Image memory, CL_RG & CL_FLOAT

Accelerated Computing CUDA CUDA Programming and Performance
1

I’m getting CL_IMAGE_FORMAT_NOT_SUPPORTED with the following code, but I can’t find anything wrong with it. CL_RGBA works for channel order, but CL_R doesn’t. Does this mean that OpenCL only supports 4 channel float images, and not 2 or 1 channel? This sounds a little silly, but it’s the best answer I can find.

Can someone confirm that I can’t use CL_RG & CL_FLOAT together in a 2D image? Otherwise, what could I be doing wrong here that would result in that error?

I’m running Mac 10.6.3 with a GTX 285 on a Mac Pro. My OpenCL code is running on the CPU here, though.

cl_image_format imgFmt;

	

	//Copy coord data to the device

	imgFmt.image_channel_order = CL_RG;

	imgFmt.image_channel_data_type = CL_FLOAT;

	(*latLong) = clCreateImage2D(warper->context,

								 CL_MEM_READ_ONLY | CL_MEM_COPY_HOST_PTR, &imgFmt,

								 (size_t) warper->srcWidth,

								 (size_t) warper->srcHeight,

								 (size_t) sizeof(float)*2*warper->srcWidth,

								 warper->latLongWork, &err);

	handleErr(err);
2

As it turns out, the format CL_RG & CL_FLOAT isn’t supported on either my CPU or GPU. Who knew? I had to write a function for finding the most appropriate channel order given the data type, and I’ve adjusted the rest of the code appropriately.