() { return m_data.array; }
};
// null matrix
template<typename T, int _Rows, int _Cols, int _Options> class DenseStorage<T, 0, _Rows, _Cols, _Options>
{
public:
DenseStorage() {}
DenseStorage(internal::constructor_without_unaligned_array_assert) {}
DenseStorage(const DenseStorage&) {}
DenseStorage& operator=(const DenseStorage&) { return *this; }
DenseStorage(DenseIndex,DenseIndex,DenseIndex) {}
void swap(DenseStorage& ) {}
static DenseIndex rows(void) {return _Rows;}
static DenseIndex cols(void) {return _Cols;}
void conservativeResize(DenseIndex,DenseIndex,DenseIndex) {}
void resize(DenseIndex,DenseIndex,DenseIndex) {}
const T *data() const { return 0; }
T *data() { return 0; }
EIGEN_DEVICE_FUNC DenseStorage() {}
EIGEN_DEVICE_FUNC explicit DenseStorage(internal::constructor_without_unaligned_array_assert) {}
EIGEN_DEVICE_FUNC DenseStorage(const DenseStorage&) {}
EIGEN_DEVICE_FUNC DenseStorage& operator=(const DenseStorage&) { return *this; }
EIGEN_DEVICE_FUNC DenseStorage(Index,Index,Index) {}
EIGEN_DEVICE_FUNC void swap(DenseStorage& ) {}
EIGEN_DEVICE_FUNC static Index rows(void) {return _Rows;}
EIGEN_DEVICE_FUNC static Index cols(void) {return _Cols;}
EIGEN_DEVICE_FUNC void conservativeResize(Index,Index,Index) {}
EIGEN_DEVICE_FUNC void resize(Index,Index,Index) {}
EIGEN_DEVICE_FUNC const T *data() const { return 0; }
EIGEN_DEVICE_FUNC T *data() { return 0; }
};
// more specializations for null matrices; these are necessary to resolve ambiguities
@@ -173,74 +248,74 @@ template<typename T, int _Cols, int _Options> class DenseStorage<T, 0, Dynamic,
template<typename T, int Size, int _Options> class DenseStorage<T, Size, Dynamic, Dynamic, _Options>
{
internal::plain_array<T,Size,_Options> m_data;
DenseIndex m_rows;
DenseIndex m_cols;
Index m_rows;
Index m_cols;
public:
DenseStorage() : m_rows(0), m_cols(0) {}
DenseStorage(internal::constructor_without_unaligned_array_assert)
EIGEN_DEVICE_FUNC DenseStorage() : m_rows(0), m_cols(0) {}
EIGEN_DEVICE_FUNC explicit DenseStorage(internal::constructor_without_unaligned_array_assert)
: m_data(internal::constructor_without_unaligned_array_assert()), m_rows(0), m_cols(0) {}
DenseStorage(const DenseStorage& other) : m_data(other.m_data), m_rows(other.m_rows), m_cols(other.m_cols) {}
DenseStorage& operator=(const DenseStorage& other)
{
EIGEN_DEVICE_FUNC DenseStorage(const DenseStorage& other) : m_data(other.m_data), m_rows(other.m_rows), m_cols(other.m_cols) {}
EIGEN_DEVICE_FUNC DenseStorage& operator=(const DenseStorage& other)
{
if (this != &other)
{
m_data = other.m_data;
m_rows = other.m_rows;
m_cols = other.m_cols;
}
return *this;
return *this;
}
DenseStorage(DenseIndex, DenseIndex nbRows, DenseIndex nbCols) : m_rows(nbRows), m_cols(nbCols) {}
void swap(DenseStorage& other)
EIGEN_DEVICE_FUNC DenseStorage(Index, Index rows, Index cols) : m_rows(rows), m_cols(cols) {}
EIGEN_DEVICE_FUNC void swap(DenseStorage& other)
{ std::swap(m_data,other.m_data); std::swap(m_rows,other.m_rows); std::swap(m_cols,other.m_cols); }
DenseIndex rows() const {return m_rows;}
DenseIndex cols() const {return m_cols;}
void conservativeResize(DenseIndex, DenseIndex nbRows, DenseIndex nbCols) { m_rows = nbRows; m_cols = nbCols; }
void resize(DenseIndex, DenseIndex nbRows, DenseIndex nbCols) { m_rows = nbRows; m_cols = nbCols; }
const T *data() const { return m_data.array; }
T *data() { return m_data.array; }
EIGEN_DEVICE_FUNC Index rows() const {return m_rows;}
EIGEN_DEVICE_FUNC Index cols() const {return m_cols;}
EIGEN_DEVICE_FUNC void conservativeResize(Index, Index rows, Index cols) { m_rows = rows; m_cols = cols; }
EIGEN_DEVICE_FUNC void resize(Index, Index rows, Index cols) { m_rows = rows; m_cols = cols; }
EIGEN_DEVICE_FUNC const T *data() const { return m_data.array; }
EIGEN_DEVICE_FUNC T *data() { return m_data.array; }
};
// dynamic-size matrix with fixed-size storage and fixed width
template<typename T, int Size, int _Cols, int _Options> class DenseStorage<T, Size, Dynamic, _Cols, _Options>
{
internal::plain_array<T,Size,_Options> m_data;
DenseIndex m_rows;
Index m_rows;
public:
DenseStorage() : m_rows(0) {}
DenseStorage(internal::constructor_without_unaligned_array_assert)
EIGEN_DEVICE_FUNC DenseStorage() : m_rows(0) {}
EIGEN_DEVICE_FUNC explicit DenseStorage(internal::constructor_without_unaligned_array_assert)
: m_data(internal::constructor_without_unaligned_array_assert()), m_rows(0) {}
DenseStorage(const DenseStorage& other) : m_data(other.m_data), m_rows(other.m_rows) {}
DenseStorage& operator=(const DenseStorage& other)
EIGEN_DEVICE_FUNC DenseStorage(const DenseStorage& other) : m_data(other.m_data), m_rows(other.m_rows) {}
EIGEN_DEVICE_FUNC DenseStorage& operator=(const DenseStorage& other)
{
if (this != &other)
{
m_data = other.m_data;
m_rows = other.m_rows;
}
return *this;
return *this;
}
DenseStorage(DenseIndex, DenseIndex nbRows, DenseIndex) : m_rows(nbRows) {}
void swap(DenseStorage& other) { std::swap(m_data,other.m_data); std::swap(m_rows,other.m_rows); }
DenseIndex rows(void) const {return m_rows;}
DenseIndex cols(void) const {return _Cols;}
void conservativeResize(DenseIndex, DenseIndex nbRows, DenseIndex) { m_rows = nbRows; }
void resize(DenseIndex, DenseIndex nbRows, DenseIndex) { m_rows = nbRows; }
const T *data() const { return m_data.array; }
T *data() { return m_data.array; }
EIGEN_DEVICE_FUNC DenseStorage(Index, Index rows, Index) : m_rows(rows) {}
EIGEN_DEVICE_FUNC void swap(DenseStorage& other) { std::swap(m_data,other.m_data); std::swap(m_rows,other.m_rows); }
EIGEN_DEVICE_FUNC Index rows(void) const {return m_rows;}
EIGEN_DEVICE_FUNC Index cols(void) const {return _Cols;}
EIGEN_DEVICE_FUNC void conservativeResize(Index, Index rows, Index) { m_rows = rows; }
EIGEN_DEVICE_FUNC void resize(Index, Index rows, Index) { m_rows = rows; }
EIGEN_DEVICE_FUNC const T *data() const { return m_data.array; }
EIGEN_DEVICE_FUNC T *data() { return m_data.array; }
};
// dynamic-size matrix with fixed-size storage and fixed height
template<typename T, int Size, int _Rows, int _Options> class DenseStorage<T, Size, _Rows, Dynamic, _Options>
{
internal::plain_array<T,Size,_Options> m_data;
DenseIndex m_cols;
Index m_cols;
public:
DenseStorage() : m_cols(0) {}
DenseStorage(internal::constructor_without_unaligned_array_assert)
EIGEN_DEVICE_FUNC DenseStorage() : m_cols(0) {}
EIGEN_DEVICE_FUNC explicit DenseStorage(internal::constructor_without_unaligned_array_assert)
: m_data(internal::constructor_without_unaligned_array_assert()), m_cols(0) {}
DenseStorage(const DenseStorage& other) : m_data(other.m_data), m_cols(other.m_cols) {}
DenseStorage& operator=(const DenseStorage& other)
EIGEN_DEVICE_FUNC DenseStorage(const DenseStorage& other) : m_data(other.m_data), m_cols(other.m_cols) {}
EIGEN_DEVICE_FUNC DenseStorage& operator=(const DenseStorage& other)
{
if (this != &other)
{
@@ -249,38 +324,62 @@ template<typename T, int Size, int _Rows, int _Options> class DenseStorage<T, Si
}
return *this;
}
DenseStorage(DenseIndex, DenseIndex, DenseIndex nbCols) : m_cols(nbCols) {}
void swap(DenseStorage& other) { std::swap(m_data,other.m_data); std::swap(m_cols,other.m_cols); }
DenseIndex rows(void) const {return _Rows;}
DenseIndex cols(void) const {return m_cols;}
void conservativeResize(DenseIndex, DenseIndex, DenseIndex nbCols) { m_cols = nbCols; }
void resize(DenseIndex, DenseIndex, DenseIndex nbCols) { m_cols = nbCols; }
const T *data() const { return m_data.array; }
T *data() { return m_data.array; }
EIGEN_DEVICE_FUNC DenseStorage(Index, Index, Index cols) : m_cols(cols) {}
EIGEN_DEVICE_FUNC void swap(DenseStorage& other) { std::swap(m_data,other.m_data); std::swap(m_cols,other.m_cols); }
EIGEN_DEVICE_FUNC Index rows(void) const {return _Rows;}
EIGEN_DEVICE_FUNC Index cols(void) const {return m_cols;}
void conservativeResize(Index, Index, Index cols) { m_cols = cols; }
void resize(Index, Index, Index cols) { m_cols = cols; }
EIGEN_DEVICE_FUNC const T *data() const { return m_data.array; }
EIGEN_DEVICE_FUNC T *data() { return m_data.array; }
};
// purely dynamic matrix.
template<typename T, int _Options> class DenseStorage<T, Dynamic, Dynamic, Dynamic, _Options>
{
T *m_data;
DenseIndex m_rows;
DenseIndex m_cols;
Index m_rows;
Index m_cols;
public:
DenseStorage() : m_data(0), m_rows(0), m_cols(0) {}
DenseStorage(internal::constructor_without_unaligned_array_assert)
EIGEN_DEVICE_FUNC DenseStorage() : m_data(0), m_rows(0), m_cols(0) {}
EIGEN_DEVICE_FUNC explicit DenseStorage(internal::constructor_without_unaligned_array_assert)
: m_data(0), m_rows(0), m_cols(0) {}
DenseStorage(DenseIndex size, DenseIndex nbRows, DenseIndex nbCols)
: m_data(internal::conditional_aligned_new_auto<T,(_Options&DontAlign)==0>(size)), m_rows(nbRows), m_cols(nbCols)
{ EIGEN_INTERNAL_DENSE_STORAGE_CTOR_PLUGIN }
#ifdef EIGEN_HAVE_RVALUE_REFERENCES
DenseStorage(DenseStorage&& other)
EIGEN_DEVICE_FUNC DenseStorage(Index size, Index rows, Index cols)
: m_data(internal::conditional_aligned_new_auto<T,(_Options&DontAlign)==0>(size)), m_rows(rows), m_cols(cols)
{
EIGEN_INTERNAL_DENSE_STORAGE_CTOR_PLUGIN({})
eigen_internal_assert(size==rows*cols && rows>=0 && cols >=0);
}
EIGEN_DEVICE_FUNC DenseStorage(const DenseStorage& other)
: m_data(internal::conditional_aligned_new_auto<T,(_Options&DontAlign)==0>(other.m_rows*other.m_cols))
, m_rows(other.m_rows)
, m_cols(other.m_cols)
{
EIGEN_INTERNAL_DENSE_STORAGE_CTOR_PLUGIN(Index size = m_rows*m_cols)
internal::smart_copy(other.m_data, other.m_data+other.m_rows*other.m_cols, m_data);
}
EIGEN_DEVICE_FUNC DenseStorage& operator=(const DenseStorage& other)
{
if (this != &other)
{
DenseStorage tmp(other);
this->swap(tmp);
}
return *this;
}
#if EIGEN_HAS_RVALUE_REFERENCES
EIGEN_DEVICE_FUNC
DenseStorage(DenseStorage&& other) EIGEN_NOEXCEPT
: m_data(std::move(other.m_data))
, m_rows(std::move(other.m_rows))
, m_cols(std::move(other.m_cols))
{
other.m_data = nullptr;
other.m_rows = 0;
other.m_cols = 0;
}
DenseStorage& operator=(DenseStorage&& other)
EIGEN_DEVICE_FUNC
DenseStorage& operator=(DenseStorage&& other) EIGEN_NOEXCEPT
{
using std::swap;
swap(m_data, other.m_data);
@@ -289,18 +388,18 @@ template<typename T, int _Options> class DenseStorage<T, Dynamic, Dynamic, Dynam
return *this;
}
#endif
~DenseStorage() { internal::conditional_aligned_delete_auto<T,(_Options&DontAlign)==0>(m_data, m_rows*m_cols); }
void swap(DenseStorage& other)
EIGEN_DEVICE_FUNC ~DenseStorage() { internal::conditional_aligned_delete_auto<T,(_Options&DontAlign)==0>(m_data, m_rows*m_cols); }
EIGEN_DEVICE_FUNC void swap(DenseStorage& other)
{ std::swap(m_data,other.m_data); std::swap(m_rows,other.m_rows); std::swap(m_cols,other.m_cols); }
DenseIndex rows(void) const {return m_rows;}
DenseIndex cols(void) const {return m_cols;}
void conservativeResize(DenseIndex size, DenseIndex nbRows, DenseIndex nbCols)
EIGEN_DEVICE_FUNC Index rows(void) const {return m_rows;}
EIGEN_DEVICE_FUNC Index cols(void) const {return m_cols;}
void conservativeResize(Index size, Index rows, Index cols)
{
m_data = internal::conditional_aligned_realloc_new_auto<T,(_Options&DontAlign)==0>(m_data, size, m_rows*m_cols);
m_rows = nbRows;
m_cols = nbCols;
m_rows = rows;
m_cols = cols;
}
void resize(DenseIndex size, DenseIndex nbRows, DenseIndex nbCols)
EIGEN_DEVICE_FUNC void resize(Index size, Index rows, Index cols)
{
if(size != m_rows*m_cols)
{
@@ -309,36 +408,56 @@ template<typename T, int _Options> class DenseStorage<T, Dynamic, Dynamic, Dynam
m_data = internal::conditional_aligned_new_auto<T,(_Options&DontAlign)==0>(size);
else
m_data = 0;
EIGEN_INTERNAL_DENSE_STORAGE_CTOR_PLUGIN
EIGEN_INTERNAL_DENSE_STORAGE_CTOR_PLUGIN({})
}
m_rows = nbRows;
m_cols = nbCols;
m_rows = rows;
m_cols = cols;
}
const T *data() const { return m_data; }
T *data() { return m_data; }
private:
DenseStorage(const DenseStorage&);
DenseStorage& operator=(const DenseStorage&);
EIGEN_DEVICE_FUNC const T *data() const { return m_data; }
EIGEN_DEVICE_FUNC T *data() { return m_data; }
};
// matrix with dynamic width and fixed height (so that matrix has dynamic size).
template<typename T, int _Rows, int _Options> class DenseStorage<T, Dynamic, _Rows, Dynamic, _Options>
{
T *m_data;
DenseIndex m_cols;
Index m_cols;
public:
DenseStorage() : m_data(0), m_cols(0) {}
DenseStorage(internal::constructor_without_unaligned_array_assert) : m_data(0), m_cols(0) {}
DenseStorage(DenseIndex size, DenseIndex, DenseIndex nbCols) : m_data(internal::conditional_aligned_new_auto<T,(_Options&DontAlign)==0>(size)), m_cols(nbCols)
{ EIGEN_INTERNAL_DENSE_STORAGE_CTOR_PLUGIN }
#ifdef EIGEN_HAVE_RVALUE_REFERENCES
DenseStorage(DenseStorage&& other)
EIGEN_DEVICE_FUNC DenseStorage() : m_data(0), m_cols(0) {}
explicit DenseStorage(internal::constructor_without_unaligned_array_assert) : m_data(0), m_cols(0) {}
EIGEN_DEVICE_FUNC DenseStorage(Index size, Index rows, Index cols) : m_data(internal::conditional_aligned_new_auto<T,(_Options&DontAlign)==0>(size)), m_cols(cols)
{
EIGEN_INTERNAL_DENSE_STORAGE_CTOR_PLUGIN({})
eigen_internal_assert(size==rows*cols && rows==_Rows && cols >=0);
EIGEN_UNUSED_VARIABLE(rows);
}
EIGEN_DEVICE_FUNC DenseStorage(const DenseStorage& other)
: m_data(internal::conditional_aligned_new_auto<T,(_Options&DontAlign)==0>(_Rows*other.m_cols))
, m_cols(other.m_cols)
{
EIGEN_INTERNAL_DENSE_STORAGE_CTOR_PLUGIN(Index size = m_cols*_Rows)
internal::smart_copy(other.m_data, other.m_data+_Rows*m_cols, m_data);
}
EIGEN_DEVICE_FUNC DenseStorage& operator=(const DenseStorage& other)
{
if (this != &other)
{
DenseStorage tmp(other);
this->swap(tmp);
}
return *this;
}
#if EIGEN_HAS_RVALUE_REFERENCES
EIGEN_DEVICE_FUNC
DenseStorage(DenseStorage&& other) EIGEN_NOEXCEPT
: m_data(std::move(other.m_data))
, m_cols(std::move(other.m_cols))
{
other.m_data = nullptr;
other.m_cols = 0;
}
DenseStorage& operator=(DenseStorage&& other)
EIGEN_DEVICE_FUNC
DenseStorage& operator=(DenseStorage&& other) EIGEN_NOEXCEPT
{
using std::swap;
swap(m_data, other.m_data);
@@ -346,16 +465,16 @@ template<typename T, int _Rows, int _Options> class DenseStorage<T, Dynamic, _Ro
return *this;
}
#endif
~DenseStorage() { internal::conditional_aligned_delete_auto<T,(_Options&DontAlign)==0>(m_data, _Rows*m_cols); }
void swap(DenseStorage& other) { std::swap(m_data,other.m_data); std::swap(m_cols,other.m_cols); }
static DenseIndex rows(void) {return _Rows;}
DenseIndex cols(void) const {return m_cols;}
void conservativeResize(DenseIndex size, DenseIndex, DenseIndex nbCols)
EIGEN_DEVICE_FUNC ~DenseStorage() { internal::conditional_aligned_delete_auto<T,(_Options&DontAlign)==0>(m_data, _Rows*m_cols); }
EIGEN_DEVICE_FUNC void swap(DenseStorage& other) { std::swap(m_data,other.m_data); std::swap(m_cols,other.m_cols); }
EIGEN_DEVICE_FUNC static Index rows(void) {return _Rows;}
EIGEN_DEVICE_FUNC Index cols(void) const {return m_cols;}
EIGEN_DEVICE_FUNC void conservativeResize(Index size, Index, Index cols)
{
m_data = internal::conditional_aligned_realloc_new_auto<T,(_Options&DontAlign)==0>(m_data, size, _Rows*m_cols);
m_cols = nbCols;
m_cols = cols;
}
EIGEN_STRONG_INLINE void resize(DenseIndex size, DenseIndex, DenseIndex nbCols)
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE void resize(Index size, Index, Index cols)
{
if(size != _Rows*m_cols)
{
@@ -364,35 +483,55 @@ template<typename T, int _Rows, int _Options> class DenseStorage<T, Dynamic, _Ro
m_data = internal::conditional_aligned_new_auto<T,(_Options&DontAlign)==0>(size);
else
m_data = 0;
EIGEN_INTERNAL_DENSE_STORAGE_CTOR_PLUGIN
EIGEN_INTERNAL_DENSE_STORAGE_CTOR_PLUGIN({})
}
m_cols = nbCols;
m_cols = cols;
}
const T *data() const { return m_data; }
T *data() { return m_data; }
private:
DenseStorage(const DenseStorage&);
DenseStorage& operator=(const DenseStorage&);
EIGEN_DEVICE_FUNC const T *data() const { return m_data; }
EIGEN_DEVICE_FUNC T *data() { return m_data; }
};
// matrix with dynamic height and fixed width (so that matrix has dynamic size).
template<typename T, int _Cols, int _Options> class DenseStorage<T, Dynamic, Dynamic, _Cols, _Options>
{
T *m_data;
DenseIndex m_rows;
Index m_rows;
public:
DenseStorage() : m_data(0), m_rows(0) {}
DenseStorage(internal::constructor_without_unaligned_array_assert) : m_data(0), m_rows(0) {}
DenseStorage(DenseIndex size, DenseIndex nbRows, DenseIndex) : m_data(internal::conditional_aligned_new_auto<T,(_Options&DontAlign)==0>(size)), m_rows(nbRows)
{ EIGEN_INTERNAL_DENSE_STORAGE_CTOR_PLUGIN }
#ifdef EIGEN_HAVE_RVALUE_REFERENCES
DenseStorage(DenseStorage&& other)
EIGEN_DEVICE_FUNC DenseStorage() : m_data(0), m_rows(0) {}
explicit DenseStorage(internal::constructor_without_unaligned_array_assert) : m_data(0), m_rows(0) {}
EIGEN_DEVICE_FUNC DenseStorage(Index size, Index rows, Index cols) : m_data(internal::conditional_aligned_new_auto<T,(_Options&DontAlign)==0>(size)), m_rows(rows)
{
EIGEN_INTERNAL_DENSE_STORAGE_CTOR_PLUGIN({})
eigen_internal_assert(size==rows*cols && rows>=0 && cols == _Cols);
EIGEN_UNUSED_VARIABLE(cols);
}
EIGEN_DEVICE_FUNC DenseStorage(const DenseStorage& other)
: m_data(internal::conditional_aligned_new_auto<T,(_Options&DontAlign)==0>(other.m_rows*_Cols))
, m_rows(other.m_rows)
{
EIGEN_INTERNAL_DENSE_STORAGE_CTOR_PLUGIN(Index size = m_rows*_Cols)
internal::smart_copy(other.m_data, other.m_data+other.m_rows*_Cols, m_data);
}
EIGEN_DEVICE_FUNC DenseStorage& operator=(const DenseStorage& other)
{
if (this != &other)
{
DenseStorage tmp(other);
this->swap(tmp);
}
return *this;
}
#if EIGEN_HAS_RVALUE_REFERENCES
EIGEN_DEVICE_FUNC
DenseStorage(DenseStorage&& other) EIGEN_NOEXCEPT
: m_data(std::move(other.m_data))
, m_rows(std::move(other.m_rows))
{
other.m_data = nullptr;
other.m_rows = 0;
}
DenseStorage& operator=(DenseStorage&& other)
EIGEN_DEVICE_FUNC
DenseStorage& operator=(DenseStorage&& other) EIGEN_NOEXCEPT
{
using std::swap;
swap(m_data, other.m_data);
@@ -400,16 +539,16 @@ template<typename T, int _Cols, int _Options> class DenseStorage<T, Dynamic, Dyn
return *this;
}
#endif
~DenseStorage() { internal::conditional_aligned_delete_auto<T,(_Options&DontAlign)==0>(m_data, _Cols*m_rows); }
void swap(DenseStorage& other) { std::swap(m_data,other.m_data); std::swap(m_rows,other.m_rows); }
DenseIndex rows(void) const {return m_rows;}
static DenseIndex cols(void) {return _Cols;}
void conservativeResize(DenseIndex size, DenseIndex nbRows, DenseIndex)
EIGEN_DEVICE_FUNC ~DenseStorage() { internal::conditional_aligned_delete_auto<T,(_Options&DontAlign)==0>(m_data, _Cols*m_rows); }
EIGEN_DEVICE_FUNC void swap(DenseStorage& other) { std::swap(m_data,other.m_data); std::swap(m_rows,other.m_rows); }
EIGEN_DEVICE_FUNC Index rows(void) const {return m_rows;}
EIGEN_DEVICE_FUNC static Index cols(void) {return _Cols;}
void conservativeResize(Index size, Index rows, Index)
{
m_data = internal::conditional_aligned_realloc_new_auto<T,(_Options&DontAlign)==0>(m_data, size, m_rows*_Cols);
m_rows = nbRows;
m_rows = rows;
}
EIGEN_STRONG_INLINE void resize(DenseIndex size, DenseIndex nbRows, DenseIndex)
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE void resize(Index size, Index rows, Index)
{
if(size != m_rows*_Cols)
{
@@ -418,15 +557,12 @@ template<typename T, int _Cols, int _Options> class DenseStorage<T, Dynamic, Dyn
m_data = internal::conditional_aligned_new_auto<T,(_Options&DontAlign)==0>(size);
else
m_data = 0;
EIGEN_INTERNAL_DENSE_STORAGE_CTOR_PLUGIN
EIGEN_INTERNAL_DENSE_STORAGE_CTOR_PLUGIN({})
}
m_rows = nbRows;
m_rows = rows;
}
const T *data() const { return m_data; }
T *data() { return m_data; }
private:
DenseStorage(const DenseStorage&);
DenseStorage& operator=(const DenseStorage&);
EIGEN_DEVICE_FUNC const T *data() const { return m_data; }
EIGEN_DEVICE_FUNC T *data() { return m_data; }
};
} // end namespace Eigen

View File

@@ -21,7 +21,7 @@ namespace Eigen {
* \param MatrixType the type of the object in which we are taking a sub/main/super diagonal
* \param DiagIndex the index of the sub/super diagonal. The default is 0 and it means the main diagonal.
* A positive value means a superdiagonal, a negative value means a subdiagonal.
* You can also use Dynamic so the index can be set at runtime.
* You can also use DynamicIndex so the index can be set at runtime.
*
* The matrix is not required to be square.
*
@@ -37,7 +37,7 @@ template<typename MatrixType, int DiagIndex>
struct traits<Diagonal<MatrixType,DiagIndex> >
: traits<MatrixType>
{
typedef typename nested<MatrixType>::type MatrixTypeNested;
typedef typename ref_selector<MatrixType>::type MatrixTypeNested;
typedef typename remove_reference<MatrixTypeNested>::type _MatrixTypeNested;
typedef typename MatrixType::StorageKind StorageKind;
enum {
@@ -52,8 +52,7 @@ struct traits<Diagonal<MatrixType,DiagIndex> >
MatrixType::MaxColsAtCompileTime - EIGEN_PLAIN_ENUM_MAX( DiagIndex, 0))),
MaxColsAtCompileTime = 1,
MaskLvalueBit = is_lvalue<MatrixType>::value ? LvalueBit : 0,
Flags = (unsigned int)_MatrixTypeNested::Flags & (HereditaryBits | LinearAccessBit | MaskLvalueBit | DirectAccessBit) & ~RowMajorBit,
CoeffReadCost = _MatrixTypeNested::CoeffReadCost,
Flags = (unsigned int)_MatrixTypeNested::Flags & (RowMajorBit | MaskLvalueBit | DirectAccessBit) & ~RowMajorBit, // FIXME DirectAccessBit should not be handled by expressions
MatrixTypeOuterStride = outer_stride_at_compile_time<MatrixType>::ret,
InnerStrideAtCompileTime = MatrixTypeOuterStride == Dynamic ? Dynamic : MatrixTypeOuterStride+1,
OuterStrideAtCompileTime = 0
@@ -70,20 +69,31 @@ template<typename MatrixType, int _DiagIndex> class Diagonal
typedef typename internal::dense_xpr_base<Diagonal>::type Base;
EIGEN_DENSE_PUBLIC_INTERFACE(Diagonal)
inline Diagonal(MatrixType& matrix, Index a_index = DiagIndex) : m_matrix(matrix), m_index(a_index) {}
EIGEN_DEVICE_FUNC
explicit inline Diagonal(MatrixType& matrix, Index a_index = DiagIndex) : m_matrix(matrix), m_index(a_index)
{
eigen_assert( a_index <= m_matrix.cols() && -a_index <= m_matrix.rows() );
}
EIGEN_INHERIT_ASSIGNMENT_OPERATORS(Diagonal)
EIGEN_DEVICE_FUNC
inline Index rows() const
{ return m_index.value()<0 ? (std::min<Index>)(m_matrix.cols(),m_matrix.rows()+m_index.value()) : (std::min<Index>)(m_matrix.rows(),m_matrix.cols()-m_index.value()); }
{
return m_index.value()<0 ? numext::mini<Index>(m_matrix.cols(),m_matrix.rows()+m_index.value())
: numext::mini<Index>(m_matrix.rows(),m_matrix.cols()-m_index.value());
}
EIGEN_DEVICE_FUNC
inline Index cols() const { return 1; }
EIGEN_DEVICE_FUNC
inline Index innerStride() const
{
return m_matrix.outerStride() + 1;
}
EIGEN_DEVICE_FUNC
inline Index outerStride() const
{
return 0;
@@ -95,62 +105,75 @@ template<typename MatrixType, int _DiagIndex> class Diagonal
const Scalar
>::type ScalarWithConstIfNotLvalue;
inline ScalarWithConstIfNotLvalue* data() { return &(m_matrix.const_cast_derived().coeffRef(rowOffset(), colOffset())); }
inline const Scalar* data() const { return &(m_matrix.const_cast_derived().coeffRef(rowOffset(), colOffset())); }
EIGEN_DEVICE_FUNC
inline ScalarWithConstIfNotLvalue* data() { return &(m_matrix.coeffRef(rowOffset(), colOffset())); }
EIGEN_DEVICE_FUNC
inline const Scalar* data() const { return &(m_matrix.coeffRef(rowOffset(), colOffset())); }
EIGEN_DEVICE_FUNC
inline Scalar& coeffRef(Index row, Index)
{
EIGEN_STATIC_ASSERT_LVALUE(MatrixType)
return m_matrix.const_cast_derived().coeffRef(row+rowOffset(), row+colOffset());
return m_matrix.coeffRef(row+rowOffset(), row+colOffset());
}
EIGEN_DEVICE_FUNC
inline const Scalar& coeffRef(Index row, Index) const
{
return m_matrix.const_cast_derived().coeffRef(row+rowOffset(), row+colOffset());
return m_matrix.coeffRef(row+rowOffset(), row+colOffset());
}
EIGEN_DEVICE_FUNC
inline CoeffReturnType coeff(Index row, Index) const
{
return m_matrix.coeff(row+rowOffset(), row+colOffset());
}
EIGEN_DEVICE_FUNC
inline Scalar& coeffRef(Index idx)
{
EIGEN_STATIC_ASSERT_LVALUE(MatrixType)
return m_matrix.const_cast_derived().coeffRef(idx+rowOffset(), idx+colOffset());
return m_matrix.coeffRef(idx+rowOffset(), idx+colOffset());
}
EIGEN_DEVICE_FUNC
inline const Scalar& coeffRef(Index idx) const
{
return m_matrix.const_cast_derived().coeffRef(idx+rowOffset(), idx+colOffset());
return m_matrix.coeffRef(idx+rowOffset(), idx+colOffset());
}
EIGEN_DEVICE_FUNC
inline CoeffReturnType coeff(Index idx) const
{
return m_matrix.coeff(idx+rowOffset(), idx+colOffset());
}
const typename internal::remove_all<typename MatrixType::Nested>::type&
EIGEN_DEVICE_FUNC
inline const typename internal::remove_all<typename MatrixType::Nested>::type&
nestedExpression() const
{
return m_matrix;
}
int index() const
EIGEN_DEVICE_FUNC
inline Index index() const
{
return m_index.value();
}
protected:
typename MatrixType::Nested m_matrix;
typename internal::ref_selector<MatrixType>::non_const_type m_matrix;
const internal::variable_if_dynamicindex<Index, DiagIndex> m_index;
private:
// some compilers may fail to optimize std::max etc in case of compile-time constants...
EIGEN_DEVICE_FUNC
EIGEN_STRONG_INLINE Index absDiagIndex() const { return m_index.value()>0 ? m_index.value() : -m_index.value(); }
EIGEN_DEVICE_FUNC
EIGEN_STRONG_INLINE Index rowOffset() const { return m_index.value()>0 ? 0 : -m_index.value(); }
EIGEN_DEVICE_FUNC
EIGEN_STRONG_INLINE Index colOffset() const { return m_index.value()>0 ? m_index.value() : 0; }
// triger a compile time error is someone try to call packet
// trigger a compile-time error if someone try to call packet
template<int LoadMode> typename MatrixType::PacketReturnType packet(Index) const;
template<int LoadMode> typename MatrixType::PacketReturnType packet(Index,Index) const;
};
@@ -167,7 +190,7 @@ template<typename Derived>
inline typename MatrixBase<Derived>::DiagonalReturnType
MatrixBase<Derived>::diagonal()
{
return derived();
return DiagonalReturnType(derived());
}
/** This is the const version of diagonal(). */
@@ -216,20 +239,20 @@ MatrixBase<Derived>::diagonal(Index index) const
*
* \sa MatrixBase::diagonal(), class Diagonal */
template<typename Derived>
template<int Index>
inline typename MatrixBase<Derived>::template DiagonalIndexReturnType<Index>::Type
template<int Index_>
inline typename MatrixBase<Derived>::template DiagonalIndexReturnType<Index_>::Type
MatrixBase<Derived>::diagonal()
{
return derived();
return typename DiagonalIndexReturnType<Index_>::Type(derived());
}
/** This is the const version of diagonal<int>(). */
template<typename Derived>
template<int Index>
inline typename MatrixBase<Derived>::template ConstDiagonalIndexReturnType<Index>::Type
template<int Index_>
inline typename MatrixBase<Derived>::template ConstDiagonalIndexReturnType<Index_>::Type
MatrixBase<Derived>::diagonal() const
{
return derived();
return typename ConstDiagonalIndexReturnType<Index_>::Type(derived());
}
} // end namespace Eigen

View File

@@ -22,7 +22,7 @@ class DiagonalBase : public EigenBase<Derived>
typedef typename DiagonalVectorType::Scalar Scalar;
typedef typename DiagonalVectorType::RealScalar RealScalar;
typedef typename internal::traits<Derived>::StorageKind StorageKind;
typedef typename internal::traits<Derived>::Index Index;
typedef typename internal::traits<Derived>::StorageIndex StorageIndex;
enum {
RowsAtCompileTime = DiagonalVectorType::SizeAtCompileTime,
@@ -30,79 +30,61 @@ class DiagonalBase : public EigenBase<Derived>
MaxRowsAtCompileTime = DiagonalVectorType::MaxSizeAtCompileTime,
MaxColsAtCompileTime = DiagonalVectorType::MaxSizeAtCompileTime,
IsVectorAtCompileTime = 0,
Flags = 0
Flags = NoPreferredStorageOrderBit
};
typedef Matrix<Scalar, RowsAtCompileTime, ColsAtCompileTime, 0, MaxRowsAtCompileTime, MaxColsAtCompileTime> DenseMatrixType;
typedef DenseMatrixType DenseType;
typedef DiagonalMatrix<Scalar,DiagonalVectorType::SizeAtCompileTime,DiagonalVectorType::MaxSizeAtCompileTime> PlainObject;
EIGEN_DEVICE_FUNC
inline const Derived& derived() const { return *static_cast<const Derived*>(this); }
EIGEN_DEVICE_FUNC
inline Derived& derived() { return *static_cast<Derived*>(this); }
EIGEN_DEVICE_FUNC
DenseMatrixType toDenseMatrix() const { return derived(); }
template<typename DenseDerived>
void evalTo(MatrixBase<DenseDerived> &other) const;
template<typename DenseDerived>
void addTo(MatrixBase<DenseDerived> &other) const
{ other.diagonal() += diagonal(); }
template<typename DenseDerived>
void subTo(MatrixBase<DenseDerived> &other) const
{ other.diagonal() -= diagonal(); }
EIGEN_DEVICE_FUNC
inline const DiagonalVectorType& diagonal() const { return derived().diagonal(); }
EIGEN_DEVICE_FUNC
inline DiagonalVectorType& diagonal() { return derived().diagonal(); }
EIGEN_DEVICE_FUNC
inline Index rows() const { return diagonal().size(); }
EIGEN_DEVICE_FUNC
inline Index cols() const { return diagonal().size(); }
/** \returns the diagonal matrix product of \c *this by the matrix \a matrix.
*/
template<typename MatrixDerived>
const DiagonalProduct<MatrixDerived, Derived, OnTheLeft>
EIGEN_DEVICE_FUNC
const Product<Derived,MatrixDerived,LazyProduct>
operator*(const MatrixBase<MatrixDerived> &matrix) const
{
return DiagonalProduct<MatrixDerived, Derived, OnTheLeft>(matrix.derived(), derived());
return Product<Derived, MatrixDerived, LazyProduct>(derived(),matrix.derived());
}
inline const DiagonalWrapper<const CwiseUnaryOp<internal::scalar_inverse_op<Scalar>, const DiagonalVectorType> >
typedef DiagonalWrapper<const CwiseUnaryOp<internal::scalar_inverse_op<Scalar>, const DiagonalVectorType> > InverseReturnType;
EIGEN_DEVICE_FUNC
inline const InverseReturnType
inverse() const
{
return diagonal().cwiseInverse();
return InverseReturnType(diagonal().cwiseInverse());
}
inline const DiagonalWrapper<const CwiseUnaryOp<internal::scalar_multiple_op<Scalar>, const DiagonalVectorType> >
EIGEN_DEVICE_FUNC
inline const DiagonalWrapper<const EIGEN_EXPR_BINARYOP_SCALAR_RETURN_TYPE(DiagonalVectorType,Scalar,product) >
operator*(const Scalar& scalar) const
{
return diagonal() * scalar;
return DiagonalWrapper<const EIGEN_EXPR_BINARYOP_SCALAR_RETURN_TYPE(DiagonalVectorType,Scalar,product) >(diagonal() * scalar);
}
friend inline const DiagonalWrapper<const CwiseUnaryOp<internal::scalar_multiple_op<Scalar>, const DiagonalVectorType> >
EIGEN_DEVICE_FUNC
friend inline const DiagonalWrapper<const EIGEN_SCALAR_BINARYOP_EXPR_RETURN_TYPE(Scalar,DiagonalVectorType,product) >
operator*(const Scalar& scalar, const DiagonalBase& other)
{
return other.diagonal() * scalar;
return DiagonalWrapper<const EIGEN_SCALAR_BINARYOP_EXPR_RETURN_TYPE(Scalar,DiagonalVectorType,product) >(scalar * other.diagonal());
}
#ifdef EIGEN2_SUPPORT
template<typename OtherDerived>
bool isApprox(const DiagonalBase<OtherDerived>& other, typename NumTraits<Scalar>::Real precision = NumTraits<Scalar>::dummy_precision()) const
{
return diagonal().isApprox(other.diagonal(), precision);
}
template<typename OtherDerived>
bool isApprox(const MatrixBase<OtherDerived>& other, typename NumTraits<Scalar>::Real precision = NumTraits<Scalar>::dummy_precision()) const
{
return toDenseMatrix().isApprox(other, precision);
}
#endif
};
template<typename Derived>
template<typename DenseDerived>
void DiagonalBase<Derived>::evalTo(MatrixBase<DenseDerived> &other) const
{
other.setZero();
other.diagonal() = diagonal();
}
#endif
/** \class DiagonalMatrix
@@ -124,10 +106,9 @@ struct traits<DiagonalMatrix<_Scalar,SizeAtCompileTime,MaxSizeAtCompileTime> >
: traits<Matrix<_Scalar,SizeAtCompileTime,SizeAtCompileTime,0,MaxSizeAtCompileTime,MaxSizeAtCompileTime> >
{
typedef Matrix<_Scalar,SizeAtCompileTime,1,0,MaxSizeAtCompileTime,1> DiagonalVectorType;
typedef Dense StorageKind;
typedef DenseIndex Index;
typedef DiagonalShape StorageKind;
enum {
Flags = LvalueBit
Flags = LvalueBit | NoPreferredStorageOrderBit
};
};
}
@@ -141,7 +122,7 @@ class DiagonalMatrix
typedef const DiagonalMatrix& Nested;
typedef _Scalar Scalar;
typedef typename internal::traits<DiagonalMatrix>::StorageKind StorageKind;
typedef typename internal::traits<DiagonalMatrix>::Index Index;
typedef typename internal::traits<DiagonalMatrix>::StorageIndex StorageIndex;
#endif
protected:
@@ -151,24 +132,31 @@ class DiagonalMatrix
public:
/** const version of diagonal(). */
EIGEN_DEVICE_FUNC
inline const DiagonalVectorType& diagonal() const { return m_diagonal; }
/** \returns a reference to the stored vector of diagonal coefficients. */
EIGEN_DEVICE_FUNC
inline DiagonalVectorType& diagonal() { return m_diagonal; }
/** Default constructor without initialization */
EIGEN_DEVICE_FUNC
inline DiagonalMatrix() {}
/** Constructs a diagonal matrix with given dimension */
inline DiagonalMatrix(Index dim) : m_diagonal(dim) {}
EIGEN_DEVICE_FUNC
explicit inline DiagonalMatrix(Index dim) : m_diagonal(dim) {}
/** 2D constructor. */
EIGEN_DEVICE_FUNC
inline DiagonalMatrix(const Scalar& x, const Scalar& y) : m_diagonal(x,y) {}
/** 3D constructor. */
EIGEN_DEVICE_FUNC
inline DiagonalMatrix(const Scalar& x, const Scalar& y, const Scalar& z) : m_diagonal(x,y,z) {}
/** Copy constructor. */
template<typename OtherDerived>
EIGEN_DEVICE_FUNC
inline DiagonalMatrix(const DiagonalBase<OtherDerived>& other) : m_diagonal(other.diagonal()) {}
#ifndef EIGEN_PARSED_BY_DOXYGEN
@@ -178,11 +166,13 @@ class DiagonalMatrix
/** generic constructor from expression of the diagonal coefficients */
template<typename OtherDerived>
EIGEN_DEVICE_FUNC
explicit inline DiagonalMatrix(const MatrixBase<OtherDerived>& other) : m_diagonal(other)
{}
/** Copy operator. */
template<typename OtherDerived>
EIGEN_DEVICE_FUNC
DiagonalMatrix& operator=(const DiagonalBase<OtherDerived>& other)
{
m_diagonal = other.diagonal();
@@ -193,6 +183,7 @@ class DiagonalMatrix
/** This is a special case of the templated operator=. Its purpose is to
* prevent a default operator= from hiding the templated operator=.
*/
EIGEN_DEVICE_FUNC
DiagonalMatrix& operator=(const DiagonalMatrix& other)
{
m_diagonal = other.diagonal();
@@ -201,14 +192,19 @@ class DiagonalMatrix
#endif
/** Resizes to given size. */
EIGEN_DEVICE_FUNC
inline void resize(Index size) { m_diagonal.resize(size); }
/** Sets all coefficients to zero. */
EIGEN_DEVICE_FUNC
inline void setZero() { m_diagonal.setZero(); }
/** Resizes and sets all coefficients to zero. */
EIGEN_DEVICE_FUNC
inline void setZero(Index size) { m_diagonal.setZero(size); }
/** Sets this matrix to be the identity matrix of the current size. */
EIGEN_DEVICE_FUNC
inline void setIdentity() { m_diagonal.setOnes(); }
/** Sets this matrix to be the identity matrix of the given size. */
EIGEN_DEVICE_FUNC
inline void setIdentity(Index size) { m_diagonal.setOnes(size); }
};
@@ -232,14 +228,15 @@ struct traits<DiagonalWrapper<_DiagonalVectorType> >
{
typedef _DiagonalVectorType DiagonalVectorType;
typedef typename DiagonalVectorType::Scalar Scalar;
typedef typename DiagonalVectorType::Index Index;
typedef typename DiagonalVectorType::StorageKind StorageKind;
typedef typename DiagonalVectorType::StorageIndex StorageIndex;
typedef DiagonalShape StorageKind;
typedef typename traits<DiagonalVectorType>::XprKind XprKind;
enum {
RowsAtCompileTime = DiagonalVectorType::SizeAtCompileTime,
ColsAtCompileTime = DiagonalVectorType::SizeAtCompileTime,
MaxRowsAtCompileTime = DiagonalVectorType::SizeAtCompileTime,
MaxColsAtCompileTime = DiagonalVectorType::SizeAtCompileTime,
Flags = traits<DiagonalVectorType>::Flags & LvalueBit
MaxRowsAtCompileTime = DiagonalVectorType::MaxSizeAtCompileTime,
MaxColsAtCompileTime = DiagonalVectorType::MaxSizeAtCompileTime,
Flags = (traits<DiagonalVectorType>::Flags & LvalueBit) | NoPreferredStorageOrderBit
};
};
}
@@ -255,9 +252,11 @@ class DiagonalWrapper
#endif
/** Constructor from expression of diagonal coefficients to wrap. */
inline DiagonalWrapper(DiagonalVectorType& a_diagonal) : m_diagonal(a_diagonal) {}
EIGEN_DEVICE_FUNC
explicit inline DiagonalWrapper(DiagonalVectorType& a_diagonal) : m_diagonal(a_diagonal) {}
/** \returns a const reference to the wrapped expression of diagonal coefficients. */
EIGEN_DEVICE_FUNC
const DiagonalVectorType& diagonal() const { return m_diagonal; }
protected:
@@ -277,7 +276,7 @@ template<typename Derived>
inline const DiagonalWrapper<const Derived>
MatrixBase<Derived>::asDiagonal() const
{
return derived();
return DiagonalWrapper<const Derived>(derived());
}
/** \returns true if *this is approximately equal to a diagonal matrix,
@@ -291,12 +290,11 @@ MatrixBase<Derived>::asDiagonal() const
template<typename Derived>
bool MatrixBase<Derived>::isDiagonal(const RealScalar& prec) const
{
using std::abs;
if(cols() != rows()) return false;
RealScalar maxAbsOnDiagonal = static_cast<RealScalar>(-1);
for(Index j = 0; j < cols(); ++j)
{
RealScalar absOnDiagonal = abs(coeff(j,j));
RealScalar absOnDiagonal = numext::abs(coeff(j,j));
if(absOnDiagonal > maxAbsOnDiagonal) maxAbsOnDiagonal = absOnDiagonal;
}
for(Index j = 0; j < cols(); ++j)
@@ -308,6 +306,38 @@ bool MatrixBase<Derived>::isDiagonal(const RealScalar& prec) const
return true;
}
namespace internal {
template<> struct storage_kind_to_shape<DiagonalShape> { typedef DiagonalShape Shape; };
struct Diagonal2Dense {};
template<> struct AssignmentKind<DenseShape,DiagonalShape> { typedef Diagonal2Dense Kind; };
// Diagonal matrix to Dense assignment
template< typename DstXprType, typename SrcXprType, typename Functor>
struct Assignment<DstXprType, SrcXprType, Functor, Diagonal2Dense>
{
static void run(DstXprType &dst, const SrcXprType &src, const internal::assign_op<typename DstXprType::Scalar,typename SrcXprType::Scalar> &/*func*/)
{
Index dstRows = src.rows();
Index dstCols = src.cols();
if((dst.rows()!=dstRows) || (dst.cols()!=dstCols))
dst.resize(dstRows, dstCols);
dst.setZero();
dst.diagonal() = src.diagonal();
}
static void run(DstXprType &dst, const SrcXprType &src, const internal::add_assign_op<typename DstXprType::Scalar,typename SrcXprType::Scalar> &/*func*/)
{ dst.diagonal() += src.diagonal(); }
static void run(DstXprType &dst, const SrcXprType &src, const internal::sub_assign_op<typename DstXprType::Scalar,typename SrcXprType::Scalar> &/*func*/)
{ dst.diagonal() -= src.diagonal(); }
};
} // namespace internal
} // end namespace Eigen
#endif // EIGEN_DIAGONALMATRIX_H

View File

@@ -13,117 +13,14 @@
namespace Eigen {
namespace internal {
template<typename MatrixType, typename DiagonalType, int ProductOrder>
struct traits<DiagonalProduct<MatrixType, DiagonalType, ProductOrder> >
: traits<MatrixType>
{
typedef typename scalar_product_traits<typename MatrixType::Scalar, typename DiagonalType::Scalar>::ReturnType Scalar;
enum {
RowsAtCompileTime = MatrixType::RowsAtCompileTime,
ColsAtCompileTime = MatrixType::ColsAtCompileTime,
MaxRowsAtCompileTime = MatrixType::MaxRowsAtCompileTime,
MaxColsAtCompileTime = MatrixType::MaxColsAtCompileTime,
_StorageOrder = MatrixType::Flags & RowMajorBit ? RowMajor : ColMajor,
_ScalarAccessOnDiag = !((int(_StorageOrder) == ColMajor && int(ProductOrder) == OnTheLeft)
||(int(_StorageOrder) == RowMajor && int(ProductOrder) == OnTheRight)),
_SameTypes = is_same<typename MatrixType::Scalar, typename DiagonalType::Scalar>::value,
// FIXME currently we need same types, but in the future the next rule should be the one
//_Vectorizable = bool(int(MatrixType::Flags)&PacketAccessBit) && ((!_PacketOnDiag) || (_SameTypes && bool(int(DiagonalType::DiagonalVectorType::Flags)&PacketAccessBit))),
_Vectorizable = bool(int(MatrixType::Flags)&PacketAccessBit) && _SameTypes && (_ScalarAccessOnDiag || (bool(int(DiagonalType::DiagonalVectorType::Flags)&PacketAccessBit))),
_LinearAccessMask = (RowsAtCompileTime==1 || ColsAtCompileTime==1) ? LinearAccessBit : 0,
Flags = ((HereditaryBits|_LinearAccessMask|AlignedBit) & (unsigned int)(MatrixType::Flags)) | (_Vectorizable ? PacketAccessBit : 0),//(int(MatrixType::Flags)&int(DiagonalType::DiagonalVectorType::Flags)&AlignedBit),
Cost0 = EIGEN_ADD_COST(NumTraits<Scalar>::MulCost, MatrixType::CoeffReadCost),
CoeffReadCost = EIGEN_ADD_COST(Cost0,DiagonalType::DiagonalVectorType::CoeffReadCost)
};
};
}
template<typename MatrixType, typename DiagonalType, int ProductOrder>
class DiagonalProduct : internal::no_assignment_operator,
public MatrixBase<DiagonalProduct<MatrixType, DiagonalType, ProductOrder> >
{
public:
typedef MatrixBase<DiagonalProduct> Base;
EIGEN_DENSE_PUBLIC_INTERFACE(DiagonalProduct)
inline DiagonalProduct(const MatrixType& matrix, const DiagonalType& diagonal)
: m_matrix(matrix), m_diagonal(diagonal)
{
eigen_assert(diagonal.diagonal().size() == (ProductOrder == OnTheLeft ? matrix.rows() : matrix.cols()));
}
EIGEN_STRONG_INLINE Index rows() const { return m_matrix.rows(); }
EIGEN_STRONG_INLINE Index cols() const { return m_matrix.cols(); }
EIGEN_STRONG_INLINE const Scalar coeff(Index row, Index col) const
{
return m_diagonal.diagonal().coeff(ProductOrder == OnTheLeft ? row : col) * m_matrix.coeff(row, col);
}
EIGEN_STRONG_INLINE const Scalar coeff(Index idx) const
{
enum {
StorageOrder = int(MatrixType::Flags) & RowMajorBit ? RowMajor : ColMajor
};
return coeff(int(StorageOrder)==ColMajor?idx:0,int(StorageOrder)==ColMajor?0:idx);
}
template<int LoadMode>
EIGEN_STRONG_INLINE PacketScalar packet(Index row, Index col) const
{
enum {
StorageOrder = Flags & RowMajorBit ? RowMajor : ColMajor
};
const Index indexInDiagonalVector = ProductOrder == OnTheLeft ? row : col;
return packet_impl<LoadMode>(row,col,indexInDiagonalVector,typename internal::conditional<
((int(StorageOrder) == RowMajor && int(ProductOrder) == OnTheLeft)
||(int(StorageOrder) == ColMajor && int(ProductOrder) == OnTheRight)), internal::true_type, internal::false_type>::type());
}
template<int LoadMode>
EIGEN_STRONG_INLINE PacketScalar packet(Index idx) const
{
enum {
StorageOrder = int(MatrixType::Flags) & RowMajorBit ? RowMajor : ColMajor
};
return packet<LoadMode>(int(StorageOrder)==ColMajor?idx:0,int(StorageOrder)==ColMajor?0:idx);
}
protected:
template<int LoadMode>
EIGEN_STRONG_INLINE PacketScalar packet_impl(Index row, Index col, Index id, internal::true_type) const
{
return internal::pmul(m_matrix.template packet<LoadMode>(row, col),
internal::pset1<PacketScalar>(m_diagonal.diagonal().coeff(id)));
}
template<int LoadMode>
EIGEN_STRONG_INLINE PacketScalar packet_impl(Index row, Index col, Index id, internal::false_type) const
{
enum {
InnerSize = (MatrixType::Flags & RowMajorBit) ? MatrixType::ColsAtCompileTime : MatrixType::RowsAtCompileTime,
DiagonalVectorPacketLoadMode = (LoadMode == Aligned && (((InnerSize%16) == 0) || (int(DiagonalType::DiagonalVectorType::Flags)&AlignedBit)==AlignedBit) ? Aligned : Unaligned)
};
return internal::pmul(m_matrix.template packet<LoadMode>(row, col),
m_diagonal.diagonal().template packet<DiagonalVectorPacketLoadMode>(id));
}
typename MatrixType::Nested m_matrix;
typename DiagonalType::Nested m_diagonal;
};
/** \returns the diagonal matrix product of \c *this by the diagonal matrix \a diagonal.
*/
template<typename Derived>
template<typename DiagonalDerived>
inline const DiagonalProduct<Derived, DiagonalDerived, OnTheRight>
inline const Product<Derived, DiagonalDerived, LazyProduct>
MatrixBase<Derived>::operator*(const DiagonalBase<DiagonalDerived> &a_diagonal) const
{
return DiagonalProduct<Derived, DiagonalDerived, OnTheRight>(derived(), a_diagonal.derived());
return Product<Derived, DiagonalDerived, LazyProduct>(derived(),a_diagonal.derived());
}
} // end namespace Eigen

View File

@@ -28,26 +28,33 @@ template<typename T, typename U,
>
struct dot_nocheck
{
typedef typename scalar_product_traits<typename traits<T>::Scalar,typename traits<U>::Scalar>::ReturnType ResScalar;
static inline ResScalar run(const MatrixBase<T>& a, const MatrixBase<U>& b)
typedef scalar_conj_product_op<typename traits<T>::Scalar,typename traits<U>::Scalar> conj_prod;
typedef typename conj_prod::result_type ResScalar;
EIGEN_DEVICE_FUNC
EIGEN_STRONG_INLINE
static ResScalar run(const MatrixBase<T>& a, const MatrixBase<U>& b)
{
return a.template binaryExpr<scalar_conj_product_op<typename traits<T>::Scalar,typename traits<U>::Scalar> >(b).sum();
return a.template binaryExpr<conj_prod>(b).sum();
}
};
template<typename T, typename U>
struct dot_nocheck<T, U, true>
{
typedef typename scalar_product_traits<typename traits<T>::Scalar,typename traits<U>::Scalar>::ReturnType ResScalar;
static inline ResScalar run(const MatrixBase<T>& a, const MatrixBase<U>& b)
typedef scalar_conj_product_op<typename traits<T>::Scalar,typename traits<U>::Scalar> conj_prod;
typedef typename conj_prod::result_type ResScalar;
EIGEN_DEVICE_FUNC
EIGEN_STRONG_INLINE
static ResScalar run(const MatrixBase<T>& a, const MatrixBase<U>& b)
{
return a.transpose().template binaryExpr<scalar_conj_product_op<typename traits<T>::Scalar,typename traits<U>::Scalar> >(b).sum();
return a.transpose().template binaryExpr<conj_prod>(b).sum();
}
};
} // end namespace internal
/** \returns the dot product of *this with other.
/** \fn MatrixBase::dot
* \returns the dot product of *this with other.
*
* \only_for_vectors
*
@@ -59,55 +66,31 @@ struct dot_nocheck<T, U, true>
*/
template<typename Derived>
template<typename OtherDerived>
typename internal::scalar_product_traits<typename internal::traits<Derived>::Scalar,typename internal::traits<OtherDerived>::Scalar>::ReturnType
EIGEN_DEVICE_FUNC
EIGEN_STRONG_INLINE
typename ScalarBinaryOpTraits<typename internal::traits<Derived>::Scalar,typename internal::traits<OtherDerived>::Scalar>::ReturnType
MatrixBase<Derived>::dot(const MatrixBase<OtherDerived>& other) const
{
EIGEN_STATIC_ASSERT_VECTOR_ONLY(Derived)
EIGEN_STATIC_ASSERT_VECTOR_ONLY(OtherDerived)
EIGEN_STATIC_ASSERT_SAME_VECTOR_SIZE(Derived,OtherDerived)
#if !(defined(EIGEN_NO_STATIC_ASSERT) && defined(EIGEN_NO_DEBUG))
typedef internal::scalar_conj_product_op<Scalar,typename OtherDerived::Scalar> func;
EIGEN_CHECK_BINARY_COMPATIBILIY(func,Scalar,typename OtherDerived::Scalar);
#endif
eigen_assert(size() == other.size());
return internal::dot_nocheck<Derived,OtherDerived>::run(*this, other);
}
#ifdef EIGEN2_SUPPORT
/** \returns the dot product of *this with other, with the Eigen2 convention that the dot product is linear in the first variable
* (conjugating the second variable). Of course this only makes a difference in the complex case.
*
* This method is only available in EIGEN2_SUPPORT mode.
*
* \only_for_vectors
*
* \sa dot()
*/
template<typename Derived>
template<typename OtherDerived>
typename internal::traits<Derived>::Scalar
MatrixBase<Derived>::eigen2_dot(const MatrixBase<OtherDerived>& other) const
{
EIGEN_STATIC_ASSERT_VECTOR_ONLY(Derived)
EIGEN_STATIC_ASSERT_VECTOR_ONLY(OtherDerived)
EIGEN_STATIC_ASSERT_SAME_VECTOR_SIZE(Derived,OtherDerived)
EIGEN_STATIC_ASSERT((internal::is_same<Scalar, typename OtherDerived::Scalar>::value),
YOU_MIXED_DIFFERENT_NUMERIC_TYPES__YOU_NEED_TO_USE_THE_CAST_METHOD_OF_MATRIXBASE_TO_CAST_NUMERIC_TYPES_EXPLICITLY)
eigen_assert(size() == other.size());
return internal::dot_nocheck<OtherDerived,Derived>::run(other,*this);
}
#endif
//---------- implementation of L2 norm and related functions ----------
/** \returns, for vectors, the squared \em l2 norm of \c *this, and for matrices the Frobenius norm.
* In both cases, it consists in the sum of the square of all the matrix entries.
* For vectors, this is also equals to the dot product of \c *this with itself.
*
* \sa dot(), norm()
* \sa dot(), norm(), lpNorm()
*/
template<typename Derived>
EIGEN_STRONG_INLINE typename NumTraits<typename internal::traits<Derived>::Scalar>::Real MatrixBase<Derived>::squaredNorm() const
@@ -119,41 +102,98 @@ EIGEN_STRONG_INLINE typename NumTraits<typename internal::traits<Derived>::Scala
* In both cases, it consists in the square root of the sum of the square of all the matrix entries.
* For vectors, this is also equals to the square root of the dot product of \c *this with itself.
*
* \sa dot(), squaredNorm()
* \sa lpNorm(), dot(), squaredNorm()
*/
template<typename Derived>
inline typename NumTraits<typename internal::traits<Derived>::Scalar>::Real MatrixBase<Derived>::norm() const
EIGEN_STRONG_INLINE typename NumTraits<typename internal::traits<Derived>::Scalar>::Real MatrixBase<Derived>::norm() const
{
using std::sqrt;
return sqrt(squaredNorm());
return numext::sqrt(squaredNorm());
}
/** \returns an expression of the quotient of *this by its own norm.
/** \returns an expression of the quotient of \c *this by its own norm.
*
* \warning If the input vector is too small (i.e., this->norm()==0),
* then this function returns a copy of the input.
*
* \only_for_vectors
*
* \sa norm(), normalize()
*/
template<typename Derived>
inline const typename MatrixBase<Derived>::PlainObject
EIGEN_STRONG_INLINE const typename MatrixBase<Derived>::PlainObject
MatrixBase<Derived>::normalized() const
{
typedef typename internal::nested<Derived>::type Nested;
typedef typename internal::remove_reference<Nested>::type _Nested;
typedef typename internal::nested_eval<Derived,2>::type _Nested;
_Nested n(derived());
return n / n.norm();
RealScalar z = n.squaredNorm();
// NOTE: after extensive benchmarking, this conditional does not impact performance, at least on recent x86 CPU
if(z>RealScalar(0))
return n / numext::sqrt(z);
else
return n;
}
/** Normalizes the vector, i.e. divides it by its own norm.
*
* \only_for_vectors
*
* \warning If the input vector is too small (i.e., this->norm()==0), then \c *this is left unchanged.
*
* \sa norm(), normalized()
*/
template<typename Derived>
inline void MatrixBase<Derived>::normalize()
EIGEN_STRONG_INLINE void MatrixBase<Derived>::normalize()
{
*this /= norm();
RealScalar z = squaredNorm();
// NOTE: after extensive benchmarking, this conditional does not impact performance, at least on recent x86 CPU
if(z>RealScalar(0))
derived() /= numext::sqrt(z);
}
/** \returns an expression of the quotient of \c *this by its own norm while avoiding underflow and overflow.
*
* \only_for_vectors
*
* This method is analogue to the normalized() method, but it reduces the risk of
* underflow and overflow when computing the norm.
*
* \warning If the input vector is too small (i.e., this->norm()==0),
* then this function returns a copy of the input.
*
* \sa stableNorm(), stableNormalize(), normalized()
*/
template<typename Derived>
EIGEN_STRONG_INLINE const typename MatrixBase<Derived>::PlainObject
MatrixBase<Derived>::stableNormalized() const
{
typedef typename internal::nested_eval<Derived,3>::type _Nested;
_Nested n(derived());
RealScalar w = n.cwiseAbs().maxCoeff();
RealScalar z = (n/w).squaredNorm();
if(z>RealScalar(0))
return n / (numext::sqrt(z)*w);
else
return n;
}
/** Normalizes the vector while avoid underflow and overflow
*
* \only_for_vectors
*
* This method is analogue to the normalize() method, but it reduces the risk of
* underflow and overflow when computing the norm.
*
* \warning If the input vector is too small (i.e., this->norm()==0), then \c *this is left unchanged.
*
* \sa stableNorm(), stableNormalized(), normalize()
*/
template<typename Derived>
EIGEN_STRONG_INLINE void MatrixBase<Derived>::stableNormalize()
{
RealScalar w = cwiseAbs().maxCoeff();
RealScalar z = (derived()/w).squaredNorm();
if(z>RealScalar(0))
derived() /= numext::sqrt(z)*w;
}
//---------- implementation of other norms ----------
@@ -164,9 +204,10 @@ template<typename Derived, int p>
struct lpNorm_selector
{
typedef typename NumTraits<typename traits<Derived>::Scalar>::Real RealScalar;
EIGEN_DEVICE_FUNC
static inline RealScalar run(const MatrixBase<Derived>& m)
{
using std::pow;
EIGEN_USING_STD_MATH(pow)
return pow(m.cwiseAbs().array().pow(p).sum(), RealScalar(1)/p);
}
};
@@ -174,6 +215,7 @@ struct lpNorm_selector
template<typename Derived>
struct lpNorm_selector<Derived, 1>
{
EIGEN_DEVICE_FUNC
static inline typename NumTraits<typename traits<Derived>::Scalar>::Real run(const MatrixBase<Derived>& m)
{
return m.cwiseAbs().sum();
@@ -183,6 +225,7 @@ struct lpNorm_selector<Derived, 1>
template<typename Derived>
struct lpNorm_selector<Derived, 2>
{
EIGEN_DEVICE_FUNC
static inline typename NumTraits<typename traits<Derived>::Scalar>::Real run(const MatrixBase<Derived>& m)
{
return m.norm();
@@ -192,23 +235,35 @@ struct lpNorm_selector<Derived, 2>
template<typename Derived>
struct lpNorm_selector<Derived, Infinity>
{
static inline typename NumTraits<typename traits<Derived>::Scalar>::Real run(const MatrixBase<Derived>& m)
typedef typename NumTraits<typename traits<Derived>::Scalar>::Real RealScalar;
EIGEN_DEVICE_FUNC
static inline RealScalar run(const MatrixBase<Derived>& m)
{
if(Derived::SizeAtCompileTime==0 || (Derived::SizeAtCompileTime==Dynamic && m.size()==0))
return RealScalar(0);
return m.cwiseAbs().maxCoeff();
}
};
} // end namespace internal
/** \returns the \f$ \ell^p \f$ norm of *this, that is, returns the p-th root of the sum of the p-th powers of the absolute values
* of the coefficients of *this. If \a p is the special value \a Eigen::Infinity, this function returns the \f$ \ell^\infty \f$
* norm, that is the maximum of the absolute values of the coefficients of *this.
/** \returns the \b coefficient-wise \f$ \ell^p \f$ norm of \c *this, that is, returns the p-th root of the sum of the p-th powers of the absolute values
* of the coefficients of \c *this. If \a p is the special value \a Eigen::Infinity, this function returns the \f$ \ell^\infty \f$
* norm, that is the maximum of the absolute values of the coefficients of \c *this.
*
* In all cases, if \c *this is empty, then the value 0 is returned.
*
* \note For matrices, this function does not compute the <a href="https://en.wikipedia.org/wiki/Operator_norm">operator-norm</a>. That is, if \c *this is a matrix, then its coefficients are interpreted as a 1D vector. Nonetheless, you can easily compute the 1-norm and \f$\infty\f$-norm matrix operator norms using \link TutorialReductionsVisitorsBroadcastingReductionsNorm partial reductions \endlink.
*
* \sa norm()
*/
template<typename Derived>
template<int p>
#ifndef EIGEN_PARSED_BY_DOXYGEN
inline typename NumTraits<typename internal::traits<Derived>::Scalar>::Real
#else
MatrixBase<Derived>::RealScalar
#endif
MatrixBase<Derived>::lpNorm() const
{
return internal::lpNorm_selector<Derived, p>::run(*this);
@@ -227,8 +282,8 @@ template<typename OtherDerived>
bool MatrixBase<Derived>::isOrthogonal
(const MatrixBase<OtherDerived>& other, const RealScalar& prec) const
{
typename internal::nested<Derived,2>::type nested(derived());
typename internal::nested<OtherDerived,2>::type otherNested(other.derived());
typename internal::nested_eval<Derived,2>::type nested(derived());
typename internal::nested_eval<OtherDerived,2>::type otherNested(other.derived());
return numext::abs2(nested.dot(otherNested)) <= prec * prec * nested.squaredNorm() * otherNested.squaredNorm();
}
@@ -246,13 +301,13 @@ bool MatrixBase<Derived>::isOrthogonal
template<typename Derived>
bool MatrixBase<Derived>::isUnitary(const RealScalar& prec) const
{
typename Derived::Nested nested(derived());
typename internal::nested_eval<Derived,1>::type self(derived());
for(Index i = 0; i < cols(); ++i)
{
if(!internal::isApprox(nested.col(i).squaredNorm(), static_cast<RealScalar>(1), prec))
if(!internal::isApprox(self.col(i).squaredNorm(), static_cast<RealScalar>(1), prec))
return false;
for(Index j = 0; j < i; ++j)
if(!internal::isMuchSmallerThan(nested.col(i).dot(nested.col(j)), static_cast<Scalar>(1), prec))
if(!internal::isMuchSmallerThan(self.col(i).dot(self.col(j)), static_cast<Scalar>(1), prec))
return false;
}
return true;

View File

@@ -13,7 +13,10 @@
namespace Eigen {
/** Common base class for all classes T such that MatrixBase has an operator=(T) and a constructor MatrixBase(T).
/** \class EigenBase
* \ingroup Core_Module
*
* Common base class for all classes T such that MatrixBase has an operator=(T) and a constructor MatrixBase(T).
*
* In other words, an EigenBase object is an object that can be copied into a MatrixBase.
*
@@ -21,39 +24,57 @@ namespace Eigen {
*
* Notice that this class is trivial, it is only used to disambiguate overloaded functions.
*
* \sa \ref TopicClassHierarchy
* \sa \blank \ref TopicClassHierarchy
*/
template<typename Derived> struct EigenBase
{
// typedef typename internal::plain_matrix_type<Derived>::type PlainObject;
/** \brief The interface type of indices
* \details To change this, \c \#define the preprocessor symbol \c EIGEN_DEFAULT_DENSE_INDEX_TYPE.
* \deprecated Since Eigen 3.3, its usage is deprecated. Use Eigen::Index instead.
* \sa StorageIndex, \ref TopicPreprocessorDirectives.
*/
typedef Eigen::Index Index;
// FIXME is it needed?
typedef typename internal::traits<Derived>::StorageKind StorageKind;
typedef typename internal::traits<Derived>::Index Index;
/** \returns a reference to the derived object */
EIGEN_DEVICE_FUNC
Derived& derived() { return *static_cast<Derived*>(this); }
/** \returns a const reference to the derived object */
EIGEN_DEVICE_FUNC
const Derived& derived() const { return *static_cast<const Derived*>(this); }
EIGEN_DEVICE_FUNC
inline Derived& const_cast_derived() const
{ return *static_cast<Derived*>(const_cast<EigenBase*>(this)); }
EIGEN_DEVICE_FUNC
inline const Derived& const_derived() const
{ return *static_cast<const Derived*>(this); }
/** \returns the number of rows. \sa cols(), RowsAtCompileTime */
EIGEN_DEVICE_FUNC
inline Index rows() const { return derived().rows(); }
/** \returns the number of columns. \sa rows(), ColsAtCompileTime*/
EIGEN_DEVICE_FUNC
inline Index cols() const { return derived().cols(); }
/** \returns the number of coefficients, which is rows()*cols().
* \sa rows(), cols(), SizeAtCompileTime. */
EIGEN_DEVICE_FUNC
inline Index size() const { return rows() * cols(); }
/** \internal Don't use it, but do the equivalent: \code dst = *this; \endcode */
template<typename Dest> inline void evalTo(Dest& dst) const
template<typename Dest>
EIGEN_DEVICE_FUNC
inline void evalTo(Dest& dst) const
{ derived().evalTo(dst); }
/** \internal Don't use it, but do the equivalent: \code dst += *this; \endcode */
template<typename Dest> inline void addTo(Dest& dst) const
template<typename Dest>
EIGEN_DEVICE_FUNC
inline void addTo(Dest& dst) const
{
// This is the default implementation,
// derived class can reimplement it in a more optimized way.
@@ -63,7 +84,9 @@ template<typename Derived> struct EigenBase
}
/** \internal Don't use it, but do the equivalent: \code dst -= *this; \endcode */
template<typename Dest> inline void subTo(Dest& dst) const
template<typename Dest>
EIGEN_DEVICE_FUNC
inline void subTo(Dest& dst) const
{
// This is the default implementation,
// derived class can reimplement it in a more optimized way.
@@ -73,7 +96,8 @@ template<typename Derived> struct EigenBase
}
/** \internal Don't use it, but do the equivalent: \code dst.applyOnTheRight(*this); \endcode */
template<typename Dest> inline void applyThisOnTheRight(Dest& dst) const
template<typename Dest>
EIGEN_DEVICE_FUNC inline void applyThisOnTheRight(Dest& dst) const
{
// This is the default implementation,
// derived class can reimplement it in a more optimized way.
@@ -81,7 +105,8 @@ template<typename Derived> struct EigenBase
}
/** \internal Don't use it, but do the equivalent: \code dst.applyOnTheLeft(*this); \endcode */
template<typename Dest> inline void applyThisOnTheLeft(Dest& dst) const
template<typename Dest>
EIGEN_DEVICE_FUNC inline void applyThisOnTheLeft(Dest& dst) const
{
// This is the default implementation,
// derived class can reimplement it in a more optimized way.
@@ -104,25 +129,28 @@ template<typename Derived> struct EigenBase
*/
template<typename Derived>
template<typename OtherDerived>
EIGEN_DEVICE_FUNC
Derived& DenseBase<Derived>::operator=(const EigenBase<OtherDerived> &other)
{
other.derived().evalTo(derived());
call_assignment(derived(), other.derived());
return derived();
}
template<typename Derived>
template<typename OtherDerived>
EIGEN_DEVICE_FUNC
Derived& DenseBase<Derived>::operator+=(const EigenBase<OtherDerived> &other)
{
other.derived().addTo(derived());
call_assignment(derived(), other.derived(), internal::add_assign_op<Scalar,typename OtherDerived::Scalar>());
return derived();
}
template<typename Derived>
template<typename OtherDerived>
EIGEN_DEVICE_FUNC
Derived& DenseBase<Derived>::operator-=(const EigenBase<OtherDerived> &other)
{
other.derived().subTo(derived());
call_assignment(derived(), other.derived(), internal::sub_assign_op<Scalar,typename OtherDerived::Scalar>());
return derived();
}

View File

@@ -39,29 +39,29 @@ template<typename ExpressionType> class ForceAlignedAccess
typedef typename internal::dense_xpr_base<ForceAlignedAccess>::type Base;
EIGEN_DENSE_PUBLIC_INTERFACE(ForceAlignedAccess)
inline ForceAlignedAccess(const ExpressionType& matrix) : m_expression(matrix) {}
EIGEN_DEVICE_FUNC explicit inline ForceAlignedAccess(const ExpressionType& matrix) : m_expression(matrix) {}
inline Index rows() const { return m_expression.rows(); }
inline Index cols() const { return m_expression.cols(); }
inline Index outerStride() const { return m_expression.outerStride(); }
inline Index innerStride() const { return m_expression.innerStride(); }
EIGEN_DEVICE_FUNC inline Index rows() const { return m_expression.rows(); }
EIGEN_DEVICE_FUNC inline Index cols() const { return m_expression.cols(); }
EIGEN_DEVICE_FUNC inline Index outerStride() const { return m_expression.outerStride(); }
EIGEN_DEVICE_FUNC inline Index innerStride() const { return m_expression.innerStride(); }
inline const CoeffReturnType coeff(Index row, Index col) const
EIGEN_DEVICE_FUNC inline const CoeffReturnType coeff(Index row, Index col) const
{
return m_expression.coeff(row, col);
}
inline Scalar& coeffRef(Index row, Index col)
EIGEN_DEVICE_FUNC inline Scalar& coeffRef(Index row, Index col)
{
return m_expression.const_cast_derived().coeffRef(row, col);
}
inline const CoeffReturnType coeff(Index index) const
EIGEN_DEVICE_FUNC inline const CoeffReturnType coeff(Index index) const
{
return m_expression.coeff(index);
}
inline Scalar& coeffRef(Index index)
EIGEN_DEVICE_FUNC inline Scalar& coeffRef(Index index)
{
return m_expression.const_cast_derived().coeffRef(index);
}
@@ -90,7 +90,7 @@ template<typename ExpressionType> class ForceAlignedAccess
m_expression.const_cast_derived().template writePacket<Aligned>(index, x);
}
operator const ExpressionType&() const { return m_expression; }
EIGEN_DEVICE_FUNC operator const ExpressionType&() const { return m_expression; }
protected:
const ExpressionType& m_expression;
@@ -127,7 +127,7 @@ template<bool Enable>
inline typename internal::add_const_on_value_type<typename internal::conditional<Enable,ForceAlignedAccess<Derived>,Derived&>::type>::type
MatrixBase<Derived>::forceAlignedAccessIf() const
{
return derived();
return derived(); // FIXME This should not work but apparently is never used
}
/** \returns an expression of *this with forced aligned access if \a Enable is true.
@@ -138,7 +138,7 @@ template<bool Enable>
inline typename internal::conditional<Enable,ForceAlignedAccess<Derived>,Derived&>::type
MatrixBase<Derived>::forceAlignedAccessIf()
{
return derived();
return derived(); // FIXME This should not work but apparently is never used
}
} // end namespace Eigen

View File

@@ -19,18 +19,19 @@ namespace internal
template<typename Derived, typename OtherDerived, bool is_integer = NumTraits<typename Derived::Scalar>::IsInteger>
struct isApprox_selector
{
EIGEN_DEVICE_FUNC
static bool run(const Derived& x, const OtherDerived& y, const typename Derived::RealScalar& prec)
{
using std::min;
typename internal::nested<Derived,2>::type nested(x);
typename internal::nested<OtherDerived,2>::type otherNested(y);
return (nested - otherNested).cwiseAbs2().sum() <= prec * prec * (min)(nested.cwiseAbs2().sum(), otherNested.cwiseAbs2().sum());
typename internal::nested_eval<Derived,2>::type nested(x);
typename internal::nested_eval<OtherDerived,2>::type otherNested(y);
return (nested - otherNested).cwiseAbs2().sum() <= prec * prec * numext::mini(nested.cwiseAbs2().sum(), otherNested.cwiseAbs2().sum());
}
};
template<typename Derived, typename OtherDerived>
struct isApprox_selector<Derived, OtherDerived, true>
{
EIGEN_DEVICE_FUNC
static bool run(const Derived& x, const OtherDerived& y, const typename Derived::RealScalar&)
{
return x.matrix() == y.matrix();
@@ -40,6 +41,7 @@ struct isApprox_selector<Derived, OtherDerived, true>
template<typename Derived, typename OtherDerived, bool is_integer = NumTraits<typename Derived::Scalar>::IsInteger>
struct isMuchSmallerThan_object_selector
{
EIGEN_DEVICE_FUNC
static bool run(const Derived& x, const OtherDerived& y, const typename Derived::RealScalar& prec)
{
return x.cwiseAbs2().sum() <= numext::abs2(prec) * y.cwiseAbs2().sum();
@@ -49,6 +51,7 @@ struct isMuchSmallerThan_object_selector
template<typename Derived, typename OtherDerived>
struct isMuchSmallerThan_object_selector<Derived, OtherDerived, true>
{
EIGEN_DEVICE_FUNC
static bool run(const Derived& x, const OtherDerived&, const typename Derived::RealScalar&)
{
return x.matrix() == Derived::Zero(x.rows(), x.cols()).matrix();
@@ -58,6 +61,7 @@ struct isMuchSmallerThan_object_selector<Derived, OtherDerived, true>
template<typename Derived, bool is_integer = NumTraits<typename Derived::Scalar>::IsInteger>
struct isMuchSmallerThan_scalar_selector
{
EIGEN_DEVICE_FUNC
static bool run(const Derived& x, const typename Derived::RealScalar& y, const typename Derived::RealScalar& prec)
{
return x.cwiseAbs2().sum() <= numext::abs2(prec * y);
@@ -67,6 +71,7 @@ struct isMuchSmallerThan_scalar_selector
template<typename Derived>
struct isMuchSmallerThan_scalar_selector<Derived, true>
{
EIGEN_DEVICE_FUNC
static bool run(const Derived& x, const typename Derived::RealScalar&, const typename Derived::RealScalar&)
{
return x.matrix() == Derived::Zero(x.rows(), x.cols()).matrix();

View File

@@ -11,29 +11,7 @@
#ifndef EIGEN_GENERAL_PRODUCT_H
#define EIGEN_GENERAL_PRODUCT_H
namespace Eigen {
/** \class GeneralProduct
* \ingroup Core_Module
*
* \brief Expression of the product of two general matrices or vectors
*
* \param LhsNested the type used to store the left-hand side
* \param RhsNested the type used to store the right-hand side
* \param ProductMode the type of the product
*
* This class represents an expression of the product of two general matrices.
* We call a general matrix, a dense matrix with full storage. For instance,
* This excludes triangular, selfadjoint, and sparse matrices.
* It is the return type of the operator* between general matrices. Its template
* arguments are determined automatically by ProductReturnType. Therefore,
* GeneralProduct should never be used direclty. To determine the result type of a
* function which involves a matrix product, use ProductReturnType::Type.
*
* \sa ProductReturnType, MatrixBase::operator*(const MatrixBase<OtherDerived>&)
*/
template<typename Lhs, typename Rhs, int ProductType = internal::product_type<Lhs,Rhs>::value>
class GeneralProduct;
namespace Eigen {
enum {
Large = 2,
@@ -46,11 +24,17 @@ template<int Rows, int Cols, int Depth> struct product_type_selector;
template<int Size, int MaxSize> struct product_size_category
{
enum { is_large = MaxSize == Dynamic ||
Size >= EIGEN_CACHEFRIENDLY_PRODUCT_THRESHOLD,
value = is_large ? Large
: Size == 1 ? 1
: Small
enum {
#ifndef EIGEN_CUDA_ARCH
is_large = MaxSize == Dynamic ||
Size >= EIGEN_CACHEFRIENDLY_PRODUCT_THRESHOLD ||
(Size==Dynamic && MaxSize>=EIGEN_CACHEFRIENDLY_PRODUCT_THRESHOLD),
#else
is_large = 0,
#endif
value = is_large ? Large
: Size == 1 ? 1
: Small
};
};
@@ -59,15 +43,14 @@ template<typename Lhs, typename Rhs> struct product_type
typedef typename remove_all<Lhs>::type _Lhs;
typedef typename remove_all<Rhs>::type _Rhs;
enum {
MaxRows = _Lhs::MaxRowsAtCompileTime,
Rows = _Lhs::RowsAtCompileTime,
MaxCols = _Rhs::MaxColsAtCompileTime,
Cols = _Rhs::ColsAtCompileTime,
MaxDepth = EIGEN_SIZE_MIN_PREFER_FIXED(_Lhs::MaxColsAtCompileTime,
_Rhs::MaxRowsAtCompileTime),
Depth = EIGEN_SIZE_MIN_PREFER_FIXED(_Lhs::ColsAtCompileTime,
_Rhs::RowsAtCompileTime),
LargeThreshold = EIGEN_CACHEFRIENDLY_PRODUCT_THRESHOLD
MaxRows = traits<_Lhs>::MaxRowsAtCompileTime,
Rows = traits<_Lhs>::RowsAtCompileTime,
MaxCols = traits<_Rhs>::MaxColsAtCompileTime,
Cols = traits<_Rhs>::ColsAtCompileTime,
MaxDepth = EIGEN_SIZE_MIN_PREFER_FIXED(traits<_Lhs>::MaxColsAtCompileTime,
traits<_Rhs>::MaxRowsAtCompileTime),
Depth = EIGEN_SIZE_MIN_PREFER_FIXED(traits<_Lhs>::ColsAtCompileTime,
traits<_Rhs>::RowsAtCompileTime)
};
// the splitting into different lines of code here, introducing the _select enums and the typedef below,
@@ -82,7 +65,8 @@ private:
public:
enum {
value = selector::ret
value = selector::ret,
ret = selector::ret
};
#ifdef EIGEN_DEBUG_PRODUCT
static void debug()
@@ -98,12 +82,13 @@ public:
#endif
};
/* The following allows to select the kind of product at compile time
* based on the three dimensions of the product.
* This is a compile time mapping from {1,Small,Large}^3 -> {product types} */
// FIXME I'm not sure the current mapping is the ideal one.
template<int M, int N> struct product_type_selector<M,N,1> { enum { ret = OuterProduct }; };
template<int M> struct product_type_selector<M, 1, 1> { enum { ret = LazyCoeffBasedProductMode }; };
template<int N> struct product_type_selector<1, N, 1> { enum { ret = LazyCoeffBasedProductMode }; };
template<int Depth> struct product_type_selector<1, 1, Depth> { enum { ret = InnerProduct }; };
template<> struct product_type_selector<1, 1, 1> { enum { ret = InnerProduct }; };
template<> struct product_type_selector<Small,1, Small> { enum { ret = CoeffBasedProductMode }; };
@@ -122,60 +107,12 @@ template<> struct product_type_selector<Small,Small,Large> { enum
template<> struct product_type_selector<Large,Small,Large> { enum { ret = GemmProduct }; };
template<> struct product_type_selector<Small,Large,Large> { enum { ret = GemmProduct }; };
template<> struct product_type_selector<Large,Large,Large> { enum { ret = GemmProduct }; };
template<> struct product_type_selector<Large,Small,Small> { enum { ret = GemmProduct }; };
template<> struct product_type_selector<Small,Large,Small> { enum { ret = GemmProduct }; };
template<> struct product_type_selector<Large,Small,Small> { enum { ret = CoeffBasedProductMode }; };
template<> struct product_type_selector<Small,Large,Small> { enum { ret = CoeffBasedProductMode }; };
template<> struct product_type_selector<Large,Large,Small> { enum { ret = GemmProduct }; };
} // end namespace internal
/** \class ProductReturnType
* \ingroup Core_Module
*
* \brief Helper class to get the correct and optimized returned type of operator*
*
* \param Lhs the type of the left-hand side
* \param Rhs the type of the right-hand side
* \param ProductMode the type of the product (determined automatically by internal::product_mode)
*
* This class defines the typename Type representing the optimized product expression
* between two matrix expressions. In practice, using ProductReturnType<Lhs,Rhs>::Type
* is the recommended way to define the result type of a function returning an expression
* which involve a matrix product. The class Product should never be
* used directly.
*
* \sa class Product, MatrixBase::operator*(const MatrixBase<OtherDerived>&)
*/
template<typename Lhs, typename Rhs, int ProductType>
struct ProductReturnType
{
// TODO use the nested type to reduce instanciations ????
// typedef typename internal::nested<Lhs,Rhs::ColsAtCompileTime>::type LhsNested;
// typedef typename internal::nested<Rhs,Lhs::RowsAtCompileTime>::type RhsNested;
typedef GeneralProduct<Lhs/*Nested*/, Rhs/*Nested*/, ProductType> Type;
};
template<typename Lhs, typename Rhs>
struct ProductReturnType<Lhs,Rhs,CoeffBasedProductMode>
{
typedef typename internal::nested<Lhs, Rhs::ColsAtCompileTime, typename internal::plain_matrix_type<Lhs>::type >::type LhsNested;
typedef typename internal::nested<Rhs, Lhs::RowsAtCompileTime, typename internal::plain_matrix_type<Rhs>::type >::type RhsNested;
typedef CoeffBasedProduct<LhsNested, RhsNested, EvalBeforeAssigningBit | EvalBeforeNestingBit> Type;
};
template<typename Lhs, typename Rhs>
struct ProductReturnType<Lhs,Rhs,LazyCoeffBasedProductMode>
{
typedef typename internal::nested<Lhs, Rhs::ColsAtCompileTime, typename internal::plain_matrix_type<Lhs>::type >::type LhsNested;
typedef typename internal::nested<Rhs, Lhs::RowsAtCompileTime, typename internal::plain_matrix_type<Rhs>::type >::type RhsNested;
typedef CoeffBasedProduct<LhsNested, RhsNested, NestByRefBit> Type;
};
// this is a workaround for sun CC
template<typename Lhs, typename Rhs>
struct LazyProductReturnType : public ProductReturnType<Lhs,Rhs,LazyCoeffBasedProductMode>
{};
/***********************************************************************
* Implementation of Inner Vector Vector Product
***********************************************************************/
@@ -187,119 +124,10 @@ struct LazyProductReturnType : public ProductReturnType<Lhs,Rhs,LazyCoeffBasedPr
// product ends up to a row-vector times col-vector product... To tackle this use
// case, we could have a specialization for Block<MatrixType,1,1> with: operator=(Scalar x);
namespace internal {
template<typename Lhs, typename Rhs>
struct traits<GeneralProduct<Lhs,Rhs,InnerProduct> >
: traits<Matrix<typename scalar_product_traits<typename Lhs::Scalar, typename Rhs::Scalar>::ReturnType,1,1> >
{};
}
template<typename Lhs, typename Rhs>
class GeneralProduct<Lhs, Rhs, InnerProduct>
: internal::no_assignment_operator,
public Matrix<typename internal::scalar_product_traits<typename Lhs::Scalar, typename Rhs::Scalar>::ReturnType,1,1>
{
typedef Matrix<typename internal::scalar_product_traits<typename Lhs::Scalar, typename Rhs::Scalar>::ReturnType,1,1> Base;
public:
GeneralProduct(const Lhs& lhs, const Rhs& rhs)
{
EIGEN_STATIC_ASSERT((internal::is_same<typename Lhs::RealScalar, typename Rhs::RealScalar>::value),
YOU_MIXED_DIFFERENT_NUMERIC_TYPES__YOU_NEED_TO_USE_THE_CAST_METHOD_OF_MATRIXBASE_TO_CAST_NUMERIC_TYPES_EXPLICITLY)
Base::coeffRef(0,0) = (lhs.transpose().cwiseProduct(rhs)).sum();
}
/** Convertion to scalar */
operator const typename Base::Scalar() const {
return Base::coeff(0,0);
}
};
/***********************************************************************
* Implementation of Outer Vector Vector Product
***********************************************************************/
namespace internal {
// Column major
template<typename ProductType, typename Dest, typename Func>
EIGEN_DONT_INLINE void outer_product_selector_run(const ProductType& prod, Dest& dest, const Func& func, const false_type&)
{
typedef typename Dest::Index Index;
// FIXME make sure lhs is sequentially stored
// FIXME not very good if rhs is real and lhs complex while alpha is real too
const Index cols = dest.cols();
for (Index j=0; j<cols; ++j)
func(dest.col(j), prod.rhs().coeff(0,j) * prod.lhs());
}
// Row major
template<typename ProductType, typename Dest, typename Func>
EIGEN_DONT_INLINE void outer_product_selector_run(const ProductType& prod, Dest& dest, const Func& func, const true_type&) {
typedef typename Dest::Index Index;
// FIXME make sure rhs is sequentially stored
// FIXME not very good if lhs is real and rhs complex while alpha is real too
const Index rows = dest.rows();
for (Index i=0; i<rows; ++i)
func(dest.row(i), prod.lhs().coeff(i,0) * prod.rhs());
}
template<typename Lhs, typename Rhs>
struct traits<GeneralProduct<Lhs,Rhs,OuterProduct> >
: traits<ProductBase<GeneralProduct<Lhs,Rhs,OuterProduct>, Lhs, Rhs> >
{};
}
template<typename Lhs, typename Rhs>
class GeneralProduct<Lhs, Rhs, OuterProduct>
: public ProductBase<GeneralProduct<Lhs,Rhs,OuterProduct>, Lhs, Rhs>
{
template<typename T> struct is_row_major : internal::conditional<(int(T::Flags)&RowMajorBit), internal::true_type, internal::false_type>::type {};
public:
EIGEN_PRODUCT_PUBLIC_INTERFACE(GeneralProduct)
GeneralProduct(const Lhs& lhs, const Rhs& rhs) : Base(lhs,rhs)
{
EIGEN_STATIC_ASSERT((internal::is_same<typename Lhs::RealScalar, typename Rhs::RealScalar>::value),
YOU_MIXED_DIFFERENT_NUMERIC_TYPES__YOU_NEED_TO_USE_THE_CAST_METHOD_OF_MATRIXBASE_TO_CAST_NUMERIC_TYPES_EXPLICITLY)
}
struct set { template<typename Dst, typename Src> void operator()(const Dst& dst, const Src& src) const { dst.const_cast_derived() = src; } };
struct add { template<typename Dst, typename Src> void operator()(const Dst& dst, const Src& src) const { dst.const_cast_derived() += src; } };
struct sub { template<typename Dst, typename Src> void operator()(const Dst& dst, const Src& src) const { dst.const_cast_derived() -= src; } };
struct adds {
Scalar m_scale;
adds(const Scalar& s) : m_scale(s) {}
template<typename Dst, typename Src> void operator()(const Dst& dst, const Src& src) const {
dst.const_cast_derived() += m_scale * src;
}
};
template<typename Dest>
inline void evalTo(Dest& dest) const {
internal::outer_product_selector_run(*this, dest, set(), is_row_major<Dest>());
}
template<typename Dest>
inline void addTo(Dest& dest) const {
internal::outer_product_selector_run(*this, dest, add(), is_row_major<Dest>());
}
template<typename Dest>
inline void subTo(Dest& dest) const {
internal::outer_product_selector_run(*this, dest, sub(), is_row_major<Dest>());
}
template<typename Dest> void scaleAndAddTo(Dest& dest, const Scalar& alpha) const
{
internal::outer_product_selector_run(*this, dest, adds(alpha), is_row_major<Dest>());
}
};
/***********************************************************************
* Implementation of General Matrix Vector Product
***********************************************************************/
@@ -313,60 +141,13 @@ class GeneralProduct<Lhs, Rhs, OuterProduct>
*/
namespace internal {
template<typename Lhs, typename Rhs>
struct traits<GeneralProduct<Lhs,Rhs,GemvProduct> >
: traits<ProductBase<GeneralProduct<Lhs,Rhs,GemvProduct>, Lhs, Rhs> >
{};
template<int Side, int StorageOrder, bool BlasCompatible>
struct gemv_selector;
struct gemv_dense_selector;
} // end namespace internal
template<typename Lhs, typename Rhs>
class GeneralProduct<Lhs, Rhs, GemvProduct>
: public ProductBase<GeneralProduct<Lhs,Rhs,GemvProduct>, Lhs, Rhs>
{
public:
EIGEN_PRODUCT_PUBLIC_INTERFACE(GeneralProduct)
typedef typename Lhs::Scalar LhsScalar;
typedef typename Rhs::Scalar RhsScalar;
GeneralProduct(const Lhs& a_lhs, const Rhs& a_rhs) : Base(a_lhs,a_rhs)
{
// EIGEN_STATIC_ASSERT((internal::is_same<typename Lhs::Scalar, typename Rhs::Scalar>::value),
// YOU_MIXED_DIFFERENT_NUMERIC_TYPES__YOU_NEED_TO_USE_THE_CAST_METHOD_OF_MATRIXBASE_TO_CAST_NUMERIC_TYPES_EXPLICITLY)
}
enum { Side = Lhs::IsVectorAtCompileTime ? OnTheLeft : OnTheRight };
typedef typename internal::conditional<int(Side)==OnTheRight,_LhsNested,_RhsNested>::type MatrixType;
template<typename Dest> void scaleAndAddTo(Dest& dst, const Scalar& alpha) const
{
eigen_assert(m_lhs.rows() == dst.rows() && m_rhs.cols() == dst.cols());
internal::gemv_selector<Side,(int(MatrixType::Flags)&RowMajorBit) ? RowMajor : ColMajor,
bool(internal::blas_traits<MatrixType>::HasUsableDirectAccess)>::run(*this, dst, alpha);
}
};
namespace internal {
// The vector is on the left => transposition
template<int StorageOrder, bool BlasCompatible>
struct gemv_selector<OnTheLeft,StorageOrder,BlasCompatible>
{
template<typename ProductType, typename Dest>
static void run(const ProductType& prod, Dest& dest, const typename ProductType::Scalar& alpha)
{
Transpose<Dest> destT(dest);
enum { OtherStorageOrder = StorageOrder == RowMajor ? ColMajor : RowMajor };
gemv_selector<OnTheRight,OtherStorageOrder,BlasCompatible>
::run(GeneralProduct<Transpose<const typename ProductType::_RhsNested>,Transpose<const typename ProductType::_LhsNested>, GemvProduct>
(prod.rhs().transpose(), prod.lhs().transpose()), destT, alpha);
}
};
template<typename Scalar,int Size,int MaxSize,bool Cond> struct gemv_static_vector_if;
template<typename Scalar,int Size,int MaxSize>
@@ -384,126 +165,161 @@ struct gemv_static_vector_if<Scalar,Size,Dynamic,true>
template<typename Scalar,int Size,int MaxSize>
struct gemv_static_vector_if<Scalar,Size,MaxSize,true>
{
#if EIGEN_ALIGN_STATICALLY
internal::plain_array<Scalar,EIGEN_SIZE_MIN_PREFER_FIXED(Size,MaxSize),0> m_data;
EIGEN_STRONG_INLINE Scalar* data() { return m_data.array; }
#else
// Some architectures cannot align on the stack,
// => let's manually enforce alignment by allocating more data and return the address of the first aligned element.
enum {
ForceAlignment = internal::packet_traits<Scalar>::Vectorizable,
PacketSize = internal::packet_traits<Scalar>::size
};
internal::plain_array<Scalar,EIGEN_SIZE_MIN_PREFER_FIXED(Size,MaxSize)+(ForceAlignment?PacketSize:0),0> m_data;
#if EIGEN_MAX_STATIC_ALIGN_BYTES!=0
internal::plain_array<Scalar,EIGEN_SIZE_MIN_PREFER_FIXED(Size,MaxSize),0,EIGEN_PLAIN_ENUM_MIN(AlignedMax,PacketSize)> m_data;
EIGEN_STRONG_INLINE Scalar* data() { return m_data.array; }
#else
// Some architectures cannot align on the stack,
// => let's manually enforce alignment by allocating more data and return the address of the first aligned element.
internal::plain_array<Scalar,EIGEN_SIZE_MIN_PREFER_FIXED(Size,MaxSize)+(ForceAlignment?EIGEN_MAX_ALIGN_BYTES:0),0> m_data;
EIGEN_STRONG_INLINE Scalar* data() {
return ForceAlignment
? reinterpret_cast<Scalar*>((reinterpret_cast<size_t>(m_data.array) & ~(size_t(15))) + 16)
? reinterpret_cast<Scalar*>((internal::UIntPtr(m_data.array) & ~(std::size_t(EIGEN_MAX_ALIGN_BYTES-1))) + EIGEN_MAX_ALIGN_BYTES)
: m_data.array;
}
#endif
};
template<> struct gemv_selector<OnTheRight,ColMajor,true>
// The vector is on the left => transposition
template<int StorageOrder, bool BlasCompatible>
struct gemv_dense_selector<OnTheLeft,StorageOrder,BlasCompatible>
{
template<typename ProductType, typename Dest>
static inline void run(const ProductType& prod, Dest& dest, const typename ProductType::Scalar& alpha)
template<typename Lhs, typename Rhs, typename Dest>
static void run(const Lhs &lhs, const Rhs &rhs, Dest& dest, const typename Dest::Scalar& alpha)
{
typedef typename ProductType::Index Index;
typedef typename ProductType::LhsScalar LhsScalar;
typedef typename ProductType::RhsScalar RhsScalar;
typedef typename ProductType::Scalar ResScalar;
typedef typename ProductType::RealScalar RealScalar;
typedef typename ProductType::ActualLhsType ActualLhsType;
typedef typename ProductType::ActualRhsType ActualRhsType;
typedef typename ProductType::LhsBlasTraits LhsBlasTraits;
typedef typename ProductType::RhsBlasTraits RhsBlasTraits;
typedef Map<Matrix<ResScalar,Dynamic,1>, Aligned> MappedDest;
Transpose<Dest> destT(dest);
enum { OtherStorageOrder = StorageOrder == RowMajor ? ColMajor : RowMajor };
gemv_dense_selector<OnTheRight,OtherStorageOrder,BlasCompatible>
::run(rhs.transpose(), lhs.transpose(), destT, alpha);
}
};
ActualLhsType actualLhs = LhsBlasTraits::extract(prod.lhs());
ActualRhsType actualRhs = RhsBlasTraits::extract(prod.rhs());
template<> struct gemv_dense_selector<OnTheRight,ColMajor,true>
{
template<typename Lhs, typename Rhs, typename Dest>
static inline void run(const Lhs &lhs, const Rhs &rhs, Dest& dest, const typename Dest::Scalar& alpha)
{
typedef typename Lhs::Scalar LhsScalar;
typedef typename Rhs::Scalar RhsScalar;
typedef typename Dest::Scalar ResScalar;
typedef typename Dest::RealScalar RealScalar;
typedef internal::blas_traits<Lhs> LhsBlasTraits;
typedef typename LhsBlasTraits::DirectLinearAccessType ActualLhsType;
typedef internal::blas_traits<Rhs> RhsBlasTraits;
typedef typename RhsBlasTraits::DirectLinearAccessType ActualRhsType;
typedef Map<Matrix<ResScalar,Dynamic,1>, EIGEN_PLAIN_ENUM_MIN(AlignedMax,internal::packet_traits<ResScalar>::size)> MappedDest;
ResScalar actualAlpha = alpha * LhsBlasTraits::extractScalarFactor(prod.lhs())
* RhsBlasTraits::extractScalarFactor(prod.rhs());
ActualLhsType actualLhs = LhsBlasTraits::extract(lhs);
ActualRhsType actualRhs = RhsBlasTraits::extract(rhs);
ResScalar actualAlpha = alpha * LhsBlasTraits::extractScalarFactor(lhs)
* RhsBlasTraits::extractScalarFactor(rhs);
// make sure Dest is a compile-time vector type (bug 1166)
typedef typename conditional<Dest::IsVectorAtCompileTime, Dest, typename Dest::ColXpr>::type ActualDest;
enum {
// FIXME find a way to allow an inner stride on the result if packet_traits<Scalar>::size==1
// on, the other hand it is good for the cache to pack the vector anyways...
EvalToDestAtCompileTime = Dest::InnerStrideAtCompileTime==1,
EvalToDestAtCompileTime = (ActualDest::InnerStrideAtCompileTime==1),
ComplexByReal = (NumTraits<LhsScalar>::IsComplex) && (!NumTraits<RhsScalar>::IsComplex),
MightCannotUseDest = (Dest::InnerStrideAtCompileTime!=1) || ComplexByReal
MightCannotUseDest = (!EvalToDestAtCompileTime) || ComplexByReal
};
gemv_static_vector_if<ResScalar,Dest::SizeAtCompileTime,Dest::MaxSizeAtCompileTime,MightCannotUseDest> static_dest;
bool alphaIsCompatible = (!ComplexByReal) || (numext::imag(actualAlpha)==RealScalar(0));
bool evalToDest = EvalToDestAtCompileTime && alphaIsCompatible;
typedef const_blas_data_mapper<LhsScalar,Index,ColMajor> LhsMapper;
typedef const_blas_data_mapper<RhsScalar,Index,RowMajor> RhsMapper;
RhsScalar compatibleAlpha = get_factor<ResScalar,RhsScalar>::run(actualAlpha);
ei_declare_aligned_stack_constructed_variable(ResScalar,actualDestPtr,dest.size(),
evalToDest ? dest.data() : static_dest.data());
if(!evalToDest)
if(!MightCannotUseDest)
{
#ifdef EIGEN_DENSE_STORAGE_CTOR_PLUGIN
int size = dest.size();
EIGEN_DENSE_STORAGE_CTOR_PLUGIN
#endif
if(!alphaIsCompatible)
{
MappedDest(actualDestPtr, dest.size()).setZero();
compatibleAlpha = RhsScalar(1);
}
else
MappedDest(actualDestPtr, dest.size()) = dest;
// shortcut if we are sure to be able to use dest directly,
// this ease the compiler to generate cleaner and more optimzized code for most common cases
general_matrix_vector_product
<Index,LhsScalar,LhsMapper,ColMajor,LhsBlasTraits::NeedToConjugate,RhsScalar,RhsMapper,RhsBlasTraits::NeedToConjugate>::run(
actualLhs.rows(), actualLhs.cols(),
LhsMapper(actualLhs.data(), actualLhs.outerStride()),
RhsMapper(actualRhs.data(), actualRhs.innerStride()),
dest.data(), 1,
compatibleAlpha);
}
general_matrix_vector_product
<Index,LhsScalar,ColMajor,LhsBlasTraits::NeedToConjugate,RhsScalar,RhsBlasTraits::NeedToConjugate>::run(
actualLhs.rows(), actualLhs.cols(),
actualLhs.data(), actualLhs.outerStride(),
actualRhs.data(), actualRhs.innerStride(),
actualDestPtr, 1,
compatibleAlpha);
if (!evalToDest)
else
{
if(!alphaIsCompatible)
dest += actualAlpha * MappedDest(actualDestPtr, dest.size());
else
dest = MappedDest(actualDestPtr, dest.size());
gemv_static_vector_if<ResScalar,ActualDest::SizeAtCompileTime,ActualDest::MaxSizeAtCompileTime,MightCannotUseDest> static_dest;
const bool alphaIsCompatible = (!ComplexByReal) || (numext::imag(actualAlpha)==RealScalar(0));
const bool evalToDest = EvalToDestAtCompileTime && alphaIsCompatible;
ei_declare_aligned_stack_constructed_variable(ResScalar,actualDestPtr,dest.size(),
evalToDest ? dest.data() : static_dest.data());
if(!evalToDest)
{
#ifdef EIGEN_DENSE_STORAGE_CTOR_PLUGIN
Index size = dest.size();
EIGEN_DENSE_STORAGE_CTOR_PLUGIN
#endif
if(!alphaIsCompatible)
{
MappedDest(actualDestPtr, dest.size()).setZero();
compatibleAlpha = RhsScalar(1);
}
else
MappedDest(actualDestPtr, dest.size()) = dest;
}
general_matrix_vector_product
<Index,LhsScalar,LhsMapper,ColMajor,LhsBlasTraits::NeedToConjugate,RhsScalar,RhsMapper,RhsBlasTraits::NeedToConjugate>::run(
actualLhs.rows(), actualLhs.cols(),
LhsMapper(actualLhs.data(), actualLhs.outerStride()),
RhsMapper(actualRhs.data(), actualRhs.innerStride()),
actualDestPtr, 1,
compatibleAlpha);
if (!evalToDest)
{
if(!alphaIsCompatible)
dest.matrix() += actualAlpha * MappedDest(actualDestPtr, dest.size());
else
dest = MappedDest(actualDestPtr, dest.size());
}
}
}
};
template<> struct gemv_selector<OnTheRight,RowMajor,true>
template<> struct gemv_dense_selector<OnTheRight,RowMajor,true>
{
template<typename ProductType, typename Dest>
static void run(const ProductType& prod, Dest& dest, const typename ProductType::Scalar& alpha)
template<typename Lhs, typename Rhs, typename Dest>
static void run(const Lhs &lhs, const Rhs &rhs, Dest& dest, const typename Dest::Scalar& alpha)
{
typedef typename ProductType::LhsScalar LhsScalar;
typedef typename ProductType::RhsScalar RhsScalar;
typedef typename ProductType::Scalar ResScalar;
typedef typename ProductType::Index Index;
typedef typename ProductType::ActualLhsType ActualLhsType;
typedef typename ProductType::ActualRhsType ActualRhsType;
typedef typename ProductType::_ActualRhsType _ActualRhsType;
typedef typename ProductType::LhsBlasTraits LhsBlasTraits;
typedef typename ProductType::RhsBlasTraits RhsBlasTraits;
typedef typename Lhs::Scalar LhsScalar;
typedef typename Rhs::Scalar RhsScalar;
typedef typename Dest::Scalar ResScalar;
typedef internal::blas_traits<Lhs> LhsBlasTraits;
typedef typename LhsBlasTraits::DirectLinearAccessType ActualLhsType;
typedef internal::blas_traits<Rhs> RhsBlasTraits;
typedef typename RhsBlasTraits::DirectLinearAccessType ActualRhsType;
typedef typename internal::remove_all<ActualRhsType>::type ActualRhsTypeCleaned;
typename add_const<ActualLhsType>::type actualLhs = LhsBlasTraits::extract(prod.lhs());
typename add_const<ActualRhsType>::type actualRhs = RhsBlasTraits::extract(prod.rhs());
typename add_const<ActualLhsType>::type actualLhs = LhsBlasTraits::extract(lhs);
typename add_const<ActualRhsType>::type actualRhs = RhsBlasTraits::extract(rhs);
ResScalar actualAlpha = alpha * LhsBlasTraits::extractScalarFactor(prod.lhs())
* RhsBlasTraits::extractScalarFactor(prod.rhs());
ResScalar actualAlpha = alpha * LhsBlasTraits::extractScalarFactor(lhs)
* RhsBlasTraits::extractScalarFactor(rhs);
enum {
// FIXME find a way to allow an inner stride on the result if packet_traits<Scalar>::size==1
// on, the other hand it is good for the cache to pack the vector anyways...
DirectlyUseRhs = _ActualRhsType::InnerStrideAtCompileTime==1
DirectlyUseRhs = ActualRhsTypeCleaned::InnerStrideAtCompileTime==1
};
gemv_static_vector_if<RhsScalar,_ActualRhsType::SizeAtCompileTime,_ActualRhsType::MaxSizeAtCompileTime,!DirectlyUseRhs> static_rhs;
gemv_static_vector_if<RhsScalar,ActualRhsTypeCleaned::SizeAtCompileTime,ActualRhsTypeCleaned::MaxSizeAtCompileTime,!DirectlyUseRhs> static_rhs;
ei_declare_aligned_stack_constructed_variable(RhsScalar,actualRhsPtr,actualRhs.size(),
DirectlyUseRhs ? const_cast<RhsScalar*>(actualRhs.data()) : static_rhs.data());
@@ -511,45 +327,48 @@ template<> struct gemv_selector<OnTheRight,RowMajor,true>
if(!DirectlyUseRhs)
{
#ifdef EIGEN_DENSE_STORAGE_CTOR_PLUGIN
int size = actualRhs.size();
Index size = actualRhs.size();
EIGEN_DENSE_STORAGE_CTOR_PLUGIN
#endif
Map<typename _ActualRhsType::PlainObject>(actualRhsPtr, actualRhs.size()) = actualRhs;
Map<typename ActualRhsTypeCleaned::PlainObject>(actualRhsPtr, actualRhs.size()) = actualRhs;
}
typedef const_blas_data_mapper<LhsScalar,Index,RowMajor> LhsMapper;
typedef const_blas_data_mapper<RhsScalar,Index,ColMajor> RhsMapper;
general_matrix_vector_product
<Index,LhsScalar,RowMajor,LhsBlasTraits::NeedToConjugate,RhsScalar,RhsBlasTraits::NeedToConjugate>::run(
<Index,LhsScalar,LhsMapper,RowMajor,LhsBlasTraits::NeedToConjugate,RhsScalar,RhsMapper,RhsBlasTraits::NeedToConjugate>::run(
actualLhs.rows(), actualLhs.cols(),
actualLhs.data(), actualLhs.outerStride(),
actualRhsPtr, 1,
dest.data(), dest.innerStride(),
LhsMapper(actualLhs.data(), actualLhs.outerStride()),
RhsMapper(actualRhsPtr, 1),
dest.data(), dest.col(0).innerStride(), //NOTE if dest is not a vector at compile-time, then dest.innerStride() might be wrong. (bug 1166)
actualAlpha);
}
};
template<> struct gemv_selector<OnTheRight,ColMajor,false>
template<> struct gemv_dense_selector<OnTheRight,ColMajor,false>
{
template<typename ProductType, typename Dest>
static void run(const ProductType& prod, Dest& dest, const typename ProductType::Scalar& alpha)
template<typename Lhs, typename Rhs, typename Dest>
static void run(const Lhs &lhs, const Rhs &rhs, Dest& dest, const typename Dest::Scalar& alpha)
{
typedef typename Dest::Index Index;
// TODO makes sure dest is sequentially stored in memory, otherwise use a temp
const Index size = prod.rhs().rows();
EIGEN_STATIC_ASSERT((!nested_eval<Lhs,1>::Evaluate),EIGEN_INTERNAL_COMPILATION_ERROR_OR_YOU_MADE_A_PROGRAMMING_MISTAKE);
// TODO if rhs is large enough it might be beneficial to make sure that dest is sequentially stored in memory, otherwise use a temp
typename nested_eval<Rhs,1>::type actual_rhs(rhs);
const Index size = rhs.rows();
for(Index k=0; k<size; ++k)
dest += (alpha*prod.rhs().coeff(k)) * prod.lhs().col(k);
dest += (alpha*actual_rhs.coeff(k)) * lhs.col(k);
}
};
template<> struct gemv_selector<OnTheRight,RowMajor,false>
template<> struct gemv_dense_selector<OnTheRight,RowMajor,false>
{
template<typename ProductType, typename Dest>
static void run(const ProductType& prod, Dest& dest, const typename ProductType::Scalar& alpha)
template<typename Lhs, typename Rhs, typename Dest>
static void run(const Lhs &lhs, const Rhs &rhs, Dest& dest, const typename Dest::Scalar& alpha)
{
typedef typename Dest::Index Index;
// TODO makes sure rhs is sequentially stored in memory, otherwise use a temp
const Index rows = prod.rows();
EIGEN_STATIC_ASSERT((!nested_eval<Lhs,1>::Evaluate),EIGEN_INTERNAL_COMPILATION_ERROR_OR_YOU_MADE_A_PROGRAMMING_MISTAKE);
typename nested_eval<Rhs,Lhs::RowsAtCompileTime>::type actual_rhs(rhs);
const Index rows = dest.rows();
for(Index i=0; i<rows; ++i)
dest.coeffRef(i) += alpha * (prod.lhs().row(i).cwiseProduct(prod.rhs().transpose())).sum();
dest.coeffRef(i) += alpha * (lhs.row(i).cwiseProduct(actual_rhs.transpose())).sum();
}
};
@@ -567,7 +386,7 @@ template<> struct gemv_selector<OnTheRight,RowMajor,false>
*/
template<typename Derived>
template<typename OtherDerived>
inline const typename ProductReturnType<Derived, OtherDerived>::Type
inline const Product<Derived, OtherDerived>
MatrixBase<Derived>::operator*(const MatrixBase<OtherDerived> &other) const
{
// A note regarding the function declaration: In MSVC, this function will sometimes
@@ -592,7 +411,8 @@ MatrixBase<Derived>::operator*(const MatrixBase<OtherDerived> &other) const
#ifdef EIGEN_DEBUG_PRODUCT
internal::product_type<Derived,OtherDerived>::debug();
#endif
return typename ProductReturnType<Derived,OtherDerived>::Type(derived(), other.derived());
return Product<Derived, OtherDerived>(derived(), other.derived());
}
/** \returns an expression of the matrix product of \c *this and \a other without implicit evaluation.
@@ -608,7 +428,7 @@ MatrixBase<Derived>::operator*(const MatrixBase<OtherDerived> &other) const
*/
template<typename Derived>
template<typename OtherDerived>
const typename LazyProductReturnType<Derived,OtherDerived>::Type
const Product<Derived,OtherDerived,LazyProduct>
MatrixBase<Derived>::lazyProduct(const MatrixBase<OtherDerived> &other) const
{
enum {
@@ -627,7 +447,7 @@ MatrixBase<Derived>::lazyProduct(const MatrixBase<OtherDerived> &other) const
INVALID_MATRIX_PRODUCT__IF_YOU_WANTED_A_COEFF_WISE_PRODUCT_YOU_MUST_USE_THE_EXPLICIT_FUNCTION)
EIGEN_STATIC_ASSERT(ProductIsValid || SameSizes, INVALID_MATRIX_PRODUCT)
return typename LazyProductReturnType<Derived,OtherDerived>::Type(derived(), other.derived());
return Product<Derived,OtherDerived,LazyProduct>(derived(), other.derived());
}
} // end namespace Eigen

View File

@@ -42,21 +42,28 @@ namespace internal {
struct default_packet_traits
{
enum {
HasHalfPacket = 0,
HasAdd = 1,
HasSub = 1,
HasMul = 1,
HasNegate = 1,
HasAbs = 1,
HasArg = 0,
HasAbs2 = 1,
HasMin = 1,
HasMax = 1,
HasConj = 1,
HasSetLinear = 1,
HasBlend = 0,
HasDiv = 0,
HasSqrt = 0,
HasRsqrt = 0,
HasExp = 0,
HasLog = 0,
HasLog1p = 0,
HasLog10 = 0,
HasPow = 0,
HasSin = 0,
@@ -64,17 +71,37 @@ struct default_packet_traits
HasTan = 0,
HasASin = 0,
HasACos = 0,
HasATan = 0
HasATan = 0,
HasSinh = 0,
HasCosh = 0,
HasTanh = 0,
HasLGamma = 0,
HasDiGamma = 0,
HasZeta = 0,
HasPolygamma = 0,
HasErf = 0,
HasErfc = 0,
HasIGamma = 0,
HasIGammac = 0,
HasBetaInc = 0,
HasRound = 0,
HasFloor = 0,
HasCeil = 0,
HasSign = 0
};
};
template<typename T> struct packet_traits : default_packet_traits
{
typedef T type;
typedef T half;
enum {
Vectorizable = 0,
size = 1,
AlignedOnScalar = 0
AlignedOnScalar = 0,
HasHalfPacket = 0
};
enum {
HasAdd = 0,
@@ -90,135 +117,239 @@ template<typename T> struct packet_traits : default_packet_traits
};
};
template<typename T> struct packet_traits<const T> : packet_traits<T> { };
template <typename Src, typename Tgt> struct type_casting_traits {
enum {
VectorizedCast = 0,
SrcCoeffRatio = 1,
TgtCoeffRatio = 1
};
};
/** \internal \returns static_cast<TgtType>(a) (coeff-wise) */
template <typename SrcPacket, typename TgtPacket>
EIGEN_DEVICE_FUNC inline TgtPacket
pcast(const SrcPacket& a) {
return static_cast<TgtPacket>(a);
}
template <typename SrcPacket, typename TgtPacket>
EIGEN_DEVICE_FUNC inline TgtPacket
pcast(const SrcPacket& a, const SrcPacket& /*b*/) {
return static_cast<TgtPacket>(a);
}
template <typename SrcPacket, typename TgtPacket>
EIGEN_DEVICE_FUNC inline TgtPacket
pcast(const SrcPacket& a, const SrcPacket& /*b*/, const SrcPacket& /*c*/, const SrcPacket& /*d*/) {
return static_cast<TgtPacket>(a);
}
/** \internal \returns a + b (coeff-wise) */
template<typename Packet> inline Packet
template<typename Packet> EIGEN_DEVICE_FUNC inline Packet
padd(const Packet& a,
const Packet& b) { return a+b; }
/** \internal \returns a - b (coeff-wise) */
template<typename Packet> inline Packet
template<typename Packet> EIGEN_DEVICE_FUNC inline Packet
psub(const Packet& a,
const Packet& b) { return a-b; }
/** \internal \returns -a (coeff-wise) */
template<typename Packet> inline Packet
template<typename Packet> EIGEN_DEVICE_FUNC inline Packet
pnegate(const Packet& a) { return -a; }
/** \internal \returns conj(a) (coeff-wise) */
template<typename Packet> inline Packet
template<typename Packet> EIGEN_DEVICE_FUNC inline Packet
pconj(const Packet& a) { return numext::conj(a); }
/** \internal \returns a * b (coeff-wise) */
template<typename Packet> inline Packet
template<typename Packet> EIGEN_DEVICE_FUNC inline Packet
pmul(const Packet& a,
const Packet& b) { return a*b; }
/** \internal \returns a / b (coeff-wise) */
template<typename Packet> inline Packet
template<typename Packet> EIGEN_DEVICE_FUNC inline Packet
pdiv(const Packet& a,
const Packet& b) { return a/b; }
/** \internal \returns the min of \a a and \a b (coeff-wise) */
template<typename Packet> inline Packet
template<typename Packet> EIGEN_DEVICE_FUNC inline Packet
pmin(const Packet& a,
const Packet& b) { using std::min; return (min)(a, b); }
const Packet& b) { return numext::mini(a, b); }
/** \internal \returns the max of \a a and \a b (coeff-wise) */
template<typename Packet> inline Packet
template<typename Packet> EIGEN_DEVICE_FUNC inline Packet
pmax(const Packet& a,
const Packet& b) { using std::max; return (max)(a, b); }
const Packet& b) { return numext::maxi(a, b); }
/** \internal \returns the absolute value of \a a */
template<typename Packet> inline Packet
template<typename Packet> EIGEN_DEVICE_FUNC inline Packet
pabs(const Packet& a) { using std::abs; return abs(a); }
/** \internal \returns the phase angle of \a a */
template<typename Packet> EIGEN_DEVICE_FUNC inline Packet
parg(const Packet& a) { using numext::arg; return arg(a); }
/** \internal \returns the bitwise and of \a a and \a b */
template<typename Packet> inline Packet
template<typename Packet> EIGEN_DEVICE_FUNC inline Packet
pand(const Packet& a, const Packet& b) { return a & b; }
/** \internal \returns the bitwise or of \a a and \a b */
template<typename Packet> inline Packet
template<typename Packet> EIGEN_DEVICE_FUNC inline Packet
por(const Packet& a, const Packet& b) { return a | b; }
/** \internal \returns the bitwise xor of \a a and \a b */
template<typename Packet> inline Packet
template<typename Packet> EIGEN_DEVICE_FUNC inline Packet
pxor(const Packet& a, const Packet& b) { return a ^ b; }
/** \internal \returns the bitwise andnot of \a a and \a b */
template<typename Packet> inline Packet
template<typename Packet> EIGEN_DEVICE_FUNC inline Packet
pandnot(const Packet& a, const Packet& b) { return a & (!b); }
/** \internal \returns a packet version of \a *from, from must be 16 bytes aligned */
template<typename Packet> inline Packet
template<typename Packet> EIGEN_DEVICE_FUNC inline Packet
pload(const typename unpacket_traits<Packet>::type* from) { return *from; }
/** \internal \returns a packet version of \a *from, (un-aligned load) */
template<typename Packet> inline Packet
template<typename Packet> EIGEN_DEVICE_FUNC inline Packet
ploadu(const typename unpacket_traits<Packet>::type* from) { return *from; }
/** \internal \returns a packet with elements of \a *from duplicated.
* For instance, for a packet of 8 elements, 4 scalar will be read from \a *from and
* duplicated to form: {from[0],from[0],from[1],from[1],,from[2],from[2],,from[3],from[3]}
* Currently, this function is only used for scalar * complex products.
*/
template<typename Packet> inline Packet
ploaddup(const typename unpacket_traits<Packet>::type* from) { return *from; }
/** \internal \returns a packet with constant coefficients \a a, e.g.: (a,a,a,a) */
template<typename Packet> inline Packet
template<typename Packet> EIGEN_DEVICE_FUNC inline Packet
pset1(const typename unpacket_traits<Packet>::type& a) { return a; }
/** \internal \returns a packet with constant coefficients \a a[0], e.g.: (a[0],a[0],a[0],a[0]) */
template<typename Packet> EIGEN_DEVICE_FUNC inline Packet
pload1(const typename unpacket_traits<Packet>::type *a) { return pset1<Packet>(*a); }
/** \internal \returns a packet with elements of \a *from duplicated.
* For instance, for a packet of 8 elements, 4 scalars will be read from \a *from and
* duplicated to form: {from[0],from[0],from[1],from[1],from[2],from[2],from[3],from[3]}
* Currently, this function is only used for scalar * complex products.
*/
template<typename Packet> EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE Packet
ploaddup(const typename unpacket_traits<Packet>::type* from) { return *from; }
/** \internal \returns a packet with elements of \a *from quadrupled.
* For instance, for a packet of 8 elements, 2 scalars will be read from \a *from and
* replicated to form: {from[0],from[0],from[0],from[0],from[1],from[1],from[1],from[1]}
* Currently, this function is only used in matrix products.
* For packet-size smaller or equal to 4, this function is equivalent to pload1
*/
template<typename Packet> EIGEN_DEVICE_FUNC inline Packet
ploadquad(const typename unpacket_traits<Packet>::type* from)
{ return pload1<Packet>(from); }
/** \internal equivalent to
* \code
* a0 = pload1(a+0);
* a1 = pload1(a+1);
* a2 = pload1(a+2);
* a3 = pload1(a+3);
* \endcode
* \sa pset1, pload1, ploaddup, pbroadcast2
*/
template<typename Packet> EIGEN_DEVICE_FUNC
inline void pbroadcast4(const typename unpacket_traits<Packet>::type *a,
Packet& a0, Packet& a1, Packet& a2, Packet& a3)
{
a0 = pload1<Packet>(a+0);
a1 = pload1<Packet>(a+1);
a2 = pload1<Packet>(a+2);
a3 = pload1<Packet>(a+3);
}
/** \internal equivalent to
* \code
* a0 = pload1(a+0);
* a1 = pload1(a+1);
* \endcode
* \sa pset1, pload1, ploaddup, pbroadcast4
*/
template<typename Packet> EIGEN_DEVICE_FUNC
inline void pbroadcast2(const typename unpacket_traits<Packet>::type *a,
Packet& a0, Packet& a1)
{
a0 = pload1<Packet>(a+0);
a1 = pload1<Packet>(a+1);
}
/** \internal \brief Returns a packet with coefficients (a,a+1,...,a+packet_size-1). */
template<typename Scalar> inline typename packet_traits<Scalar>::type
plset(const Scalar& a) { return a; }
template<typename Packet> EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE Packet
plset(const typename unpacket_traits<Packet>::type& a) { return a; }
/** \internal copy the packet \a from to \a *to, \a to must be 16 bytes aligned */
template<typename Scalar, typename Packet> inline void pstore(Scalar* to, const Packet& from)
template<typename Scalar, typename Packet> EIGEN_DEVICE_FUNC inline void pstore(Scalar* to, const Packet& from)
{ (*to) = from; }
/** \internal copy the packet \a from to \a *to, (un-aligned store) */
template<typename Scalar, typename Packet> inline void pstoreu(Scalar* to, const Packet& from)
{ (*to) = from; }
template<typename Scalar, typename Packet> EIGEN_DEVICE_FUNC inline void pstoreu(Scalar* to, const Packet& from)
{ (*to) = from; }
template<typename Scalar, typename Packet> EIGEN_DEVICE_FUNC inline Packet pgather(const Scalar* from, Index /*stride*/)
{ return ploadu<Packet>(from); }
template<typename Scalar, typename Packet> EIGEN_DEVICE_FUNC inline void pscatter(Scalar* to, const Packet& from, Index /*stride*/)
{ pstore(to, from); }
/** \internal tries to do cache prefetching of \a addr */
template<typename Scalar> inline void prefetch(const Scalar* addr)
template<typename Scalar> EIGEN_DEVICE_FUNC inline void prefetch(const Scalar* addr)
{
#if !defined(_MSC_VER)
__builtin_prefetch(addr);
#ifdef __CUDA_ARCH__
#if defined(__LP64__)
// 64-bit pointer operand constraint for inlined asm
asm(" prefetch.L1 [ %1 ];" : "=l"(addr) : "l"(addr));
#else
// 32-bit pointer operand constraint for inlined asm
asm(" prefetch.L1 [ %1 ];" : "=r"(addr) : "r"(addr));
#endif
#elif (!EIGEN_COMP_MSVC) && (EIGEN_COMP_GNUC || EIGEN_COMP_CLANG || EIGEN_COMP_ICC)
__builtin_prefetch(addr);
#endif
}
/** \internal \returns the first element of a packet */
template<typename Packet> inline typename unpacket_traits<Packet>::type pfirst(const Packet& a)
template<typename Packet> EIGEN_DEVICE_FUNC inline typename unpacket_traits<Packet>::type pfirst(const Packet& a)
{ return a; }
/** \internal \returns a packet where the element i contains the sum of the packet of \a vec[i] */
template<typename Packet> inline Packet
template<typename Packet> EIGEN_DEVICE_FUNC inline Packet
preduxp(const Packet* vecs) { return vecs[0]; }
/** \internal \returns the sum of the elements of \a a*/
template<typename Packet> inline typename unpacket_traits<Packet>::type predux(const Packet& a)
template<typename Packet> EIGEN_DEVICE_FUNC inline typename unpacket_traits<Packet>::type predux(const Packet& a)
{ return a; }
/** \internal \returns the sum of the elements of \a a by block of 4 elements.
* For a packet {a0, a1, a2, a3, a4, a5, a6, a7}, it returns a half packet {a0+a4, a1+a5, a2+a6, a3+a7}
* For packet-size smaller or equal to 4, this boils down to a noop.
*/
template<typename Packet> EIGEN_DEVICE_FUNC inline
typename conditional<(unpacket_traits<Packet>::size%8)==0,typename unpacket_traits<Packet>::half,Packet>::type
predux_downto4(const Packet& a)
{ return a; }
/** \internal \returns the product of the elements of \a a*/
template<typename Packet> inline typename unpacket_traits<Packet>::type predux_mul(const Packet& a)
template<typename Packet> EIGEN_DEVICE_FUNC inline typename unpacket_traits<Packet>::type predux_mul(const Packet& a)
{ return a; }
/** \internal \returns the min of the elements of \a a*/
template<typename Packet> inline typename unpacket_traits<Packet>::type predux_min(const Packet& a)
template<typename Packet> EIGEN_DEVICE_FUNC inline typename unpacket_traits<Packet>::type predux_min(const Packet& a)
{ return a; }
/** \internal \returns the max of the elements of \a a*/
template<typename Packet> inline typename unpacket_traits<Packet>::type predux_max(const Packet& a)
template<typename Packet> EIGEN_DEVICE_FUNC inline typename unpacket_traits<Packet>::type predux_max(const Packet& a)
{ return a; }
/** \internal \returns the reversed elements of \a a*/
template<typename Packet> inline Packet preverse(const Packet& a)
template<typename Packet> EIGEN_DEVICE_FUNC inline Packet preverse(const Packet& a)
{ return a; }
/** \internal \returns \a a with real and imaginary part flipped (for complex type only) */
template<typename Packet> inline Packet pcplxflip(const Packet& a)
template<typename Packet> EIGEN_DEVICE_FUNC inline Packet pcplxflip(const Packet& a)
{
// FIXME: uncomment the following in case we drop the internal imag and real functions.
// using std::imag;
@@ -250,6 +381,22 @@ Packet pasin(const Packet& a) { using std::asin; return asin(a); }
template<typename Packet> EIGEN_DECLARE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS
Packet pacos(const Packet& a) { using std::acos; return acos(a); }
/** \internal \returns the arc tangent of \a a (coeff-wise) */
template<typename Packet> EIGEN_DECLARE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS
Packet patan(const Packet& a) { using std::atan; return atan(a); }
/** \internal \returns the hyperbolic sine of \a a (coeff-wise) */
template<typename Packet> EIGEN_DECLARE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS
Packet psinh(const Packet& a) { using std::sinh; return sinh(a); }
/** \internal \returns the hyperbolic cosine of \a a (coeff-wise) */
template<typename Packet> EIGEN_DECLARE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS
Packet pcosh(const Packet& a) { using std::cosh; return cosh(a); }
/** \internal \returns the hyperbolic tan of \a a (coeff-wise) */
template<typename Packet> EIGEN_DECLARE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS
Packet ptanh(const Packet& a) { using std::tanh; return tanh(a); }
/** \internal \returns the exp of \a a (coeff-wise) */
template<typename Packet> EIGEN_DECLARE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS
Packet pexp(const Packet& a) { using std::exp; return exp(a); }
@@ -258,10 +405,36 @@ Packet pexp(const Packet& a) { using std::exp; return exp(a); }
template<typename Packet> EIGEN_DECLARE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS
Packet plog(const Packet& a) { using std::log; return log(a); }
/** \internal \returns the log1p of \a a (coeff-wise) */
template<typename Packet> EIGEN_DECLARE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS
Packet plog1p(const Packet& a) { return numext::log1p(a); }
/** \internal \returns the log10 of \a a (coeff-wise) */
template<typename Packet> EIGEN_DECLARE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS
Packet plog10(const Packet& a) { using std::log10; return log10(a); }
/** \internal \returns the square-root of \a a (coeff-wise) */
template<typename Packet> EIGEN_DECLARE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS
Packet psqrt(const Packet& a) { using std::sqrt; return sqrt(a); }
/** \internal \returns the reciprocal square-root of \a a (coeff-wise) */
template<typename Packet> EIGEN_DECLARE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS
Packet prsqrt(const Packet& a) {
return pdiv(pset1<Packet>(1), psqrt(a));
}
/** \internal \returns the rounded value of \a a (coeff-wise) */
template<typename Packet> EIGEN_DECLARE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS
Packet pround(const Packet& a) { using numext::round; return round(a); }
/** \internal \returns the floor of \a a (coeff-wise) */
template<typename Packet> EIGEN_DECLARE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS
Packet pfloor(const Packet& a) { using numext::floor; return floor(a); }
/** \internal \returns the ceil of \a a (coeff-wise) */
template<typename Packet> EIGEN_DECLARE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS
Packet pceil(const Packet& a) { using numext::ceil; return ceil(a); }
/***************************************************************************
* The following functions might not have to be overwritten for vectorized types
***************************************************************************/
@@ -275,34 +448,45 @@ inline void pstore1(typename unpacket_traits<Packet>::type* to, const typename u
}
/** \internal \returns a * b + c (coeff-wise) */
template<typename Packet> inline Packet
template<typename Packet> EIGEN_DEVICE_FUNC inline Packet
pmadd(const Packet& a,
const Packet& b,
const Packet& c)
{ return padd(pmul(a, b),c); }
/** \internal \returns a packet version of \a *from.
* If LoadMode equals #Aligned, \a from must be 16 bytes aligned */
template<typename Packet, int LoadMode>
inline Packet ploadt(const typename unpacket_traits<Packet>::type* from)
* The pointer \a from must be aligned on a \a Alignment bytes boundary. */
template<typename Packet, int Alignment>
EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE Packet ploadt(const typename unpacket_traits<Packet>::type* from)
{
if(LoadMode == Aligned)
if(Alignment >= unpacket_traits<Packet>::alignment)
return pload<Packet>(from);
else
return ploadu<Packet>(from);
}
/** \internal copy the packet \a from to \a *to.
* If StoreMode equals #Aligned, \a to must be 16 bytes aligned */
template<typename Scalar, typename Packet, int LoadMode>
inline void pstoret(Scalar* to, const Packet& from)
* The pointer \a from must be aligned on a \a Alignment bytes boundary. */
template<typename Scalar, typename Packet, int Alignment>
EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE void pstoret(Scalar* to, const Packet& from)
{
if(LoadMode == Aligned)
if(Alignment >= unpacket_traits<Packet>::alignment)
pstore(to, from);
else
pstoreu(to, from);
}
/** \internal \returns a packet version of \a *from.
* Unlike ploadt, ploadt_ro takes advantage of the read-only memory path on the
* hardware if available to speedup the loading of data that won't be modified
* by the current computation.
*/
template<typename Packet, int LoadMode>
EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE Packet ploadt_ro(const typename unpacket_traits<Packet>::type* from)
{
return ploadt<Packet, LoadMode>(from);
}
/** \internal default implementation of palign() allowing partial specialization */
template<int Offset,typename PacketType>
struct palign_impl
@@ -336,15 +520,74 @@ inline void palign(PacketType& first, const PacketType& second)
* Fast complex products (GCC generates a function call which is very slow)
***************************************************************************/
// Eigen+CUDA does not support complexes.
#ifndef __CUDACC__
template<> inline std::complex<float> pmul(const std::complex<float>& a, const std::complex<float>& b)
{ return std::complex<float>(real(a)*real(b) - imag(a)*imag(b), imag(a)*real(b) + real(a)*imag(b)); }
template<> inline std::complex<double> pmul(const std::complex<double>& a, const std::complex<double>& b)
{ return std::complex<double>(real(a)*real(b) - imag(a)*imag(b), imag(a)*real(b) + real(a)*imag(b)); }
#endif
/***************************************************************************
* PacketBlock, that is a collection of N packets where the number of words
* in the packet is a multiple of N.
***************************************************************************/
template <typename Packet,int N=unpacket_traits<Packet>::size> struct PacketBlock {
Packet packet[N];
};
template<typename Packet> EIGEN_DEVICE_FUNC inline void
ptranspose(PacketBlock<Packet,1>& /*kernel*/) {
// Nothing to do in the scalar case, i.e. a 1x1 matrix.
}
/***************************************************************************
* Selector, i.e. vector of N boolean values used to select (i.e. blend)
* words from 2 packets.
***************************************************************************/
template <size_t N> struct Selector {
bool select[N];
};
template<typename Packet> EIGEN_DEVICE_FUNC inline Packet
pblend(const Selector<unpacket_traits<Packet>::size>& ifPacket, const Packet& thenPacket, const Packet& elsePacket) {
return ifPacket.select[0] ? thenPacket : elsePacket;
}
/** \internal \returns \a a with the first coefficient replaced by the scalar b */
template<typename Packet> EIGEN_DEVICE_FUNC inline Packet
pinsertfirst(const Packet& a, typename unpacket_traits<Packet>::type b)
{
// Default implementation based on pblend.
// It must be specialized for higher performance.
Selector<unpacket_traits<Packet>::size> mask;
mask.select[0] = true;
// This for loop should be optimized away by the compiler.
for(Index i=1; i<unpacket_traits<Packet>::size; ++i)
mask.select[i] = false;
return pblend(mask, pset1<Packet>(b), a);
}
/** \internal \returns \a a with the last coefficient replaced by the scalar b */
template<typename Packet> EIGEN_DEVICE_FUNC inline Packet
pinsertlast(const Packet& a, typename unpacket_traits<Packet>::type b)
{
// Default implementation based on pblend.
// It must be specialized for higher performance.
Selector<unpacket_traits<Packet>::size> mask;
// This for loop should be optimized away by the compiler.
for(Index i=0; i<unpacket_traits<Packet>::size-1; ++i)
mask.select[i] = false;
mask.select[unpacket_traits<Packet>::size-1] = true;
return pblend(mask, pset1<Packet>(b), a);
}
} // end namespace internal
} // end namespace Eigen
#endif // EIGEN_GENERIC_PACKET_MATH_H

View File

@@ -1,7 +1,7 @@
// This file is part of Eigen, a lightweight C++ template library
// for linear algebra.
//
// Copyright (C) 2010-2012 Gael Guennebaud <gael.guennebaud@inria.fr>
// Copyright (C) 2010-2016 Gael Guennebaud <gael.guennebaud@inria.fr>
// Copyright (C) 2010 Benoit Jacob <jacob.benoit.1@gmail.com>
//
// This Source Code Form is subject to the terms of the Mozilla
@@ -11,13 +11,30 @@
#ifndef EIGEN_GLOBAL_FUNCTIONS_H
#define EIGEN_GLOBAL_FUNCTIONS_H
#define EIGEN_ARRAY_DECLARE_GLOBAL_UNARY(NAME,FUNCTOR) \
#ifdef EIGEN_PARSED_BY_DOXYGEN
#define EIGEN_ARRAY_DECLARE_GLOBAL_UNARY(NAME,FUNCTOR,DOC_OP,DOC_DETAILS) \
/** \returns an expression of the coefficient-wise DOC_OP of \a x
DOC_DETAILS
\sa <a href="group__CoeffwiseMathFunctions.html#cwisetable_##NAME">Math functions</a>, class CwiseUnaryOp
*/ \
template<typename Derived> \
inline const Eigen::CwiseUnaryOp<Eigen::internal::FUNCTOR<typename Derived::Scalar>, const Derived> \
NAME(const Eigen::ArrayBase<Derived>& x) { \
return x.derived(); \
NAME(const Eigen::ArrayBase<Derived>& x);
#else
#define EIGEN_ARRAY_DECLARE_GLOBAL_UNARY(NAME,FUNCTOR,DOC_OP,DOC_DETAILS) \
template<typename Derived> \
inline const Eigen::CwiseUnaryOp<Eigen::internal::FUNCTOR<typename Derived::Scalar>, const Derived> \
(NAME)(const Eigen::ArrayBase<Derived>& x) { \
return Eigen::CwiseUnaryOp<Eigen::internal::FUNCTOR<typename Derived::Scalar>, const Derived>(x.derived()); \
}
#endif // EIGEN_PARSED_BY_DOXYGEN
#define EIGEN_ARRAY_DECLARE_GLOBAL_EIGEN_UNARY(NAME,FUNCTOR) \
\
template<typename Derived> \
@@ -30,55 +47,133 @@
{ \
static inline typename NAME##_retval<ArrayBase<Derived> >::type run(const Eigen::ArrayBase<Derived>& x) \
{ \
return x.derived(); \
return typename NAME##_retval<ArrayBase<Derived> >::type(x.derived()); \
} \
};
namespace Eigen
{
EIGEN_ARRAY_DECLARE_GLOBAL_UNARY(real,scalar_real_op)
EIGEN_ARRAY_DECLARE_GLOBAL_UNARY(imag,scalar_imag_op)
EIGEN_ARRAY_DECLARE_GLOBAL_UNARY(conj,scalar_conjugate_op)
EIGEN_ARRAY_DECLARE_GLOBAL_UNARY(sin,scalar_sin_op)
EIGEN_ARRAY_DECLARE_GLOBAL_UNARY(cos,scalar_cos_op)
EIGEN_ARRAY_DECLARE_GLOBAL_UNARY(asin,scalar_asin_op)
EIGEN_ARRAY_DECLARE_GLOBAL_UNARY(acos,scalar_acos_op)
EIGEN_ARRAY_DECLARE_GLOBAL_UNARY(tan,scalar_tan_op)
EIGEN_ARRAY_DECLARE_GLOBAL_UNARY(exp,scalar_exp_op)
EIGEN_ARRAY_DECLARE_GLOBAL_UNARY(log,scalar_log_op)
EIGEN_ARRAY_DECLARE_GLOBAL_UNARY(abs,scalar_abs_op)
EIGEN_ARRAY_DECLARE_GLOBAL_UNARY(sqrt,scalar_sqrt_op)
EIGEN_ARRAY_DECLARE_GLOBAL_UNARY(real,scalar_real_op,real part,\sa ArrayBase::real)
EIGEN_ARRAY_DECLARE_GLOBAL_UNARY(imag,scalar_imag_op,imaginary part,\sa ArrayBase::imag)
EIGEN_ARRAY_DECLARE_GLOBAL_UNARY(conj,scalar_conjugate_op,complex conjugate,\sa ArrayBase::conjugate)
EIGEN_ARRAY_DECLARE_GLOBAL_UNARY(inverse,scalar_inverse_op,inverse,\sa ArrayBase::inverse)
EIGEN_ARRAY_DECLARE_GLOBAL_UNARY(sin,scalar_sin_op,sine,\sa ArrayBase::sin)
EIGEN_ARRAY_DECLARE_GLOBAL_UNARY(cos,scalar_cos_op,cosine,\sa ArrayBase::cos)
EIGEN_ARRAY_DECLARE_GLOBAL_UNARY(tan,scalar_tan_op,tangent,\sa ArrayBase::tan)
EIGEN_ARRAY_DECLARE_GLOBAL_UNARY(atan,scalar_atan_op,arc-tangent,\sa ArrayBase::atan)
EIGEN_ARRAY_DECLARE_GLOBAL_UNARY(asin,scalar_asin_op,arc-sine,\sa ArrayBase::asin)
EIGEN_ARRAY_DECLARE_GLOBAL_UNARY(acos,scalar_acos_op,arc-consine,\sa ArrayBase::acos)
EIGEN_ARRAY_DECLARE_GLOBAL_UNARY(sinh,scalar_sinh_op,hyperbolic sine,\sa ArrayBase::sinh)
EIGEN_ARRAY_DECLARE_GLOBAL_UNARY(cosh,scalar_cosh_op,hyperbolic cosine,\sa ArrayBase::cosh)
EIGEN_ARRAY_DECLARE_GLOBAL_UNARY(tanh,scalar_tanh_op,hyperbolic tangent,\sa ArrayBase::tanh)
EIGEN_ARRAY_DECLARE_GLOBAL_UNARY(lgamma,scalar_lgamma_op,natural logarithm of the gamma function,\sa ArrayBase::lgamma)
EIGEN_ARRAY_DECLARE_GLOBAL_UNARY(digamma,scalar_digamma_op,derivative of lgamma,\sa ArrayBase::digamma)
EIGEN_ARRAY_DECLARE_GLOBAL_UNARY(erf,scalar_erf_op,error function,\sa ArrayBase::erf)
EIGEN_ARRAY_DECLARE_GLOBAL_UNARY(erfc,scalar_erfc_op,complement error function,\sa ArrayBase::erfc)
EIGEN_ARRAY_DECLARE_GLOBAL_UNARY(exp,scalar_exp_op,exponential,\sa ArrayBase::exp)
EIGEN_ARRAY_DECLARE_GLOBAL_UNARY(log,scalar_log_op,natural logarithm,\sa Eigen::log10 DOXCOMMA ArrayBase::log)
EIGEN_ARRAY_DECLARE_GLOBAL_UNARY(log1p,scalar_log1p_op,natural logarithm of 1 plus the value,\sa ArrayBase::log1p)
EIGEN_ARRAY_DECLARE_GLOBAL_UNARY(log10,scalar_log10_op,base 10 logarithm,\sa Eigen::log DOXCOMMA ArrayBase::log)
EIGEN_ARRAY_DECLARE_GLOBAL_UNARY(abs,scalar_abs_op,absolute value,\sa ArrayBase::abs DOXCOMMA MatrixBase::cwiseAbs)
EIGEN_ARRAY_DECLARE_GLOBAL_UNARY(abs2,scalar_abs2_op,squared absolute value,\sa ArrayBase::abs2 DOXCOMMA MatrixBase::cwiseAbs2)
EIGEN_ARRAY_DECLARE_GLOBAL_UNARY(arg,scalar_arg_op,complex argument,\sa ArrayBase::arg)
EIGEN_ARRAY_DECLARE_GLOBAL_UNARY(sqrt,scalar_sqrt_op,square root,\sa ArrayBase::sqrt DOXCOMMA MatrixBase::cwiseSqrt)
EIGEN_ARRAY_DECLARE_GLOBAL_UNARY(rsqrt,scalar_rsqrt_op,reciprocal square root,\sa ArrayBase::rsqrt)
EIGEN_ARRAY_DECLARE_GLOBAL_UNARY(square,scalar_square_op,square (power 2),\sa Eigen::abs2 DOXCOMMA Eigen::pow DOXCOMMA ArrayBase::square)
EIGEN_ARRAY_DECLARE_GLOBAL_UNARY(cube,scalar_cube_op,cube (power 3),\sa Eigen::pow DOXCOMMA ArrayBase::cube)
EIGEN_ARRAY_DECLARE_GLOBAL_UNARY(round,scalar_round_op,nearest integer,\sa Eigen::floor DOXCOMMA Eigen::ceil DOXCOMMA ArrayBase::round)
EIGEN_ARRAY_DECLARE_GLOBAL_UNARY(floor,scalar_floor_op,nearest integer not greater than the giben value,\sa Eigen::ceil DOXCOMMA ArrayBase::floor)
EIGEN_ARRAY_DECLARE_GLOBAL_UNARY(ceil,scalar_ceil_op,nearest integer not less than the giben value,\sa Eigen::floor DOXCOMMA ArrayBase::ceil)
EIGEN_ARRAY_DECLARE_GLOBAL_UNARY(isnan,scalar_isnan_op,not-a-number test,\sa Eigen::isinf DOXCOMMA Eigen::isfinite DOXCOMMA ArrayBase::isnan)
EIGEN_ARRAY_DECLARE_GLOBAL_UNARY(isinf,scalar_isinf_op,infinite value test,\sa Eigen::isnan DOXCOMMA Eigen::isfinite DOXCOMMA ArrayBase::isinf)
EIGEN_ARRAY_DECLARE_GLOBAL_UNARY(isfinite,scalar_isfinite_op,finite value test,\sa Eigen::isinf DOXCOMMA Eigen::isnan DOXCOMMA ArrayBase::isfinite)
EIGEN_ARRAY_DECLARE_GLOBAL_UNARY(sign,scalar_sign_op,sign (or 0),\sa ArrayBase::sign)
template<typename Derived>
inline const Eigen::CwiseUnaryOp<Eigen::internal::scalar_pow_op<typename Derived::Scalar>, const Derived>
pow(const Eigen::ArrayBase<Derived>& x, const typename Derived::Scalar& exponent) {
/** \returns an expression of the coefficient-wise power of \a x to the given constant \a exponent.
*
* \tparam ScalarExponent is the scalar type of \a exponent. It must be compatible with the scalar type of the given expression (\c Derived::Scalar).
*
* \sa ArrayBase::pow()
*
* \relates ArrayBase
*/
#ifdef EIGEN_PARSED_BY_DOXYGEN
template<typename Derived,typename ScalarExponent>
inline const CwiseBinaryOp<internal::scalar_pow_op<Derived::Scalar,ScalarExponent>,Derived,Constant<ScalarExponent> >
pow(const Eigen::ArrayBase<Derived>& x, const ScalarExponent& exponent);
#else
template<typename Derived,typename ScalarExponent>
inline typename internal::enable_if< !(internal::is_same<typename Derived::Scalar,ScalarExponent>::value) && EIGEN_SCALAR_BINARY_SUPPORTED(pow,typename Derived::Scalar,ScalarExponent),
const EIGEN_EXPR_BINARYOP_SCALAR_RETURN_TYPE(Derived,ScalarExponent,pow) >::type
pow(const Eigen::ArrayBase<Derived>& x, const ScalarExponent& exponent) {
return x.derived().pow(exponent);
}
template<typename Derived>
inline const Eigen::CwiseBinaryOp<Eigen::internal::scalar_binary_pow_op<typename Derived::Scalar, typename Derived::Scalar>, const Derived, const Derived>
pow(const Eigen::ArrayBase<Derived>& x, const Eigen::ArrayBase<Derived>& exponents)
inline const EIGEN_EXPR_BINARYOP_SCALAR_RETURN_TYPE(Derived,typename Derived::Scalar,pow)
pow(const Eigen::ArrayBase<Derived>& x, const typename Derived::Scalar& exponent) {
return x.derived().pow(exponent);
}
#endif
/** \returns an expression of the coefficient-wise power of \a x to the given array of \a exponents.
*
* This function computes the coefficient-wise power.
*
* Example: \include Cwise_array_power_array.cpp
* Output: \verbinclude Cwise_array_power_array.out
*
* \sa ArrayBase::pow()
*
* \relates ArrayBase
*/
template<typename Derived,typename ExponentDerived>
inline const Eigen::CwiseBinaryOp<Eigen::internal::scalar_pow_op<typename Derived::Scalar, typename ExponentDerived::Scalar>, const Derived, const ExponentDerived>
pow(const Eigen::ArrayBase<Derived>& x, const Eigen::ArrayBase<ExponentDerived>& exponents)
{
return Eigen::CwiseBinaryOp<Eigen::internal::scalar_binary_pow_op<typename Derived::Scalar, typename Derived::Scalar>, const Derived, const Derived>(
return Eigen::CwiseBinaryOp<Eigen::internal::scalar_pow_op<typename Derived::Scalar, typename ExponentDerived::Scalar>, const Derived, const ExponentDerived>(
x.derived(),
exponents.derived()
);
}
/**
* \brief Component-wise division of a scalar by array elements.
**/
template <typename Derived>
inline const Eigen::CwiseUnaryOp<Eigen::internal::scalar_inverse_mult_op<typename Derived::Scalar>, const Derived>
operator/(const typename Derived::Scalar& s, const Eigen::ArrayBase<Derived>& a)
/** \returns an expression of the coefficient-wise power of the scalar \a x to the given array of \a exponents.
*
* This function computes the coefficient-wise power between a scalar and an array of exponents.
*
* \tparam Scalar is the scalar type of \a x. It must be compatible with the scalar type of the given array expression (\c Derived::Scalar).
*
* Example: \include Cwise_scalar_power_array.cpp
* Output: \verbinclude Cwise_scalar_power_array.out
*
* \sa ArrayBase::pow()
*
* \relates ArrayBase
*/
#ifdef EIGEN_PARSED_BY_DOXYGEN
template<typename Scalar,typename Derived>
inline const CwiseBinaryOp<internal::scalar_pow_op<Scalar,Derived::Scalar>,Constant<Scalar>,Derived>
pow(const Scalar& x,const Eigen::ArrayBase<Derived>& x);
#else
template<typename Scalar, typename Derived>
inline typename internal::enable_if< !(internal::is_same<typename Derived::Scalar,Scalar>::value) && EIGEN_SCALAR_BINARY_SUPPORTED(pow,Scalar,typename Derived::Scalar),
const EIGEN_SCALAR_BINARYOP_EXPR_RETURN_TYPE(Scalar,Derived,pow) >::type
pow(const Scalar& x, const Eigen::ArrayBase<Derived>& exponents)
{
return Eigen::CwiseUnaryOp<Eigen::internal::scalar_inverse_mult_op<typename Derived::Scalar>, const Derived>(
a.derived(),
Eigen::internal::scalar_inverse_mult_op<typename Derived::Scalar>(s)
);
return EIGEN_SCALAR_BINARYOP_EXPR_RETURN_TYPE(Scalar,Derived,pow)(
typename internal::plain_constant_type<Derived,Scalar>::type(exponents.rows(), exponents.cols(), x), exponents.derived() );
}
template<typename Derived>
inline const EIGEN_SCALAR_BINARYOP_EXPR_RETURN_TYPE(typename Derived::Scalar,Derived,pow)
pow(const typename Derived::Scalar& x, const Eigen::ArrayBase<Derived>& exponents)
{
return EIGEN_SCALAR_BINARYOP_EXPR_RETURN_TYPE(typename Derived::Scalar,Derived,pow)(
typename internal::plain_constant_type<Derived,typename Derived::Scalar>::type(exponents.rows(), exponents.cols(), x), exponents.derived() );
}
#endif
namespace internal
{
EIGEN_ARRAY_DECLARE_GLOBAL_EIGEN_UNARY(real,scalar_real_op)

View File

@@ -49,7 +49,7 @@ std::ostream & print_matrix(std::ostream & s, const Derived& _m, const IOFormat&
*/
struct IOFormat
{
/** Default contructor, see class IOFormat for the meaning of the parameters */
/** Default constructor, see class IOFormat for the meaning of the parameters */
IOFormat(int _precision = StreamPrecision, int _flags = 0,
const std::string& _coeffSeparator = " ",
const std::string& _rowSeparator = "\n", const std::string& _rowPrefix="", const std::string& _rowSuffix="",
@@ -57,6 +57,10 @@ struct IOFormat
: matPrefix(_matPrefix), matSuffix(_matSuffix), rowPrefix(_rowPrefix), rowSuffix(_rowSuffix), rowSeparator(_rowSeparator),
rowSpacer(""), coeffSeparator(_coeffSeparator), precision(_precision), flags(_flags)
{
// TODO check if rowPrefix, rowSuffix or rowSeparator contains a newline
// don't add rowSpacer if columns are not to be aligned
if((flags & DontAlignCols))
return;
int i = int(matSuffix.length())-1;
while (i>=0 && matSuffix[i]!='\n')
{
@@ -76,7 +80,7 @@ struct IOFormat
*
* \brief Pseudo expression providing matrix output with given format
*
* \param ExpressionType the type of the object on which IO stream operations are performed
* \tparam ExpressionType the type of the object on which IO stream operations are performed
*
* This class represents an expression with stream operators controlled by a given IOFormat.
* It is the return type of DenseBase::format()
@@ -101,51 +105,23 @@ class WithFormat
}
protected:
const typename ExpressionType::Nested m_matrix;
typename ExpressionType::Nested m_matrix;
IOFormat m_format;
};
/** \returns a WithFormat proxy object allowing to print a matrix the with given
* format \a fmt.
*
* See class IOFormat for some examples.
*
* \sa class IOFormat, class WithFormat
*/
template<typename Derived>
inline const WithFormat<Derived>
DenseBase<Derived>::format(const IOFormat& fmt) const
{
return WithFormat<Derived>(derived(), fmt);
}
namespace internal {
template<typename Scalar, bool IsInteger>
struct significant_decimals_default_impl
{
typedef typename NumTraits<Scalar>::Real RealScalar;
static inline int run()
{
using std::ceil;
using std::log;
return cast<RealScalar,int>(ceil(-log(NumTraits<RealScalar>::epsilon())/log(RealScalar(10))));
}
};
template<typename Scalar>
struct significant_decimals_default_impl<Scalar, true>
{
static inline int run()
{
return 0;
}
};
// NOTE: This helper is kept for backward compatibility with previous code specializing
// this internal::significant_decimals_impl structure. In the future we should directly
// call digits10() which has been introduced in July 2016 in 3.3.
template<typename Scalar>
struct significant_decimals_impl
: significant_decimals_default_impl<Scalar, NumTraits<Scalar>::IsInteger>
{};
{
static inline int run()
{
return NumTraits<Scalar>::digits10();
}
};
/** \internal
* print the matrix \a _m to the output stream \a s using the output format \a fmt */
@@ -160,7 +136,6 @@ std::ostream & print_matrix(std::ostream & s, const Derived& _m, const IOFormat&
typename Derived::Nested m = _m;
typedef typename Derived::Scalar Scalar;
typedef typename Derived::Index Index;
Index width = 0;

118
extern/Eigen3/Eigen/src/Core/Inverse.h vendored Normal file
View File

@@ -0,0 +1,118 @@
// This file is part of Eigen, a lightweight C++ template library
// for linear algebra.
//
// Copyright (C) 2014 Gael Guennebaud <gael.guennebaud@inria.fr>
//
// This Source Code Form is subject to the terms of the Mozilla
// Public License v. 2.0. If a copy of the MPL was not distributed
// with this file, You can obtain one at http://mozilla.org/MPL/2.0/.
#ifndef EIGEN_INVERSE_H
#define EIGEN_INVERSE_H
namespace Eigen {
template<typename XprType,typename StorageKind> class InverseImpl;
namespace internal {
template<typename XprType>
struct traits<Inverse<XprType> >
: traits<typename XprType::PlainObject>
{
typedef typename XprType::PlainObject PlainObject;
typedef traits<PlainObject> BaseTraits;
enum {
Flags = BaseTraits::Flags & RowMajorBit
};
};
} // end namespace internal
/** \class Inverse
*
* \brief Expression of the inverse of another expression
*
* \tparam XprType the type of the expression we are taking the inverse
*
* This class represents an abstract expression of A.inverse()
* and most of the time this is the only way it is used.
*
*/
template<typename XprType>
class Inverse : public InverseImpl<XprType,typename internal::traits<XprType>::StorageKind>
{
public:
typedef typename XprType::StorageIndex StorageIndex;
typedef typename XprType::PlainObject PlainObject;
typedef typename XprType::Scalar Scalar;
typedef typename internal::ref_selector<XprType>::type XprTypeNested;
typedef typename internal::remove_all<XprTypeNested>::type XprTypeNestedCleaned;
typedef typename internal::ref_selector<Inverse>::type Nested;
typedef typename internal::remove_all<XprType>::type NestedExpression;
explicit EIGEN_DEVICE_FUNC Inverse(const XprType &xpr)
: m_xpr(xpr)
{}
EIGEN_DEVICE_FUNC Index rows() const { return m_xpr.rows(); }
EIGEN_DEVICE_FUNC Index cols() const { return m_xpr.cols(); }
EIGEN_DEVICE_FUNC const XprTypeNestedCleaned& nestedExpression() const { return m_xpr; }
protected:
XprTypeNested m_xpr;
};
// Generic API dispatcher
template<typename XprType, typename StorageKind>
class InverseImpl
: public internal::generic_xpr_base<Inverse<XprType> >::type
{
public:
typedef typename internal::generic_xpr_base<Inverse<XprType> >::type Base;
typedef typename XprType::Scalar Scalar;
private:
Scalar coeff(Index row, Index col) const;
Scalar coeff(Index i) const;
};
namespace internal {
/** \internal
* \brief Default evaluator for Inverse expression.
*
* This default evaluator for Inverse expression simply evaluate the inverse into a temporary
* by a call to internal::call_assignment_no_alias.
* Therefore, inverse implementers only have to specialize Assignment<Dst,Inverse<...>, ...> for
* there own nested expression.
*
* \sa class Inverse
*/
template<typename ArgType>
struct unary_evaluator<Inverse<ArgType> >
: public evaluator<typename Inverse<ArgType>::PlainObject>
{
typedef Inverse<ArgType> InverseType;
typedef typename InverseType::PlainObject PlainObject;
typedef evaluator<PlainObject> Base;
enum { Flags = Base::Flags | EvalBeforeNestingBit };
unary_evaluator(const InverseType& inv_xpr)
: m_result(inv_xpr.rows(), inv_xpr.cols())
{
::new (static_cast<Base*>(this)) Base(m_result);
internal::call_assignment_no_alias(m_result, inv_xpr);
}
protected:
PlainObject m_result;
};
} // end namespace internal
} // end namespace Eigen
#endif // EIGEN_INVERSE_H

View File

@@ -13,13 +13,41 @@
namespace Eigen {
namespace internal {
template<typename PlainObjectType, int MapOptions, typename StrideType>
struct traits<Map<PlainObjectType, MapOptions, StrideType> >
: public traits<PlainObjectType>
{
typedef traits<PlainObjectType> TraitsBase;
enum {
PlainObjectTypeInnerSize = ((traits<PlainObjectType>::Flags&RowMajorBit)==RowMajorBit)
? PlainObjectType::ColsAtCompileTime
: PlainObjectType::RowsAtCompileTime,
InnerStrideAtCompileTime = StrideType::InnerStrideAtCompileTime == 0
? int(PlainObjectType::InnerStrideAtCompileTime)
: int(StrideType::InnerStrideAtCompileTime),
OuterStrideAtCompileTime = StrideType::OuterStrideAtCompileTime == 0
? (InnerStrideAtCompileTime==Dynamic || PlainObjectTypeInnerSize==Dynamic
? Dynamic
: int(InnerStrideAtCompileTime) * int(PlainObjectTypeInnerSize))
: int(StrideType::OuterStrideAtCompileTime),
Alignment = int(MapOptions)&int(AlignedMask),
Flags0 = TraitsBase::Flags & (~NestByRefBit),
Flags = is_lvalue<PlainObjectType>::value ? int(Flags0) : (int(Flags0) & ~LvalueBit)
};
private:
enum { Options }; // Expressions don't have Options
};
}
/** \class Map
* \ingroup Core_Module
*
* \brief A matrix or vector expression mapping an existing array of data.
*
* \tparam PlainObjectType the equivalent matrix type of the mapped data
* \tparam MapOptions specifies whether the pointer is \c #Aligned, or \c #Unaligned.
* \tparam MapOptions specifies the pointer alignment in bytes. It can be: \c #Aligned128, , \c #Aligned64, \c #Aligned32, \c #Aligned16, \c #Aligned8 or \c #Unaligned.
* The default is \c #Unaligned.
* \tparam StrideType optionally specifies strides. By default, Map assumes the memory layout
* of an ordinary, contiguous array. This can be overridden by specifying strides.
@@ -63,44 +91,6 @@ namespace Eigen {
*
* \sa PlainObjectBase::Map(), \ref TopicStorageOrders
*/
namespace internal {
template<typename PlainObjectType, int MapOptions, typename StrideType>
struct traits<Map<PlainObjectType, MapOptions, StrideType> >
: public traits<PlainObjectType>
{
typedef traits<PlainObjectType> TraitsBase;
typedef typename PlainObjectType::Index Index;
typedef typename PlainObjectType::Scalar Scalar;
enum {
InnerStrideAtCompileTime = StrideType::InnerStrideAtCompileTime == 0
? int(PlainObjectType::InnerStrideAtCompileTime)
: int(StrideType::InnerStrideAtCompileTime),
OuterStrideAtCompileTime = StrideType::OuterStrideAtCompileTime == 0
? int(PlainObjectType::OuterStrideAtCompileTime)
: int(StrideType::OuterStrideAtCompileTime),
HasNoInnerStride = InnerStrideAtCompileTime == 1,
HasNoOuterStride = StrideType::OuterStrideAtCompileTime == 0,
HasNoStride = HasNoInnerStride && HasNoOuterStride,
IsAligned = bool(EIGEN_ALIGN) && ((int(MapOptions)&Aligned)==Aligned),
IsDynamicSize = PlainObjectType::SizeAtCompileTime==Dynamic,
KeepsPacketAccess = bool(HasNoInnerStride)
&& ( bool(IsDynamicSize)
|| HasNoOuterStride
|| ( OuterStrideAtCompileTime!=Dynamic
&& ((static_cast<int>(sizeof(Scalar))*OuterStrideAtCompileTime)%16)==0 ) ),
Flags0 = TraitsBase::Flags & (~NestByRefBit),
Flags1 = IsAligned ? (int(Flags0) | AlignedBit) : (int(Flags0) & ~AlignedBit),
Flags2 = (bool(HasNoStride) || bool(PlainObjectType::IsVectorAtCompileTime))
? int(Flags1) : int(Flags1 & ~LinearAccessBit),
Flags3 = is_lvalue<PlainObjectType>::value ? int(Flags2) : (int(Flags2) & ~LvalueBit),
Flags = KeepsPacketAccess ? int(Flags3) : (int(Flags3) & ~PacketAccessBit)
};
private:
enum { Options }; // Expressions don't have Options
};
}
template<typename PlainObjectType, int MapOptions, typename StrideType> class Map
: public MapBase<Map<PlainObjectType, MapOptions, StrideType> >
{
@@ -110,34 +100,34 @@ template<typename PlainObjectType, int MapOptions, typename StrideType> class Ma
EIGEN_DENSE_PUBLIC_INTERFACE(Map)
typedef typename Base::PointerType PointerType;
#if EIGEN2_SUPPORT_STAGE <= STAGE30_FULL_EIGEN3_API
typedef const Scalar* PointerArgType;
inline PointerType cast_to_pointer_type(PointerArgType ptr) { return const_cast<PointerType>(ptr); }
#else
typedef PointerType PointerArgType;
EIGEN_DEVICE_FUNC
inline PointerType cast_to_pointer_type(PointerArgType ptr) { return ptr; }
#endif
EIGEN_DEVICE_FUNC
inline Index innerStride() const
{
return StrideType::InnerStrideAtCompileTime != 0 ? m_stride.inner() : 1;
}
EIGEN_DEVICE_FUNC
inline Index outerStride() const
{
return StrideType::OuterStrideAtCompileTime != 0 ? m_stride.outer()
: IsVectorAtCompileTime ? this->size()
: int(Flags)&RowMajorBit ? this->cols()
: this->rows();
return int(StrideType::OuterStrideAtCompileTime) != 0 ? m_stride.outer()
: int(internal::traits<Map>::OuterStrideAtCompileTime) != Dynamic ? Index(internal::traits<Map>::OuterStrideAtCompileTime)
: IsVectorAtCompileTime ? (this->size() * innerStride())
: (int(Flags)&RowMajorBit) ? (this->cols() * innerStride())
: (this->rows() * innerStride());
}
/** Constructor in the fixed-size case.
*
* \param dataPtr pointer to the array to map
* \param a_stride optional Stride object, passing the strides.
* \param stride optional Stride object, passing the strides.
*/
inline Map(PointerArgType dataPtr, const StrideType& a_stride = StrideType())
: Base(cast_to_pointer_type(dataPtr)), m_stride(a_stride)
EIGEN_DEVICE_FUNC
explicit inline Map(PointerArgType dataPtr, const StrideType& stride = StrideType())
: Base(cast_to_pointer_type(dataPtr)), m_stride(stride)
{
PlainObjectType::Base::_check_template_params();
}
@@ -145,11 +135,12 @@ template<typename PlainObjectType, int MapOptions, typename StrideType> class Ma
/** Constructor in the dynamic-size vector case.
*
* \param dataPtr pointer to the array to map
* \param a_size the size of the vector expression
* \param a_stride optional Stride object, passing the strides.
* \param size the size of the vector expression
* \param stride optional Stride object, passing the strides.
*/
inline Map(PointerArgType dataPtr, Index a_size, const StrideType& a_stride = StrideType())
: Base(cast_to_pointer_type(dataPtr), a_size), m_stride(a_stride)
EIGEN_DEVICE_FUNC
inline Map(PointerArgType dataPtr, Index size, const StrideType& stride = StrideType())
: Base(cast_to_pointer_type(dataPtr), size), m_stride(stride)
{
PlainObjectType::Base::_check_template_params();
}
@@ -157,12 +148,13 @@ template<typename PlainObjectType, int MapOptions, typename StrideType> class Ma
/** Constructor in the dynamic-size matrix case.
*
* \param dataPtr pointer to the array to map
* \param nbRows the number of rows of the matrix expression
* \param nbCols the number of columns of the matrix expression
* \param a_stride optional Stride object, passing the strides.
* \param rows the number of rows of the matrix expression
* \param cols the number of columns of the matrix expression
* \param stride optional Stride object, passing the strides.
*/
inline Map(PointerArgType dataPtr, Index nbRows, Index nbCols, const StrideType& a_stride = StrideType())
: Base(cast_to_pointer_type(dataPtr), nbRows, nbCols), m_stride(a_stride)
EIGEN_DEVICE_FUNC
inline Map(PointerArgType dataPtr, Index rows, Index cols, const StrideType& stride = StrideType())
: Base(cast_to_pointer_type(dataPtr), rows, cols), m_stride(stride)
{
PlainObjectType::Base::_check_template_params();
}
@@ -173,19 +165,6 @@ template<typename PlainObjectType, int MapOptions, typename StrideType> class Ma
StrideType m_stride;
};
template<typename _Scalar, int _Rows, int _Cols, int _Options, int _MaxRows, int _MaxCols>
inline Array<_Scalar, _Rows, _Cols, _Options, _MaxRows, _MaxCols>
::Array(const Scalar *data)
{
this->_set_noalias(Eigen::Map<const Array>(data));
}
template<typename _Scalar, int _Rows, int _Cols, int _Options, int _MaxRows, int _MaxCols>
inline Matrix<_Scalar, _Rows, _Cols, _Options, _MaxRows, _MaxCols>
::Matrix(const Scalar *data)
{
this->_set_noalias(Eigen::Map<const Matrix>(data));
}
} // end namespace Eigen

View File

@@ -12,15 +12,25 @@
#define EIGEN_MAPBASE_H
#define EIGEN_STATIC_ASSERT_INDEX_BASED_ACCESS(Derived) \
EIGEN_STATIC_ASSERT((int(internal::traits<Derived>::Flags) & LinearAccessBit) || Derived::IsVectorAtCompileTime, \
EIGEN_STATIC_ASSERT((int(internal::evaluator<Derived>::Flags) & LinearAccessBit) || Derived::IsVectorAtCompileTime, \
YOU_ARE_TRYING_TO_USE_AN_INDEX_BASED_ACCESSOR_ON_AN_EXPRESSION_THAT_DOES_NOT_SUPPORT_THAT)
namespace Eigen {
/** \class MapBase
* \ingroup Core_Module
/** \ingroup Core_Module
*
* \brief Base class for Map and Block expression with direct access
* \brief Base class for dense Map and Block expression with direct access
*
* This base class provides the const low-level accessors (e.g. coeff, coeffRef) of dense
* Map and Block objects with direct access.
* Typical users do not have to directly deal with this class.
*
* This class can be extended by through the macro plugin \c EIGEN_MAPBASE_PLUGIN.
* See \link TopicCustomizing_Plugins customizing Eigen \endlink for details.
*
* The \c Derived class has to provide the following two methods describing the memory layout:
* \code Index innerStride() const; \endcode
* \code Index outerStride() const; \endcode
*
* \sa class Map, class Block
*/
@@ -33,11 +43,11 @@ template<typename Derived> class MapBase<Derived, ReadOnlyAccessors>
enum {
RowsAtCompileTime = internal::traits<Derived>::RowsAtCompileTime,
ColsAtCompileTime = internal::traits<Derived>::ColsAtCompileTime,
InnerStrideAtCompileTime = internal::traits<Derived>::InnerStrideAtCompileTime,
SizeAtCompileTime = Base::SizeAtCompileTime
};
typedef typename internal::traits<Derived>::StorageKind StorageKind;
typedef typename internal::traits<Derived>::Index Index;
typedef typename internal::traits<Derived>::Scalar Scalar;
typedef typename internal::packet_traits<Scalar>::type PacketScalar;
typedef typename NumTraits<Scalar>::Real RealScalar;
@@ -76,8 +86,10 @@ template<typename Derived> class MapBase<Derived, ReadOnlyAccessors>
typedef typename Base::CoeffReturnType CoeffReturnType;
inline Index rows() const { return m_rows.value(); }
inline Index cols() const { return m_cols.value(); }
/** \copydoc DenseBase::rows() */
EIGEN_DEVICE_FUNC inline Index rows() const { return m_rows.value(); }
/** \copydoc DenseBase::cols() */
EIGEN_DEVICE_FUNC inline Index cols() const { return m_cols.value(); }
/** Returns a pointer to the first coefficient of the matrix or vector.
*
@@ -85,30 +97,39 @@ template<typename Derived> class MapBase<Derived, ReadOnlyAccessors>
*
* \sa innerStride(), outerStride()
*/
inline const Scalar* data() const { return m_data; }
EIGEN_DEVICE_FUNC inline const Scalar* data() const { return m_data; }
/** \copydoc PlainObjectBase::coeff(Index,Index) const */
EIGEN_DEVICE_FUNC
inline const Scalar& coeff(Index rowId, Index colId) const
{
return m_data[colId * colStride() + rowId * rowStride()];
}
/** \copydoc PlainObjectBase::coeff(Index) const */
EIGEN_DEVICE_FUNC
inline const Scalar& coeff(Index index) const
{
EIGEN_STATIC_ASSERT_INDEX_BASED_ACCESS(Derived)
return m_data[index * innerStride()];
}
/** \copydoc PlainObjectBase::coeffRef(Index,Index) const */
EIGEN_DEVICE_FUNC
inline const Scalar& coeffRef(Index rowId, Index colId) const
{
return this->m_data[colId * colStride() + rowId * rowStride()];
}
/** \copydoc PlainObjectBase::coeffRef(Index) const */
EIGEN_DEVICE_FUNC
inline const Scalar& coeffRef(Index index) const
{
EIGEN_STATIC_ASSERT_INDEX_BASED_ACCESS(Derived)
return this->m_data[index * innerStride()];
}
/** \internal */
template<int LoadMode>
inline PacketScalar packet(Index rowId, Index colId) const
{
@@ -116,6 +137,7 @@ template<typename Derived> class MapBase<Derived, ReadOnlyAccessors>
(m_data + (colId * colStride() + rowId * rowStride()));
}
/** \internal */
template<int LoadMode>
inline PacketScalar packet(Index index) const
{
@@ -123,12 +145,16 @@ template<typename Derived> class MapBase<Derived, ReadOnlyAccessors>
return internal::ploadt<PacketScalar, LoadMode>(m_data + index * innerStride());
}
/** \internal Constructor for fixed size matrices or vectors */
EIGEN_DEVICE_FUNC
explicit inline MapBase(PointerType dataPtr) : m_data(dataPtr), m_rows(RowsAtCompileTime), m_cols(ColsAtCompileTime)
{
EIGEN_STATIC_ASSERT_FIXED_SIZE(Derived)
checkSanity();
checkSanity<Derived>();
}
/** \internal Constructor for dynamically sized vectors */
EIGEN_DEVICE_FUNC
inline MapBase(PointerType dataPtr, Index vecSize)
: m_data(dataPtr),
m_rows(RowsAtCompileTime == Dynamic ? vecSize : Index(RowsAtCompileTime)),
@@ -137,34 +163,59 @@ template<typename Derived> class MapBase<Derived, ReadOnlyAccessors>
EIGEN_STATIC_ASSERT_VECTOR_ONLY(Derived)
eigen_assert(vecSize >= 0);
eigen_assert(dataPtr == 0 || SizeAtCompileTime == Dynamic || SizeAtCompileTime == vecSize);
checkSanity();
checkSanity<Derived>();
}
inline MapBase(PointerType dataPtr, Index nbRows, Index nbCols)
: m_data(dataPtr), m_rows(nbRows), m_cols(nbCols)
/** \internal Constructor for dynamically sized matrices */
EIGEN_DEVICE_FUNC
inline MapBase(PointerType dataPtr, Index rows, Index cols)
: m_data(dataPtr), m_rows(rows), m_cols(cols)
{
eigen_assert( (dataPtr == 0)
|| ( nbRows >= 0 && (RowsAtCompileTime == Dynamic || RowsAtCompileTime == nbRows)
&& nbCols >= 0 && (ColsAtCompileTime == Dynamic || ColsAtCompileTime == nbCols)));
checkSanity();
|| ( rows >= 0 && (RowsAtCompileTime == Dynamic || RowsAtCompileTime == rows)
&& cols >= 0 && (ColsAtCompileTime == Dynamic || ColsAtCompileTime == cols)));
checkSanity<Derived>();
}
#ifdef EIGEN_MAPBASE_PLUGIN
#include EIGEN_MAPBASE_PLUGIN
#endif
protected:
void checkSanity() const
template<typename T>
EIGEN_DEVICE_FUNC
void checkSanity(typename internal::enable_if<(internal::traits<T>::Alignment>0),void*>::type = 0) const
{
EIGEN_STATIC_ASSERT(EIGEN_IMPLIES(internal::traits<Derived>::Flags&PacketAccessBit,
internal::inner_stride_at_compile_time<Derived>::ret==1),
PACKET_ACCESS_REQUIRES_TO_HAVE_INNER_STRIDE_FIXED_TO_1);
eigen_assert(EIGEN_IMPLIES(internal::traits<Derived>::Flags&AlignedBit, (size_t(m_data) % 16) == 0)
&& "input pointer is not aligned on a 16 byte boundary");
#if EIGEN_MAX_ALIGN_BYTES>0
// innerStride() is not set yet when this function is called, so we optimistically assume the lowest plausible value:
const Index minInnerStride = InnerStrideAtCompileTime == Dynamic ? 1 : Index(InnerStrideAtCompileTime);
EIGEN_ONLY_USED_FOR_DEBUG(minInnerStride);
eigen_assert(( ((internal::UIntPtr(m_data) % internal::traits<Derived>::Alignment) == 0)
|| (cols() * rows() * minInnerStride * sizeof(Scalar)) < internal::traits<Derived>::Alignment ) && "data is not aligned");
#endif
}
template<typename T>
EIGEN_DEVICE_FUNC
void checkSanity(typename internal::enable_if<internal::traits<T>::Alignment==0,void*>::type = 0) const
{}
PointerType m_data;
const internal::variable_if_dynamic<Index, RowsAtCompileTime> m_rows;
const internal::variable_if_dynamic<Index, ColsAtCompileTime> m_cols;
};
/** \ingroup Core_Module
*
* \brief Base class for non-const dense Map and Block expression with direct access
*
* This base class provides the non-const low-level accessors (e.g. coeff and coeffRef) of
* dense Map and Block objects with direct access.
* It inherits MapBase<Derived, ReadOnlyAccessors> which defines the const variant for reading specific entries.
*
* \sa class Map, class Block
*/
template<typename Derived> class MapBase<Derived, WriteAccessors>
: public MapBase<Derived, ReadOnlyAccessors>
{
@@ -175,7 +226,7 @@ template<typename Derived> class MapBase<Derived, WriteAccessors>
typedef typename Base::Scalar Scalar;
typedef typename Base::PacketScalar PacketScalar;
typedef typename Base::Index Index;
typedef typename Base::StorageIndex StorageIndex;
typedef typename Base::PointerType PointerType;
using Base::derived;
@@ -196,14 +247,18 @@ template<typename Derived> class MapBase<Derived, WriteAccessors>
const Scalar
>::type ScalarWithConstIfNotLvalue;
EIGEN_DEVICE_FUNC
inline const Scalar* data() const { return this->m_data; }
EIGEN_DEVICE_FUNC
inline ScalarWithConstIfNotLvalue* data() { return this->m_data; } // no const-cast here so non-const-correct code will give a compile error
EIGEN_DEVICE_FUNC
inline ScalarWithConstIfNotLvalue& coeffRef(Index row, Index col)
{
return this->m_data[col * colStride() + row * rowStride()];
}
EIGEN_DEVICE_FUNC
inline ScalarWithConstIfNotLvalue& coeffRef(Index index)
{
EIGEN_STATIC_ASSERT_INDEX_BASED_ACCESS(Derived)
@@ -225,10 +280,11 @@ template<typename Derived> class MapBase<Derived, WriteAccessors>
(this->m_data + index * innerStride(), val);
}
explicit inline MapBase(PointerType dataPtr) : Base(dataPtr) {}
inline MapBase(PointerType dataPtr, Index vecSize) : Base(dataPtr, vecSize) {}
inline MapBase(PointerType dataPtr, Index nbRows, Index nbCols) : Base(dataPtr, nbRows, nbCols) {}
EIGEN_DEVICE_FUNC explicit inline MapBase(PointerType dataPtr) : Base(dataPtr) {}
EIGEN_DEVICE_FUNC inline MapBase(PointerType dataPtr, Index vecSize) : Base(dataPtr, vecSize) {}
EIGEN_DEVICE_FUNC inline MapBase(PointerType dataPtr, Index rows, Index cols) : Base(dataPtr, rows, cols) {}
EIGEN_DEVICE_FUNC
Derived& operator=(const MapBase& other)
{
ReadOnlyMapBase::Base::operator=(other);

File diff suppressed because it is too large Load Diff

View File

@@ -0,0 +1,101 @@
// This file is part of Eigen, a lightweight C++ template library
// for linear algebra.
//
// Copyright (C) 2014 Pedro Gonnet (pedro.gonnet@gmail.com)
// Copyright (C) 2016 Gael Guennebaud <gael.guennebaud@inria.fr>
//
// This Source Code Form is subject to the terms of the Mozilla
// Public License v. 2.0. If a copy of the MPL was not distributed
// with this file, You can obtain one at http://mozilla.org/MPL/2.0/.
#ifndef EIGEN_MATHFUNCTIONSIMPL_H
#define EIGEN_MATHFUNCTIONSIMPL_H
namespace Eigen {
namespace internal {
/** \internal \returns the hyperbolic tan of \a a (coeff-wise)
Doesn't do anything fancy, just a 13/6-degree rational interpolant which
is accurate up to a couple of ulp in the range [-9, 9], outside of which
the tanh(x) = +/-1.
This implementation works on both scalars and packets.
*/
template<typename T>
T generic_fast_tanh_float(const T& a_x)
{
// Clamp the inputs to the range [-9, 9] since anything outside
// this range is +/-1.0f in single-precision.
const T plus_9 = pset1<T>(9.f);
const T minus_9 = pset1<T>(-9.f);
// NOTE GCC prior to 6.3 might improperly optimize this max/min
// step such that if a_x is nan, x will be either 9 or -9,
// and tanh will return 1 or -1 instead of nan.
// This is supposed to be fixed in gcc6.3,
// see: https://gcc.gnu.org/bugzilla/show_bug.cgi?id=72867
const T x = pmax(minus_9,pmin(plus_9,a_x));
// The monomial coefficients of the numerator polynomial (odd).
const T alpha_1 = pset1<T>(4.89352455891786e-03f);
const T alpha_3 = pset1<T>(6.37261928875436e-04f);
const T alpha_5 = pset1<T>(1.48572235717979e-05f);
const T alpha_7 = pset1<T>(5.12229709037114e-08f);
const T alpha_9 = pset1<T>(-8.60467152213735e-11f);
const T alpha_11 = pset1<T>(2.00018790482477e-13f);
const T alpha_13 = pset1<T>(-2.76076847742355e-16f);
// The monomial coefficients of the denominator polynomial (even).
const T beta_0 = pset1<T>(4.89352518554385e-03f);
const T beta_2 = pset1<T>(2.26843463243900e-03f);
const T beta_4 = pset1<T>(1.18534705686654e-04f);
const T beta_6 = pset1<T>(1.19825839466702e-06f);
// Since the polynomials are odd/even, we need x^2.
const T x2 = pmul(x, x);
// Evaluate the numerator polynomial p.
T p = pmadd(x2, alpha_13, alpha_11);
p = pmadd(x2, p, alpha_9);
p = pmadd(x2, p, alpha_7);
p = pmadd(x2, p, alpha_5);
p = pmadd(x2, p, alpha_3);
p = pmadd(x2, p, alpha_1);
p = pmul(x, p);
// Evaluate the denominator polynomial p.
T q = pmadd(x2, beta_6, beta_4);
q = pmadd(x2, q, beta_2);
q = pmadd(x2, q, beta_0);
// Divide the numerator by the denominator.
return pdiv(p, q);
}
template<typename RealScalar>
EIGEN_STRONG_INLINE
RealScalar positive_real_hypot(const RealScalar& x, const RealScalar& y)
{
EIGEN_USING_STD_MATH(sqrt);
RealScalar p, qp;
p = numext::maxi(x,y);
if(p==RealScalar(0)) return RealScalar(0);
qp = numext::mini(y,x) / p;
return p * sqrt(RealScalar(1) + qp*qp);
}
template<typename Scalar>
struct hypot_impl
{
typedef typename NumTraits<Scalar>::Real RealScalar;
static inline RealScalar run(const Scalar& x, const Scalar& y)
{
EIGEN_USING_STD_MATH(abs);
return positive_real_hypot<RealScalar>(abs(x), abs(y));
}
};
} // end namespace internal
} // end namespace Eigen
#endif // EIGEN_MATHFUNCTIONSIMPL_H

View File

@@ -13,6 +13,45 @@
namespace Eigen {
namespace internal {
template<typename _Scalar, int _Rows, int _Cols, int _Options, int _MaxRows, int _MaxCols>
struct traits<Matrix<_Scalar, _Rows, _Cols, _Options, _MaxRows, _MaxCols> >
{
private:
enum { size = internal::size_at_compile_time<_Rows,_Cols>::ret };
typedef typename find_best_packet<_Scalar,size>::type PacketScalar;
enum {
row_major_bit = _Options&RowMajor ? RowMajorBit : 0,
is_dynamic_size_storage = _MaxRows==Dynamic || _MaxCols==Dynamic,
max_size = is_dynamic_size_storage ? Dynamic : _MaxRows*_MaxCols,
default_alignment = compute_default_alignment<_Scalar,max_size>::value,
actual_alignment = ((_Options&DontAlign)==0) ? default_alignment : 0,
required_alignment = unpacket_traits<PacketScalar>::alignment,
packet_access_bit = (packet_traits<_Scalar>::Vectorizable && (EIGEN_UNALIGNED_VECTORIZE || (actual_alignment>=required_alignment))) ? PacketAccessBit : 0
};
public:
typedef _Scalar Scalar;
typedef Dense StorageKind;
typedef Eigen::Index StorageIndex;
typedef MatrixXpr XprKind;
enum {
RowsAtCompileTime = _Rows,
ColsAtCompileTime = _Cols,
MaxRowsAtCompileTime = _MaxRows,
MaxColsAtCompileTime = _MaxCols,
Flags = compute_matrix_flags<_Scalar, _Rows, _Cols, _Options, _MaxRows, _MaxCols>::ret,
Options = _Options,
InnerStrideAtCompileTime = 1,
OuterStrideAtCompileTime = (Options&RowMajor) ? ColsAtCompileTime : RowsAtCompileTime,
// FIXME, the following flag in only used to define NeedsToAlign in PlainObjectBase
EvaluatorFlags = LinearAccessBit | DirectAccessBit | packet_access_bit | row_major_bit,
Alignment = actual_alignment
};
};
}
/** \class Matrix
* \ingroup Core_Module
*
@@ -24,13 +63,13 @@ namespace Eigen {
* The %Matrix class encompasses \em both fixed-size and dynamic-size objects (\ref fixedsize "note").
*
* The first three template parameters are required:
* \tparam _Scalar \anchor matrix_tparam_scalar Numeric type, e.g. float, double, int or std::complex<float>.
* User defined sclar types are supported as well (see \ref user_defined_scalars "here").
* \tparam _Scalar Numeric type, e.g. float, double, int or std::complex<float>.
* User defined scalar types are supported as well (see \ref user_defined_scalars "here").
* \tparam _Rows Number of rows, or \b Dynamic
* \tparam _Cols Number of columns, or \b Dynamic
*
* The remaining template parameters are optional -- in most cases you don't have to worry about them.
* \tparam _Options \anchor matrix_tparam_options A combination of either \b #RowMajor or \b #ColMajor, and of either
* \tparam _Options A combination of either \b #RowMajor or \b #ColMajor, and of either
* \b #AutoAlign or \b #DontAlign.
* The former controls \ref TopicStorageOrders "storage order", and defaults to column-major. The latter controls alignment, which is required
* for vectorization. It defaults to aligning matrices except for fixed sizes that aren't a multiple of the packet size.
@@ -67,7 +106,7 @@ namespace Eigen {
* \endcode
*
* This class can be extended with the help of the plugin mechanism described on the page
* \ref TopicCustomizingEigen by defining the preprocessor symbol \c EIGEN_MATRIX_PLUGIN.
* \ref TopicCustomizing_Plugins by defining the preprocessor symbol \c EIGEN_MATRIX_PLUGIN.
*
* <i><b>Some notes:</b></i>
*
@@ -97,32 +136,44 @@ namespace Eigen {
* are the dimensions of the original matrix, while _Rows and _Cols are Dynamic.</dd>
* </dl>
*
* \see MatrixBase for the majority of the API methods for matrices, \ref TopicClassHierarchy,
* \ref TopicStorageOrders
* <i><b>ABI and storage layout</b></i>
*
* The table below summarizes the ABI of some possible Matrix instances which is fixed thorough the lifetime of Eigen 3.
* <table class="manual">
* <tr><th>Matrix type</th><th>Equivalent C structure</th></tr>
* <tr><td>\code Matrix<T,Dynamic,Dynamic> \endcode</td><td>\code
* struct {
* T *data; // with (size_t(data)%EIGEN_MAX_ALIGN_BYTES)==0
* Eigen::Index rows, cols;
* };
* \endcode</td></tr>
* <tr class="alt"><td>\code
* Matrix<T,Dynamic,1>
* Matrix<T,1,Dynamic> \endcode</td><td>\code
* struct {
* T *data; // with (size_t(data)%EIGEN_MAX_ALIGN_BYTES)==0
* Eigen::Index size;
* };
* \endcode</td></tr>
* <tr><td>\code Matrix<T,Rows,Cols> \endcode</td><td>\code
* struct {
* T data[Rows*Cols]; // with (size_t(data)%A(Rows*Cols*sizeof(T)))==0
* };
* \endcode</td></tr>
* <tr class="alt"><td>\code Matrix<T,Dynamic,Dynamic,0,MaxRows,MaxCols> \endcode</td><td>\code
* struct {
* T data[MaxRows*MaxCols]; // with (size_t(data)%A(MaxRows*MaxCols*sizeof(T)))==0
* Eigen::Index rows, cols;
* };
* \endcode</td></tr>
* </table>
* Note that in this table Rows, Cols, MaxRows and MaxCols are all positive integers. A(S) is defined to the largest possible power-of-two
* smaller to EIGEN_MAX_STATIC_ALIGN_BYTES.
*
* \see MatrixBase for the majority of the API methods for matrices, \ref TopicClassHierarchy,
* \ref TopicStorageOrders
*/
namespace internal {
template<typename _Scalar, int _Rows, int _Cols, int _Options, int _MaxRows, int _MaxCols>
struct traits<Matrix<_Scalar, _Rows, _Cols, _Options, _MaxRows, _MaxCols> >
{
typedef _Scalar Scalar;
typedef Dense StorageKind;
typedef DenseIndex Index;
typedef MatrixXpr XprKind;
enum {
RowsAtCompileTime = _Rows,
ColsAtCompileTime = _Cols,
MaxRowsAtCompileTime = _MaxRows,
MaxColsAtCompileTime = _MaxCols,
Flags = compute_matrix_flags<_Scalar, _Rows, _Cols, _Options, _MaxRows, _MaxCols>::ret,
CoeffReadCost = NumTraits<Scalar>::ReadCost,
Options = _Options,
InnerStrideAtCompileTime = 1,
OuterStrideAtCompileTime = (Options&RowMajor) ? ColsAtCompileTime : RowsAtCompileTime
};
};
}
template<typename _Scalar, int _Rows, int _Cols, int _Options, int _MaxRows, int _MaxCols>
class Matrix
: public PlainObjectBase<Matrix<_Scalar, _Rows, _Cols, _Options, _MaxRows, _MaxCols> >
@@ -151,6 +202,7 @@ class Matrix
*
* \callgraph
*/
EIGEN_DEVICE_FUNC
EIGEN_STRONG_INLINE Matrix& operator=(const Matrix& other)
{
return Base::_set(other);
@@ -167,7 +219,8 @@ class Matrix
* remain row-vectors and vectors remain vectors.
*/
template<typename OtherDerived>
EIGEN_STRONG_INLINE Matrix& operator=(const MatrixBase<OtherDerived>& other)
EIGEN_DEVICE_FUNC
EIGEN_STRONG_INLINE Matrix& operator=(const DenseBase<OtherDerived>& other)
{
return Base::_set(other);
}
@@ -179,12 +232,14 @@ class Matrix
* \copydetails DenseBase::operator=(const EigenBase<OtherDerived> &other)
*/
template<typename OtherDerived>
EIGEN_DEVICE_FUNC
EIGEN_STRONG_INLINE Matrix& operator=(const EigenBase<OtherDerived> &other)
{
return Base::operator=(other);
}
template<typename OtherDerived>
EIGEN_DEVICE_FUNC
EIGEN_STRONG_INLINE Matrix& operator=(const ReturnByValue<OtherDerived>& func)
{
return Base::operator=(func);
@@ -200,6 +255,7 @@ class Matrix
*
* \sa resize(Index,Index)
*/
EIGEN_DEVICE_FUNC
EIGEN_STRONG_INLINE Matrix() : Base()
{
Base::_check_template_params();
@@ -207,60 +263,85 @@ class Matrix
}
// FIXME is it still needed
Matrix(internal::constructor_without_unaligned_array_assert)
EIGEN_DEVICE_FUNC
explicit Matrix(internal::constructor_without_unaligned_array_assert)
: Base(internal::constructor_without_unaligned_array_assert())
{ Base::_check_template_params(); EIGEN_INITIALIZE_COEFFS_IF_THAT_OPTION_IS_ENABLED }
#ifdef EIGEN_HAVE_RVALUE_REFERENCES
Matrix(Matrix&& other)
#if EIGEN_HAS_RVALUE_REFERENCES
EIGEN_DEVICE_FUNC
Matrix(Matrix&& other) EIGEN_NOEXCEPT_IF(std::is_nothrow_move_constructible<Scalar>::value)
: Base(std::move(other))
{
Base::_check_template_params();
if (RowsAtCompileTime!=Dynamic && ColsAtCompileTime!=Dynamic)
Base::_set_noalias(other);
}
Matrix& operator=(Matrix&& other)
EIGEN_DEVICE_FUNC
Matrix& operator=(Matrix&& other) EIGEN_NOEXCEPT_IF(std::is_nothrow_move_assignable<Scalar>::value)
{
other.swap(*this);
return *this;
}
#endif
/** \brief Constructs a vector or row-vector with given dimension. \only_for_vectors
*
* Note that this is only useful for dynamic-size vectors. For fixed-size vectors,
* it is redundant to pass the dimension here, so it makes more sense to use the default
* constructor Matrix() instead.
*/
EIGEN_STRONG_INLINE explicit Matrix(Index dim)
: Base(dim, RowsAtCompileTime == 1 ? 1 : dim, ColsAtCompileTime == 1 ? 1 : dim)
#ifndef EIGEN_PARSED_BY_DOXYGEN
// This constructor is for both 1x1 matrices and dynamic vectors
template<typename T>
EIGEN_DEVICE_FUNC
EIGEN_STRONG_INLINE explicit Matrix(const T& x)
{
Base::_check_template_params();
EIGEN_STATIC_ASSERT_VECTOR_ONLY(Matrix)
eigen_assert(dim >= 0);
eigen_assert(SizeAtCompileTime == Dynamic || SizeAtCompileTime == dim);
EIGEN_INITIALIZE_COEFFS_IF_THAT_OPTION_IS_ENABLED
Base::template _init1<T>(x);
}
#ifndef EIGEN_PARSED_BY_DOXYGEN
template<typename T0, typename T1>
EIGEN_DEVICE_FUNC
EIGEN_STRONG_INLINE Matrix(const T0& x, const T1& y)
{
Base::_check_template_params();
Base::template _init2<T0,T1>(x, y);
}
#else
/** \brief Constructs a fixed-sized matrix initialized with coefficients starting at \a data */
EIGEN_DEVICE_FUNC
explicit Matrix(const Scalar *data);
/** \brief Constructs a vector or row-vector with given dimension. \only_for_vectors
*
* This is useful for dynamic-size vectors. For fixed-size vectors,
* it is redundant to pass these parameters, so one should use the default constructor
* Matrix() instead.
*
* \warning This constructor is disabled for fixed-size \c 1x1 matrices. For instance,
* calling Matrix<double,1,1>(1) will call the initialization constructor: Matrix(const Scalar&).
* For fixed-size \c 1x1 matrices it is therefore recommended to use the default
* constructor Matrix() instead, especially when using one of the non standard
* \c EIGEN_INITIALIZE_MATRICES_BY_{ZERO,\c NAN} macros (see \ref TopicPreprocessorDirectives).
*/
EIGEN_STRONG_INLINE explicit Matrix(Index dim);
/** \brief Constructs an initialized 1x1 matrix with the given coefficient */
Matrix(const Scalar& x);
/** \brief Constructs an uninitialized matrix with \a rows rows and \a cols columns.
*
* This is useful for dynamic-size matrices. For fixed-size matrices,
* it is redundant to pass these parameters, so one should use the default constructor
* Matrix() instead. */
* Matrix() instead.
*
* \warning This constructor is disabled for fixed-size \c 1x2 and \c 2x1 vectors. For instance,
* calling Matrix2f(2,1) will call the initialization constructor: Matrix(const Scalar& x, const Scalar& y).
* For fixed-size \c 1x2 or \c 2x1 vectors it is therefore recommended to use the default
* constructor Matrix() instead, especially when using one of the non standard
* \c EIGEN_INITIALIZE_MATRICES_BY_{ZERO,\c NAN} macros (see \ref TopicPreprocessorDirectives).
*/
EIGEN_DEVICE_FUNC
Matrix(Index rows, Index cols);
/** \brief Constructs an initialized 2D vector with given coefficients */
Matrix(const Scalar& x, const Scalar& y);
#endif
/** \brief Constructs an initialized 3D vector with given coefficients */
EIGEN_DEVICE_FUNC
EIGEN_STRONG_INLINE Matrix(const Scalar& x, const Scalar& y, const Scalar& z)
{
Base::_check_template_params();
@@ -270,6 +351,7 @@ class Matrix
m_storage.data()[2] = z;
}
/** \brief Constructs an initialized 4D vector with given coefficients */
EIGEN_DEVICE_FUNC
EIGEN_STRONG_INLINE Matrix(const Scalar& x, const Scalar& y, const Scalar& z, const Scalar& w)
{
Base::_check_template_params();
@@ -280,76 +362,33 @@ class Matrix
m_storage.data()[3] = w;
}
explicit Matrix(const Scalar *data);
/** \brief Constructor copying the value of the expression \a other */
template<typename OtherDerived>
EIGEN_STRONG_INLINE Matrix(const MatrixBase<OtherDerived>& other)
: Base(other.rows() * other.cols(), other.rows(), other.cols())
{
// This test resides here, to bring the error messages closer to the user. Normally, these checks
// are performed deeply within the library, thus causing long and scary error traces.
EIGEN_STATIC_ASSERT((internal::is_same<Scalar, typename OtherDerived::Scalar>::value),
YOU_MIXED_DIFFERENT_NUMERIC_TYPES__YOU_NEED_TO_USE_THE_CAST_METHOD_OF_MATRIXBASE_TO_CAST_NUMERIC_TYPES_EXPLICITLY)
Base::_check_template_params();
Base::_set_noalias(other);
}
/** \brief Copy constructor */
EIGEN_STRONG_INLINE Matrix(const Matrix& other)
: Base(other.rows() * other.cols(), other.rows(), other.cols())
{
Base::_check_template_params();
Base::_set_noalias(other);
}
/** \brief Copy constructor with in-place evaluation */
template<typename OtherDerived>
EIGEN_STRONG_INLINE Matrix(const ReturnByValue<OtherDerived>& other)
{
Base::_check_template_params();
Base::resize(other.rows(), other.cols());
other.evalTo(*this);
}
EIGEN_DEVICE_FUNC
EIGEN_STRONG_INLINE Matrix(const Matrix& other) : Base(other)
{ }
/** \brief Copy constructor for generic expressions.
* \sa MatrixBase::operator=(const EigenBase<OtherDerived>&)
*/
template<typename OtherDerived>
EIGEN_DEVICE_FUNC
EIGEN_STRONG_INLINE Matrix(const EigenBase<OtherDerived> &other)
: Base(other.derived().rows() * other.derived().cols(), other.derived().rows(), other.derived().cols())
{
Base::_check_template_params();
Base::_resize_to_match(other);
// FIXME/CHECK: isn't *this = other.derived() more efficient. it allows to
// go for pure _set() implementations, right?
*this = other;
}
: Base(other.derived())
{ }
/** \internal
* \brief Override MatrixBase::swap() since for dynamic-sized matrices
* of same type it is enough to swap the data pointers.
*/
template<typename OtherDerived>
void swap(MatrixBase<OtherDerived> const & other)
{ this->_swap(other.derived()); }
inline Index innerStride() const { return 1; }
inline Index outerStride() const { return this->innerSize(); }
EIGEN_DEVICE_FUNC inline Index innerStride() const { return 1; }
EIGEN_DEVICE_FUNC inline Index outerStride() const { return this->innerSize(); }
/////////// Geometry module ///////////
template<typename OtherDerived>
EIGEN_DEVICE_FUNC
explicit Matrix(const RotationBase<OtherDerived,ColsAtCompileTime>& r);
template<typename OtherDerived>
EIGEN_DEVICE_FUNC
Matrix& operator=(const RotationBase<OtherDerived,ColsAtCompileTime>& r);
#ifdef EIGEN2_SUPPORT
template<typename OtherDerived>
explicit Matrix(const eigen2_RotationBase<OtherDerived,ColsAtCompileTime>& r);
template<typename OtherDerived>
Matrix& operator=(const eigen2_RotationBase<OtherDerived,ColsAtCompileTime>& r);
#endif
// allow to extend Matrix outside Eigen
#ifdef EIGEN_MATRIX_PLUGIN
#include EIGEN_MATRIX_PLUGIN

View File

@@ -41,9 +41,9 @@ namespace Eigen {
* \endcode
*
* This class can be extended with the help of the plugin mechanism described on the page
* \ref TopicCustomizingEigen by defining the preprocessor symbol \c EIGEN_MATRIXBASE_PLUGIN.
* \ref TopicCustomizing_Plugins by defining the preprocessor symbol \c EIGEN_MATRIXBASE_PLUGIN.
*
* \sa \ref TopicClassHierarchy
* \sa \blank \ref TopicClassHierarchy
*/
template<typename Derived> class MatrixBase
: public DenseBase<Derived>
@@ -52,7 +52,7 @@ template<typename Derived> class MatrixBase
#ifndef EIGEN_PARSED_BY_DOXYGEN
typedef MatrixBase StorageBaseType;
typedef typename internal::traits<Derived>::StorageKind StorageKind;
typedef typename internal::traits<Derived>::Index Index;
typedef typename internal::traits<Derived>::StorageIndex StorageIndex;
typedef typename internal::traits<Derived>::Scalar Scalar;
typedef typename internal::packet_traits<Scalar>::type PacketScalar;
typedef typename NumTraits<Scalar>::Real RealScalar;
@@ -66,7 +66,6 @@ template<typename Derived> class MatrixBase
using Base::MaxSizeAtCompileTime;
using Base::IsVectorAtCompileTime;
using Base::Flags;
using Base::CoeffReadCost;
using Base::derived;
using Base::const_cast_derived;
@@ -98,25 +97,14 @@ template<typename Derived> class MatrixBase
/** \returns the size of the main diagonal, which is min(rows(),cols()).
* \sa rows(), cols(), SizeAtCompileTime. */
inline Index diagonalSize() const { return (std::min)(rows(),cols()); }
EIGEN_DEVICE_FUNC
inline Index diagonalSize() const { return (numext::mini)(rows(),cols()); }
/** \brief The plain matrix type corresponding to this expression.
*
* This is not necessarily exactly the return type of eval(). In the case of plain matrices,
* the return type of eval() is a const reference to a matrix, not a matrix! It is however guaranteed
* that the return type of eval() is either PlainObject or const PlainObject&.
*/
typedef Matrix<typename internal::traits<Derived>::Scalar,
internal::traits<Derived>::RowsAtCompileTime,
internal::traits<Derived>::ColsAtCompileTime,
AutoAlign | (internal::traits<Derived>::Flags&RowMajorBit ? RowMajor : ColMajor),
internal::traits<Derived>::MaxRowsAtCompileTime,
internal::traits<Derived>::MaxColsAtCompileTime
> PlainObject;
typedef typename Base::PlainObject PlainObject;
#ifndef EIGEN_PARSED_BY_DOXYGEN
/** \internal Represents a matrix with all coefficients equal to one another*/
typedef CwiseNullaryOp<internal::scalar_constant_op<Scalar>,Derived> ConstantReturnType;
typedef CwiseNullaryOp<internal::scalar_constant_op<Scalar>,PlainObject> ConstantReturnType;
/** \internal the return type of MatrixBase::adjoint() */
typedef typename internal::conditional<NumTraits<Scalar>::IsComplex,
CwiseUnaryOp<internal::scalar_conjugate_op<Scalar>, ConstTransposeReturnType>,
@@ -125,7 +113,7 @@ template<typename Derived> class MatrixBase
/** \internal Return type of eigenvalues() */
typedef Matrix<std::complex<RealScalar>, internal::traits<Derived>::ColsAtCompileTime, 1, ColMajor> EigenvaluesReturnType;
/** \internal the return type of identity */
typedef CwiseNullaryOp<internal::scalar_identity_op<Scalar>,Derived> IdentityReturnType;
typedef CwiseNullaryOp<internal::scalar_identity_op<Scalar>,PlainObject> IdentityReturnType;
/** \internal the return type of unit vectors */
typedef Block<const CwiseNullaryOp<internal::scalar_identity_op<Scalar>, SquareMatrixType>,
internal::traits<Derived>::RowsAtCompileTime,
@@ -133,6 +121,7 @@ template<typename Derived> class MatrixBase
#endif // not EIGEN_PARSED_BY_DOXYGEN
#define EIGEN_CURRENT_STORAGE_BASE_CLASS Eigen::MatrixBase
#define EIGEN_DOC_UNARY_ADDONS(X,Y)
# include "../plugins/CommonCwiseUnaryOps.h"
# include "../plugins/CommonCwiseBinaryOps.h"
# include "../plugins/MatrixCwiseUnaryOps.h"
@@ -141,41 +130,44 @@ template<typename Derived> class MatrixBase
# include EIGEN_MATRIXBASE_PLUGIN
# endif
#undef EIGEN_CURRENT_STORAGE_BASE_CLASS
#undef EIGEN_DOC_UNARY_ADDONS
/** Special case of the template operator=, in order to prevent the compiler
* from generating a default operator= (issue hit with g++ 4.1)
*/
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE
Derived& operator=(const MatrixBase& other);
// We cannot inherit here via Base::operator= since it is causing
// trouble with MSVC.
template <typename OtherDerived>
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE
Derived& operator=(const DenseBase<OtherDerived>& other);
template <typename OtherDerived>
EIGEN_DEVICE_FUNC
Derived& operator=(const EigenBase<OtherDerived>& other);
template<typename OtherDerived>
EIGEN_DEVICE_FUNC
Derived& operator=(const ReturnByValue<OtherDerived>& other);
template<typename ProductDerived, typename Lhs, typename Rhs>
Derived& lazyAssign(const ProductBase<ProductDerived, Lhs,Rhs>& other);
template<typename MatrixPower, typename Lhs, typename Rhs>
Derived& lazyAssign(const MatrixPowerProduct<MatrixPower, Lhs,Rhs>& other);
template<typename OtherDerived>
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE
Derived& operator+=(const MatrixBase<OtherDerived>& other);
template<typename OtherDerived>
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE
Derived& operator-=(const MatrixBase<OtherDerived>& other);
template<typename OtherDerived>
const typename ProductReturnType<Derived,OtherDerived>::Type
EIGEN_DEVICE_FUNC
const Product<Derived,OtherDerived>
operator*(const MatrixBase<OtherDerived> &other) const;
template<typename OtherDerived>
const typename LazyProductReturnType<Derived,OtherDerived>::Type
EIGEN_DEVICE_FUNC
const Product<Derived,OtherDerived,LazyProduct>
lazyProduct(const MatrixBase<OtherDerived> &other) const;
template<typename OtherDerived>
@@ -188,84 +180,93 @@ template<typename Derived> class MatrixBase
void applyOnTheRight(const EigenBase<OtherDerived>& other);
template<typename DiagonalDerived>
const DiagonalProduct<Derived, DiagonalDerived, OnTheRight>
EIGEN_DEVICE_FUNC
const Product<Derived, DiagonalDerived, LazyProduct>
operator*(const DiagonalBase<DiagonalDerived> &diagonal) const;
template<typename OtherDerived>
typename internal::scalar_product_traits<typename internal::traits<Derived>::Scalar,typename internal::traits<OtherDerived>::Scalar>::ReturnType
EIGEN_DEVICE_FUNC
typename ScalarBinaryOpTraits<typename internal::traits<Derived>::Scalar,typename internal::traits<OtherDerived>::Scalar>::ReturnType
dot(const MatrixBase<OtherDerived>& other) const;
#ifdef EIGEN2_SUPPORT
template<typename OtherDerived>
Scalar eigen2_dot(const MatrixBase<OtherDerived>& other) const;
#endif
RealScalar squaredNorm() const;
RealScalar norm() const;
EIGEN_DEVICE_FUNC RealScalar squaredNorm() const;
EIGEN_DEVICE_FUNC RealScalar norm() const;
RealScalar stableNorm() const;
RealScalar blueNorm() const;
RealScalar hypotNorm() const;
const PlainObject normalized() const;
void normalize();
EIGEN_DEVICE_FUNC const PlainObject normalized() const;
EIGEN_DEVICE_FUNC const PlainObject stableNormalized() const;
EIGEN_DEVICE_FUNC void normalize();
EIGEN_DEVICE_FUNC void stableNormalize();
const AdjointReturnType adjoint() const;
void adjointInPlace();
EIGEN_DEVICE_FUNC const AdjointReturnType adjoint() const;
EIGEN_DEVICE_FUNC void adjointInPlace();
typedef Diagonal<Derived> DiagonalReturnType;
EIGEN_DEVICE_FUNC
DiagonalReturnType diagonal();
typedef typename internal::add_const<Diagonal<const Derived> >::type ConstDiagonalReturnType;
EIGEN_DEVICE_FUNC
ConstDiagonalReturnType diagonal() const;
template<int Index> struct DiagonalIndexReturnType { typedef Diagonal<Derived,Index> Type; };
template<int Index> struct ConstDiagonalIndexReturnType { typedef const Diagonal<const Derived,Index> Type; };
template<int Index> typename DiagonalIndexReturnType<Index>::Type diagonal();
template<int Index> typename ConstDiagonalIndexReturnType<Index>::Type diagonal() const;
template<int Index>
EIGEN_DEVICE_FUNC
typename DiagonalIndexReturnType<Index>::Type diagonal();
template<int Index>
EIGEN_DEVICE_FUNC
typename ConstDiagonalIndexReturnType<Index>::Type diagonal() const;
typedef Diagonal<Derived,DynamicIndex> DiagonalDynamicIndexReturnType;
typedef typename internal::add_const<Diagonal<const Derived,DynamicIndex> >::type ConstDiagonalDynamicIndexReturnType;
EIGEN_DEVICE_FUNC
DiagonalDynamicIndexReturnType diagonal(Index index);
EIGEN_DEVICE_FUNC
ConstDiagonalDynamicIndexReturnType diagonal(Index index) const;
#ifdef EIGEN2_SUPPORT
template<unsigned int Mode> typename internal::eigen2_part_return_type<Derived, Mode>::type part();
template<unsigned int Mode> const typename internal::eigen2_part_return_type<Derived, Mode>::type part() const;
// huuuge hack. make Eigen2's matrix.part<Diagonal>() work in eigen3. Problem: Diagonal is now a class template instead
// of an integer constant. Solution: overload the part() method template wrt template parameters list.
template<template<typename T, int N> class U>
const DiagonalWrapper<ConstDiagonalReturnType> part() const
{ return diagonal().asDiagonal(); }
#endif // EIGEN2_SUPPORT
template<unsigned int Mode> struct TriangularViewReturnType { typedef TriangularView<Derived, Mode> Type; };
template<unsigned int Mode> struct ConstTriangularViewReturnType { typedef const TriangularView<const Derived, Mode> Type; };
template<unsigned int Mode> typename TriangularViewReturnType<Mode>::Type triangularView();
template<unsigned int Mode> typename ConstTriangularViewReturnType<Mode>::Type triangularView() const;
template<unsigned int Mode>
EIGEN_DEVICE_FUNC
typename TriangularViewReturnType<Mode>::Type triangularView();
template<unsigned int Mode>
EIGEN_DEVICE_FUNC
typename ConstTriangularViewReturnType<Mode>::Type triangularView() const;
template<unsigned int UpLo> struct SelfAdjointViewReturnType { typedef SelfAdjointView<Derived, UpLo> Type; };
template<unsigned int UpLo> struct ConstSelfAdjointViewReturnType { typedef const SelfAdjointView<const Derived, UpLo> Type; };
template<unsigned int UpLo> typename SelfAdjointViewReturnType<UpLo>::Type selfadjointView();
template<unsigned int UpLo> typename ConstSelfAdjointViewReturnType<UpLo>::Type selfadjointView() const;
template<unsigned int UpLo>
EIGEN_DEVICE_FUNC
typename SelfAdjointViewReturnType<UpLo>::Type selfadjointView();
template<unsigned int UpLo>
EIGEN_DEVICE_FUNC
typename ConstSelfAdjointViewReturnType<UpLo>::Type selfadjointView() const;
const SparseView<Derived> sparseView(const Scalar& m_reference = Scalar(0),
const typename NumTraits<Scalar>::Real& m_epsilon = NumTraits<Scalar>::dummy_precision()) const;
static const IdentityReturnType Identity();
static const IdentityReturnType Identity(Index rows, Index cols);
static const BasisReturnType Unit(Index size, Index i);
static const BasisReturnType Unit(Index i);
static const BasisReturnType UnitX();
static const BasisReturnType UnitY();
static const BasisReturnType UnitZ();
static const BasisReturnType UnitW();
EIGEN_DEVICE_FUNC static const IdentityReturnType Identity();
EIGEN_DEVICE_FUNC static const IdentityReturnType Identity(Index rows, Index cols);
EIGEN_DEVICE_FUNC static const BasisReturnType Unit(Index size, Index i);
EIGEN_DEVICE_FUNC static const BasisReturnType Unit(Index i);
EIGEN_DEVICE_FUNC static const BasisReturnType UnitX();
EIGEN_DEVICE_FUNC static const BasisReturnType UnitY();
EIGEN_DEVICE_FUNC static const BasisReturnType UnitZ();
EIGEN_DEVICE_FUNC static const BasisReturnType UnitW();
EIGEN_DEVICE_FUNC
const DiagonalWrapper<const Derived> asDiagonal() const;
const PermutationWrapper<const Derived> asPermutation() const;
EIGEN_DEVICE_FUNC
Derived& setIdentity();
EIGEN_DEVICE_FUNC
Derived& setIdentity(Index rows, Index cols);
bool isIdentity(const RealScalar& prec = NumTraits<Scalar>::dummy_precision()) const;
@@ -284,7 +285,7 @@ template<typename Derived> class MatrixBase
* fuzzy comparison such as isApprox()
* \sa isApprox(), operator!= */
template<typename OtherDerived>
inline bool operator==(const MatrixBase<OtherDerived>& other) const
EIGEN_DEVICE_FUNC inline bool operator==(const MatrixBase<OtherDerived>& other) const
{ return cwiseEqual(other).all(); }
/** \returns true if at least one pair of coefficients of \c *this and \a other are not exactly equal to each other.
@@ -292,64 +293,50 @@ template<typename Derived> class MatrixBase
* fuzzy comparison such as isApprox()
* \sa isApprox(), operator== */
template<typename OtherDerived>
inline bool operator!=(const MatrixBase<OtherDerived>& other) const
EIGEN_DEVICE_FUNC inline bool operator!=(const MatrixBase<OtherDerived>& other) const
{ return cwiseNotEqual(other).any(); }
NoAlias<Derived,Eigen::MatrixBase > noalias();
inline const ForceAlignedAccess<Derived> forceAlignedAccess() const;
inline ForceAlignedAccess<Derived> forceAlignedAccess();
template<bool Enable> inline typename internal::add_const_on_value_type<typename internal::conditional<Enable,ForceAlignedAccess<Derived>,Derived&>::type>::type forceAlignedAccessIf() const;
template<bool Enable> inline typename internal::conditional<Enable,ForceAlignedAccess<Derived>,Derived&>::type forceAlignedAccessIf();
// TODO forceAlignedAccess is temporarily disabled
// Need to find a nicer workaround.
inline const Derived& forceAlignedAccess() const { return derived(); }
inline Derived& forceAlignedAccess() { return derived(); }
template<bool Enable> inline const Derived& forceAlignedAccessIf() const { return derived(); }
template<bool Enable> inline Derived& forceAlignedAccessIf() { return derived(); }
Scalar trace() const;
EIGEN_DEVICE_FUNC Scalar trace() const;
/////////// Array module ///////////
template<int p> EIGEN_DEVICE_FUNC RealScalar lpNorm() const;
template<int p> RealScalar lpNorm() const;
MatrixBase<Derived>& matrix() { return *this; }
const MatrixBase<Derived>& matrix() const { return *this; }
EIGEN_DEVICE_FUNC MatrixBase<Derived>& matrix() { return *this; }
EIGEN_DEVICE_FUNC const MatrixBase<Derived>& matrix() const { return *this; }
/** \returns an \link Eigen::ArrayBase Array \endlink expression of this matrix
* \sa ArrayBase::matrix() */
ArrayWrapper<Derived> array() { return derived(); }
const ArrayWrapper<const Derived> array() const { return derived(); }
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE ArrayWrapper<Derived> array() { return ArrayWrapper<Derived>(derived()); }
/** \returns a const \link Eigen::ArrayBase Array \endlink expression of this matrix
* \sa ArrayBase::matrix() */
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const ArrayWrapper<const Derived> array() const { return ArrayWrapper<const Derived>(derived()); }
/////////// LU module ///////////
const FullPivLU<PlainObject> fullPivLu() const;
const PartialPivLU<PlainObject> partialPivLu() const;
inline const FullPivLU<PlainObject> fullPivLu() const;
inline const PartialPivLU<PlainObject> partialPivLu() const;
#if EIGEN2_SUPPORT_STAGE < STAGE20_RESOLVE_API_CONFLICTS
const LU<PlainObject> lu() const;
#endif
inline const PartialPivLU<PlainObject> lu() const;
#ifdef EIGEN2_SUPPORT
const LU<PlainObject> eigen2_lu() const;
#endif
inline const Inverse<Derived> inverse() const;
#if EIGEN2_SUPPORT_STAGE > STAGE20_RESOLVE_API_CONFLICTS
const PartialPivLU<PlainObject> lu() const;
#endif
#ifdef EIGEN2_SUPPORT
template<typename ResultType>
void computeInverse(MatrixBase<ResultType> *result) const {
*result = this->inverse();
}
#endif
const internal::inverse_impl<Derived> inverse() const;
template<typename ResultType>
void computeInverseAndDetWithCheck(
inline void computeInverseAndDetWithCheck(
ResultType& inverse,
typename ResultType::Scalar& determinant,
bool& invertible,
const RealScalar& absDeterminantThreshold = NumTraits<Scalar>::dummy_precision()
) const;
template<typename ResultType>
void computeInverseWithCheck(
inline void computeInverseWithCheck(
ResultType& inverse,
bool& invertible,
const RealScalar& absDeterminantThreshold = NumTraits<Scalar>::dummy_precision()
@@ -358,65 +345,70 @@ template<typename Derived> class MatrixBase
/////////// Cholesky module ///////////
const LLT<PlainObject> llt() const;
const LDLT<PlainObject> ldlt() const;
inline const LLT<PlainObject> llt() const;
inline const LDLT<PlainObject> ldlt() const;
/////////// QR module ///////////
const HouseholderQR<PlainObject> householderQr() const;
const ColPivHouseholderQR<PlainObject> colPivHouseholderQr() const;
const FullPivHouseholderQR<PlainObject> fullPivHouseholderQr() const;
#ifdef EIGEN2_SUPPORT
const QR<PlainObject> qr() const;
#endif
inline const HouseholderQR<PlainObject> householderQr() const;
inline const ColPivHouseholderQR<PlainObject> colPivHouseholderQr() const;
inline const FullPivHouseholderQR<PlainObject> fullPivHouseholderQr() const;
inline const CompleteOrthogonalDecomposition<PlainObject> completeOrthogonalDecomposition() const;
EigenvaluesReturnType eigenvalues() const;
RealScalar operatorNorm() const;
/////////// Eigenvalues module ///////////
inline EigenvaluesReturnType eigenvalues() const;
inline RealScalar operatorNorm() const;
/////////// SVD module ///////////
JacobiSVD<PlainObject> jacobiSvd(unsigned int computationOptions = 0) const;
#ifdef EIGEN2_SUPPORT
SVD<PlainObject> svd() const;
#endif
inline JacobiSVD<PlainObject> jacobiSvd(unsigned int computationOptions = 0) const;
inline BDCSVD<PlainObject> bdcSvd(unsigned int computationOptions = 0) const;
/////////// Geometry module ///////////
#ifndef EIGEN_PARSED_BY_DOXYGEN
/// \internal helper struct to form the return type of the cross product
template<typename OtherDerived> struct cross_product_return_type {
typedef typename internal::scalar_product_traits<typename internal::traits<Derived>::Scalar,typename internal::traits<OtherDerived>::Scalar>::ReturnType Scalar;
typedef typename ScalarBinaryOpTraits<typename internal::traits<Derived>::Scalar,typename internal::traits<OtherDerived>::Scalar>::ReturnType Scalar;
typedef Matrix<Scalar,MatrixBase::RowsAtCompileTime,MatrixBase::ColsAtCompileTime> type;
};
#endif // EIGEN_PARSED_BY_DOXYGEN
template<typename OtherDerived>
typename cross_product_return_type<OtherDerived>::type
EIGEN_DEVICE_FUNC
#ifndef EIGEN_PARSED_BY_DOXYGEN
inline typename cross_product_return_type<OtherDerived>::type
#else
inline PlainObject
#endif
cross(const MatrixBase<OtherDerived>& other) const;
template<typename OtherDerived>
PlainObject cross3(const MatrixBase<OtherDerived>& other) const;
PlainObject unitOrthogonal(void) const;
Matrix<Scalar,3,1> eulerAngles(Index a0, Index a1, Index a2) const;
#if EIGEN2_SUPPORT_STAGE > STAGE20_RESOLVE_API_CONFLICTS
ScalarMultipleReturnType operator*(const UniformScaling<Scalar>& s) const;
EIGEN_DEVICE_FUNC
inline PlainObject cross3(const MatrixBase<OtherDerived>& other) const;
EIGEN_DEVICE_FUNC
inline PlainObject unitOrthogonal(void) const;
EIGEN_DEVICE_FUNC
inline Matrix<Scalar,3,1> eulerAngles(Index a0, Index a1, Index a2) const;
// put this as separate enum value to work around possible GCC 4.3 bug (?)
enum { HomogeneousReturnTypeDirection = ColsAtCompileTime==1?Vertical:Horizontal };
enum { HomogeneousReturnTypeDirection = ColsAtCompileTime==1&&RowsAtCompileTime==1 ? ((internal::traits<Derived>::Flags&RowMajorBit)==RowMajorBit ? Horizontal : Vertical)
: ColsAtCompileTime==1 ? Vertical : Horizontal };
typedef Homogeneous<Derived, HomogeneousReturnTypeDirection> HomogeneousReturnType;
HomogeneousReturnType homogeneous() const;
#endif
EIGEN_DEVICE_FUNC
inline HomogeneousReturnType homogeneous() const;
enum {
SizeMinusOne = SizeAtCompileTime==Dynamic ? Dynamic : SizeAtCompileTime-1
};
typedef Block<const Derived,
internal::traits<Derived>::ColsAtCompileTime==1 ? SizeMinusOne : 1,
internal::traits<Derived>::ColsAtCompileTime==1 ? 1 : SizeMinusOne> ConstStartMinusOne;
typedef CwiseUnaryOp<internal::scalar_quotient1_op<typename internal::traits<Derived>::Scalar>,
const ConstStartMinusOne > HNormalizedReturnType;
const HNormalizedReturnType hnormalized() const;
typedef EIGEN_EXPR_BINARYOP_SCALAR_RETURN_TYPE(ConstStartMinusOne,Scalar,quotient) HNormalizedReturnType;
EIGEN_DEVICE_FUNC
inline const HNormalizedReturnType hnormalized() const;
////////// Householder module ///////////
@@ -452,58 +444,32 @@ template<typename Derived> class MatrixBase
///////// MatrixFunctions module /////////
typedef typename internal::stem_function<Scalar>::type StemFunction;
const MatrixExponentialReturnValue<Derived> exp() const;
#define EIGEN_MATRIX_FUNCTION(ReturnType, Name, Description) \
/** \returns an expression of the matrix Description of \c *this. \brief This function requires the <a href="unsupported/group__MatrixFunctions__Module.html"> unsupported MatrixFunctions module</a>. To compute the coefficient-wise Description use ArrayBase::##Name . */ \
const ReturnType<Derived> Name() const;
#define EIGEN_MATRIX_FUNCTION_1(ReturnType, Name, Description, Argument) \
/** \returns an expression of the matrix Description of \c *this. \brief This function requires the <a href="unsupported/group__MatrixFunctions__Module.html"> unsupported MatrixFunctions module</a>. To compute the coefficient-wise Description use ArrayBase::##Name . */ \
const ReturnType<Derived> Name(Argument) const;
EIGEN_MATRIX_FUNCTION(MatrixExponentialReturnValue, exp, exponential)
/** \brief Helper function for the <a href="unsupported/group__MatrixFunctions__Module.html"> unsupported MatrixFunctions module</a>.*/
const MatrixFunctionReturnValue<Derived> matrixFunction(StemFunction f) const;
const MatrixFunctionReturnValue<Derived> cosh() const;
const MatrixFunctionReturnValue<Derived> sinh() const;
const MatrixFunctionReturnValue<Derived> cos() const;
const MatrixFunctionReturnValue<Derived> sin() const;
const MatrixSquareRootReturnValue<Derived> sqrt() const;
const MatrixLogarithmReturnValue<Derived> log() const;
const MatrixPowerReturnValue<Derived> pow(const RealScalar& p) const;
#ifdef EIGEN2_SUPPORT
template<typename ProductDerived, typename Lhs, typename Rhs>
Derived& operator+=(const Flagged<ProductBase<ProductDerived, Lhs,Rhs>, 0,
EvalBeforeAssigningBit>& other);
template<typename ProductDerived, typename Lhs, typename Rhs>
Derived& operator-=(const Flagged<ProductBase<ProductDerived, Lhs,Rhs>, 0,
EvalBeforeAssigningBit>& other);
/** \deprecated because .lazy() is deprecated
* Overloaded for cache friendly product evaluation */
template<typename OtherDerived>
Derived& lazyAssign(const Flagged<OtherDerived, 0, EvalBeforeAssigningBit>& other)
{ return lazyAssign(other._expression()); }
template<unsigned int Added>
const Flagged<Derived, Added, 0> marked() const;
const Flagged<Derived, 0, EvalBeforeAssigningBit> lazy() const;
inline const Cwise<Derived> cwise() const;
inline Cwise<Derived> cwise();
VectorBlock<Derived> start(Index size);
const VectorBlock<const Derived> start(Index size) const;
VectorBlock<Derived> end(Index size);
const VectorBlock<const Derived> end(Index size) const;
template<int Size> VectorBlock<Derived,Size> start();
template<int Size> const VectorBlock<const Derived,Size> start() const;
template<int Size> VectorBlock<Derived,Size> end();
template<int Size> const VectorBlock<const Derived,Size> end() const;
Minor<Derived> minor(Index row, Index col);
const Minor<Derived> minor(Index row, Index col) const;
#endif
EIGEN_MATRIX_FUNCTION(MatrixFunctionReturnValue, cosh, hyperbolic cosine)
EIGEN_MATRIX_FUNCTION(MatrixFunctionReturnValue, sinh, hyperbolic sine)
EIGEN_MATRIX_FUNCTION(MatrixFunctionReturnValue, cos, cosine)
EIGEN_MATRIX_FUNCTION(MatrixFunctionReturnValue, sin, sine)
EIGEN_MATRIX_FUNCTION(MatrixSquareRootReturnValue, sqrt, square root)
EIGEN_MATRIX_FUNCTION(MatrixLogarithmReturnValue, log, logarithm)
EIGEN_MATRIX_FUNCTION_1(MatrixPowerReturnValue, pow, power to \c p, const RealScalar& p)
EIGEN_MATRIX_FUNCTION_1(MatrixComplexPowerReturnValue, pow, power to \c p, const std::complex<RealScalar>& p)
protected:
MatrixBase() : Base() {}
EIGEN_DEVICE_FUNC MatrixBase() : Base() {}
private:
explicit MatrixBase(int);
MatrixBase(int,int);
template<typename OtherDerived> explicit MatrixBase(const MatrixBase<OtherDerived>&);
EIGEN_DEVICE_FUNC explicit MatrixBase(int);
EIGEN_DEVICE_FUNC MatrixBase(int,int);
template<typename OtherDerived> EIGEN_DEVICE_FUNC explicit MatrixBase(const MatrixBase<OtherDerived>&);
protected:
// mixing arrays and matrices is not legal
template<typename OtherDerived> Derived& operator+=(const ArrayBase<OtherDerived>& )

View File

@@ -13,25 +13,24 @@
namespace Eigen {
/** \class NestByValue
* \ingroup Core_Module
*
* \brief Expression which must be nested by value
*
* \param ExpressionType the type of the object of which we are requiring nesting-by-value
*
* This class is the return type of MatrixBase::nestByValue()
* and most of the time this is the only way it is used.
*
* \sa MatrixBase::nestByValue()
*/
namespace internal {
template<typename ExpressionType>
struct traits<NestByValue<ExpressionType> > : public traits<ExpressionType>
{};
}
/** \class NestByValue
* \ingroup Core_Module
*
* \brief Expression which must be nested by value
*
* \tparam ExpressionType the type of the object of which we are requiring nesting-by-value
*
* This class is the return type of MatrixBase::nestByValue()
* and most of the time this is the only way it is used.
*
* \sa MatrixBase::nestByValue()
*/
template<typename ExpressionType> class NestByValue
: public internal::dense_xpr_base< NestByValue<ExpressionType> >::type
{
@@ -40,29 +39,29 @@ template<typename ExpressionType> class NestByValue
typedef typename internal::dense_xpr_base<NestByValue>::type Base;
EIGEN_DENSE_PUBLIC_INTERFACE(NestByValue)
inline NestByValue(const ExpressionType& matrix) : m_expression(matrix) {}
EIGEN_DEVICE_FUNC explicit inline NestByValue(const ExpressionType& matrix) : m_expression(matrix) {}
inline Index rows() const { return m_expression.rows(); }
inline Index cols() const { return m_expression.cols(); }
inline Index outerStride() const { return m_expression.outerStride(); }
inline Index innerStride() const { return m_expression.innerStride(); }
EIGEN_DEVICE_FUNC inline Index rows() const { return m_expression.rows(); }
EIGEN_DEVICE_FUNC inline Index cols() const { return m_expression.cols(); }
EIGEN_DEVICE_FUNC inline Index outerStride() const { return m_expression.outerStride(); }
EIGEN_DEVICE_FUNC inline Index innerStride() const { return m_expression.innerStride(); }
inline const CoeffReturnType coeff(Index row, Index col) const
EIGEN_DEVICE_FUNC inline const CoeffReturnType coeff(Index row, Index col) const
{
return m_expression.coeff(row, col);
}
inline Scalar& coeffRef(Index row, Index col)
EIGEN_DEVICE_FUNC inline Scalar& coeffRef(Index row, Index col)
{
return m_expression.const_cast_derived().coeffRef(row, col);
}
inline const CoeffReturnType coeff(Index index) const
EIGEN_DEVICE_FUNC inline const CoeffReturnType coeff(Index index) const
{
return m_expression.coeff(index);
}
inline Scalar& coeffRef(Index index)
EIGEN_DEVICE_FUNC inline Scalar& coeffRef(Index index)
{
return m_expression.const_cast_derived().coeffRef(index);
}
@@ -91,7 +90,7 @@ template<typename ExpressionType> class NestByValue
m_expression.const_cast_derived().template writePacket<LoadMode>(index, x);
}
operator const ExpressionType&() const { return m_expression; }
EIGEN_DEVICE_FUNC operator const ExpressionType&() const { return m_expression; }
protected:
const ExpressionType m_expression;

View File

@@ -17,7 +17,7 @@ namespace Eigen {
*
* \brief Pseudo expression providing an operator = assuming no aliasing
*
* \param ExpressionType the type of the object on which to do the lazy assignment
* \tparam ExpressionType the type of the object on which to do the lazy assignment
*
* This class represents an expression with special assignment operators
* assuming no aliasing between the target expression and the source expression.
@@ -30,62 +30,36 @@ namespace Eigen {
template<typename ExpressionType, template <typename> class StorageBase>
class NoAlias
{
typedef typename ExpressionType::Scalar Scalar;
public:
NoAlias(ExpressionType& expression) : m_expression(expression) {}
/** Behaves like MatrixBase::lazyAssign(other)
* \sa MatrixBase::lazyAssign() */
template<typename OtherDerived>
EIGEN_STRONG_INLINE ExpressionType& operator=(const StorageBase<OtherDerived>& other)
{ return internal::assign_selector<ExpressionType,OtherDerived,false>::run(m_expression,other.derived()); }
/** \sa MatrixBase::operator+= */
template<typename OtherDerived>
EIGEN_STRONG_INLINE ExpressionType& operator+=(const StorageBase<OtherDerived>& other)
{
typedef SelfCwiseBinaryOp<internal::scalar_sum_op<Scalar>, ExpressionType, OtherDerived> SelfAdder;
SelfAdder tmp(m_expression);
typedef typename internal::nested<OtherDerived>::type OtherDerivedNested;
typedef typename internal::remove_all<OtherDerivedNested>::type _OtherDerivedNested;
internal::assign_selector<SelfAdder,_OtherDerivedNested,false>::run(tmp,OtherDerivedNested(other.derived()));
return m_expression;
}
/** \sa MatrixBase::operator-= */
template<typename OtherDerived>
EIGEN_STRONG_INLINE ExpressionType& operator-=(const StorageBase<OtherDerived>& other)
{
typedef SelfCwiseBinaryOp<internal::scalar_difference_op<Scalar>, ExpressionType, OtherDerived> SelfAdder;
SelfAdder tmp(m_expression);
typedef typename internal::nested<OtherDerived>::type OtherDerivedNested;
typedef typename internal::remove_all<OtherDerivedNested>::type _OtherDerivedNested;
internal::assign_selector<SelfAdder,_OtherDerivedNested,false>::run(tmp,OtherDerivedNested(other.derived()));
return m_expression;
}
#ifndef EIGEN_PARSED_BY_DOXYGEN
template<typename ProductDerived, typename Lhs, typename Rhs>
EIGEN_STRONG_INLINE ExpressionType& operator+=(const ProductBase<ProductDerived, Lhs,Rhs>& other)
{ other.derived().addTo(m_expression); return m_expression; }
template<typename ProductDerived, typename Lhs, typename Rhs>
EIGEN_STRONG_INLINE ExpressionType& operator-=(const ProductBase<ProductDerived, Lhs,Rhs>& other)
{ other.derived().subTo(m_expression); return m_expression; }
template<typename Lhs, typename Rhs, int NestingFlags>
EIGEN_STRONG_INLINE ExpressionType& operator+=(const CoeffBasedProduct<Lhs,Rhs,NestingFlags>& other)
{ return m_expression.derived() += CoeffBasedProduct<Lhs,Rhs,NestByRefBit>(other.lhs(), other.rhs()); }
template<typename Lhs, typename Rhs, int NestingFlags>
EIGEN_STRONG_INLINE ExpressionType& operator-=(const CoeffBasedProduct<Lhs,Rhs,NestingFlags>& other)
{ return m_expression.derived() -= CoeffBasedProduct<Lhs,Rhs,NestByRefBit>(other.lhs(), other.rhs()); }
typedef typename ExpressionType::Scalar Scalar;
explicit NoAlias(ExpressionType& expression) : m_expression(expression) {}
template<typename OtherDerived>
ExpressionType& operator=(const ReturnByValue<OtherDerived>& func)
{ return m_expression = func; }
#endif
EIGEN_DEVICE_FUNC
EIGEN_STRONG_INLINE ExpressionType& operator=(const StorageBase<OtherDerived>& other)
{
call_assignment_no_alias(m_expression, other.derived(), internal::assign_op<Scalar,typename OtherDerived::Scalar>());
return m_expression;
}
template<typename OtherDerived>
EIGEN_DEVICE_FUNC
EIGEN_STRONG_INLINE ExpressionType& operator+=(const StorageBase<OtherDerived>& other)
{
call_assignment_no_alias(m_expression, other.derived(), internal::add_assign_op<Scalar,typename OtherDerived::Scalar>());
return m_expression;
}
template<typename OtherDerived>
EIGEN_DEVICE_FUNC
EIGEN_STRONG_INLINE ExpressionType& operator-=(const StorageBase<OtherDerived>& other)
{
call_assignment_no_alias(m_expression, other.derived(), internal::sub_assign_op<Scalar,typename OtherDerived::Scalar>());
return m_expression;
}
EIGEN_DEVICE_FUNC
ExpressionType& expression() const
{
return m_expression;
@@ -126,7 +100,7 @@ class NoAlias
template<typename Derived>
NoAlias<Derived,MatrixBase> MatrixBase<Derived>::noalias()
{
return derived();
return NoAlias<Derived, Eigen::MatrixBase >(derived());
}
} // end namespace Eigen

View File

@@ -12,24 +12,57 @@
namespace Eigen {
namespace internal {
// default implementation of digits10(), based on numeric_limits if specialized,
// 0 for integer types, and log10(epsilon()) otherwise.
template< typename T,
bool use_numeric_limits = std::numeric_limits<T>::is_specialized,
bool is_integer = NumTraits<T>::IsInteger>
struct default_digits10_impl
{
static int run() { return std::numeric_limits<T>::digits10; }
};
template<typename T>
struct default_digits10_impl<T,false,false> // Floating point
{
static int run() {
using std::log10;
using std::ceil;
typedef typename NumTraits<T>::Real Real;
return int(ceil(-log10(NumTraits<Real>::epsilon())));
}
};
template<typename T>
struct default_digits10_impl<T,false,true> // Integer
{
static int run() { return 0; }
};
} // end namespace internal
/** \class NumTraits
* \ingroup Core_Module
*
* \brief Holds information about the various numeric (i.e. scalar) types allowed by Eigen.
*
* \param T the numeric type at hand
* \tparam T the numeric type at hand
*
* This class stores enums, typedefs and static methods giving information about a numeric type.
*
* The provided data consists of:
* \li A typedef \a Real, giving the "real part" type of \a T. If \a T is already real,
* then \a Real is just a typedef to \a T. If \a T is \c std::complex<U> then \a Real
* \li A typedef \c Real, giving the "real part" type of \a T. If \a T is already real,
* then \c Real is just a typedef to \a T. If \a T is \c std::complex<U> then \c Real
* is a typedef to \a U.
* \li A typedef \a NonInteger, giving the type that should be used for operations producing non-integral values,
* \li A typedef \c NonInteger, giving the type that should be used for operations producing non-integral values,
* such as quotients, square roots, etc. If \a T is a floating-point type, then this typedef just gives
* \a T again. Note however that many Eigen functions such as internal::sqrt simply refuse to
* take integers. Outside of a few cases, Eigen doesn't do automatic type promotion. Thus, this typedef is
* only intended as a helper for code that needs to explicitly promote types.
* \li A typedef \c Literal giving the type to use for numeric literals such as "2" or "0.5". For instance, for \c std::complex<U>, Literal is defined as \c U.
* Of course, this type must be fully compatible with \a T. In doubt, just use \a T here.
* \li A typedef \a Nested giving the type to use to nest a value inside of the expression tree. If you don't know what
* this means, just use \a T here.
* \li An enum value \a IsComplex. It is equal to 1 if \a T is a \c std::complex
@@ -42,10 +75,14 @@ namespace Eigen {
* \li An enum value \a IsSigned. It is equal to \c 1 if \a T is a signed type and to 0 if \a T is unsigned.
* \li An enum value \a RequireInitialization. It is equal to \c 1 if the constructor of the numeric type \a T must
* be called, and to 0 if it is safe not to call it. Default is 0 if \a T is an arithmetic type, and 1 otherwise.
* \li An epsilon() function which, unlike std::numeric_limits::epsilon(), returns a \a Real instead of a \a T.
* \li An epsilon() function which, unlike <a href="http://en.cppreference.com/w/cpp/types/numeric_limits/epsilon">std::numeric_limits::epsilon()</a>,
* it returns a \a Real instead of a \a T.
* \li A dummy_precision() function returning a weak epsilon value. It is mainly used as a default
* value by the fuzzy comparison operators.
* \li highest() and lowest() functions returning the highest and lowest possible values respectively.
* \li digits10() function returning the number of decimal digits that can be represented without change. This is
* the analogue of <a href="http://en.cppreference.com/w/cpp/types/numeric_limits/digits10">std::numeric_limits<T>::digits10</a>
* which is used as the default implementation if specialized.
*/
template<typename T> struct GenericNumTraits
@@ -67,22 +104,47 @@ template<typename T> struct GenericNumTraits
T
>::type NonInteger;
typedef T Nested;
typedef T Literal;
static inline Real epsilon() { return std::numeric_limits<T>::epsilon(); }
EIGEN_DEVICE_FUNC
static inline Real epsilon()
{
return numext::numeric_limits<T>::epsilon();
}
EIGEN_DEVICE_FUNC
static inline int digits10()
{
return internal::default_digits10_impl<T>::run();
}
EIGEN_DEVICE_FUNC
static inline Real dummy_precision()
{
// make sure to override this for floating-point types
return Real(0);
}
static inline T highest() { return (std::numeric_limits<T>::max)(); }
static inline T lowest() { return IsInteger ? (std::numeric_limits<T>::min)() : (-(std::numeric_limits<T>::max)()); }
#ifdef EIGEN2_SUPPORT
enum {
HasFloatingPoint = !IsInteger
};
typedef NonInteger FloatingPoint;
#endif
EIGEN_DEVICE_FUNC
static inline T highest() {
return (numext::numeric_limits<T>::max)();
}
EIGEN_DEVICE_FUNC
static inline T lowest() {
return IsInteger ? (numext::numeric_limits<T>::min)() : (-(numext::numeric_limits<T>::max)());
}
EIGEN_DEVICE_FUNC
static inline T infinity() {
return numext::numeric_limits<T>::infinity();
}
EIGEN_DEVICE_FUNC
static inline T quiet_NaN() {
return numext::numeric_limits<T>::quiet_NaN();
}
};
template<typename T> struct NumTraits : GenericNumTraits<T>
@@ -91,11 +153,13 @@ template<typename T> struct NumTraits : GenericNumTraits<T>
template<> struct NumTraits<float>
: GenericNumTraits<float>
{
EIGEN_DEVICE_FUNC
static inline float dummy_precision() { return 1e-5f; }
};
template<> struct NumTraits<double> : GenericNumTraits<double>
{
EIGEN_DEVICE_FUNC
static inline double dummy_precision() { return 1e-12; }
};
@@ -109,6 +173,7 @@ template<typename _Real> struct NumTraits<std::complex<_Real> >
: GenericNumTraits<std::complex<_Real> >
{
typedef _Real Real;
typedef typename NumTraits<_Real>::Literal Literal;
enum {
IsComplex = 1,
RequireInitialization = NumTraits<_Real>::RequireInitialization,
@@ -117,8 +182,12 @@ template<typename _Real> struct NumTraits<std::complex<_Real> >
MulCost = 4 * NumTraits<Real>::MulCost + 2 * NumTraits<Real>::AddCost
};
EIGEN_DEVICE_FUNC
static inline Real epsilon() { return NumTraits<Real>::epsilon(); }
EIGEN_DEVICE_FUNC
static inline Real dummy_precision() { return NumTraits<Real>::dummy_precision(); }
EIGEN_DEVICE_FUNC
static inline int digits10() { return NumTraits<Real>::digits10(); }
};
template<typename Scalar, int Rows, int Cols, int Options, int MaxRows, int MaxCols>
@@ -130,21 +199,50 @@ struct NumTraits<Array<Scalar, Rows, Cols, Options, MaxRows, MaxCols> >
typedef typename NumTraits<Scalar>::NonInteger NonIntegerScalar;
typedef Array<NonIntegerScalar, Rows, Cols, Options, MaxRows, MaxCols> NonInteger;
typedef ArrayType & Nested;
typedef typename NumTraits<Scalar>::Literal Literal;
enum {
IsComplex = NumTraits<Scalar>::IsComplex,
IsInteger = NumTraits<Scalar>::IsInteger,
IsSigned = NumTraits<Scalar>::IsSigned,
RequireInitialization = 1,
ReadCost = ArrayType::SizeAtCompileTime==Dynamic ? Dynamic : ArrayType::SizeAtCompileTime * NumTraits<Scalar>::ReadCost,
AddCost = ArrayType::SizeAtCompileTime==Dynamic ? Dynamic : ArrayType::SizeAtCompileTime * NumTraits<Scalar>::AddCost,
MulCost = ArrayType::SizeAtCompileTime==Dynamic ? Dynamic : ArrayType::SizeAtCompileTime * NumTraits<Scalar>::MulCost
ReadCost = ArrayType::SizeAtCompileTime==Dynamic ? HugeCost : ArrayType::SizeAtCompileTime * NumTraits<Scalar>::ReadCost,
AddCost = ArrayType::SizeAtCompileTime==Dynamic ? HugeCost : ArrayType::SizeAtCompileTime * NumTraits<Scalar>::AddCost,
MulCost = ArrayType::SizeAtCompileTime==Dynamic ? HugeCost : ArrayType::SizeAtCompileTime * NumTraits<Scalar>::MulCost
};
EIGEN_DEVICE_FUNC
static inline RealScalar epsilon() { return NumTraits<RealScalar>::epsilon(); }
EIGEN_DEVICE_FUNC
static inline RealScalar dummy_precision() { return NumTraits<RealScalar>::dummy_precision(); }
static inline int digits10() { return NumTraits<Scalar>::digits10(); }
};
template<> struct NumTraits<std::string>
: GenericNumTraits<std::string>
{
enum {
RequireInitialization = 1,
ReadCost = HugeCost,
AddCost = HugeCost,
MulCost = HugeCost
};
static inline int digits10() { return 0; }
private:
static inline std::string epsilon();
static inline std::string dummy_precision();
static inline std::string lowest();
static inline std::string highest();
static inline std::string infinity();
static inline std::string quiet_NaN();
};
// Empty specialization for void to allow template specialization based on NumTraits<T>::Real with T==void and SFINAE.
template<> struct NumTraits<void> {};
} // end namespace Eigen
#endif // EIGEN_NUMTRAITS_H

View File

@@ -2,7 +2,7 @@
// for linear algebra.
//
// Copyright (C) 2009 Benoit Jacob <jacob.benoit.1@gmail.com>
// Copyright (C) 2009-2011 Gael Guennebaud <gael.guennebaud@inria.fr>
// Copyright (C) 2009-2015 Gael Guennebaud <gael.guennebaud@inria.fr>
//
// This Source Code Form is subject to the terms of the Mozilla
// Public License v. 2.0. If a copy of the MPL was not distributed
@@ -13,14 +13,18 @@
namespace Eigen {
template<int RowCol,typename IndicesType,typename MatrixType, typename StorageKind> class PermutedImpl;
namespace internal {
enum PermPermProduct_t {PermPermProduct};
} // end namespace internal
/** \class PermutationBase
* \ingroup Core_Module
*
* \brief Base class for permutations
*
* \param Derived the derived class
* \tparam Derived the derived class
*
* This class is the base class for all expressions representing a permutation matrix,
* internally stored as a vector of integers.
@@ -38,17 +42,6 @@ template<int RowCol,typename IndicesType,typename MatrixType, typename StorageKi
*
* \sa class PermutationMatrix, class PermutationWrapper
*/
namespace internal {
template<typename PermutationType, typename MatrixType, int Side, bool Transposed=false>
struct permut_matrix_product_retval;
template<typename PermutationType, typename MatrixType, int Side, bool Transposed=false>
struct permut_sparsematrix_product_retval;
enum PermPermProduct_t {PermPermProduct};
} // end namespace internal
template<typename Derived>
class PermutationBase : public EigenBase<Derived>
{
@@ -60,19 +53,20 @@ class PermutationBase : public EigenBase<Derived>
typedef typename Traits::IndicesType IndicesType;
enum {
Flags = Traits::Flags,
CoeffReadCost = Traits::CoeffReadCost,
RowsAtCompileTime = Traits::RowsAtCompileTime,
ColsAtCompileTime = Traits::ColsAtCompileTime,
MaxRowsAtCompileTime = Traits::MaxRowsAtCompileTime,
MaxColsAtCompileTime = Traits::MaxColsAtCompileTime
};
typedef typename Traits::Scalar Scalar;
typedef typename Traits::Index Index;
typedef Matrix<Scalar,RowsAtCompileTime,ColsAtCompileTime,0,MaxRowsAtCompileTime,MaxColsAtCompileTime>
typedef typename Traits::StorageIndex StorageIndex;
typedef Matrix<StorageIndex,RowsAtCompileTime,ColsAtCompileTime,0,MaxRowsAtCompileTime,MaxColsAtCompileTime>
DenseMatrixType;
typedef PermutationMatrix<IndicesType::SizeAtCompileTime,IndicesType::MaxSizeAtCompileTime,Index>
typedef PermutationMatrix<IndicesType::SizeAtCompileTime,IndicesType::MaxSizeAtCompileTime,StorageIndex>
PlainPermutationType;
typedef PlainPermutationType PlainObject;
using Base::derived;
typedef Inverse<Derived> InverseReturnType;
typedef void Scalar;
#endif
/** Copies the other permutation into *this */
@@ -118,7 +112,7 @@ class PermutationBase : public EigenBase<Derived>
void evalTo(MatrixBase<DenseDerived>& other) const
{
other.setZero();
for (int i=0; i<rows();++i)
for (Index i=0; i<rows(); ++i)
other.coeffRef(indices().coeff(i),i) = typename DenseDerived::Scalar(1);
}
#endif
@@ -147,7 +141,8 @@ class PermutationBase : public EigenBase<Derived>
/** Sets *this to be the identity permutation matrix */
void setIdentity()
{
for(Index i = 0; i < size(); ++i)
StorageIndex n = StorageIndex(size());
for(StorageIndex i = 0; i < n; ++i)
indices().coeffRef(i) = i;
}
@@ -163,18 +158,18 @@ class PermutationBase : public EigenBase<Derived>
*
* \returns a reference to *this.
*
* \warning This is much slower than applyTranspositionOnTheRight(int,int):
* \warning This is much slower than applyTranspositionOnTheRight(Index,Index):
* this has linear complexity and requires a lot of branching.
*
* \sa applyTranspositionOnTheRight(int,int)
* \sa applyTranspositionOnTheRight(Index,Index)
*/
Derived& applyTranspositionOnTheLeft(Index i, Index j)
{
eigen_assert(i>=0 && j>=0 && i<size() && j<size());
for(Index k = 0; k < size(); ++k)
{
if(indices().coeff(k) == i) indices().coeffRef(k) = j;
else if(indices().coeff(k) == j) indices().coeffRef(k) = i;
if(indices().coeff(k) == i) indices().coeffRef(k) = StorageIndex(j);
else if(indices().coeff(k) == j) indices().coeffRef(k) = StorageIndex(i);
}
return derived();
}
@@ -185,7 +180,7 @@ class PermutationBase : public EigenBase<Derived>
*
* This is a fast operation, it only consists in swapping two indices.
*
* \sa applyTranspositionOnTheLeft(int,int)
* \sa applyTranspositionOnTheLeft(Index,Index)
*/
Derived& applyTranspositionOnTheRight(Index i, Index j)
{
@@ -196,16 +191,16 @@ class PermutationBase : public EigenBase<Derived>
/** \returns the inverse permutation matrix.
*
* \note \note_try_to_help_rvo
* \note \blank \note_try_to_help_rvo
*/
inline Transpose<PermutationBase> inverse() const
{ return derived(); }
inline InverseReturnType inverse() const
{ return InverseReturnType(derived()); }
/** \returns the tranpose permutation matrix.
*
* \note \note_try_to_help_rvo
* \note \blank \note_try_to_help_rvo
*/
inline Transpose<PermutationBase> transpose() const
{ return derived(); }
inline InverseReturnType transpose() const
{ return InverseReturnType(derived()); }
/**** multiplication helpers to hopefully get RVO ****/
@@ -215,13 +210,13 @@ class PermutationBase : public EigenBase<Derived>
template<typename OtherDerived>
void assignTranspose(const PermutationBase<OtherDerived>& other)
{
for (int i=0; i<rows();++i) indices().coeffRef(other.indices().coeff(i)) = i;
for (Index i=0; i<rows();++i) indices().coeffRef(other.indices().coeff(i)) = i;
}
template<typename Lhs,typename Rhs>
void assignProduct(const Lhs& lhs, const Rhs& rhs)
{
eigen_assert(lhs.cols() == rhs.rows());
for (int i=0; i<rows();++i) indices().coeffRef(i) = lhs.indices().coeff(rhs.indices().coeff(i));
for (Index i=0; i<rows();++i) indices().coeffRef(i) = lhs.indices().coeff(rhs.indices().coeff(i));
}
#endif
@@ -229,7 +224,7 @@ class PermutationBase : public EigenBase<Derived>
/** \returns the product permutation matrix.
*
* \note \note_try_to_help_rvo
* \note \blank \note_try_to_help_rvo
*/
template<typename Other>
inline PlainPermutationType operator*(const PermutationBase<Other>& other) const
@@ -237,18 +232,18 @@ class PermutationBase : public EigenBase<Derived>
/** \returns the product of a permutation with another inverse permutation.
*
* \note \note_try_to_help_rvo
* \note \blank \note_try_to_help_rvo
*/
template<typename Other>
inline PlainPermutationType operator*(const Transpose<PermutationBase<Other> >& other) const
inline PlainPermutationType operator*(const InverseImpl<Other,PermutationStorage>& other) const
{ return PlainPermutationType(internal::PermPermProduct, *this, other.eval()); }
/** \returns the product of an inverse permutation with another permutation.
*
* \note \note_try_to_help_rvo
* \note \blank \note_try_to_help_rvo
*/
template<typename Other> friend
inline PlainPermutationType operator*(const Transpose<PermutationBase<Other> >& other, const PermutationBase& perm)
inline PlainPermutationType operator*(const InverseImpl<Other, PermutationStorage>& other, const PermutationBase& perm)
{ return PlainPermutationType(internal::PermPermProduct, other.eval(), perm); }
/** \returns the determinant of the permutation matrix, which is either 1 or -1 depending on the parity of the permutation.
@@ -284,39 +279,43 @@ class PermutationBase : public EigenBase<Derived>
};
namespace internal {
template<int SizeAtCompileTime, int MaxSizeAtCompileTime, typename _StorageIndex>
struct traits<PermutationMatrix<SizeAtCompileTime, MaxSizeAtCompileTime, _StorageIndex> >
: traits<Matrix<_StorageIndex,SizeAtCompileTime,SizeAtCompileTime,0,MaxSizeAtCompileTime,MaxSizeAtCompileTime> >
{
typedef PermutationStorage StorageKind;
typedef Matrix<_StorageIndex, SizeAtCompileTime, 1, 0, MaxSizeAtCompileTime, 1> IndicesType;
typedef _StorageIndex StorageIndex;
typedef void Scalar;
};
}
/** \class PermutationMatrix
* \ingroup Core_Module
*
* \brief Permutation matrix
*
* \param SizeAtCompileTime the number of rows/cols, or Dynamic
* \param MaxSizeAtCompileTime the maximum number of rows/cols, or Dynamic. This optional parameter defaults to SizeAtCompileTime. Most of the time, you should not have to specify it.
* \param IndexType the interger type of the indices
* \tparam SizeAtCompileTime the number of rows/cols, or Dynamic
* \tparam MaxSizeAtCompileTime the maximum number of rows/cols, or Dynamic. This optional parameter defaults to SizeAtCompileTime. Most of the time, you should not have to specify it.
* \tparam _StorageIndex the integer type of the indices
*
* This class represents a permutation matrix, internally stored as a vector of integers.
*
* \sa class PermutationBase, class PermutationWrapper, class DiagonalMatrix
*/
namespace internal {
template<int SizeAtCompileTime, int MaxSizeAtCompileTime, typename IndexType>
struct traits<PermutationMatrix<SizeAtCompileTime, MaxSizeAtCompileTime, IndexType> >
: traits<Matrix<IndexType,SizeAtCompileTime,SizeAtCompileTime,0,MaxSizeAtCompileTime,MaxSizeAtCompileTime> >
{
typedef IndexType Index;
typedef Matrix<IndexType, SizeAtCompileTime, 1, 0, MaxSizeAtCompileTime, 1> IndicesType;
};
}
template<int SizeAtCompileTime, int MaxSizeAtCompileTime, typename IndexType>
class PermutationMatrix : public PermutationBase<PermutationMatrix<SizeAtCompileTime, MaxSizeAtCompileTime, IndexType> >
template<int SizeAtCompileTime, int MaxSizeAtCompileTime, typename _StorageIndex>
class PermutationMatrix : public PermutationBase<PermutationMatrix<SizeAtCompileTime, MaxSizeAtCompileTime, _StorageIndex> >
{
typedef PermutationBase<PermutationMatrix> Base;
typedef internal::traits<PermutationMatrix> Traits;
public:
typedef const PermutationMatrix& Nested;
#ifndef EIGEN_PARSED_BY_DOXYGEN
typedef typename Traits::IndicesType IndicesType;
typedef typename Traits::StorageIndex StorageIndex;
#endif
inline PermutationMatrix()
@@ -324,8 +323,10 @@ class PermutationMatrix : public PermutationBase<PermutationMatrix<SizeAtCompile
/** Constructs an uninitialized permutation matrix of given size.
*/
inline PermutationMatrix(int size) : m_indices(size)
{}
explicit inline PermutationMatrix(Index size) : m_indices(size)
{
eigen_internal_assert(size <= NumTraits<StorageIndex>::highest());
}
/** Copy constructor. */
template<typename OtherDerived>
@@ -346,7 +347,7 @@ class PermutationMatrix : public PermutationBase<PermutationMatrix<SizeAtCompile
* array's size.
*/
template<typename Other>
explicit inline PermutationMatrix(const MatrixBase<Other>& a_indices) : m_indices(a_indices)
explicit inline PermutationMatrix(const MatrixBase<Other>& indices) : m_indices(indices)
{}
/** Convert the Transpositions \a tr to a permutation matrix */
@@ -393,10 +394,13 @@ class PermutationMatrix : public PermutationBase<PermutationMatrix<SizeAtCompile
#ifndef EIGEN_PARSED_BY_DOXYGEN
template<typename Other>
PermutationMatrix(const Transpose<PermutationBase<Other> >& other)
: m_indices(other.nestedPermutation().size())
PermutationMatrix(const InverseImpl<Other,PermutationStorage>& other)
: m_indices(other.derived().nestedExpression().size())
{
for (int i=0; i<m_indices.size();++i) m_indices.coeffRef(other.nestedPermutation().indices().coeff(i)) = i;
eigen_internal_assert(m_indices.size() <= NumTraits<StorageIndex>::highest());
StorageIndex end = StorageIndex(m_indices.size());
for (StorageIndex i=0; i<end;++i)
m_indices.coeffRef(other.derived().nestedExpression().indices().coeff(i)) = i;
}
template<typename Lhs,typename Rhs>
PermutationMatrix(internal::PermPermProduct_t, const Lhs& lhs, const Rhs& rhs)
@@ -413,18 +417,20 @@ class PermutationMatrix : public PermutationBase<PermutationMatrix<SizeAtCompile
namespace internal {
template<int SizeAtCompileTime, int MaxSizeAtCompileTime, typename IndexType, int _PacketAccess>
struct traits<Map<PermutationMatrix<SizeAtCompileTime, MaxSizeAtCompileTime, IndexType>,_PacketAccess> >
: traits<Matrix<IndexType,SizeAtCompileTime,SizeAtCompileTime,0,MaxSizeAtCompileTime,MaxSizeAtCompileTime> >
template<int SizeAtCompileTime, int MaxSizeAtCompileTime, typename _StorageIndex, int _PacketAccess>
struct traits<Map<PermutationMatrix<SizeAtCompileTime, MaxSizeAtCompileTime, _StorageIndex>,_PacketAccess> >
: traits<Matrix<_StorageIndex,SizeAtCompileTime,SizeAtCompileTime,0,MaxSizeAtCompileTime,MaxSizeAtCompileTime> >
{
typedef IndexType Index;
typedef Map<const Matrix<IndexType, SizeAtCompileTime, 1, 0, MaxSizeAtCompileTime, 1>, _PacketAccess> IndicesType;
typedef PermutationStorage StorageKind;
typedef Map<const Matrix<_StorageIndex, SizeAtCompileTime, 1, 0, MaxSizeAtCompileTime, 1>, _PacketAccess> IndicesType;
typedef _StorageIndex StorageIndex;
typedef void Scalar;
};
}
template<int SizeAtCompileTime, int MaxSizeAtCompileTime, typename IndexType, int _PacketAccess>
class Map<PermutationMatrix<SizeAtCompileTime, MaxSizeAtCompileTime, IndexType>,_PacketAccess>
: public PermutationBase<Map<PermutationMatrix<SizeAtCompileTime, MaxSizeAtCompileTime, IndexType>,_PacketAccess> >
template<int SizeAtCompileTime, int MaxSizeAtCompileTime, typename _StorageIndex, int _PacketAccess>
class Map<PermutationMatrix<SizeAtCompileTime, MaxSizeAtCompileTime, _StorageIndex>,_PacketAccess>
: public PermutationBase<Map<PermutationMatrix<SizeAtCompileTime, MaxSizeAtCompileTime, _StorageIndex>,_PacketAccess> >
{
typedef PermutationBase<Map> Base;
typedef internal::traits<Map> Traits;
@@ -432,14 +438,14 @@ class Map<PermutationMatrix<SizeAtCompileTime, MaxSizeAtCompileTime, IndexType>,
#ifndef EIGEN_PARSED_BY_DOXYGEN
typedef typename Traits::IndicesType IndicesType;
typedef typename IndicesType::Scalar Index;
typedef typename IndicesType::Scalar StorageIndex;
#endif
inline Map(const Index* indicesPtr)
inline Map(const StorageIndex* indicesPtr)
: m_indices(indicesPtr)
{}
inline Map(const Index* indicesPtr, Index size)
inline Map(const StorageIndex* indicesPtr, Index size)
: m_indices(indicesPtr,size)
{}
@@ -474,40 +480,36 @@ class Map<PermutationMatrix<SizeAtCompileTime, MaxSizeAtCompileTime, IndexType>,
IndicesType m_indices;
};
/** \class PermutationWrapper
* \ingroup Core_Module
*
* \brief Class to view a vector of integers as a permutation matrix
*
* \param _IndicesType the type of the vector of integer (can be any compatible expression)
*
* This class allows to view any vector expression of integers as a permutation matrix.
*
* \sa class PermutationBase, class PermutationMatrix
*/
struct PermutationStorage {};
template<typename _IndicesType> class TranspositionsWrapper;
namespace internal {
template<typename _IndicesType>
struct traits<PermutationWrapper<_IndicesType> >
{
typedef PermutationStorage StorageKind;
typedef typename _IndicesType::Scalar Scalar;
typedef typename _IndicesType::Scalar Index;
typedef void Scalar;
typedef typename _IndicesType::Scalar StorageIndex;
typedef _IndicesType IndicesType;
enum {
RowsAtCompileTime = _IndicesType::SizeAtCompileTime,
ColsAtCompileTime = _IndicesType::SizeAtCompileTime,
MaxRowsAtCompileTime = IndicesType::MaxRowsAtCompileTime,
MaxColsAtCompileTime = IndicesType::MaxColsAtCompileTime,
Flags = 0,
CoeffReadCost = _IndicesType::CoeffReadCost
MaxRowsAtCompileTime = IndicesType::MaxSizeAtCompileTime,
MaxColsAtCompileTime = IndicesType::MaxSizeAtCompileTime,
Flags = 0
};
};
}
/** \class PermutationWrapper
* \ingroup Core_Module
*
* \brief Class to view a vector of integers as a permutation matrix
*
* \tparam _IndicesType the type of the vector of integer (can be any compatible expression)
*
* This class allows to view any vector expression of integers as a permutation matrix.
*
* \sa class PermutationBase, class PermutationMatrix
*/
template<typename _IndicesType>
class PermutationWrapper : public PermutationBase<PermutationWrapper<_IndicesType> >
{
@@ -519,8 +521,8 @@ class PermutationWrapper : public PermutationBase<PermutationWrapper<_IndicesTyp
typedef typename Traits::IndicesType IndicesType;
#endif
inline PermutationWrapper(const IndicesType& a_indices)
: m_indices(a_indices)
inline PermutationWrapper(const IndicesType& indices)
: m_indices(indices)
{}
/** const version of indices(). */
@@ -532,182 +534,86 @@ class PermutationWrapper : public PermutationBase<PermutationWrapper<_IndicesTyp
typename IndicesType::Nested m_indices;
};
/** \returns the matrix with the permutation applied to the columns.
*/
template<typename Derived, typename PermutationDerived>
inline const internal::permut_matrix_product_retval<PermutationDerived, Derived, OnTheRight>
operator*(const MatrixBase<Derived>& matrix,
const PermutationBase<PermutationDerived> &permutation)
template<typename MatrixDerived, typename PermutationDerived>
EIGEN_DEVICE_FUNC
const Product<MatrixDerived, PermutationDerived, AliasFreeProduct>
operator*(const MatrixBase<MatrixDerived> &matrix,
const PermutationBase<PermutationDerived>& permutation)
{
return internal::permut_matrix_product_retval
<PermutationDerived, Derived, OnTheRight>
(permutation.derived(), matrix.derived());
return Product<MatrixDerived, PermutationDerived, AliasFreeProduct>
(matrix.derived(), permutation.derived());
}
/** \returns the matrix with the permutation applied to the rows.
*/
template<typename Derived, typename PermutationDerived>
inline const internal::permut_matrix_product_retval
<PermutationDerived, Derived, OnTheLeft>
template<typename PermutationDerived, typename MatrixDerived>
EIGEN_DEVICE_FUNC
const Product<PermutationDerived, MatrixDerived, AliasFreeProduct>
operator*(const PermutationBase<PermutationDerived> &permutation,
const MatrixBase<Derived>& matrix)
const MatrixBase<MatrixDerived>& matrix)
{
return internal::permut_matrix_product_retval
<PermutationDerived, Derived, OnTheLeft>
(permutation.derived(), matrix.derived());
return Product<PermutationDerived, MatrixDerived, AliasFreeProduct>
(permutation.derived(), matrix.derived());
}
namespace internal {
template<typename PermutationType, typename MatrixType, int Side, bool Transposed>
struct traits<permut_matrix_product_retval<PermutationType, MatrixType, Side, Transposed> >
template<typename PermutationType>
class InverseImpl<PermutationType, PermutationStorage>
: public EigenBase<Inverse<PermutationType> >
{
typedef typename MatrixType::PlainObject ReturnType;
};
template<typename PermutationType, typename MatrixType, int Side, bool Transposed>
struct permut_matrix_product_retval
: public ReturnByValue<permut_matrix_product_retval<PermutationType, MatrixType, Side, Transposed> >
{
typedef typename remove_all<typename MatrixType::Nested>::type MatrixTypeNestedCleaned;
typedef typename MatrixType::Index Index;
permut_matrix_product_retval(const PermutationType& perm, const MatrixType& matrix)
: m_permutation(perm), m_matrix(matrix)
{}
inline Index rows() const { return m_matrix.rows(); }
inline Index cols() const { return m_matrix.cols(); }
template<typename Dest> inline void evalTo(Dest& dst) const
{
const Index n = Side==OnTheLeft ? rows() : cols();
// FIXME we need an is_same for expression that is not sensitive to constness. For instance
// is_same_xpr<Block<const Matrix>, Block<Matrix> >::value should be true.
if( is_same<MatrixTypeNestedCleaned,Dest>::value
&& blas_traits<MatrixTypeNestedCleaned>::HasUsableDirectAccess
&& blas_traits<Dest>::HasUsableDirectAccess
&& extract_data(dst) == extract_data(m_matrix))
{
// apply the permutation inplace
Matrix<bool,PermutationType::RowsAtCompileTime,1,0,PermutationType::MaxRowsAtCompileTime> mask(m_permutation.size());
mask.fill(false);
Index r = 0;
while(r < m_permutation.size())
{
// search for the next seed
while(r<m_permutation.size() && mask[r]) r++;
if(r>=m_permutation.size())
break;
// we got one, let's follow it until we are back to the seed
Index k0 = r++;
Index kPrev = k0;
mask.coeffRef(k0) = true;
for(Index k=m_permutation.indices().coeff(k0); k!=k0; k=m_permutation.indices().coeff(k))
{
Block<Dest, Side==OnTheLeft ? 1 : Dest::RowsAtCompileTime, Side==OnTheRight ? 1 : Dest::ColsAtCompileTime>(dst, k)
.swap(Block<Dest, Side==OnTheLeft ? 1 : Dest::RowsAtCompileTime, Side==OnTheRight ? 1 : Dest::ColsAtCompileTime>
(dst,((Side==OnTheLeft) ^ Transposed) ? k0 : kPrev));
mask.coeffRef(k) = true;
kPrev = k;
}
}
}
else
{
for(int i = 0; i < n; ++i)
{
Block<Dest, Side==OnTheLeft ? 1 : Dest::RowsAtCompileTime, Side==OnTheRight ? 1 : Dest::ColsAtCompileTime>
(dst, ((Side==OnTheLeft) ^ Transposed) ? m_permutation.indices().coeff(i) : i)
=
Block<const MatrixTypeNestedCleaned,Side==OnTheLeft ? 1 : MatrixType::RowsAtCompileTime,Side==OnTheRight ? 1 : MatrixType::ColsAtCompileTime>
(m_matrix, ((Side==OnTheRight) ^ Transposed) ? m_permutation.indices().coeff(i) : i);
}
}
}
protected:
const PermutationType& m_permutation;
typename MatrixType::Nested m_matrix;
};
/* Template partial specialization for transposed/inverse permutations */
template<typename Derived>
struct traits<Transpose<PermutationBase<Derived> > >
: traits<Derived>
{};
} // end namespace internal
template<typename Derived>
class Transpose<PermutationBase<Derived> >
: public EigenBase<Transpose<PermutationBase<Derived> > >
{
typedef Derived PermutationType;
typedef typename PermutationType::IndicesType IndicesType;
typedef typename PermutationType::PlainPermutationType PlainPermutationType;
typedef internal::traits<PermutationType> PermTraits;
protected:
InverseImpl() {}
public:
typedef Inverse<PermutationType> InverseType;
using EigenBase<Inverse<PermutationType> >::derived;
#ifndef EIGEN_PARSED_BY_DOXYGEN
typedef internal::traits<PermutationType> Traits;
typedef typename Derived::DenseMatrixType DenseMatrixType;
typedef typename PermutationType::DenseMatrixType DenseMatrixType;
enum {
Flags = Traits::Flags,
CoeffReadCost = Traits::CoeffReadCost,
RowsAtCompileTime = Traits::RowsAtCompileTime,
ColsAtCompileTime = Traits::ColsAtCompileTime,
MaxRowsAtCompileTime = Traits::MaxRowsAtCompileTime,
MaxColsAtCompileTime = Traits::MaxColsAtCompileTime
RowsAtCompileTime = PermTraits::RowsAtCompileTime,
ColsAtCompileTime = PermTraits::ColsAtCompileTime,
MaxRowsAtCompileTime = PermTraits::MaxRowsAtCompileTime,
MaxColsAtCompileTime = PermTraits::MaxColsAtCompileTime
};
typedef typename Traits::Scalar Scalar;
#endif
Transpose(const PermutationType& p) : m_permutation(p) {}
inline int rows() const { return m_permutation.rows(); }
inline int cols() const { return m_permutation.cols(); }
#ifndef EIGEN_PARSED_BY_DOXYGEN
template<typename DenseDerived>
void evalTo(MatrixBase<DenseDerived>& other) const
{
other.setZero();
for (int i=0; i<rows();++i)
other.coeffRef(i, m_permutation.indices().coeff(i)) = typename DenseDerived::Scalar(1);
for (Index i=0; i<derived().rows();++i)
other.coeffRef(i, derived().nestedExpression().indices().coeff(i)) = typename DenseDerived::Scalar(1);
}
#endif
/** \return the equivalent permutation matrix */
PlainPermutationType eval() const { return *this; }
PlainPermutationType eval() const { return derived(); }
DenseMatrixType toDenseMatrix() const { return *this; }
DenseMatrixType toDenseMatrix() const { return derived(); }
/** \returns the matrix with the inverse permutation applied to the columns.
*/
template<typename OtherDerived> friend
inline const internal::permut_matrix_product_retval<PermutationType, OtherDerived, OnTheRight, true>
operator*(const MatrixBase<OtherDerived>& matrix, const Transpose& trPerm)
const Product<OtherDerived, InverseType, AliasFreeProduct>
operator*(const MatrixBase<OtherDerived>& matrix, const InverseType& trPerm)
{
return internal::permut_matrix_product_retval<PermutationType, OtherDerived, OnTheRight, true>(trPerm.m_permutation, matrix.derived());
return Product<OtherDerived, InverseType, AliasFreeProduct>(matrix.derived(), trPerm.derived());
}
/** \returns the matrix with the inverse permutation applied to the rows.
*/
template<typename OtherDerived>
inline const internal::permut_matrix_product_retval<PermutationType, OtherDerived, OnTheLeft, true>
const Product<InverseType, OtherDerived, AliasFreeProduct>
operator*(const MatrixBase<OtherDerived>& matrix) const
{
return internal::permut_matrix_product_retval<PermutationType, OtherDerived, OnTheLeft, true>(m_permutation, matrix.derived());
return Product<InverseType, OtherDerived, AliasFreeProduct>(derived(), matrix.derived());
}
const PermutationType& nestedPermutation() const { return m_permutation; }
protected:
const PermutationType& m_permutation;
};
template<typename Derived>
@@ -716,6 +622,12 @@ const PermutationWrapper<const Derived> MatrixBase<Derived>::asPermutation() con
return derived();
}
namespace internal {
template<> struct AssignmentKind<DenseShape,PermutationShape> { typedef EigenBase2EigenBase Kind; };
} // end namespace internal
} // end namespace Eigen
#endif // EIGEN_PERMUTATIONMATRIX_H

View File

@@ -28,6 +28,7 @@ namespace internal {
template<int MaxSizeAtCompileTime> struct check_rows_cols_for_overflow {
template<typename Index>
EIGEN_DEVICE_FUNC
static EIGEN_ALWAYS_INLINE void run(Index, Index)
{
}
@@ -35,11 +36,12 @@ template<int MaxSizeAtCompileTime> struct check_rows_cols_for_overflow {
template<> struct check_rows_cols_for_overflow<Dynamic> {
template<typename Index>
EIGEN_DEVICE_FUNC
static EIGEN_ALWAYS_INLINE void run(Index rows, Index cols)
{
// http://hg.mozilla.org/mozilla-central/file/6c8a909977d3/xpcom/ds/CheckedInt.h#l242
// we assume Index is signed
Index max_index = (size_t(1) << (8 * sizeof(Index) - 1)) - 1; // assume Index is signed
Index max_index = (std::size_t(1) << (8 * sizeof(Index) - 1)) - 1; // assume Index is signed
bool error = (rows == 0 || cols == 0) ? false
: (rows > max_index / cols);
if (error)
@@ -56,33 +58,41 @@ template<typename MatrixTypeA, typename MatrixTypeB, bool SwapPointers> struct m
} // end namespace internal
#ifdef EIGEN_PARSED_BY_DOXYGEN
namespace doxygen {
// This is a workaround to doxygen not being able to understand the inheritance logic
// when it is hidden by the dense_xpr_base helper struct.
// Moreover, doxygen fails to include members that are not documented in the declaration body of
// MatrixBase if we inherits MatrixBase<Matrix<_Scalar, _Rows, _Cols, _Options, _MaxRows, _MaxCols> >,
// this is why we simply inherits MatrixBase, though this does not make sense.
/** This class is just a workaround for Doxygen and it does not not actually exist. */
template<typename Derived> struct dense_xpr_base_dispatcher;
/** This class is just a workaround for Doxygen and it does not not actually exist. */
template<typename _Scalar, int _Rows, int _Cols, int _Options, int _MaxRows, int _MaxCols>
struct dense_xpr_base_dispatcher<Matrix<_Scalar, _Rows, _Cols, _Options, _MaxRows, _MaxCols> >
: public MatrixBase {};
/** This class is just a workaround for Doxygen and it does not not actually exist. */
template<typename _Scalar, int _Rows, int _Cols, int _Options, int _MaxRows, int _MaxCols>
struct dense_xpr_base_dispatcher<Array<_Scalar, _Rows, _Cols, _Options, _MaxRows, _MaxCols> >
: public ArrayBase {};
} // namespace doxygen
/** \class PlainObjectBase
* \ingroup Core_Module
* \brief %Dense storage base class for matrices and arrays.
*
* This class can be extended with the help of the plugin mechanism described on the page
* \ref TopicCustomizingEigen by defining the preprocessor symbol \c EIGEN_PLAINOBJECTBASE_PLUGIN.
* \ref TopicCustomizing_Plugins by defining the preprocessor symbol \c EIGEN_PLAINOBJECTBASE_PLUGIN.
*
* \tparam Derived is the derived type, e.g., a Matrix or Array
*
* \sa \ref TopicClassHierarchy
*/
#ifdef EIGEN_PARSED_BY_DOXYGEN
namespace internal {
// this is a warkaround to doxygen not being able to understand the inheritence logic
// when it is hidden by the dense_xpr_base helper struct.
template<typename Derived> struct dense_xpr_base_dispatcher_for_doxygen;// : public MatrixBase<Derived> {};
/** This class is just a workaround for Doxygen and it does not not actually exist. */
template<typename _Scalar, int _Rows, int _Cols, int _Options, int _MaxRows, int _MaxCols>
struct dense_xpr_base_dispatcher_for_doxygen<Matrix<_Scalar, _Rows, _Cols, _Options, _MaxRows, _MaxCols> >
: public MatrixBase<Matrix<_Scalar, _Rows, _Cols, _Options, _MaxRows, _MaxCols> > {};
/** This class is just a workaround for Doxygen and it does not not actually exist. */
template<typename _Scalar, int _Rows, int _Cols, int _Options, int _MaxRows, int _MaxCols>
struct dense_xpr_base_dispatcher_for_doxygen<Array<_Scalar, _Rows, _Cols, _Options, _MaxRows, _MaxCols> >
: public ArrayBase<Array<_Scalar, _Rows, _Cols, _Options, _MaxRows, _MaxCols> > {};
} // namespace internal
template<typename Derived>
class PlainObjectBase : public internal::dense_xpr_base_dispatcher_for_doxygen<Derived>
class PlainObjectBase : public doxygen::dense_xpr_base_dispatcher<Derived>
#else
template<typename Derived>
class PlainObjectBase : public internal::dense_xpr_base<Derived>::type
@@ -93,8 +103,8 @@ class PlainObjectBase : public internal::dense_xpr_base<Derived>::type
typedef typename internal::dense_xpr_base<Derived>::type Base;
typedef typename internal::traits<Derived>::StorageKind StorageKind;
typedef typename internal::traits<Derived>::Index Index;
typedef typename internal::traits<Derived>::Scalar Scalar;
typedef typename internal::packet_traits<Scalar>::type PacketScalar;
typedef typename NumTraits<Scalar>::Real RealScalar;
typedef Derived DenseType;
@@ -113,28 +123,40 @@ class PlainObjectBase : public internal::dense_xpr_base<Derived>::type
typedef Eigen::Map<Derived, Unaligned> MapType;
friend class Eigen::Map<const Derived, Unaligned>;
typedef const Eigen::Map<const Derived, Unaligned> ConstMapType;
friend class Eigen::Map<Derived, Aligned>;
typedef Eigen::Map<Derived, Aligned> AlignedMapType;
friend class Eigen::Map<const Derived, Aligned>;
typedef const Eigen::Map<const Derived, Aligned> ConstAlignedMapType;
#if EIGEN_MAX_ALIGN_BYTES>0
// for EIGEN_MAX_ALIGN_BYTES==0, AlignedMax==Unaligned, and many compilers generate warnings for friend-ing a class twice.
friend class Eigen::Map<Derived, AlignedMax>;
friend class Eigen::Map<const Derived, AlignedMax>;
#endif
typedef Eigen::Map<Derived, AlignedMax> AlignedMapType;
typedef const Eigen::Map<const Derived, AlignedMax> ConstAlignedMapType;
template<typename StrideType> struct StridedMapType { typedef Eigen::Map<Derived, Unaligned, StrideType> type; };
template<typename StrideType> struct StridedConstMapType { typedef Eigen::Map<const Derived, Unaligned, StrideType> type; };
template<typename StrideType> struct StridedAlignedMapType { typedef Eigen::Map<Derived, Aligned, StrideType> type; };
template<typename StrideType> struct StridedConstAlignedMapType { typedef Eigen::Map<const Derived, Aligned, StrideType> type; };
template<typename StrideType> struct StridedAlignedMapType { typedef Eigen::Map<Derived, AlignedMax, StrideType> type; };
template<typename StrideType> struct StridedConstAlignedMapType { typedef Eigen::Map<const Derived, AlignedMax, StrideType> type; };
protected:
DenseStorage<Scalar, Base::MaxSizeAtCompileTime, Base::RowsAtCompileTime, Base::ColsAtCompileTime, Options> m_storage;
public:
enum { NeedsToAlign = SizeAtCompileTime != Dynamic && (internal::traits<Derived>::Flags & AlignedBit) != 0 };
enum { NeedsToAlign = (SizeAtCompileTime != Dynamic) && (internal::traits<Derived>::Alignment>0) };
EIGEN_MAKE_ALIGNED_OPERATOR_NEW_IF(NeedsToAlign)
EIGEN_DEVICE_FUNC
Base& base() { return *static_cast<Base*>(this); }
EIGEN_DEVICE_FUNC
const Base& base() const { return *static_cast<const Base*>(this); }
EIGEN_DEVICE_FUNC
EIGEN_STRONG_INLINE Index rows() const { return m_storage.rows(); }
EIGEN_DEVICE_FUNC
EIGEN_STRONG_INLINE Index cols() const { return m_storage.cols(); }
/** This is an overloaded version of DenseCoeffsBase<Derived,ReadOnlyAccessors>::coeff(Index,Index) const
* provided to by-pass the creation of an evaluator of the expression, thus saving compilation efforts.
*
* See DenseCoeffsBase<Derived,ReadOnlyAccessors>::coeff(Index) const for details. */
EIGEN_DEVICE_FUNC
EIGEN_STRONG_INLINE const Scalar& coeff(Index rowId, Index colId) const
{
if(Flags & RowMajorBit)
@@ -143,11 +165,21 @@ class PlainObjectBase : public internal::dense_xpr_base<Derived>::type
return m_storage.data()[rowId + colId * m_storage.rows()];
}
/** This is an overloaded version of DenseCoeffsBase<Derived,ReadOnlyAccessors>::coeff(Index) const
* provided to by-pass the creation of an evaluator of the expression, thus saving compilation efforts.
*
* See DenseCoeffsBase<Derived,ReadOnlyAccessors>::coeff(Index) const for details. */
EIGEN_DEVICE_FUNC
EIGEN_STRONG_INLINE const Scalar& coeff(Index index) const
{
return m_storage.data()[index];
}
/** This is an overloaded version of DenseCoeffsBase<Derived,WriteAccessors>::coeffRef(Index,Index) const
* provided to by-pass the creation of an evaluator of the expression, thus saving compilation efforts.
*
* See DenseCoeffsBase<Derived,WriteAccessors>::coeffRef(Index,Index) const for details. */
EIGEN_DEVICE_FUNC
EIGEN_STRONG_INLINE Scalar& coeffRef(Index rowId, Index colId)
{
if(Flags & RowMajorBit)
@@ -156,11 +188,19 @@ class PlainObjectBase : public internal::dense_xpr_base<Derived>::type
return m_storage.data()[rowId + colId * m_storage.rows()];
}
/** This is an overloaded version of DenseCoeffsBase<Derived,WriteAccessors>::coeffRef(Index) const
* provided to by-pass the creation of an evaluator of the expression, thus saving compilation efforts.
*
* See DenseCoeffsBase<Derived,WriteAccessors>::coeffRef(Index) const for details. */
EIGEN_DEVICE_FUNC
EIGEN_STRONG_INLINE Scalar& coeffRef(Index index)
{
return m_storage.data()[index];
}
/** This is the const version of coeffRef(Index,Index) which is thus synonym of coeff(Index,Index).
* It is provided for convenience. */
EIGEN_DEVICE_FUNC
EIGEN_STRONG_INLINE const Scalar& coeffRef(Index rowId, Index colId) const
{
if(Flags & RowMajorBit)
@@ -169,6 +209,9 @@ class PlainObjectBase : public internal::dense_xpr_base<Derived>::type
return m_storage.data()[rowId + colId * m_storage.rows()];
}
/** This is the const version of coeffRef(Index) which is thus synonym of coeff(Index).
* It is provided for convenience. */
EIGEN_DEVICE_FUNC
EIGEN_STRONG_INLINE const Scalar& coeffRef(Index index) const
{
return m_storage.data()[index];
@@ -209,11 +252,11 @@ class PlainObjectBase : public internal::dense_xpr_base<Derived>::type
}
/** \returns a const pointer to the data array of this matrix */
EIGEN_STRONG_INLINE const Scalar *data() const
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE const Scalar *data() const
{ return m_storage.data(); }
/** \returns a pointer to the data array of this matrix */
EIGEN_STRONG_INLINE Scalar *data()
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE Scalar *data()
{ return m_storage.data(); }
/** Resizes \c *this to a \a rows x \a cols matrix.
@@ -232,22 +275,22 @@ class PlainObjectBase : public internal::dense_xpr_base<Derived>::type
*
* \sa resize(Index) for vectors, resize(NoChange_t, Index), resize(Index, NoChange_t)
*/
EIGEN_STRONG_INLINE void resize(Index nbRows, Index nbCols)
EIGEN_DEVICE_FUNC
EIGEN_STRONG_INLINE void resize(Index rows, Index cols)
{
eigen_assert( EIGEN_IMPLIES(RowsAtCompileTime!=Dynamic,nbRows==RowsAtCompileTime)
&& EIGEN_IMPLIES(ColsAtCompileTime!=Dynamic,nbCols==ColsAtCompileTime)
&& EIGEN_IMPLIES(RowsAtCompileTime==Dynamic && MaxRowsAtCompileTime!=Dynamic,nbRows<=MaxRowsAtCompileTime)
&& EIGEN_IMPLIES(ColsAtCompileTime==Dynamic && MaxColsAtCompileTime!=Dynamic,nbCols<=MaxColsAtCompileTime)
&& nbRows>=0 && nbCols>=0 && "Invalid sizes when resizing a matrix or array.");
internal::check_rows_cols_for_overflow<MaxSizeAtCompileTime>::run(nbRows, nbCols);
eigen_assert( EIGEN_IMPLIES(RowsAtCompileTime!=Dynamic,rows==RowsAtCompileTime)
&& EIGEN_IMPLIES(ColsAtCompileTime!=Dynamic,cols==ColsAtCompileTime)
&& EIGEN_IMPLIES(RowsAtCompileTime==Dynamic && MaxRowsAtCompileTime!=Dynamic,rows<=MaxRowsAtCompileTime)
&& EIGEN_IMPLIES(ColsAtCompileTime==Dynamic && MaxColsAtCompileTime!=Dynamic,cols<=MaxColsAtCompileTime)
&& rows>=0 && cols>=0 && "Invalid sizes when resizing a matrix or array.");
internal::check_rows_cols_for_overflow<MaxSizeAtCompileTime>::run(rows, cols);
#ifdef EIGEN_INITIALIZE_COEFFS
Index size = nbRows*nbCols;
Index size = rows*cols;
bool size_changed = size != this->size();
m_storage.resize(size, nbRows, nbCols);
m_storage.resize(size, rows, cols);
if(size_changed) EIGEN_INITIALIZE_COEFFS_IF_THAT_OPTION_IS_ENABLED
#else
internal::check_rows_cols_for_overflow<MaxSizeAtCompileTime>::run(nbRows, nbCols);
m_storage.resize(nbRows*nbCols, nbRows, nbCols);
m_storage.resize(rows*cols, rows, cols);
#endif
}
@@ -262,6 +305,7 @@ class PlainObjectBase : public internal::dense_xpr_base<Derived>::type
*
* \sa resize(Index,Index), resize(NoChange_t, Index), resize(Index, NoChange_t)
*/
EIGEN_DEVICE_FUNC
inline void resize(Index size)
{
EIGEN_STATIC_ASSERT_VECTOR_ONLY(PlainObjectBase)
@@ -286,9 +330,10 @@ class PlainObjectBase : public internal::dense_xpr_base<Derived>::type
*
* \sa resize(Index,Index)
*/
inline void resize(NoChange_t, Index nbCols)
EIGEN_DEVICE_FUNC
inline void resize(NoChange_t, Index cols)
{
resize(rows(), nbCols);
resize(rows(), cols);
}
/** Resizes the matrix, changing only the number of rows. For the parameter of type NoChange_t, just pass the special value \c NoChange
@@ -299,9 +344,10 @@ class PlainObjectBase : public internal::dense_xpr_base<Derived>::type
*
* \sa resize(Index,Index)
*/
inline void resize(Index nbRows, NoChange_t)
EIGEN_DEVICE_FUNC
inline void resize(Index rows, NoChange_t)
{
resize(nbRows, cols());
resize(rows, cols());
}
/** Resizes \c *this to have the same dimensions as \a other.
@@ -312,6 +358,7 @@ class PlainObjectBase : public internal::dense_xpr_base<Derived>::type
* remain row-vectors and vectors remain vectors.
*/
template<typename OtherDerived>
EIGEN_DEVICE_FUNC
EIGEN_STRONG_INLINE void resizeLike(const EigenBase<OtherDerived>& _other)
{
const OtherDerived& other = _other.derived();
@@ -339,9 +386,10 @@ class PlainObjectBase : public internal::dense_xpr_base<Derived>::type
* Matrices are resized relative to the top-left element. In case values need to be
* appended to the matrix they will be uninitialized.
*/
EIGEN_STRONG_INLINE void conservativeResize(Index nbRows, Index nbCols)
EIGEN_DEVICE_FUNC
EIGEN_STRONG_INLINE void conservativeResize(Index rows, Index cols)
{
internal::conservative_resize_like_impl<Derived>::run(*this, nbRows, nbCols);
internal::conservative_resize_like_impl<Derived>::run(*this, rows, cols);
}
/** Resizes the matrix to \a rows x \a cols while leaving old values untouched.
@@ -351,10 +399,11 @@ class PlainObjectBase : public internal::dense_xpr_base<Derived>::type
*
* In case the matrix is growing, new rows will be uninitialized.
*/
EIGEN_STRONG_INLINE void conservativeResize(Index nbRows, NoChange_t)
EIGEN_DEVICE_FUNC
EIGEN_STRONG_INLINE void conservativeResize(Index rows, NoChange_t)
{
// Note: see the comment in conservativeResize(Index,Index)
conservativeResize(nbRows, cols());
conservativeResize(rows, cols());
}
/** Resizes the matrix to \a rows x \a cols while leaving old values untouched.
@@ -364,10 +413,11 @@ class PlainObjectBase : public internal::dense_xpr_base<Derived>::type
*
* In case the matrix is growing, new columns will be uninitialized.
*/
EIGEN_STRONG_INLINE void conservativeResize(NoChange_t, Index nbCols)
EIGEN_DEVICE_FUNC
EIGEN_STRONG_INLINE void conservativeResize(NoChange_t, Index cols)
{
// Note: see the comment in conservativeResize(Index,Index)
conservativeResize(rows(), nbCols);
conservativeResize(rows(), cols);
}
/** Resizes the vector to \a size while retaining old values.
@@ -378,6 +428,7 @@ class PlainObjectBase : public internal::dense_xpr_base<Derived>::type
*
* When values are appended, they will be uninitialized.
*/
EIGEN_DEVICE_FUNC
EIGEN_STRONG_INLINE void conservativeResize(Index size)
{
internal::conservative_resize_like_impl<Derived>::run(*this, size);
@@ -393,6 +444,7 @@ class PlainObjectBase : public internal::dense_xpr_base<Derived>::type
* appended to the matrix they will copied from \c other.
*/
template<typename OtherDerived>
EIGEN_DEVICE_FUNC
EIGEN_STRONG_INLINE void conservativeResizeLike(const DenseBase<OtherDerived>& other)
{
internal::conservative_resize_like_impl<Derived,OtherDerived>::run(*this, other);
@@ -401,6 +453,7 @@ class PlainObjectBase : public internal::dense_xpr_base<Derived>::type
/** This is a special case of the templated operator=. Its purpose is to
* prevent a default operator= from hiding the templated operator=.
*/
EIGEN_DEVICE_FUNC
EIGEN_STRONG_INLINE Derived& operator=(const PlainObjectBase& other)
{
return _set(other);
@@ -408,6 +461,7 @@ class PlainObjectBase : public internal::dense_xpr_base<Derived>::type
/** \sa MatrixBase::lazyAssign() */
template<typename OtherDerived>
EIGEN_DEVICE_FUNC
EIGEN_STRONG_INLINE Derived& lazyAssign(const DenseBase<OtherDerived>& other)
{
_resize_to_match(other);
@@ -415,12 +469,18 @@ class PlainObjectBase : public internal::dense_xpr_base<Derived>::type
}
template<typename OtherDerived>
EIGEN_DEVICE_FUNC
EIGEN_STRONG_INLINE Derived& operator=(const ReturnByValue<OtherDerived>& func)
{
resize(func.rows(), func.cols());
return Base::operator=(func);
}
// Prevent user from trying to instantiate PlainObjectBase objects
// by making all its constructor protected. See bug 1074.
protected:
EIGEN_DEVICE_FUNC
EIGEN_STRONG_INLINE PlainObjectBase() : m_storage()
{
// _check_template_params();
@@ -430,20 +490,23 @@ class PlainObjectBase : public internal::dense_xpr_base<Derived>::type
#ifndef EIGEN_PARSED_BY_DOXYGEN
// FIXME is it still needed ?
/** \internal */
PlainObjectBase(internal::constructor_without_unaligned_array_assert)
EIGEN_DEVICE_FUNC
explicit PlainObjectBase(internal::constructor_without_unaligned_array_assert)
: m_storage(internal::constructor_without_unaligned_array_assert())
{
// _check_template_params(); EIGEN_INITIALIZE_COEFFS_IF_THAT_OPTION_IS_ENABLED
}
#endif
#ifdef EIGEN_HAVE_RVALUE_REFERENCES
PlainObjectBase(PlainObjectBase&& other)
#if EIGEN_HAS_RVALUE_REFERENCES
EIGEN_DEVICE_FUNC
PlainObjectBase(PlainObjectBase&& other) EIGEN_NOEXCEPT
: m_storage( std::move(other.m_storage) )
{
}
PlainObjectBase& operator=(PlainObjectBase&& other)
EIGEN_DEVICE_FUNC
PlainObjectBase& operator=(PlainObjectBase&& other) EIGEN_NOEXCEPT
{
using std::swap;
swap(m_storage, other.m_storage);
@@ -452,31 +515,56 @@ class PlainObjectBase : public internal::dense_xpr_base<Derived>::type
#endif
/** Copy constructor */
EIGEN_DEVICE_FUNC
EIGEN_STRONG_INLINE PlainObjectBase(const PlainObjectBase& other)
: m_storage()
{
_check_template_params();
lazyAssign(other);
}
template<typename OtherDerived>
EIGEN_STRONG_INLINE PlainObjectBase(const DenseBase<OtherDerived> &other)
: m_storage()
{
_check_template_params();
lazyAssign(other);
}
EIGEN_STRONG_INLINE PlainObjectBase(Index a_size, Index nbRows, Index nbCols)
: m_storage(a_size, nbRows, nbCols)
: Base(), m_storage(other.m_storage) { }
EIGEN_DEVICE_FUNC
EIGEN_STRONG_INLINE PlainObjectBase(Index size, Index rows, Index cols)
: m_storage(size, rows, cols)
{
// _check_template_params();
// EIGEN_INITIALIZE_COEFFS_IF_THAT_OPTION_IS_ENABLED
}
/** \copydoc MatrixBase::operator=(const EigenBase<OtherDerived>&)
/** \sa PlainObjectBase::operator=(const EigenBase<OtherDerived>&) */
template<typename OtherDerived>
EIGEN_DEVICE_FUNC
EIGEN_STRONG_INLINE PlainObjectBase(const DenseBase<OtherDerived> &other)
: m_storage()
{
_check_template_params();
resizeLike(other);
_set_noalias(other);
}
/** \sa PlainObjectBase::operator=(const EigenBase<OtherDerived>&) */
template<typename OtherDerived>
EIGEN_DEVICE_FUNC
EIGEN_STRONG_INLINE PlainObjectBase(const EigenBase<OtherDerived> &other)
: m_storage()
{
_check_template_params();
resizeLike(other);
*this = other.derived();
}
/** \brief Copy constructor with in-place evaluation */
template<typename OtherDerived>
EIGEN_DEVICE_FUNC
EIGEN_STRONG_INLINE PlainObjectBase(const ReturnByValue<OtherDerived>& other)
{
_check_template_params();
// FIXME this does not automatically transpose vectors if necessary
resize(other.rows(), other.cols());
other.evalTo(this->derived());
}
public:
/** \brief Copies the generic expression \a other into *this.
* \copydetails DenseBase::operator=(const EigenBase<OtherDerived> &other)
*/
template<typename OtherDerived>
EIGEN_DEVICE_FUNC
EIGEN_STRONG_INLINE Derived& operator=(const EigenBase<OtherDerived> &other)
{
_resize_to_match(other);
@@ -484,21 +572,15 @@ class PlainObjectBase : public internal::dense_xpr_base<Derived>::type
return this->derived();
}
/** \sa MatrixBase::operator=(const EigenBase<OtherDerived>&) */
template<typename OtherDerived>
EIGEN_STRONG_INLINE PlainObjectBase(const EigenBase<OtherDerived> &other)
: m_storage(other.derived().rows() * other.derived().cols(), other.derived().rows(), other.derived().cols())
{
_check_template_params();
internal::check_rows_cols_for_overflow<MaxSizeAtCompileTime>::run(other.derived().rows(), other.derived().cols());
Base::operator=(other.derived());
}
/** \name Map
* These are convenience functions returning Map objects. The Map() static functions return unaligned Map objects,
* while the AlignedMap() functions return aligned Map objects and thus should be called only with 16-byte-aligned
* \a data pointers.
*
* Here is an example using strides:
* \include Matrix_Map_stride.cpp
* Output: \verbinclude Matrix_Map_stride.out
*
* \see class Map
*/
//@{
@@ -568,16 +650,16 @@ class PlainObjectBase : public internal::dense_xpr_base<Derived>::type
//@}
using Base::setConstant;
Derived& setConstant(Index size, const Scalar& value);
Derived& setConstant(Index rows, Index cols, const Scalar& value);
EIGEN_DEVICE_FUNC Derived& setConstant(Index size, const Scalar& val);
EIGEN_DEVICE_FUNC Derived& setConstant(Index rows, Index cols, const Scalar& val);
using Base::setZero;
Derived& setZero(Index size);
Derived& setZero(Index rows, Index cols);
EIGEN_DEVICE_FUNC Derived& setZero(Index size);
EIGEN_DEVICE_FUNC Derived& setZero(Index rows, Index cols);
using Base::setOnes;
Derived& setOnes(Index size);
Derived& setOnes(Index rows, Index cols);
EIGEN_DEVICE_FUNC Derived& setOnes(Index size);
EIGEN_DEVICE_FUNC Derived& setOnes(Index rows, Index cols);
using Base::setRandom;
Derived& setRandom(Index size);
@@ -596,6 +678,7 @@ class PlainObjectBase : public internal::dense_xpr_base<Derived>::type
* remain row-vectors and vectors remain vectors.
*/
template<typename OtherDerived>
EIGEN_DEVICE_FUNC
EIGEN_STRONG_INLINE void _resize_to_match(const EigenBase<OtherDerived>& other)
{
#ifdef EIGEN_NO_AUTOMATIC_RESIZING
@@ -603,8 +686,6 @@ class PlainObjectBase : public internal::dense_xpr_base<Derived>::type
: (rows() == other.rows() && cols() == other.cols())))
&& "Size mismatch. Automatic resizing is disabled because EIGEN_NO_AUTOMATIC_RESIZING is defined");
EIGEN_ONLY_USED_FOR_DEBUG(other);
if(this->size()==0)
resizeLike(other);
#else
resizeLike(other);
#endif
@@ -624,25 +705,23 @@ class PlainObjectBase : public internal::dense_xpr_base<Derived>::type
*
* \internal
*/
// aliasing is dealt once in internall::call_assignment
// so at this stage we have to assume aliasing... and resising has to be done later.
template<typename OtherDerived>
EIGEN_DEVICE_FUNC
EIGEN_STRONG_INLINE Derived& _set(const DenseBase<OtherDerived>& other)
{
_set_selector(other.derived(), typename internal::conditional<static_cast<bool>(int(OtherDerived::Flags) & EvalBeforeAssigningBit), internal::true_type, internal::false_type>::type());
internal::call_assignment(this->derived(), other.derived());
return this->derived();
}
template<typename OtherDerived>
EIGEN_STRONG_INLINE void _set_selector(const OtherDerived& other, const internal::true_type&) { _set_noalias(other.eval()); }
template<typename OtherDerived>
EIGEN_STRONG_INLINE void _set_selector(const OtherDerived& other, const internal::false_type&) { _set_noalias(other); }
/** \internal Like _set() but additionally makes the assumption that no aliasing effect can happen (which
* is the case when creating a new matrix) so one can enforce lazy evaluation.
*
* \sa operator=(const MatrixBase<OtherDerived>&), _set()
*/
template<typename OtherDerived>
EIGEN_DEVICE_FUNC
EIGEN_STRONG_INLINE Derived& _set_noalias(const DenseBase<OtherDerived>& other)
{
// I don't think we need this resize call since the lazyAssign will anyways resize
@@ -650,40 +729,175 @@ class PlainObjectBase : public internal::dense_xpr_base<Derived>::type
//_resize_to_match(other);
// the 'false' below means to enforce lazy evaluation. We don't use lazyAssign() because
// it wouldn't allow to copy a row-vector into a column-vector.
return internal::assign_selector<Derived,OtherDerived,false>::run(this->derived(), other.derived());
internal::call_assignment_no_alias(this->derived(), other.derived(), internal::assign_op<Scalar,typename OtherDerived::Scalar>());
return this->derived();
}
template<typename T0, typename T1>
EIGEN_STRONG_INLINE void _init2(Index nbRows, Index nbCols, typename internal::enable_if<Base::SizeAtCompileTime!=2,T0>::type* = 0)
EIGEN_DEVICE_FUNC
EIGEN_STRONG_INLINE void _init2(Index rows, Index cols, typename internal::enable_if<Base::SizeAtCompileTime!=2,T0>::type* = 0)
{
EIGEN_STATIC_ASSERT(bool(NumTraits<T0>::IsInteger) &&
bool(NumTraits<T1>::IsInteger),
FLOATING_POINT_ARGUMENT_PASSED__INTEGER_WAS_EXPECTED)
resize(nbRows,nbCols);
resize(rows,cols);
}
template<typename T0, typename T1>
EIGEN_STRONG_INLINE void _init2(const Scalar& val0, const Scalar& val1, typename internal::enable_if<Base::SizeAtCompileTime==2,T0>::type* = 0)
EIGEN_DEVICE_FUNC
EIGEN_STRONG_INLINE void _init2(const T0& val0, const T1& val1, typename internal::enable_if<Base::SizeAtCompileTime==2,T0>::type* = 0)
{
EIGEN_STATIC_ASSERT_VECTOR_SPECIFIC_SIZE(PlainObjectBase, 2)
m_storage.data()[0] = val0;
m_storage.data()[1] = val1;
m_storage.data()[0] = Scalar(val0);
m_storage.data()[1] = Scalar(val1);
}
template<typename T0, typename T1>
EIGEN_DEVICE_FUNC
EIGEN_STRONG_INLINE void _init2(const Index& val0, const Index& val1,
typename internal::enable_if< (!internal::is_same<Index,Scalar>::value)
&& (internal::is_same<T0,Index>::value)
&& (internal::is_same<T1,Index>::value)
&& Base::SizeAtCompileTime==2,T1>::type* = 0)
{
EIGEN_STATIC_ASSERT_VECTOR_SPECIFIC_SIZE(PlainObjectBase, 2)
m_storage.data()[0] = Scalar(val0);
m_storage.data()[1] = Scalar(val1);
}
// The argument is convertible to the Index type and we either have a non 1x1 Matrix, or a dynamic-sized Array,
// then the argument is meant to be the size of the object.
template<typename T>
EIGEN_DEVICE_FUNC
EIGEN_STRONG_INLINE void _init1(Index size, typename internal::enable_if< (Base::SizeAtCompileTime!=1 || !internal::is_convertible<T, Scalar>::value)
&& ((!internal::is_same<typename internal::traits<Derived>::XprKind,ArrayXpr>::value || Base::SizeAtCompileTime==Dynamic)),T>::type* = 0)
{
// NOTE MSVC 2008 complains if we directly put bool(NumTraits<T>::IsInteger) as the EIGEN_STATIC_ASSERT argument.
const bool is_integer = NumTraits<T>::IsInteger;
EIGEN_UNUSED_VARIABLE(is_integer);
EIGEN_STATIC_ASSERT(is_integer,
FLOATING_POINT_ARGUMENT_PASSED__INTEGER_WAS_EXPECTED)
resize(size);
}
// We have a 1x1 matrix/array => the argument is interpreted as the value of the unique coefficient (case where scalar type can be implicitely converted)
template<typename T>
EIGEN_DEVICE_FUNC
EIGEN_STRONG_INLINE void _init1(const Scalar& val0, typename internal::enable_if<Base::SizeAtCompileTime==1 && internal::is_convertible<T, Scalar>::value,T>::type* = 0)
{
EIGEN_STATIC_ASSERT_VECTOR_SPECIFIC_SIZE(PlainObjectBase, 1)
m_storage.data()[0] = val0;
}
// We have a 1x1 matrix/array => the argument is interpreted as the value of the unique coefficient (case where scalar type match the index type)
template<typename T>
EIGEN_DEVICE_FUNC
EIGEN_STRONG_INLINE void _init1(const Index& val0,
typename internal::enable_if< (!internal::is_same<Index,Scalar>::value)
&& (internal::is_same<Index,T>::value)
&& Base::SizeAtCompileTime==1
&& internal::is_convertible<T, Scalar>::value,T*>::type* = 0)
{
EIGEN_STATIC_ASSERT_VECTOR_SPECIFIC_SIZE(PlainObjectBase, 1)
m_storage.data()[0] = Scalar(val0);
}
// Initialize a fixed size matrix from a pointer to raw data
template<typename T>
EIGEN_DEVICE_FUNC
EIGEN_STRONG_INLINE void _init1(const Scalar* data){
this->_set_noalias(ConstMapType(data));
}
// Initialize an arbitrary matrix from a dense expression
template<typename T, typename OtherDerived>
EIGEN_DEVICE_FUNC
EIGEN_STRONG_INLINE void _init1(const DenseBase<OtherDerived>& other){
this->_set_noalias(other);
}
// Initialize an arbitrary matrix from an object convertible to the Derived type.
template<typename T>
EIGEN_DEVICE_FUNC
EIGEN_STRONG_INLINE void _init1(const Derived& other){
this->_set_noalias(other);
}
// Initialize an arbitrary matrix from a generic Eigen expression
template<typename T, typename OtherDerived>
EIGEN_DEVICE_FUNC
EIGEN_STRONG_INLINE void _init1(const EigenBase<OtherDerived>& other){
this->derived() = other;
}
template<typename T, typename OtherDerived>
EIGEN_DEVICE_FUNC
EIGEN_STRONG_INLINE void _init1(const ReturnByValue<OtherDerived>& other)
{
resize(other.rows(), other.cols());
other.evalTo(this->derived());
}
template<typename T, typename OtherDerived, int ColsAtCompileTime>
EIGEN_DEVICE_FUNC
EIGEN_STRONG_INLINE void _init1(const RotationBase<OtherDerived,ColsAtCompileTime>& r)
{
this->derived() = r;
}
// For fixed-size Array<Scalar,...>
template<typename T>
EIGEN_DEVICE_FUNC
EIGEN_STRONG_INLINE void _init1(const Scalar& val0,
typename internal::enable_if< Base::SizeAtCompileTime!=Dynamic
&& Base::SizeAtCompileTime!=1
&& internal::is_convertible<T, Scalar>::value
&& internal::is_same<typename internal::traits<Derived>::XprKind,ArrayXpr>::value,T>::type* = 0)
{
Base::setConstant(val0);
}
// For fixed-size Array<Index,...>
template<typename T>
EIGEN_DEVICE_FUNC
EIGEN_STRONG_INLINE void _init1(const Index& val0,
typename internal::enable_if< (!internal::is_same<Index,Scalar>::value)
&& (internal::is_same<Index,T>::value)
&& Base::SizeAtCompileTime!=Dynamic
&& Base::SizeAtCompileTime!=1
&& internal::is_convertible<T, Scalar>::value
&& internal::is_same<typename internal::traits<Derived>::XprKind,ArrayXpr>::value,T*>::type* = 0)
{
Base::setConstant(val0);
}
template<typename MatrixTypeA, typename MatrixTypeB, bool SwapPointers>
friend struct internal::matrix_swap_impl;
/** \internal generic implementation of swap for dense storage since for dynamic-sized matrices of same type it is enough to swap the
* data pointers.
public:
#ifndef EIGEN_PARSED_BY_DOXYGEN
/** \internal
* \brief Override DenseBase::swap() since for dynamic-sized matrices
* of same type it is enough to swap the data pointers.
*/
template<typename OtherDerived>
void _swap(DenseBase<OtherDerived> const & other)
EIGEN_DEVICE_FUNC
void swap(DenseBase<OtherDerived> & other)
{
enum { SwapPointers = internal::is_same<Derived, OtherDerived>::value && Base::SizeAtCompileTime==Dynamic };
internal::matrix_swap_impl<Derived, OtherDerived, bool(SwapPointers)>::run(this->derived(), other.const_cast_derived());
internal::matrix_swap_impl<Derived, OtherDerived, bool(SwapPointers)>::run(this->derived(), other.derived());
}
public:
#ifndef EIGEN_PARSED_BY_DOXYGEN
/** \internal
* \brief const version forwarded to DenseBase::swap
*/
template<typename OtherDerived>
EIGEN_DEVICE_FUNC
void swap(DenseBase<OtherDerived> const & other)
{ Base::swap(other.derived()); }
EIGEN_DEVICE_FUNC
static EIGEN_STRONG_INLINE void _check_template_params()
{
EIGEN_STATIC_ASSERT((EIGEN_IMPLIES(MaxRowsAtCompileTime==1 && MaxColsAtCompileTime!=1, (Options&RowMajor)==RowMajor)
@@ -697,10 +911,9 @@ class PlainObjectBase : public internal::dense_xpr_base<Derived>::type
&& (Options & (DontAlign|RowMajor)) == Options),
INVALID_MATRIX_TEMPLATE_PARAMETERS)
}
#endif
private:
enum { ThisConstantIsPrivateInPlainObjectBase };
enum { IsPlainObjectBase = 1 };
#endif
};
namespace internal {
@@ -708,7 +921,6 @@ namespace internal {
template <typename Derived, typename OtherDerived, bool IsVector>
struct conservative_resize_like_impl
{
typedef typename Derived::Index Index;
static void run(DenseBase<Derived>& _this, Index rows, Index cols)
{
if (_this.rows() == rows && _this.cols() == cols) return;
@@ -724,8 +936,8 @@ struct conservative_resize_like_impl
{
// The storage order does not allow us to use reallocation.
typename Derived::PlainObject tmp(rows,cols);
const Index common_rows = (std::min)(rows, _this.rows());
const Index common_cols = (std::min)(cols, _this.cols());
const Index common_rows = numext::mini(rows, _this.rows());
const Index common_cols = numext::mini(cols, _this.cols());
tmp.block(0,0,common_rows,common_cols) = _this.block(0,0,common_rows,common_cols);
_this.derived().swap(tmp);
}
@@ -758,8 +970,8 @@ struct conservative_resize_like_impl
{
// The storage order does not allow us to use reallocation.
typename Derived::PlainObject tmp(other);
const Index common_rows = (std::min)(tmp.rows(), _this.rows());
const Index common_cols = (std::min)(tmp.cols(), _this.cols());
const Index common_rows = numext::mini(tmp.rows(), _this.rows());
const Index common_cols = numext::mini(tmp.cols(), _this.cols());
tmp.block(0,0,common_rows,common_cols) = _this.block(0,0,common_rows,common_cols);
_this.derived().swap(tmp);
}
@@ -774,7 +986,6 @@ struct conservative_resize_like_impl<Derived,OtherDerived,true>
{
using conservative_resize_like_impl<Derived,OtherDerived,false>::run;
typedef typename Derived::Index Index;
static void run(DenseBase<Derived>& _this, Index size)
{
const Index new_rows = Derived::RowsAtCompileTime==1 ? 1 : size;
@@ -800,6 +1011,7 @@ struct conservative_resize_like_impl<Derived,OtherDerived,true>
template<typename MatrixTypeA, typename MatrixTypeB, bool SwapPointers>
struct matrix_swap_impl
{
EIGEN_DEVICE_FUNC
static inline void run(MatrixTypeA& a, MatrixTypeB& b)
{
a.base().swap(b);
@@ -809,6 +1021,7 @@ struct matrix_swap_impl
template<typename MatrixTypeA, typename MatrixTypeB>
struct matrix_swap_impl<MatrixTypeA, MatrixTypeB, true>
{
EIGEN_DEVICE_FUNC
static inline void run(MatrixTypeA& a, MatrixTypeB& b)
{
static_cast<typename MatrixTypeA::Base&>(a).m_storage.swap(static_cast<typename MatrixTypeB::Base&>(b).m_storage);

View File

@@ -3,96 +3,184 @@
//
// Copyright (C) 2008-2011 Gael Guennebaud <gael.guennebaud@inria.fr>
//
// This Source Code Form is subject to the terms of the Mozilla Public
// License, v. 2.0. If a copy of the MPL was not distributed with this
// file, You can obtain one at http://mozilla.org/MPL/2.0/.
// This Source Code Form is subject to the terms of the Mozilla
// Public License v. 2.0. If a copy of the MPL was not distributed
// with this file, You can obtain one at http://mozilla.org/MPL/2.0/.
#ifndef EIGEN_PRODUCT_H
#define EIGEN_PRODUCT_H
template<typename Lhs, typename Rhs> class Product;
template<typename Lhs, typename Rhs, typename StorageKind> class ProductImpl;
namespace Eigen {
template<typename Lhs, typename Rhs, int Option, typename StorageKind> class ProductImpl;
namespace internal {
template<typename Lhs, typename Rhs, int Option>
struct traits<Product<Lhs, Rhs, Option> >
{
typedef typename remove_all<Lhs>::type LhsCleaned;
typedef typename remove_all<Rhs>::type RhsCleaned;
typedef traits<LhsCleaned> LhsTraits;
typedef traits<RhsCleaned> RhsTraits;
typedef MatrixXpr XprKind;
typedef typename ScalarBinaryOpTraits<typename traits<LhsCleaned>::Scalar, typename traits<RhsCleaned>::Scalar>::ReturnType Scalar;
typedef typename product_promote_storage_type<typename LhsTraits::StorageKind,
typename RhsTraits::StorageKind,
internal::product_type<Lhs,Rhs>::ret>::ret StorageKind;
typedef typename promote_index_type<typename LhsTraits::StorageIndex,
typename RhsTraits::StorageIndex>::type StorageIndex;
enum {
RowsAtCompileTime = LhsTraits::RowsAtCompileTime,
ColsAtCompileTime = RhsTraits::ColsAtCompileTime,
MaxRowsAtCompileTime = LhsTraits::MaxRowsAtCompileTime,
MaxColsAtCompileTime = RhsTraits::MaxColsAtCompileTime,
// FIXME: only needed by GeneralMatrixMatrixTriangular
InnerSize = EIGEN_SIZE_MIN_PREFER_FIXED(LhsTraits::ColsAtCompileTime, RhsTraits::RowsAtCompileTime),
// The storage order is somewhat arbitrary here. The correct one will be determined through the evaluator.
Flags = (MaxRowsAtCompileTime==1 && MaxColsAtCompileTime!=1) ? RowMajorBit
: (MaxColsAtCompileTime==1 && MaxRowsAtCompileTime!=1) ? 0
: ( ((LhsTraits::Flags&NoPreferredStorageOrderBit) && (RhsTraits::Flags&RowMajorBit))
|| ((RhsTraits::Flags&NoPreferredStorageOrderBit) && (LhsTraits::Flags&RowMajorBit)) ) ? RowMajorBit
: NoPreferredStorageOrderBit
};
};
} // end namespace internal
/** \class Product
* \ingroup Core_Module
*
* \brief Expression of the product of two arbitrary matrices or vectors
*
* \param Lhs the type of the left-hand side expression
* \param Rhs the type of the right-hand side expression
* \tparam _Lhs the type of the left-hand side expression
* \tparam _Rhs the type of the right-hand side expression
*
* This class represents an expression of the product of two arbitrary matrices.
*
* The other template parameters are:
* \tparam Option can be DefaultProduct, AliasFreeProduct, or LazyProduct
*
*/
namespace internal {
template<typename Lhs, typename Rhs>
struct traits<Product<Lhs, Rhs> >
{
typedef MatrixXpr XprKind;
typedef typename remove_all<Lhs>::type LhsCleaned;
typedef typename remove_all<Rhs>::type RhsCleaned;
typedef typename scalar_product_traits<typename traits<LhsCleaned>::Scalar, typename traits<RhsCleaned>::Scalar>::ReturnType Scalar;
typedef typename promote_storage_type<typename traits<LhsCleaned>::StorageKind,
typename traits<RhsCleaned>::StorageKind>::ret StorageKind;
typedef typename promote_index_type<typename traits<LhsCleaned>::Index,
typename traits<RhsCleaned>::Index>::type Index;
enum {
RowsAtCompileTime = LhsCleaned::RowsAtCompileTime,
ColsAtCompileTime = RhsCleaned::ColsAtCompileTime,
MaxRowsAtCompileTime = LhsCleaned::MaxRowsAtCompileTime,
MaxColsAtCompileTime = RhsCleaned::MaxColsAtCompileTime,
Flags = (MaxRowsAtCompileTime==1 ? RowMajorBit : 0), // TODO should be no storage order
CoeffReadCost = 0 // TODO CoeffReadCost should not be part of the expression traits
};
};
} // end namespace internal
template<typename Lhs, typename Rhs>
class Product : public ProductImpl<Lhs,Rhs,typename internal::promote_storage_type<typename internal::traits<Lhs>::StorageKind,
typename internal::traits<Rhs>::StorageKind>::ret>
template<typename _Lhs, typename _Rhs, int Option>
class Product : public ProductImpl<_Lhs,_Rhs,Option,
typename internal::product_promote_storage_type<typename internal::traits<_Lhs>::StorageKind,
typename internal::traits<_Rhs>::StorageKind,
internal::product_type<_Lhs,_Rhs>::ret>::ret>
{
public:
typedef _Lhs Lhs;
typedef _Rhs Rhs;
typedef typename ProductImpl<
Lhs, Rhs,
typename internal::promote_storage_type<typename Lhs::StorageKind,
typename Rhs::StorageKind>::ret>::Base Base;
Lhs, Rhs, Option,
typename internal::product_promote_storage_type<typename internal::traits<Lhs>::StorageKind,
typename internal::traits<Rhs>::StorageKind,
internal::product_type<Lhs,Rhs>::ret>::ret>::Base Base;
EIGEN_GENERIC_PUBLIC_INTERFACE(Product)
typedef typename Lhs::Nested LhsNested;
typedef typename Rhs::Nested RhsNested;
typedef typename internal::ref_selector<Lhs>::type LhsNested;
typedef typename internal::ref_selector<Rhs>::type RhsNested;
typedef typename internal::remove_all<LhsNested>::type LhsNestedCleaned;
typedef typename internal::remove_all<RhsNested>::type RhsNestedCleaned;
Product(const Lhs& lhs, const Rhs& rhs) : m_lhs(lhs), m_rhs(rhs)
EIGEN_DEVICE_FUNC Product(const Lhs& lhs, const Rhs& rhs) : m_lhs(lhs), m_rhs(rhs)
{
eigen_assert(lhs.cols() == rhs.rows()
&& "invalid matrix product"
&& "if you wanted a coeff-wise or a dot product use the respective explicit functions");
}
inline Index rows() const { return m_lhs.rows(); }
inline Index cols() const { return m_rhs.cols(); }
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE Index rows() const { return m_lhs.rows(); }
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE Index cols() const { return m_rhs.cols(); }
const LhsNestedCleaned& lhs() const { return m_lhs; }
const RhsNestedCleaned& rhs() const { return m_rhs; }
EIGEN_DEVICE_FUNC const LhsNestedCleaned& lhs() const { return m_lhs; }
EIGEN_DEVICE_FUNC const RhsNestedCleaned& rhs() const { return m_rhs; }
protected:
const LhsNested m_lhs;
const RhsNested m_rhs;
LhsNested m_lhs;
RhsNested m_rhs;
};
template<typename Lhs, typename Rhs>
class ProductImpl<Lhs,Rhs,Dense> : public internal::dense_xpr_base<Product<Lhs,Rhs> >::type
namespace internal {
template<typename Lhs, typename Rhs, int Option, int ProductTag = internal::product_type<Lhs,Rhs>::ret>
class dense_product_base
: public internal::dense_xpr_base<Product<Lhs,Rhs,Option> >::type
{};
/** Convertion to scalar for inner-products */
template<typename Lhs, typename Rhs, int Option>
class dense_product_base<Lhs, Rhs, Option, InnerProduct>
: public internal::dense_xpr_base<Product<Lhs,Rhs,Option> >::type
{
typedef Product<Lhs, Rhs> Derived;
public:
typedef typename internal::dense_xpr_base<Product<Lhs, Rhs> >::type Base;
EIGEN_DENSE_PUBLIC_INTERFACE(Derived)
typedef Product<Lhs,Rhs,Option> ProductXpr;
typedef typename internal::dense_xpr_base<ProductXpr>::type Base;
public:
using Base::derived;
typedef typename Base::Scalar Scalar;
EIGEN_STRONG_INLINE operator const Scalar() const
{
return internal::evaluator<ProductXpr>(derived()).coeff(0,0);
}
};
} // namespace internal
// Generic API dispatcher
template<typename Lhs, typename Rhs, int Option, typename StorageKind>
class ProductImpl : public internal::generic_xpr_base<Product<Lhs,Rhs,Option>, MatrixXpr, StorageKind>::type
{
public:
typedef typename internal::generic_xpr_base<Product<Lhs,Rhs,Option>, MatrixXpr, StorageKind>::type Base;
};
template<typename Lhs, typename Rhs, int Option>
class ProductImpl<Lhs,Rhs,Option,Dense>
: public internal::dense_product_base<Lhs,Rhs,Option>
{
typedef Product<Lhs, Rhs, Option> Derived;
public:
typedef typename internal::dense_product_base<Lhs, Rhs, Option> Base;
EIGEN_DENSE_PUBLIC_INTERFACE(Derived)
protected:
enum {
IsOneByOne = (RowsAtCompileTime == 1 || RowsAtCompileTime == Dynamic) &&
(ColsAtCompileTime == 1 || ColsAtCompileTime == Dynamic),
EnableCoeff = IsOneByOne || Option==LazyProduct
};
public:
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE Scalar coeff(Index row, Index col) const
{
EIGEN_STATIC_ASSERT(EnableCoeff, THIS_METHOD_IS_ONLY_FOR_INNER_OR_LAZY_PRODUCTS);
eigen_assert( (Option==LazyProduct) || (this->rows() == 1 && this->cols() == 1) );
return internal::evaluator<Derived>(derived()).coeff(row,col);
}
EIGEN_DEVICE_FUNC EIGEN_STRONG_INLINE Scalar coeff(Index i) const
{
EIGEN_STATIC_ASSERT(EnableCoeff, THIS_METHOD_IS_ONLY_FOR_INNER_OR_LAZY_PRODUCTS);
eigen_assert( (Option==LazyProduct) || (this->rows() == 1 && this->cols() == 1) );
return internal::evaluator<Derived>(derived()).coeff(i);
}
};
} // end namespace Eigen
#endif // EIGEN_PRODUCT_H

File diff suppressed because it is too large Load Diff

View File

@@ -16,8 +16,7 @@ namespace internal {
template<typename Scalar> struct scalar_random_op {
EIGEN_EMPTY_STRUCT_CTOR(scalar_random_op)
template<typename Index>
inline const Scalar operator() (Index, Index = 0) const { return random<Scalar>(); }
inline const Scalar operator() () const { return random<Scalar>(); }
};
template<typename Scalar>
@@ -28,12 +27,18 @@ struct functor_traits<scalar_random_op<Scalar> >
/** \returns a random matrix expression
*
* Numbers are uniformly spread through their whole definition range for integer types,
* and in the [-1:1] range for floating point scalar types.
*
* The parameters \a rows and \a cols are the number of rows and of columns of
* the returned matrix. Must be compatible with this MatrixBase type.
*
* \not_reentrant
*
* This variant is meant to be used for dynamic-size matrix types. For fixed-size types,
* it is redundant to pass \a rows and \a cols as arguments, so Random() should be used
* instead.
*
*
* Example: \include MatrixBase_random_int_int.cpp
* Output: \verbinclude MatrixBase_random_int_int.out
@@ -41,22 +46,28 @@ struct functor_traits<scalar_random_op<Scalar> >
* This expression has the "evaluate before nesting" flag so that it will be evaluated into
* a temporary matrix whenever it is nested in a larger expression. This prevents unexpected
* behavior with expressions involving random matrices.
*
* See DenseBase::NullaryExpr(Index, const CustomNullaryOp&) for an example using C++11 random generators.
*
* \sa MatrixBase::setRandom(), MatrixBase::Random(Index), MatrixBase::Random()
* \sa DenseBase::setRandom(), DenseBase::Random(Index), DenseBase::Random()
*/
template<typename Derived>
inline const CwiseNullaryOp<internal::scalar_random_op<typename internal::traits<Derived>::Scalar>, Derived>
inline const typename DenseBase<Derived>::RandomReturnType
DenseBase<Derived>::Random(Index rows, Index cols)
{
return NullaryExpr(rows, cols, internal::scalar_random_op<Scalar>());
}
/** \returns a random vector expression
*
* Numbers are uniformly spread through their whole definition range for integer types,
* and in the [-1:1] range for floating point scalar types.
*
* The parameter \a size is the size of the returned vector.
* Must be compatible with this MatrixBase type.
*
* \only_for_vectors
* \not_reentrant
*
* This variant is meant to be used for dynamic-size vector types. For fixed-size types,
* it is redundant to pass \a size as argument, so Random() should be used
@@ -69,10 +80,10 @@ DenseBase<Derived>::Random(Index rows, Index cols)
* a temporary vector whenever it is nested in a larger expression. This prevents unexpected
* behavior with expressions involving random matrices.
*
* \sa MatrixBase::setRandom(), MatrixBase::Random(Index,Index), MatrixBase::Random()
* \sa DenseBase::setRandom(), DenseBase::Random(Index,Index), DenseBase::Random()
*/
template<typename Derived>
inline const CwiseNullaryOp<internal::scalar_random_op<typename internal::traits<Derived>::Scalar>, Derived>
inline const typename DenseBase<Derived>::RandomReturnType
DenseBase<Derived>::Random(Index size)
{
return NullaryExpr(size, internal::scalar_random_op<Scalar>());
@@ -80,6 +91,9 @@ DenseBase<Derived>::Random(Index size)
/** \returns a fixed-size random matrix or vector expression
*
* Numbers are uniformly spread through their whole definition range for integer types,
* and in the [-1:1] range for floating point scalar types.
*
* This variant is only for fixed-size MatrixBase types. For dynamic-size types, you
* need to use the variants taking size arguments.
*
@@ -89,11 +103,13 @@ DenseBase<Derived>::Random(Index size)
* This expression has the "evaluate before nesting" flag so that it will be evaluated into
* a temporary matrix whenever it is nested in a larger expression. This prevents unexpected
* behavior with expressions involving random matrices.
*
* \not_reentrant
*
* \sa MatrixBase::setRandom(), MatrixBase::Random(Index,Index), MatrixBase::Random(Index)
* \sa DenseBase::setRandom(), DenseBase::Random(Index,Index), DenseBase::Random(Index)
*/
template<typename Derived>
inline const CwiseNullaryOp<internal::scalar_random_op<typename internal::traits<Derived>::Scalar>, Derived>
inline const typename DenseBase<Derived>::RandomReturnType
DenseBase<Derived>::Random()
{
return NullaryExpr(RowsAtCompileTime, ColsAtCompileTime, internal::scalar_random_op<Scalar>());
@@ -101,6 +117,11 @@ DenseBase<Derived>::Random()
/** Sets all coefficients in this expression to random values.
*
* Numbers are uniformly spread through their whole definition range for integer types,
* and in the [-1:1] range for floating point scalar types.
*
* \not_reentrant
*
* Example: \include MatrixBase_setRandom.cpp
* Output: \verbinclude MatrixBase_setRandom.out
*
@@ -114,12 +135,16 @@ inline Derived& DenseBase<Derived>::setRandom()
/** Resizes to the given \a newSize, and sets all coefficients in this expression to random values.
*
* Numbers are uniformly spread through their whole definition range for integer types,
* and in the [-1:1] range for floating point scalar types.
*
* \only_for_vectors
* \not_reentrant
*
* Example: \include Matrix_setRandom_int.cpp
* Output: \verbinclude Matrix_setRandom_int.out
*
* \sa MatrixBase::setRandom(), setRandom(Index,Index), class CwiseNullaryOp, MatrixBase::Random()
* \sa DenseBase::setRandom(), setRandom(Index,Index), class CwiseNullaryOp, DenseBase::Random()
*/
template<typename Derived>
EIGEN_STRONG_INLINE Derived&
@@ -131,19 +156,24 @@ PlainObjectBase<Derived>::setRandom(Index newSize)
/** Resizes to the given size, and sets all coefficients in this expression to random values.
*
* \param nbRows the new number of rows
* \param nbCols the new number of columns
* Numbers are uniformly spread through their whole definition range for integer types,
* and in the [-1:1] range for floating point scalar types.
*
* \not_reentrant
*
* \param rows the new number of rows
* \param cols the new number of columns
*
* Example: \include Matrix_setRandom_int_int.cpp
* Output: \verbinclude Matrix_setRandom_int_int.out
*
* \sa MatrixBase::setRandom(), setRandom(Index), class CwiseNullaryOp, MatrixBase::Random()
* \sa DenseBase::setRandom(), setRandom(Index), class CwiseNullaryOp, DenseBase::Random()
*/
template<typename Derived>
EIGEN_STRONG_INLINE Derived&
PlainObjectBase<Derived>::setRandom(Index nbRows, Index nbCols)
PlainObjectBase<Derived>::setRandom(Index rows, Index cols)
{
resize(nbRows, nbCols);
resize(rows, cols);
return setRandom();
}

View File

@@ -27,8 +27,9 @@ template<typename Func, typename Derived>
struct redux_traits
{
public:
typedef typename find_best_packet<typename Derived::Scalar,Derived::SizeAtCompileTime>::type PacketType;
enum {
PacketSize = packet_traits<typename Derived::Scalar>::size,
PacketSize = unpacket_traits<PacketType>::size,
InnerMaxSize = int(Derived::IsRowMajor)
? Derived::MaxColsAtCompileTime
: Derived::MaxRowsAtCompileTime
@@ -37,8 +38,8 @@ public:
enum {
MightVectorize = (int(Derived::Flags)&ActualPacketAccessBit)
&& (functor_traits<Func>::PacketAccess),
MayLinearVectorize = MightVectorize && (int(Derived::Flags)&LinearAccessBit),
MaySliceVectorize = MightVectorize && int(InnerMaxSize)>=3*PacketSize
MayLinearVectorize = bool(MightVectorize) && (int(Derived::Flags)&LinearAccessBit),
MaySliceVectorize = bool(MightVectorize) && int(InnerMaxSize)>=3*PacketSize
};
public:
@@ -50,21 +51,34 @@ public:
public:
enum {
Cost = ( Derived::SizeAtCompileTime == Dynamic
|| Derived::CoeffReadCost == Dynamic
|| (Derived::SizeAtCompileTime!=1 && functor_traits<Func>::Cost == Dynamic)
) ? Dynamic
: Derived::SizeAtCompileTime * Derived::CoeffReadCost
+ (Derived::SizeAtCompileTime-1) * functor_traits<Func>::Cost,
Cost = Derived::SizeAtCompileTime == Dynamic ? HugeCost
: Derived::SizeAtCompileTime * Derived::CoeffReadCost + (Derived::SizeAtCompileTime-1) * functor_traits<Func>::Cost,
UnrollingLimit = EIGEN_UNROLLING_LIMIT * (int(Traversal) == int(DefaultTraversal) ? 1 : int(PacketSize))
};
public:
enum {
Unrolling = Cost != Dynamic && Cost <= UnrollingLimit
? CompleteUnrolling
: NoUnrolling
Unrolling = Cost <= UnrollingLimit ? CompleteUnrolling : NoUnrolling
};
#ifdef EIGEN_DEBUG_ASSIGN
static void debug()
{
std::cerr << "Xpr: " << typeid(typename Derived::XprType).name() << std::endl;
std::cerr.setf(std::ios::hex, std::ios::basefield);
EIGEN_DEBUG_VAR(Derived::Flags)
std::cerr.unsetf(std::ios::hex);
EIGEN_DEBUG_VAR(InnerMaxSize)
EIGEN_DEBUG_VAR(PacketSize)
EIGEN_DEBUG_VAR(MightVectorize)
EIGEN_DEBUG_VAR(MayLinearVectorize)
EIGEN_DEBUG_VAR(MaySliceVectorize)
EIGEN_DEBUG_VAR(Traversal)
EIGEN_DEBUG_VAR(UnrollingLimit)
EIGEN_DEBUG_VAR(Unrolling)
std::cerr << std::endl;
}
#endif
};
/***************************************************************************
@@ -82,6 +96,7 @@ struct redux_novec_unroller
typedef typename Derived::Scalar Scalar;
EIGEN_DEVICE_FUNC
static EIGEN_STRONG_INLINE Scalar run(const Derived &mat, const Func& func)
{
return func(redux_novec_unroller<Func, Derived, Start, HalfLength>::run(mat,func),
@@ -99,6 +114,7 @@ struct redux_novec_unroller<Func, Derived, Start, 1>
typedef typename Derived::Scalar Scalar;
EIGEN_DEVICE_FUNC
static EIGEN_STRONG_INLINE Scalar run(const Derived &mat, const Func&)
{
return mat.coeffByOuterInner(outer, inner);
@@ -112,6 +128,7 @@ template<typename Func, typename Derived, int Start>
struct redux_novec_unroller<Func, Derived, Start, 0>
{
typedef typename Derived::Scalar Scalar;
EIGEN_DEVICE_FUNC
static EIGEN_STRONG_INLINE Scalar run(const Derived&, const Func&) { return Scalar(); }
};
@@ -121,12 +138,12 @@ template<typename Func, typename Derived, int Start, int Length>
struct redux_vec_unroller
{
enum {
PacketSize = packet_traits<typename Derived::Scalar>::size,
PacketSize = redux_traits<Func, Derived>::PacketSize,
HalfLength = Length/2
};
typedef typename Derived::Scalar Scalar;
typedef typename packet_traits<Scalar>::type PacketScalar;
typedef typename redux_traits<Func, Derived>::PacketType PacketScalar;
static EIGEN_STRONG_INLINE PacketScalar run(const Derived &mat, const Func& func)
{
@@ -140,18 +157,18 @@ template<typename Func, typename Derived, int Start>
struct redux_vec_unroller<Func, Derived, Start, 1>
{
enum {
index = Start * packet_traits<typename Derived::Scalar>::size,
index = Start * redux_traits<Func, Derived>::PacketSize,
outer = index / int(Derived::InnerSizeAtCompileTime),
inner = index % int(Derived::InnerSizeAtCompileTime),
alignment = (Derived::Flags & AlignedBit) ? Aligned : Unaligned
alignment = Derived::Alignment
};
typedef typename Derived::Scalar Scalar;
typedef typename packet_traits<Scalar>::type PacketScalar;
typedef typename redux_traits<Func, Derived>::PacketType PacketScalar;
static EIGEN_STRONG_INLINE PacketScalar run(const Derived &mat, const Func&)
{
return mat.template packetByOuterInner<alignment>(outer, inner);
return mat.template packetByOuterInner<alignment,PacketScalar>(outer, inner);
}
};
@@ -169,8 +186,8 @@ template<typename Func, typename Derived>
struct redux_impl<Func, Derived, DefaultTraversal, NoUnrolling>
{
typedef typename Derived::Scalar Scalar;
typedef typename Derived::Index Index;
static EIGEN_STRONG_INLINE Scalar run(const Derived& mat, const Func& func)
EIGEN_DEVICE_FUNC
static EIGEN_STRONG_INLINE Scalar run(const Derived &mat, const Func& func)
{
eigen_assert(mat.rows()>0 && mat.cols()>0 && "you are using an empty matrix");
Scalar res;
@@ -193,19 +210,19 @@ template<typename Func, typename Derived>
struct redux_impl<Func, Derived, LinearVectorizedTraversal, NoUnrolling>
{
typedef typename Derived::Scalar Scalar;
typedef typename packet_traits<Scalar>::type PacketScalar;
typedef typename Derived::Index Index;
typedef typename redux_traits<Func, Derived>::PacketType PacketScalar;
static Scalar run(const Derived& mat, const Func& func)
static Scalar run(const Derived &mat, const Func& func)
{
const Index size = mat.size();
eigen_assert(size && "you are using an empty matrix");
const Index packetSize = packet_traits<Scalar>::size;
const Index alignedStart = internal::first_aligned(mat);
const Index packetSize = redux_traits<Func, Derived>::PacketSize;
const int packetAlignment = unpacket_traits<PacketScalar>::alignment;
enum {
alignment = bool(Derived::Flags & DirectAccessBit) || bool(Derived::Flags & AlignedBit)
? Aligned : Unaligned
alignment0 = (bool(Derived::Flags & DirectAccessBit) && bool(packet_traits<Scalar>::AlignedOnScalar)) ? int(packetAlignment) : int(Unaligned),
alignment = EIGEN_PLAIN_ENUM_MAX(alignment0, Derived::Alignment)
};
const Index alignedStart = internal::first_default_aligned(mat.nestedExpression());
const Index alignedSize2 = ((size-alignedStart)/(2*packetSize))*(2*packetSize);
const Index alignedSize = ((size-alignedStart)/(packetSize))*(packetSize);
const Index alignedEnd2 = alignedStart + alignedSize2;
@@ -213,19 +230,19 @@ struct redux_impl<Func, Derived, LinearVectorizedTraversal, NoUnrolling>
Scalar res;
if(alignedSize)
{
PacketScalar packet_res0 = mat.template packet<alignment>(alignedStart);
PacketScalar packet_res0 = mat.template packet<alignment,PacketScalar>(alignedStart);
if(alignedSize>packetSize) // we have at least two packets to partly unroll the loop
{
PacketScalar packet_res1 = mat.template packet<alignment>(alignedStart+packetSize);
PacketScalar packet_res1 = mat.template packet<alignment,PacketScalar>(alignedStart+packetSize);
for(Index index = alignedStart + 2*packetSize; index < alignedEnd2; index += 2*packetSize)
{
packet_res0 = func.packetOp(packet_res0, mat.template packet<alignment>(index));
packet_res1 = func.packetOp(packet_res1, mat.template packet<alignment>(index+packetSize));
packet_res0 = func.packetOp(packet_res0, mat.template packet<alignment,PacketScalar>(index));
packet_res1 = func.packetOp(packet_res1, mat.template packet<alignment,PacketScalar>(index+packetSize));
}
packet_res0 = func.packetOp(packet_res0,packet_res1);
if(alignedEnd>alignedEnd2)
packet_res0 = func.packetOp(packet_res0, mat.template packet<alignment>(alignedEnd2));
packet_res0 = func.packetOp(packet_res0, mat.template packet<alignment,PacketScalar>(alignedEnd2));
}
res = func.predux(packet_res0);
@@ -252,25 +269,24 @@ template<typename Func, typename Derived, int Unrolling>
struct redux_impl<Func, Derived, SliceVectorizedTraversal, Unrolling>
{
typedef typename Derived::Scalar Scalar;
typedef typename packet_traits<Scalar>::type PacketScalar;
typedef typename Derived::Index Index;
typedef typename redux_traits<Func, Derived>::PacketType PacketType;
static Scalar run(const Derived& mat, const Func& func)
EIGEN_DEVICE_FUNC static Scalar run(const Derived &mat, const Func& func)
{
eigen_assert(mat.rows()>0 && mat.cols()>0 && "you are using an empty matrix");
const Index innerSize = mat.innerSize();
const Index outerSize = mat.outerSize();
enum {
packetSize = packet_traits<Scalar>::size
packetSize = redux_traits<Func, Derived>::PacketSize
};
const Index packetedInnerSize = ((innerSize)/packetSize)*packetSize;
Scalar res;
if(packetedInnerSize)
{
PacketScalar packet_res = mat.template packet<Unaligned>(0,0);
PacketType packet_res = mat.template packet<Unaligned,PacketType>(0,0);
for(Index j=0; j<outerSize; ++j)
for(Index i=(j==0?packetSize:0); i<packetedInnerSize; i+=Index(packetSize))
packet_res = func.packetOp(packet_res, mat.template packetByOuterInner<Unaligned>(j,i));
packet_res = func.packetOp(packet_res, mat.template packetByOuterInner<Unaligned,PacketType>(j,i));
res = func.predux(packet_res);
for(Index j=0; j<outerSize; ++j)
@@ -291,22 +307,90 @@ template<typename Func, typename Derived>
struct redux_impl<Func, Derived, LinearVectorizedTraversal, CompleteUnrolling>
{
typedef typename Derived::Scalar Scalar;
typedef typename packet_traits<Scalar>::type PacketScalar;
typedef typename redux_traits<Func, Derived>::PacketType PacketScalar;
enum {
PacketSize = packet_traits<Scalar>::size,
PacketSize = redux_traits<Func, Derived>::PacketSize,
Size = Derived::SizeAtCompileTime,
VectorizedSize = (Size / PacketSize) * PacketSize
};
static EIGEN_STRONG_INLINE Scalar run(const Derived& mat, const Func& func)
EIGEN_DEVICE_FUNC static EIGEN_STRONG_INLINE Scalar run(const Derived &mat, const Func& func)
{
eigen_assert(mat.rows()>0 && mat.cols()>0 && "you are using an empty matrix");
Scalar res = func.predux(redux_vec_unroller<Func, Derived, 0, Size / PacketSize>::run(mat,func));
if (VectorizedSize != Size)
res = func(res,redux_novec_unroller<Func, Derived, VectorizedSize, Size-VectorizedSize>::run(mat,func));
return res;
if (VectorizedSize > 0) {
Scalar res = func.predux(redux_vec_unroller<Func, Derived, 0, Size / PacketSize>::run(mat,func));
if (VectorizedSize != Size)
res = func(res,redux_novec_unroller<Func, Derived, VectorizedSize, Size-VectorizedSize>::run(mat,func));
return res;
}
else {
return redux_novec_unroller<Func, Derived, 0, Size>::run(mat,func);
}
}
};
// evaluator adaptor
template<typename _XprType>
class redux_evaluator
{
public:
typedef _XprType XprType;
EIGEN_DEVICE_FUNC explicit redux_evaluator(const XprType &xpr) : m_evaluator(xpr), m_xpr(xpr) {}
typedef typename XprType::Scalar Scalar;
typedef typename XprType::CoeffReturnType CoeffReturnType;
typedef typename XprType::PacketScalar PacketScalar;
typedef typename XprType::PacketReturnType PacketReturnType;
enum {
MaxRowsAtCompileTime = XprType::MaxRowsAtCompileTime,
MaxColsAtCompileTime = XprType::MaxColsAtCompileTime,
// TODO we should not remove DirectAccessBit and rather find an elegant way to query the alignment offset at runtime from the evaluator
Flags = evaluator<XprType>::Flags & ~DirectAccessBit,
IsRowMajor = XprType::IsRowMajor,
SizeAtCompileTime = XprType::SizeAtCompileTime,
InnerSizeAtCompileTime = XprType::InnerSizeAtCompileTime,
CoeffReadCost = evaluator<XprType>::CoeffReadCost,
Alignment = evaluator<XprType>::Alignment
};
EIGEN_DEVICE_FUNC Index rows() const { return m_xpr.rows(); }
EIGEN_DEVICE_FUNC Index cols() const { return m_xpr.cols(); }
EIGEN_DEVICE_FUNC Index size() const { return m_xpr.size(); }
EIGEN_DEVICE_FUNC Index innerSize() const { return m_xpr.innerSize(); }
EIGEN_DEVICE_FUNC Index outerSize() const { return m_xpr.outerSize(); }
EIGEN_DEVICE_FUNC
CoeffReturnType coeff(Index row, Index col) const
{ return m_evaluator.coeff(row, col); }
EIGEN_DEVICE_FUNC
CoeffReturnType coeff(Index index) const
{ return m_evaluator.coeff(index); }
template<int LoadMode, typename PacketType>
PacketType packet(Index row, Index col) const
{ return m_evaluator.template packet<LoadMode,PacketType>(row, col); }
template<int LoadMode, typename PacketType>
PacketType packet(Index index) const
{ return m_evaluator.template packet<LoadMode,PacketType>(index); }
EIGEN_DEVICE_FUNC
CoeffReturnType coeffByOuterInner(Index outer, Index inner) const
{ return m_evaluator.coeff(IsRowMajor ? outer : inner, IsRowMajor ? inner : outer); }
template<int LoadMode, typename PacketType>
PacketType packetByOuterInner(Index outer, Index inner) const
{ return m_evaluator.template packet<LoadMode,PacketType>(IsRowMajor ? outer : inner, IsRowMajor ? inner : outer); }
const XprType & nestedExpression() const { return m_xpr; }
protected:
internal::evaluator<XprType> m_evaluator;
const XprType &m_xpr;
};
} // end namespace internal
/***************************************************************************
@@ -317,18 +401,21 @@ struct redux_impl<Func, Derived, LinearVectorizedTraversal, CompleteUnrolling>
/** \returns the result of a full redux operation on the whole matrix or vector using \a func
*
* The template parameter \a BinaryOp is the type of the functor \a func which must be
* an associative operator. Both current STL and TR1 functor styles are handled.
* an associative operator. Both current C++98 and C++11 functor styles are handled.
*
* \sa DenseBase::sum(), DenseBase::minCoeff(), DenseBase::maxCoeff(), MatrixBase::colwise(), MatrixBase::rowwise()
*/
template<typename Derived>
template<typename Func>
EIGEN_STRONG_INLINE typename internal::result_of<Func(typename internal::traits<Derived>::Scalar)>::type
EIGEN_STRONG_INLINE typename internal::traits<Derived>::Scalar
DenseBase<Derived>::redux(const Func& func) const
{
typedef typename internal::remove_all<typename Derived::Nested>::type ThisNested;
return internal::redux_impl<Func, ThisNested>
::run(derived(), func);
eigen_assert(this->rows()>0 && this->cols()>0 && "you are using an empty matrix");
typedef typename internal::redux_evaluator<Derived> ThisEvaluator;
ThisEvaluator thisEval(derived());
return internal::redux_impl<Func, ThisEvaluator>::run(thisEval, func);
}
/** \returns the minimum of all coefficients of \c *this.
@@ -338,7 +425,7 @@ template<typename Derived>
EIGEN_STRONG_INLINE typename internal::traits<Derived>::Scalar
DenseBase<Derived>::minCoeff() const
{
return this->redux(Eigen::internal::scalar_min_op<Scalar>());
return derived().redux(Eigen::internal::scalar_min_op<Scalar,Scalar>());
}
/** \returns the maximum of all coefficients of \c *this.
@@ -348,10 +435,12 @@ template<typename Derived>
EIGEN_STRONG_INLINE typename internal::traits<Derived>::Scalar
DenseBase<Derived>::maxCoeff() const
{
return this->redux(Eigen::internal::scalar_max_op<Scalar>());
return derived().redux(Eigen::internal::scalar_max_op<Scalar,Scalar>());
}
/** \returns the sum of all coefficients of *this
/** \returns the sum of all coefficients of \c *this
*
* If \c *this is empty, then the value 0 is returned.
*
* \sa trace(), prod(), mean()
*/
@@ -361,7 +450,7 @@ DenseBase<Derived>::sum() const
{
if(SizeAtCompileTime==0 || (SizeAtCompileTime==Dynamic && size()==0))
return Scalar(0);
return this->redux(Eigen::internal::scalar_sum_op<Scalar>());
return derived().redux(Eigen::internal::scalar_sum_op<Scalar,Scalar>());
}
/** \returns the mean of all coefficients of *this
@@ -372,7 +461,14 @@ template<typename Derived>
EIGEN_STRONG_INLINE typename internal::traits<Derived>::Scalar
DenseBase<Derived>::mean() const
{
return Scalar(this->redux(Eigen::internal::scalar_sum_op<Scalar>())) / Scalar(this->size());
#ifdef __INTEL_COMPILER
#pragma warning push
#pragma warning ( disable : 2259 )
#endif
return Scalar(derived().redux(Eigen::internal::scalar_sum_op<Scalar,Scalar>())) / Scalar(this->size());
#ifdef __INTEL_COMPILER
#pragma warning pop
#endif
}
/** \returns the product of all coefficients of *this
@@ -388,7 +484,7 @@ DenseBase<Derived>::prod() const
{
if(SizeAtCompileTime==0 || (SizeAtCompileTime==Dynamic && size()==0))
return Scalar(1);
return this->redux(Eigen::internal::scalar_product_op<Scalar>());
return derived().redux(Eigen::internal::scalar_product_op<Scalar>());
}
/** \returns the trace of \c *this, i.e. the sum of the coefficients on the main diagonal.

View File

@@ -12,79 +12,6 @@
namespace Eigen {
template<typename Derived> class RefBase;
template<typename PlainObjectType, int Options = 0,
typename StrideType = typename internal::conditional<PlainObjectType::IsVectorAtCompileTime,InnerStride<1>,OuterStride<> >::type > class Ref;
/** \class Ref
* \ingroup Core_Module
*
* \brief A matrix or vector expression mapping an existing expressions
*
* \tparam PlainObjectType the equivalent matrix type of the mapped data
* \tparam Options specifies whether the pointer is \c #Aligned, or \c #Unaligned.
* The default is \c #Unaligned.
* \tparam StrideType optionally specifies strides. By default, Ref implies a contiguous storage along the inner dimension (inner stride==1),
* but accept a variable outer stride (leading dimension).
* This can be overridden by specifying strides.
* The type passed here must be a specialization of the Stride template, see examples below.
*
* This class permits to write non template functions taking Eigen's object as parameters while limiting the number of copies.
* A Ref<> object can represent either a const expression or a l-value:
* \code
* // in-out argument:
* void foo1(Ref<VectorXf> x);
*
* // read-only const argument:
* void foo2(const Ref<const VectorXf>& x);
* \endcode
*
* In the in-out case, the input argument must satisfies the constraints of the actual Ref<> type, otherwise a compilation issue will be triggered.
* By default, a Ref<VectorXf> can reference any dense vector expression of float having a contiguous memory layout.
* Likewise, a Ref<MatrixXf> can reference any column major dense matrix expression of float whose column's elements are contiguously stored with
* the possibility to have a constant space inbetween each column, i.e.: the inner stride mmust be equal to 1, but the outer-stride (or leading dimension),
* can be greater than the number of rows.
*
* In the const case, if the input expression does not match the above requirement, then it is evaluated into a temporary before being passed to the function.
* Here are some examples:
* \code
* MatrixXf A;
* VectorXf a;
* foo1(a.head()); // OK
* foo1(A.col()); // OK
* foo1(A.row()); // compilation error because here innerstride!=1
* foo2(A.row()); // The row is copied into a contiguous temporary
* foo2(2*a); // The expression is evaluated into a temporary
* foo2(A.col().segment(2,4)); // No temporary
* \endcode
*
* The range of inputs that can be referenced without temporary can be enlarged using the last two template parameter.
* Here is an example accepting an innerstride!=1:
* \code
* // in-out argument:
* void foo3(Ref<VectorXf,0,InnerStride<> > x);
* foo3(A.row()); // OK
* \endcode
* The downside here is that the function foo3 might be significantly slower than foo1 because it won't be able to exploit vectorization, and will involved more
* expensive address computations even if the input is contiguously stored in memory. To overcome this issue, one might propose to overloads internally calling a
* template function, e.g.:
* \code
* // in the .h:
* void foo(const Ref<MatrixXf>& A);
* void foo(const Ref<MatrixXf,0,Stride<> >& A);
*
* // in the .cpp:
* template<typename TypeOfA> void foo_impl(const TypeOfA& A) {
* ... // crazy code goes here
* }
* void foo(const Ref<MatrixXf>& A) { foo_impl(A); }
* void foo(const Ref<MatrixXf,0,Stride<> >& A) { foo_impl(A); }
* \endcode
*
*
* \sa PlainObjectBase::Map(), \ref TopicStorageOrders
*/
namespace internal {
template<typename _PlainObjectType, int _Options, typename _StrideType>
@@ -95,7 +22,8 @@ struct traits<Ref<_PlainObjectType, _Options, _StrideType> >
typedef _StrideType StrideType;
enum {
Options = _Options,
Flags = traits<Map<_PlainObjectType, _Options, _StrideType> >::Flags | NestByRefBit
Flags = traits<Map<_PlainObjectType, _Options, _StrideType> >::Flags | NestByRefBit,
Alignment = traits<Map<_PlainObjectType, _Options, _StrideType> >::Alignment
};
template<typename Derived> struct match {
@@ -107,7 +35,13 @@ struct traits<Ref<_PlainObjectType, _Options, _StrideType> >
|| (int(StrideType::InnerStrideAtCompileTime)==0 && int(Derived::InnerStrideAtCompileTime)==1),
OuterStrideMatch = Derived::IsVectorAtCompileTime
|| int(StrideType::OuterStrideAtCompileTime)==int(Dynamic) || int(StrideType::OuterStrideAtCompileTime)==int(Derived::OuterStrideAtCompileTime),
AlignmentMatch = (_Options!=Aligned) || ((PlainObjectType::Flags&AlignedBit)==0) || ((traits<Derived>::Flags&AlignedBit)==AlignedBit),
// NOTE, this indirection of evaluator<Derived>::Alignment is needed
// to workaround a very strange bug in MSVC related to the instantiation
// of has_*ary_operator in evaluator<CwiseNullaryOp>.
// This line is surprisingly very sensitive. For instance, simply adding parenthesis
// as "DerivedAlignment = (int(evaluator<Derived>::Alignment))," will make MSVC fail...
DerivedAlignment = int(evaluator<Derived>::Alignment),
AlignmentMatch = (int(traits<PlainObjectType>::Alignment)==int(Unaligned)) || (DerivedAlignment >= int(Alignment)), // FIXME the first condition is not very clear, it should be replaced by the required alignment
ScalarTypeMatch = internal::is_same<typename PlainObjectType::Scalar, typename Derived::Scalar>::value,
MatchAtCompileTime = HasDirectAccess && StorageOrderMatch && InnerStrideMatch && OuterStrideMatch && AlignmentMatch && ScalarTypeMatch
};
@@ -132,12 +66,12 @@ public:
typedef MapBase<Derived> Base;
EIGEN_DENSE_PUBLIC_INTERFACE(RefBase)
inline Index innerStride() const
EIGEN_DEVICE_FUNC inline Index innerStride() const
{
return StrideType::InnerStrideAtCompileTime != 0 ? m_stride.inner() : 1;
}
inline Index outerStride() const
EIGEN_DEVICE_FUNC inline Index outerStride() const
{
return StrideType::OuterStrideAtCompileTime != 0 ? m_stride.outer()
: IsVectorAtCompileTime ? this->size()
@@ -145,7 +79,7 @@ public:
: this->rows();
}
RefBase()
EIGEN_DEVICE_FUNC RefBase()
: Base(0,RowsAtCompileTime==Dynamic?0:RowsAtCompileTime,ColsAtCompileTime==Dynamic?0:ColsAtCompileTime),
// Stride<> does not allow default ctor for Dynamic strides, so let' initialize it with dummy values:
m_stride(StrideType::OuterStrideAtCompileTime==Dynamic?0:StrideType::OuterStrideAtCompileTime,
@@ -159,8 +93,10 @@ protected:
typedef Stride<StrideType::OuterStrideAtCompileTime,StrideType::InnerStrideAtCompileTime> StrideBase;
template<typename Expression>
void construct(Expression& expr)
EIGEN_DEVICE_FUNC void construct(Expression& expr)
{
EIGEN_STATIC_ASSERT_SAME_MATRIX_SIZE(PlainObjectType,Expression);
if(PlainObjectType::RowsAtCompileTime==1)
{
eigen_assert(expr.rows()==1 || expr.cols()==1);
@@ -184,15 +120,83 @@ protected:
StrideBase m_stride;
};
/** \class Ref
* \ingroup Core_Module
*
* \brief A matrix or vector expression mapping an existing expression
*
* \tparam PlainObjectType the equivalent matrix type of the mapped data
* \tparam Options specifies the pointer alignment in bytes. It can be: \c #Aligned128, , \c #Aligned64, \c #Aligned32, \c #Aligned16, \c #Aligned8 or \c #Unaligned.
* The default is \c #Unaligned.
* \tparam StrideType optionally specifies strides. By default, Ref implies a contiguous storage along the inner dimension (inner stride==1),
* but accepts a variable outer stride (leading dimension).
* This can be overridden by specifying strides.
* The type passed here must be a specialization of the Stride template, see examples below.
*
* This class provides a way to write non-template functions taking Eigen objects as parameters while limiting the number of copies.
* A Ref<> object can represent either a const expression or a l-value:
* \code
* // in-out argument:
* void foo1(Ref<VectorXf> x);
*
* // read-only const argument:
* void foo2(const Ref<const VectorXf>& x);
* \endcode
*
* In the in-out case, the input argument must satisfy the constraints of the actual Ref<> type, otherwise a compilation issue will be triggered.
* By default, a Ref<VectorXf> can reference any dense vector expression of float having a contiguous memory layout.
* Likewise, a Ref<MatrixXf> can reference any column-major dense matrix expression of float whose column's elements are contiguously stored with
* the possibility to have a constant space in-between each column, i.e. the inner stride must be equal to 1, but the outer stride (or leading dimension)
* can be greater than the number of rows.
*
* In the const case, if the input expression does not match the above requirement, then it is evaluated into a temporary before being passed to the function.
* Here are some examples:
* \code
* MatrixXf A;
* VectorXf a;
* foo1(a.head()); // OK
* foo1(A.col()); // OK
* foo1(A.row()); // Compilation error because here innerstride!=1
* foo2(A.row()); // Compilation error because A.row() is a 1xN object while foo2 is expecting a Nx1 object
* foo2(A.row().transpose()); // The row is copied into a contiguous temporary
* foo2(2*a); // The expression is evaluated into a temporary
* foo2(A.col().segment(2,4)); // No temporary
* \endcode
*
* The range of inputs that can be referenced without temporary can be enlarged using the last two template parameters.
* Here is an example accepting an innerstride!=1:
* \code
* // in-out argument:
* void foo3(Ref<VectorXf,0,InnerStride<> > x);
* foo3(A.row()); // OK
* \endcode
* The downside here is that the function foo3 might be significantly slower than foo1 because it won't be able to exploit vectorization, and will involve more
* expensive address computations even if the input is contiguously stored in memory. To overcome this issue, one might propose to overload internally calling a
* template function, e.g.:
* \code
* // in the .h:
* void foo(const Ref<MatrixXf>& A);
* void foo(const Ref<MatrixXf,0,Stride<> >& A);
*
* // in the .cpp:
* template<typename TypeOfA> void foo_impl(const TypeOfA& A) {
* ... // crazy code goes here
* }
* void foo(const Ref<MatrixXf>& A) { foo_impl(A); }
* void foo(const Ref<MatrixXf,0,Stride<> >& A) { foo_impl(A); }
* \endcode
*
*
* \sa PlainObjectBase::Map(), \ref TopicStorageOrders
*/
template<typename PlainObjectType, int Options, typename StrideType> class Ref
: public RefBase<Ref<PlainObjectType, Options, StrideType> >
{
private:
typedef internal::traits<Ref> Traits;
template<typename Derived>
inline Ref(const PlainObjectBase<Derived>& expr,
typename internal::enable_if<bool(Traits::template match<Derived>::MatchAtCompileTime),Derived>::type* = 0);
EIGEN_DEVICE_FUNC inline Ref(const PlainObjectBase<Derived>& expr,
typename internal::enable_if<bool(Traits::template match<Derived>::MatchAtCompileTime),Derived>::type* = 0);
public:
typedef RefBase<Ref> Base;
@@ -201,23 +205,24 @@ template<typename PlainObjectType, int Options, typename StrideType> class Ref
#ifndef EIGEN_PARSED_BY_DOXYGEN
template<typename Derived>
inline Ref(PlainObjectBase<Derived>& expr,
typename internal::enable_if<bool(Traits::template match<Derived>::MatchAtCompileTime),Derived>::type* = 0)
EIGEN_DEVICE_FUNC inline Ref(PlainObjectBase<Derived>& expr,
typename internal::enable_if<bool(Traits::template match<Derived>::MatchAtCompileTime),Derived>::type* = 0)
{
EIGEN_STATIC_ASSERT(static_cast<bool>(Traits::template match<Derived>::MatchAtCompileTime), STORAGE_LAYOUT_DOES_NOT_MATCH);
EIGEN_STATIC_ASSERT(bool(Traits::template match<Derived>::MatchAtCompileTime), STORAGE_LAYOUT_DOES_NOT_MATCH);
Base::construct(expr.derived());
}
template<typename Derived>
inline Ref(const DenseBase<Derived>& expr,
typename internal::enable_if<bool(Traits::template match<Derived>::MatchAtCompileTime),Derived>::type* = 0)
EIGEN_DEVICE_FUNC inline Ref(const DenseBase<Derived>& expr,
typename internal::enable_if<bool(Traits::template match<Derived>::MatchAtCompileTime),Derived>::type* = 0)
#else
/** Implicit constructor from any dense expression */
template<typename Derived>
inline Ref(DenseBase<Derived>& expr)
#endif
{
EIGEN_STATIC_ASSERT(static_cast<bool>(internal::is_lvalue<Derived>::value), THIS_EXPRESSION_IS_NOT_A_LVALUE__IT_IS_READ_ONLY);
EIGEN_STATIC_ASSERT(static_cast<bool>(Traits::template match<Derived>::MatchAtCompileTime), STORAGE_LAYOUT_DOES_NOT_MATCH);
enum { THIS_EXPRESSION_IS_NOT_A_LVALUE__IT_IS_READ_ONLY = Derived::ThisConstantIsPrivateInPlainObjectBase};
EIGEN_STATIC_ASSERT(bool(internal::is_lvalue<Derived>::value), THIS_EXPRESSION_IS_NOT_A_LVALUE__IT_IS_READ_ONLY);
EIGEN_STATIC_ASSERT(bool(Traits::template match<Derived>::MatchAtCompileTime), STORAGE_LAYOUT_DOES_NOT_MATCH);
EIGEN_STATIC_ASSERT(!Derived::IsPlainObjectBase,THIS_EXPRESSION_IS_NOT_A_LVALUE__IT_IS_READ_ONLY);
Base::construct(expr.const_cast_derived());
}
@@ -236,36 +241,36 @@ template<typename TPlainObjectType, int Options, typename StrideType> class Ref<
EIGEN_DENSE_PUBLIC_INTERFACE(Ref)
template<typename Derived>
inline Ref(const DenseBase<Derived>& expr,
typename internal::enable_if<bool(Traits::template match<Derived>::ScalarTypeMatch),Derived>::type* = 0)
EIGEN_DEVICE_FUNC inline Ref(const DenseBase<Derived>& expr,
typename internal::enable_if<bool(Traits::template match<Derived>::ScalarTypeMatch),Derived>::type* = 0)
{
// std::cout << match_helper<Derived>::HasDirectAccess << "," << match_helper<Derived>::OuterStrideMatch << "," << match_helper<Derived>::InnerStrideMatch << "\n";
// std::cout << int(StrideType::OuterStrideAtCompileTime) << " - " << int(Derived::OuterStrideAtCompileTime) << "\n";
// std::cout << int(StrideType::InnerStrideAtCompileTime) << " - " << int(Derived::InnerStrideAtCompileTime) << "\n";
construct(expr.derived(), typename Traits::template match<Derived>::type());
}
inline Ref(const Ref& other) : Base(other) {
EIGEN_DEVICE_FUNC inline Ref(const Ref& other) : Base(other) {
// copy constructor shall not copy the m_object, to avoid unnecessary malloc and copy
}
template<typename OtherRef>
inline Ref(const RefBase<OtherRef>& other) {
EIGEN_DEVICE_FUNC inline Ref(const RefBase<OtherRef>& other) {
construct(other.derived(), typename Traits::template match<OtherRef>::type());
}
protected:
template<typename Expression>
void construct(const Expression& expr,internal::true_type)
EIGEN_DEVICE_FUNC void construct(const Expression& expr,internal::true_type)
{
Base::construct(expr);
}
template<typename Expression>
void construct(const Expression& expr, internal::false_type)
EIGEN_DEVICE_FUNC void construct(const Expression& expr, internal::false_type)
{
m_object.lazyAssign(expr);
internal::call_assignment_no_alias(m_object,expr,internal::assign_op<Scalar,Scalar>());
Base::construct(m_object);
}

Some files were not shown because too many files have changed in this diff Show More