Hi there, 牛延强 <[email protected]> writes: > Platform: AMD APU C60 > Once the local_size(0) bigger than 48, there are several results > error after evaluated multiple times. > > my english is too poor,I'm so sorry!
one thing to take into account is that older AMD GPUs/APUs do not support barrier() in a kernel if the work group size exceeds the wavefront size. (64) HTH, Andreas _______________________________________________ PyOpenCL mailing list [email protected] http://lists.tiker.net/listinfo/pyopencl
