They now also store a list of samples per pixel, and then get
shaded together with the ztransp samples. This comes with a
slight speed hit, but mainly memory might be a concern. However,
testing some peach scenes I haven't problems.
This completes the pipeline make-over, as started in 2006. With this
option, during rendering, each sample for every layer and pass is being
saved on disk (looks like non-antialiased images). Then the composite
and color correction happens, then a clip to 0-1 range, and only in end
all samples get combined - using sampling filters such as gauss/mitch/catmul.
This results in artefact-free antialiased images. Even Z-combine or
ID masks now work perfect for it!
This is an unfinished commit btw; Brecht will finish this for strands.
Also Halo doesnt work yet.
To activate FSA: press "Save Buffers" and the new button next to it. :)
Problem: artist wants character to walk in grass, but still have all rendered
in seperate render-layers, for postpro effects and vblur. How to efficiently
create a mask image you can put *over* the character for the grass?
Solution has two parts; this commits allows any layer inside of the renderlayers
to become a Z-mask (Z values for solid gets filled in, but not rendered).
Second part of commit is render option "Only render stuff that's in front of
a zbuffer value that was filled in (saves render time)
of strands changing between frames, vector blur couldn't work. Now
speed vectors are interpolated from the surface. This also means
child particles don't have to be computed in the previous and next
frames, so saves time too.
Also, duplis are now taking into account, the proper way to exclude
them is to set the material to be not traceable.
Removed an unnecessary pointer from the VlakRen struct to save some
memory, not really that significant, but still, saves 70 mb for 10
million faces.
=============================
A new approximate ambient occlusion method has been added, next to the
existing one based on raytracing. This method is specifically targetted
at use in animations, since it is inherently noise free, and so will
not flicker across frames.
http://www.blender.org/development/current-projects/changes-since-244/approximate-ambient-occlusion/http://peach.blender.org/index.php/approximate-ambient-occlusion/
Further improvements are still needed, but it can be tested already. There
are still a number of known issues:
- Bias errors on backfaces.
- For performance, instanced object do not occlude currently.
- Sky textures don't work well, the derivatives for texture evaluation
are not correct.
- Multiple passes do not work entirely correct (they are not accurate
to begin with, but could be better).
Lampbuffers require painful bias tweaking (to prevent aliasing or to
get shadow detail). Sometimes you want this different per object, like
for gras you want less shadow detail, but for the ground you want high
detail. This feature allows to tweak it.
The new "LBias" slider is in shader panel, bottom. Ugly! But, thats for
later...
It seems everytime I try to fix something here I break something
else, but anyway, another try at getting this to work properly.
Fixes for:
- Getting dupliverts/faces derivedmesh with orco caused wrong results
on meshes in linked dupligroups with proxy, because modifier stack
was revaluated with wrong object matrix, now gets orco another way.
- Fix render instances being hidden when original object was not added
to the object render list.
- Changed the way object instances find their original objects, now
works the other way around, original objects look for their instances
instead.
There's probably issues still with recursive dupligroups..
====================
- "From Dupli" option for orco and uv texture coordinates. For dupliverts,
duplifaces and dupli particles, this uses the orco and uv at the point
on the parent surface. Can for example be used for texturing feathers
and leafs. Note that uv only works for duplifaces and particles emitted
from faces, these are not defined at vertices.
- "Width Fade" option for strand render, to fade out along the width of the
strand. Committing this so it can be tested, might be changed or removed
even, if it doesn't give nice results.
============================
- Strand render now has options to remove child strands as
the object's faces becomes smaller, in the Simplification
particle panel.
- "Reference Size" is the approximate size of the object on
screen, after which simplification starts.
- "Rate" is how fast strands are removed.
- "Transition" is the percentage of strands being faded out as
they are removed.
- Another "Viewport" option removes strands on faces that are
outside of the viewport. "Rate" again controls how fast these
are removed.
- Strand render in Blender Units now has an adjustable minimum
width. Below this minimum width, strands start fading out
instead of getting smaller.
===============================
- One shadowbuffer per thread.
- Added more break tests in shadow buffer code.
- Removed R.clipcrop global, solution is not nice yet, but at
least threadsafe.
- Fixed bug in strand render shadow buffer code.
=================
Big commit, but little user visible changes.
- Dupliverts and duplifaces are now rendered as instances, instead
of storing all of the geometry for each dupli, now an instance is
created with a matrix transform refering to the source object.
This should allow us to render tree leaves more memory efficient.
- Radiosity and to some degree raytracing of such objects is not
really efficient still. For radiosity this is fundamentally hard
to solve, but raytracing an octree could be created for each object,
but the current octree code with it's fixed size doesn't allow this
efficiently.
- The regression tests survived, but with I expect that some bugs will
pop up .. hopefully not too many :).
Implementation Notes
====================
- Dupligroups and linked meshes are not rendered as instances yet,
since they can in fact be different due to various reasons,
instancing of these types of duplis that are the same can be added
for them at a later point.
- Each ObjectRen now stores it's own database, instead of there being
one big databases of faces, verts, .. . Which objects that are actually
rendered are defined by the list of ObjectRenInstances, which all refer
to an ObjectRen.
- Homogeneous coordinatess and clipping is now not stored in vertices
anymore, but instead computed on the fly. This couldn't work for
instances. That does mean some extra computation has to be done, but
memory lookups can be slow too, and this saves some memory. Overall
I didn't find a significant speed impact.
- OSA rendering for solid and ztransp now is different. Instead of e.g.
going 8 times over the databases times and rendering the z-buffer, it
now goes over the database once and renders each polygon 8 times. That
was necessary to keep instances efficient, and can also give some
performance improvement without instances.
- There was already instancing support in the yafray export code, now it
uses Blender's render instances for export.
- UV and color layer storage in the render was a bit messy before, now
should be easier to understand.
- convertblender.c was reorganized somewhat. Regular render, speedvector
and baking now use a single function to create the database, previously
there was code duplicated for it.
- Some of these changes were done with future multithreading of scene
and shadow buffer creation in mind, though especially for scene creation
much work remains to be done to make it threadsafe, since it also involves
a lot of code from blenkernel, and there is an ugly conflict with the way
dupli groups work here .. though in the render code itself it's almost there.
=========
- Fix crash in particle transform with the particle system not editable.
- Particle child distribution and caching is now multithreaded.
- Child particles now have a separate Render Amount next to the existing
Amount. The render amount particles are now only distributed and cached
at render time, which should make editing with child particles faster.
- Two new options for diffuse strand shading:
- Surface Diffuse: computes the strand normal taking the normal at
the surface into account.
- Blending Distance: the distance in Blender units over which to
blend in the normal at the surface.
- Special strand rendering for more memory efficient and faster hair and
grass. This is a work in progress, and has a number of known issues,
don't report bugs to me for this feature yet.
More info:
http://www.blender.org/development/current-projects/changes-since-244/particles/
=============
A new "Selected to Active" option in the Bake panel, to (typically) bake
a high poly object onto a low poly object. Code based on patch #7339 by
Frank Richter (Crystal Space developer), thanks!.
Normal Mapping
==============
Camera, World, Object and Tangent space is now supported for baking, and
for material textures. The "NMap TS" setting is replaced with a dropdown
of the four choices in the image texture buttons.
http://www.blender.org/development/current-projects/changes-since-244/render-baking/
=========
Merge of the famous particle patch by Janne Karhu, a full rewrite
of the Blender particle system. This includes:
- Emitter, Hair and Reactor particle types.
- Newtonian, Keyed and Boids physics.
- Various particle visualisation and rendering types.
- Vertex group and texture control for various properties.
- Interpolated child particles from parents.
- Hair editing with combing, growing, cutting, .. .
- Explode modifier.
- Harmonic, Magnetic fields, and multiple falloff types.
.. and lots of other things, some more info is here:
http://wiki.blender.org/index.php/BlenderDev/Particles_Rewritehttp://wiki.blender.org/index.php/BlenderDev/Particles_Rewrite_Doc
The new particle system cannot be backwards compatible. Old particle
systems are being converted to the new system, but will require
tweaking to get them looking the same as before.
Point Cache
===========
The new system to replace manual baking, based on automatic caching
on disk. This is currently used by softbodies and the particle system.
See the Cache API section on:
http://wiki.blender.org/index.php/BlenderDev/PhysicsSprint
Documentation
=============
These new features still need good docs for the release logs, help
for this is appreciated.
two separate files, raytrace.c and rayshade.c. The tracing code can now
be used separately from the renderer (will be used in a later commit),
and the raytracing acceleration structure can now also be easily replaced,
if someone wants to experiment with that.
Added support for multiple UVs in the render engine. This also involved
changing the way faces are stored, to allow data to be added optionally
per 256 faces, same as the existing system for vertices.
A UV layer can be specified in the Map Input panel and the Geometry node
by name. Leaving this field blank will default to the active UV layer.
Also added sharing of face selection and hiding between UV layers, and at
the same time improved syncing with editmode selection and hiding.
Still to do:
- Multi UV support for fastshade.
- Multires and NMesh preservation of multiple UV sets.
Please read:
http://www.blender3d.org/cms/Imaging.834.0.html
Or in short:
- adding MultiLayer Image support
- recoded entire Image API
- better integration of movie/sequence Images
Was a whole load of work... went down for a week to do this. So, will need
a lot of testing! Will be in irc all evening.
Functionality was a bit limited still; now added the option to bind
a group of Lamps entirely to a Material, excluding them from any other
Material. Note that Lamp visibility layer options still work as for
Lamps usually.
In preview.blend: made a lightgroup for all lamps, and set them to be
the "RenderLayer override". That way it will ignore local material
lightgroups for previews.
- Bug: material emit was ignored (showed in preview render backdrop)
- Bug: world exposure was ignored
- Bug: lamp halo was ignoring 'render layer light override'.
Further reshuffled the way shadows are being pre-calculated, this to enable
more advanced (and faster) usage of Material lightgroups. Now shadows are
being cached in lamps, using a per-sample counter to check if a recalc is
needed. Will also work (later) for Raytracing node shaders.
- New: Material LightGroup option "Always", which always shades the lights
in the group, independent of visibility layer. (so it allows to move such
lights to hidden layer, not influencing anything).
Removed all limitations from render code for maximum threads. The only
define for this now is in BLI_threads.h, and currently set to 8.
Note that each thread renders an entire tile, and also allocates the
buffers for the tiles, so; more threads might work better with smaller
tiles.
IMPORTANT: node system won't work yet with more than 2 threads! So, don't
try material nodes or compositing with over 2 threads. That I'll commit
later today.
What does work (should work :) is AO and soft shadow now.
All data layers, including MVert/MEdge/MFace, are now managed as custom
data layers. The pointers like Mesh.mvert, Mesh.dvert or Mesh.mcol are
still used of course, but allocating, copying or freeing these arrays
should be done through the CustomData API.
Work in progress documentation on this is here:
http://mediawiki.blender.org/index.php/BlenderDev/BlenderArchitecture/CustomData
Replaced TFace by MTFace:
This is the same struct, except that it does not contain color, that now
always stays separated in MCol. This was not a good design decision to
begin with, and it is needed for adding multiple color layers later. Note
that this does mean older Blender versions will not be able to read UV
coordinates from the next release, due to an SDNA limitation.
Removed DispListMesh:
This now fully replaced by DerivedMesh. To provide access to arrays of
vertices, edges and faces, like DispListMesh does. The semantics of the
DerivedMesh.getVertArray() and similar functions were changed to return
a pointer to an array if one exists, or otherwise allocate a temporary
one. On releasing the DerivedMesh, this temporary array will be removed
automatically.
Removed ssDM and meshDM DerivedMesh backends:
The ssDM backend was for DispListMesh, so that became obsolete automatically.
The meshDM backend was replaced by the custom data backend, that now figures
out which layers need to be modified, and only duplicates those.
This changes code in many places, and overall removes 2514 lines of code.
So, there's a good chance this might break some stuff, although I've been
testing it for a few days now. The good news is, adding multiple color and
uv layers should now become easy.
Here's the full release log with example file.
http://www.blender3d.org/cms/Render_Baking.827.0.html
For people who don't read docs; just press ALT+CTRL+B on a Mesh
with texture faces!
Todos:
- maybe some filter options extra?
- Make normal maps in Tangent space
Irregular shadow buffer, for solid OSA case, suffered aliasing. This caused
by optimized storage of shadow factors in 1 pixel (averaged for all sub
samples). That didn't work well really...
Now the code uses same method as for transparent faces, storing results for
each sample.
http://www.blender3d.org/cms/Shadow_buffer__Halfway.786.0.html
Simply said: by using the average of the nearest and 2nd nearest Z value
in Shadowbuffers you can reduce bias errors very well.
For backwards compatibility it is a new buffer type though.
Full log:
http://www.blender3d.org/cms/Irregular_Shadow_Buffe.785.0.html
In short: this is a shadow buffer approach that always results in crispy
shadows, independent of lamp buffer size or zoom level. This shadow buffer
system also supports transparent shadow.
This is part of work on refreshing Shadow Buffers in Blender. You now can
choose of two types (Classical, Irregular). More types will follow. Also
quality issues for Classical shadow buffers are going to be reviewed,
especially to solve the lousy Biasing.
For the CVS log record; it is based on articles:
Gregory Johnson et al, University of Texas, Austin. (Regular grid method).
Timo Aila and Samuli Laine, Helsinki University of Technology. (BSP method).
Auto-clip Lamp Buffer
Setting a good range for clipping start/end for Lamp buffers is important
for good quality shadow (depth details). That can be quite a hassle, when
many lamps or animated objects are involved.
This option allows to have the clipping range calculated based on the
visible vertices in the spotbundle. For clip start and clip end it can
be set individually. Typically the clip start defines quality most.
The shadow buffer 'bias' value is corrected for this automatic clipping
too, to ensure that ranges that differ give same biasing appearance.
(If this wasn't done, you would see lighted areas become darker or
lighter, or even artefacts, when the range changes suddenly
NOTE: since it only checks for vertices, be aware that large planes can
give errors.
Implementation note: zbuffer values are non-linear (1/z) because that
allows linear interpolation while filling scanlines. That's the main
reason for the precision to be highest close to the eye (or lamp).
It's even a useful feature, since you want details to be precise
when they're closeby.
Since these values are also in the -1 to 1 range, using floats here
wouldn't solve problems a lot. Maybe trying a 64 bits Z once might
be an interesting coding job.
images (like used for rendering icons).
When working during Orange on new render pipeline, I've left this topic
alone for a while... subpixel precision testing is very time consuming and
needs concentration for a while. :)
This commit brings back precision as it was for 2.41. Below a short
explanation of the solved issues.
- the window matrix for rendering is kept constant during all OSA passes,
this to ensure clipping happens for each pass identically.
- a subpixel offset is only applied on filling in the z-buffer
- this offset is inverse corrected for shadepixel code, only on 2 places
Another nasty issue is that for filtered rendering (gauss etc), the tiles
(or entire image) is temporally increased 2 pixel in size. This caused a
'dark' (or sky color) edge on the rendering. During Orange that was solved
with a hardcoded clipping offset value, which only corrected for larger
pictures (like > 500 pixels in size).
Now this clipping offset is correctly calculated, based on render size.
Last issue: the view border in 3d window was calculated using integers,
giving small errors in display too. Now it uses float, so visually the
view border is more close to what a render shows.
- thread render stores per thread and per lamp the last intersected
shadow face
- the Isect struct, for intersections, could use a couple of extra
zeros on start.
Thread render error: a flag was stored in read-only data to indicate
whether top or bottom hemisphere of sky was rendered. That can't work
with tiles (but did work when scanlines were threads).
Code to allow "Env" material to mask out ztransp gave bad AA on edges of
solid faces, when transparant was behind it.
Recoded "Env" to use index -1 in the polygon index buffer, and restored
code that caused the bad AA.
"Backbuf" image render option is back :)
Nicer coded, using Image texture functions.
If you want 100% reliable pixel to pixel accuracy you have to use
compositing. (Old Backbuf didn't do this accuracy either btw).
This commit brings back:
- Field Render
- MBlur Render (old style)
- Border render with or without cropping
Note: Field Render is not supported in Compositor yet. Blurring or filter
will destroy field information.
Both MotionBlur as Field render are done before Compositing happens.
Fixes:
- The "Save Buffers" option only worked on single frame renders, not for
Anim render.
- Found an un-initalized variable in Render initialize... this might have
caused the unknown random crashes with render.
Code restructure:
Cleaned up names and calls throughout the pipeline, more clearly telling
what goes on in functions.
This is visible in the updated first image of the Wiki doc:
http://mediawiki.blender.org/index.php/BlenderDev/RenderPipeline
newline missing at end of file fixes:
TypedConstraint.h WheelInfo.h RaycastVehicle.h VehicleRaycaster.h
CcdPhysicsEnvironment.cpp
radiance_hdr.c fixed the following warning by changing type of local variable:
radiance_hdr.c:357: warning: pointer targets in passing argument 3 of
‘fwritecol rs’ differ in signedness
edgeRender.c,edgeRender.h same thing changed type of local vars to get rid
of warnings about signedness
RAS_OpenGLRasterizer.cpp removed unused variable unit line 1295
Kent
- added a zoom factor for Planar Environment Map, allowing to zoom in or
out on the reflected object. The zoom is corrected when rendering, so it
just gives control over the resolution.
- Planar maps don't cycle anymore, but only render the front plane, using
standard image texture extend mode.
- threaded rendering is back in envmap render
