2009-01-07, 21:15
malloc Wrote:Could just have GL render to texture of appropriate size and then read the texture to local memory. Of course then we just write it back to the card for VDPAU to use.
I'm not saying VDPAU should be dropped like a hot overloaded software decoding CPU trying to also do memory swaps....but there is another way that involve less invasive surgery on XBMC's OpenGL approach.
• OpenCL can efficiently share resources with OpenGL
- Textures, Buffer Objects and Renderbuffers
- Data is shared, not copied
- OpenCL objects are created from OpenGL objects
- clCreateFromGLBuffer(), clCreateFromGLTexture2D(), clCreateFromGLRenderbuffer()
OK forum moderators...you can say this is irrelevant and move this post (but why the insistence on pursuing a beta Nvidia technology that's so kludgey for XBMC?)