- Use BGL buffer instead of string for image data.
- Add buffer interface to image source.
- Allow customization of pixel format.
- Add valid property to check if the image data is available.
The image property of all Image source objects will now
return a BGL 'buffer' object. Previously it was returning
a string, which was not working at all with Python 3.1.
The BGL buffer type allows sequence access to bytes and
is directly usable in BGL OpenGL wrapper functions.
The buffer is formated as a 1 dimensional array of bytes
with 4 bytes per pixel in RGBA order.
BGL buffers will also be accepted in the ImageBuff load()
and plot() functions.
It is possible to customize the pixel format by using
the VideoTexture.imageToArray(image, mode) function:
the first argument is a Image source object, the second
optional argument is a format string using the R, G, B,
A, 0 and 1 characters. For example "BGR" means that each
pixel will be 3 bytes, corresponding to the Blue, Green
and Red channel in that order. Use 0 for a fixed hex 00
value, 1 for hex FF. The default mode is "RGBA".
All Image source objects now support the buffer interface
which allows to create memoryview objects for direct access
to the image internal buffer without memory copy. The buffer
format is one dimensional array of bytes with 4 bytes per
pixel in RGBA order. The buffer is writable, which allows
custom modifications of the image data.
v = memoryview(source)
A bug in the Python 3.1 buffer API will cause a crash if
the memoryview object cannot be created. Therefore, you
must always check first that an image data is available
before creating a memoryview object. Use the new valid
attribute for that:
if source.valid:
v = memoryview(source)
...
Note: the BGL buffer object itself does not yet support
the buffer interface.
Note: the valid attribute makes sense only if you use
image source in conjunction with texture object like this:
# refresh texture but keep image data in memory
texture.refresh(False)
if texture.source.valid:
v = memoryview(texture.source)
# process image
...
# invalidate image for next texture refresh
texture.source.refresh()
Limitation: While memoryview objects exist, the image cannot be
resized. Resizing occurs with ImageViewport objects when the
viewport size is changed or with ImageFFmpeg when a new image
is reloaded for example. Any attempt to resize will cause a
runtime error. Delete the memoryview objects is you want to
resize an image source object.
Add optional parameter to VideoTexture.Texture refresh() method
to specify timestamp (in seconds from start of movie) of the frame
to be loaded. This value is passed down to image source and for
VideoFFmpeg source, it is used instead of current time to load
the frame from the video file.
When combined with an audio actuator, it can be used to synchronize
the sound and the image: specify the same video file in the sound
actuator and use the KX_SoundActuator time attribute as timestamp
to refresh: the frame corresponding to the sound will be loaded:
GameLogic.video.refresh(True, soundAct.time)
Synopsis: plot(brush,width,height,x,y,mode)
plot(imgbuff,x,y,mode)
The first form uses a byte array containing the brush shape.
The second form uses another ImageBuff object as a brush.
The ImageBuff object must be initialized before you can call
these methods. Use load(rgb_buffer,sizex,sizey) method to create
an image buffer of given size (with alpha channel set to 255).
The brush is plotted directly in the image buffer. The texture
is updated only when the VideoTexture.Texture parent object is
refreshed: this will download the image buffer to the GPU.
brush: Byte array containing RGBA data to be plotted in image buffer.
The data must be continuous in memory, organized row by row
starting from lower left corner of the image. Each pixel is
4 bytes representing RGBA data in that order.
width: Horizontal size in pixels of image in brush.
height: Vertical size in pixels of the image in brush.
imgbuff:Another ImageBuff object that is used as a brush. The object
must have been initialized first with load().
x: Horizontal position in pixel from left side of the image buffer
where the brush will be plotted. The brush is plotted on pixels
positions x->x+width-1. Clipping is performed if the brush falls
partially outside the image buffer.
y: Vertical position in pixel from bottom side of the image buffer
where the brush will be plotted.
mode: Mode of drawing. Use one of the following value:
0 : MIX
1 : ADD
2 : SUB
3 : MUL
4 : LIGHTEN
5 : DARKEN
6 : ERASE ALPHA
7 : ADD ALPHA
1000 : COPY RGBA (default)
1001 : COPY RGB
1002 : COPY ALPHA
Modes 0 to 7 are 'blend' modes: the brush pixels are combined
with the image pixel in various ways. Refer to Blender documentation
to learn more about these modes.
* Convert all code to use new functions.
* Branch maintainers may want to skip this commit, and run this
conversion script instead, if they use a lot of math functions
in new code:
http://www.pasteall.org/9052/python
* bring back 'player' libtype, after investigation with ideasman.
scons/mingw works nicely, for some reason msvc fails to link still, will look further into it.
* further cleaning of 'player' stuff. Now only 3 libs are remaining, of which ideally the stubs lib will be fixed at some point, fading away into the dark history of not-so-nice code. The current blenderplayer part is still a little bit hackish, I'll see if I can find a better alternative, for now it works good enough.
* first working changes to get blenderplayer linking
* blenderplayer/ moved into source/ (CMakeLists.txt changed for that too)
* added externs for bprogname to gp_ghost, so that it links properly
* Add BGE_CXXFLAGS so we can get rid of hard-coded BGE compiler settings. This was only done for windows, but now linuxers and osxers should be able to set BGE-specific optimisation too. See the windows default configs for example.
* ensure all SConscripts are ready for win64-vc (where necessary).
* ensure we have proper _DEBUG flag for Python when we're doing a debug build.
* some cleaning up of linking etc.
* ensure /EHsc is there for game engine modules.
* Removed modules Expression and CValue, neither were ever available.
* Added GameLogic.EvalExpression(exp) from the Expression module, evaluates an expression like the expression controller (not sure if this is really that useful since python is far more advanced).
* resetting the original blend file path didint work (own fault == -> =)
* Py3.x PyModule_Create didnt allow importing since it didn't add to sys.modules,
Looks like they want us to use init-tab array, but this doesn't suit us since
it needs to be setup before python is initialized.
* Documented GameLogic.globalDict
svn merge https://svn.blender.org/svnroot/bf-blender/trunk/blender -r19820:HEAD
Notes:
* Game and sequencer RNA, and sequencer header are now out of date
a bit after changes in trunk.
* I didn't know how to port these bugfixes, most likely they are
not needed anymore.
* Fix "duplicate strip" always increase the user count for ipo.
* IPO pinning on sequencer strips was lost during Undo.
- Removed/Commented some unused vars
- CValue::GetPropertyText() could return a temp reference to a variable on the stack, option wasnt used anywhere so removed.
- KX_ConstraintWrapper::GetConstraintId allows args but ignored them
- KX_ConstraintWrapper::PySetParam didnt return NULL on an error (messing up pythons exceptions).
- BLI_natstrcmp didnt return 0 when the while loop exited
SCA_RandomActuator: The random generator was shared between replicas and not deleted. Added ref counting between replicas to allow deletion at the end.
KX_Camera: The scenegraph node was not deleted for temporary cameras (ImageMirror and shadow), causing 500 bytes leak per frame and per shadow light.
KX_GameActuator: Global dictionary buffer was not deleted after saving.
KX_MotionState: The motion state for compound child was not deleted
KX_ReplaceMeshActuator: The mesh was unnecessarily converted for each actuator and not deleted, causing large memleak.
After these fix, YoFrankie runs without memleak.
This commit extend the technique of dynamic linked list to the mesh
slots so as to eliminate dumb scan or map lookup. It provides massive
performance improvement in the culling and in the rasterizer when
the majority of objects are static.
Other improvements:
- Compute the opengl matrix only for objects that are visible.
- Simplify hash function for GEN_HasedPtr
- Scan light list instead of general object list to render shadows
- Remove redundant opengl calls to set specularity, shinyness and diffuse
between each mesh slots.
- Cache GPU material to avoid frequent call to GPU_material_from_blender
- Only set once the fixed elements of mesh slot
- Use more inline function
The following table shows the performance increase between 2.48, 1st round
and this round of improvement. The test was done with a scene containing
40000 objects, of which 1000 are in the view frustrum approximately. The
object are simple textured cube to make sure the GPU is not the bottleneck.
As some of the rasterizer processing time has moved under culling, I present
the sum of scenegraph(includes culling)+rasterizer time
Scenegraph+rasterizer(ms) 2.48 1st round 3rd round
All objects static, 323.0 86.0 7.2
all visible, 1000 in
the view frustrum
All objects static, 219.0 49.7 N/A(*)
all invisible.
All objects moving, 323.0 105.6 34.7
all visible, 1000 in
the view frustrum
Scene destruction 40min 40min 4s
(*) : this time is not representative because the frame rate was at 60fps.
In that case, the GPU holds down the GE by frame sync. By design, the
overhead of the rasterizer is 0 when the the objects are invisible.
This table shows a global speed up between 9x and 45x compared to 2.48a
for scenegraph, culling and rasterizer overhead. The speed up goes much
higher when objects are invisible.
An additional 2-4x speed up is possible in the scenegraph by upgrading
the Moto library to use Eigen2 BLAS library instead of C++ classes but
the scenegraph is already so fast that it is not a priority right now.
Next speed up in logic: many things to do there...
I added module clearing before there was checks for invalid python objects, so now its not needed for BGE Builtin types at least.
also made the builtin modules get re-used if they already exist and clear all user modules when the game engine finishes so with Module-Py-Controllers the referenced modules are at least up to date when pressing Pkey.
- More verbose error messages.
- BL_Shader wasnt setting error messages on some errors
- FilterNormal depth attribute was checking for float which is bad because scripts often expect ints assigned to float attributes.
- Added a check to PyVecTo for a tuple rather then always using a generic python sequence. On my system this is over 2x faster with an optmized build.
This changes how the BGE classes and Python work together, which hasnt changed since blender went opensource.
The main difference is PyObjectPlus - the base class for most game engine classes, no longer inherit from PyObject, and cannot be cast to a PyObject.
This has the advantage that the BGE does not have to keep 2 reference counts valid for C++ and Python.
Previously C++ classes would never be freed while python held a reference, however this reference could be problematic eg: a GameObject that isnt in a scene anymore should not be used by python, doing so could even crash blender in some cases.
Instead PyObjectPlus has a member "PyObject *m_proxy" which is lazily initialized when python needs it. m_proxy reference counts are managed by python, though it should never be freed while the C++ class exists since it holds a reference to avoid making and freeing it all the time.
When the C++ class is free'd it sets the m_proxy reference to NULL, If python accesses this variable it will raise a RuntimeError, (check the isValid attribute to see if its valid without raising an error).
- This replaces the m_zombie bool and IsZombie() tests added recently.
In python return values that used to be..
return value->AddRef();
Are now
return value->GetProxy();
or...
return value->NewProxy(true); // true means python owns this C++ value which will be deleted when the PyObject is freed
Other small changes...
- KX_Camera and KX_Light didnt have get/setitem access in their PyType definition.
- CList.from_id() error checking for a long was checking for -1 against an unsigned value (own fault)
- CValue::SpecialRelease was incrementing an int for no reason.
- renamed m_attrlist to m_attr_dict since its a PyDict type.
- removed custom getattro/setattro functions for KX_Scene and KX_GameObject, use py_base_getattro, py_base_setattro for all subclasses of PyObjectPlus.
- lowercase windows.h in VideoBase.cpp for cross compiling.
Added occlusion culling capability in the BGE.
More info: http://wiki.blender.org/index.php/Dev:Ref/Release_Notes/2.49/Game_Engine#BGE_Scenegraph_improvement
MSVC, scons, cmake, Makefile updated.
Other minor performance improvements:
- The rasterizer was computing the openGL model matrix of the objects too many times
- DBVT view frustrum culling was not properly culling behind the near plane:
Large objects behind the camera were sent to the GPU
- Remove all references to mesh split/join feature as it is not yet functional
Notes:
* Sequence transform strip uses G.scene global, this is commented
out now, should be fixed.
* Etch-a-ton code was most difficult to merge. The files already in
2.5 got merged, but no new files were added. Calls to these files
are commented out with "XXX etch-a-ton". editarmature.c and
transform_snap.c were complex to merge. Martin, please check?
* Game engine compiles and links again here for scons/make/cmake
(player still fails to link).
The multi-thread cache service is activated only on multi-core processors.
It consists in loading, decoding and caching the video frames in a
separate thread. The cache size is 5 decoded frames and 30 raw frames.
Note that the opening of video file/stream/camera is not multi-thread:
you will still experience a delay at the VideoFFmpeg object creation.
Processing of the video frame (resize, loading to texture) is still done
in the main thread. Caching is automatically enabled for video file,
video streaming and video camera.
Video streaming now works correctly: the videos frames are loaded
at the correct rate. Network delays and frequency drifts are automatically
compensated.
Note: an http video source is always treated as a streaming source,
even though the http protocol allows seeking. For the user it means that
he cannot define start/stop range and cannot restart the video except
by reopening the source. Pause/play is however possible.
Video camera is now correctly handled on Linux: it will not slow down the BGE.
A video camera is treated as a streaming source.
- variables that shadow vers declared earlier
- Py_Fatal print an error to the stderr
- gcc was complaining about the order of initialized vars (for classes)
- const return values for ints and bools didnt do anything.
- braces for ambiguous if statements
* giving compileflags, cc_compileflags and cxx_compileflags to BlenderLib() now actually overrides any other setting (so there's no unclarity when ie. conflicting options are being specified in REL_CFLAGS et al). These are set after either release or debug flags, but before any *_WARN flags (so those stay maintained).
* add cxx_compileflags for GE parts on win32-vc to have better performance.
* NOTE: if platform maintainers (OSX and Linux) could check and do the same for their systems. Not vital, but probably very, very much welcomed by GE users.
