Hi Jerome, you are right, there are memory restrictions. Changing the size of the arrays and everything works fine.
Thanks a lot, Lars Am Freitag, den 28.11.2014, 21:59 +0100 schrieb Jerome Kieffer: > On Fri, 28 Nov 2014 19:06:19 +0100 > Lars Callenbach <[email protected]> wrote: > > > > Everything works fine for AMD (CPU) but for NVIDIA I get a compilation > > Apparently it is not at compilation but at execution. You are probably > going beyond the limits of the hardware. How many register do you use ? > how many workgroup do you launch ? > > Since Nvidia dropped the support for their profiler (2 years ago), it > is harder not to get actual indication on that. > > Is you kernel complicated ? > > Cheers, _______________________________________________ PyOpenCL mailing list [email protected] http://lists.tiker.net/listinfo/pyopencl
