Hi,
I am running the opencv-cl performance test suite, for some tests assertions is coming because of hard coded local threads. the condition is localthread[0]localthread[1]localthread[2] should be less than or equal to cl_kernel_work_group_size. As mentioned in opencl spec, the cl_kernel_work_group_size is the maximum work_group_size that can be used to execute kernel on a specific device. but here the local threads are hard coded so am getting multiplication of localthreads as 256, and am getting cl_kernel_work_group_size for some functions less than 256, (ie., 128, 64 etc.,). can any one tell me, that whether my understanding is correct or not, if possible can suggest me the solution to solve this issue.
Thanks, Shabuddin.