clEnqueueNDRangeKernel failure causes VPU recovery on next run

Discussion created by plap on Feb 13, 2010
Latest reply on Feb 15, 2010 by MicahVillmow

I just tried running my kernel with different local sizes and when I tried 8 by 4 by 4 (256), I received a CL_INVALID_WORK_GROUP_SIZE error by clEnqueueNDRangeKernel . I reduced to 8,2,4 and ran again and then VPU recovery was triggered within 4 seconds. I ran the exact same kernel/setup again and it completed without problems after 6 seconds.

Global work size was 48 by 48 by 48.