Archives Discussions

[joeapisit]

Hi all,

I cannot use GPU device of A6-3650 to run double precision floating-point kernel code. I try to use

#pragma OPENCL EXTENSION cl_amd_fp64 : enable

and

#pragma OPENCL EXTENSION cl_khr_fp64 : enable

but it show me "build error" when compile kernel code

Please give me any advice. Thanks.

#pragma OPENCL EXTENSION cl_khr_fp64 : enable

[nou]

always check if device support such extension. for example in clinfo. A6-3650 AFAIK don't support double.

[Meteorhead]

Double-precision capable IGP-s are not likely to come even with Trinity, or not even with generation after that.

It would really rock, but unfortunately gaming industry is still the main drive of GPU development, and there nobody sees the unavoidable need of DP on a GPU, not to mention if it comes at the cost of heavy SP peroformance decrease.