Does anyone know how the “maximum texture image units” and “maximum texture coordinate sets” limits (MAX_TEXTURE_IMAGE_UNITS and MAX_TEXTURE_COORDS in OpenGL) affect CUDA programs using textures? The values are 32 and 8 for a 8800 GTX. From what I can gather, MAX_TEXTURE_IMAGE_UNITS limits the number of texture units you can have in a cubin file (seems reasonable), but I’m unsure about what MAX_TEXTURE_COORDS does.

Any help would be appreciated.


The hardware supports up to 128 bound textures, although I’m not sure if this many are supported by CUDA currently.

CUDA doesn’t have interpolators (or rasterization for that matter) so the number of texture coordinates is irrelevant.