CUDA与OpenGL交互问题，求救

siheng303

貌似必须进行拷贝，不能直接把RGB数据绑定给一个二维纹理

qet123

貌似这个问题很普遍啊，同求解答

swordwest

答案保密？直接贴在后面可否？

huawenguang

你如果用的PBO,应该要有一个把数据从PBO拷贝到纹理的一个过程吧。

1. // "index" is used to copy pixels from a PBO to a texture object
   
2. // "nextIndex" is used to update pixels in the other PBO
   
3. index = (index + 1) % 2;
   
4. if(pboMode == 1)                // with 1 PBO
   
5.     nextIndex = index;
   
6. else if(pboMode == 2)           // with 2 PBOs
   
7.     nextIndex = (index + 1) % 2;
   
8. 
   
9. // bind the texture and PBO
   
10. glBindTexture(GL_TEXTURE_2D, textureId);
    
11. glBindBufferARB(GL_PIXEL_UNPACK_BUFFER_ARB, pboIds[index]);
    
12. 
    
13. // copy pixels from PBO to texture object
    
14. // Use offset instead of ponter.
    
15. glTexSubImage2D(GL_TEXTURE_2D, 0, 0, 0, WIDTH, HEIGHT,
    
16.                 GL_BGRA, GL_UNSIGNED_BYTE, 0);
    
17. 
    
18. 
    
19. // bind PBO to update texture source
    
20. glBindBufferARB(GL_PIXEL_UNPACK_BUFFER_ARB, pboIds[nextIndex]);
    
21. 
    
22. // Note that glMapBufferARB() causes sync issue.
    
23. // If GPU is working with this buffer, glMapBufferARB() will wait(stall)
    
24. // until GPU to finish its job. To avoid waiting (idle), you can call
    
25. // first glBufferDataARB() with NULL pointer before glMapBufferARB().
    
26. // If you do that, the previous data in PBO will be discarded and
    
27. // glMapBufferARB() returns a new allocated pointer immediately
    
28. // even if GPU is still working with the previous data.
    
29. glBufferDataARB(GL_PIXEL_UNPACK_BUFFER_ARB, DATA_SIZE, 0, GL_STREAM_DRAW_ARB);
    
30. 
    
31. // map the buffer object into client's memory
    
32. GLubyte* ptr = (GLubyte*)glMapBufferARB(GL_PIXEL_UNPACK_BUFFER_ARB,
    
33.                                         GL_WRITE_ONLY_ARB);
    
34. if(ptr)
    
35. {
    
36.     // update data directly on the mapped buffer
    
37.     updatePixels(ptr, DATA_SIZE);
    
38.     glUnmapBufferARB(GL_PIXEL_UNPACK_BUFFER_ARB); // release the mapped buffer
    
39. }
    
40. 
    
41. // it is good idea to release PBOs with ID 0 after use.
    
42. // Once bound with 0, all pixel operations are back to normal ways.
    
43. glBindBufferARB(GL_PIXEL_UNPACK_BUFFER_ARB, 0);
    

复制代码

huawenguang

我觉得可以尝试一个先把RGB数据拷贝到CUDA的数组：cudaMemcpy2DToArray()
然后把这个数组绑定到一个纹理：cudaBindTextureToArray()
这可能比较方便些。