Hi.
clGetDeviceInfo call with parameter CL_DEVICE_MAX_WORK_ITEM_SIZES returns:
1024 * 1024 * 1024 on the CPU;
256 * 256 * 256 on the Radeon;
1024 * 1024 * 64 on the GeForce;
Can I run more than 1024 1D-kernels as follows?
size_t global[] = { 1000000 };
clEnqueueNDRangeKernel ( cmd, kernel, 1, NULL, global, NULL, 0, NULL, NULL );
And… how to calculate the limit of 1D-kernels?
Just multiply the available dimension?
102410241024 = 1 073 741 824
256256256 = 16 777 216
1024102464 = 67 108 864
Is that correct?