Blender Git Loki
Git Commits -> Revision 0ff79c8
Revision 0ff79c8 by Benoit Bolsee (decklink) December 20, 2015, 10:32 (GMT) |
Decklink: use floating point texture format when possible. Previous implementation was using GL_RED_INTEGER texture format for all non-RGBA pixel format. This isn't supported by intel GPU (although the GL_texture_rg extension is declared supported). Floating RGBA texture are now used for the following pixel formats: 8BitYUV: Cb->B, Y0->G, Cr->R, Y1->A 10BitYUV: no fixed mapping between CbCrY and RGB, a shader is required 8BitARGB: direct mapping 8BitBGRA: direct mapping 10BitRGBXLE: direct mapping (A is undefined) Other pixel formats are mapped to GL_RED_INTEGER (GL_R32UI internal format, usampler2D must be used in the shader). Note: the 10BitYUV, 10BitRGBXLE and 8BitARGB mapping only works on little- endian host CPU. For big endian CPU, other formats must be used (not yet implemented). The texture MIN/MAX_FILTER is now set to NEAREST. Previously it was set to LINEAR, which was detremental on multibytes pixel format. The sample shader in the documentation for the above formats will be changed to reflect the new mapping. |
Commit Details:
Full Hash: 0ff79c87a440d886cbaed3f2753a7a8cb466ad60
Parent Commit: 054d871
Lines Changed: +45, -14