Jerome Kieffer <Jerome.Kieffer@...> writes:
On Mon, 8 Jun 2015 12:37:47 +0000 (UTC)
Riccardo De Maria <riccardodemaria <at> gmail.com> wrote:
> I am just starting with opencl. I have notebook with an Intel HD BroadWell
> U-Processor GT2 on Intel Gen OCL driver on Linux.
> The following kernel get and return non-sense as input and output. Please
> note the cl_khr_fp64: enable. The code runs fine on nvidia tesla in another
> machine. Everything works fine with float32. Do you have any clue why fp64
> is broken? Is it a driver issue? How can I dig further?
The question should be "is fp64 supported" (by your hardware)?
I never used it on Beignet.
Thanks a lot for the hint.
From clinfo cl_khr_fp64 the extension is not listed. I now read here
Double/FP64 is still not natively supported in OpenCL under Broadwell which
is a disappointment but not a surprise; again Intel should provide their own
FP64 extension like AMD if they cannot get Khronos certification.
Double/FP64 is supported under DirectX thus some native FP64 is supported by
the hardware. """
The bottom line, is that pyopecl/beignet does not complain if the hardware
is not compatible with the code, right?