On Fri, Nov 29, 2013 at 09:57:53PM -0500, Jan Vesely wrote:
> Hi,
> 
> these two patches attempt to improve the situation with store-double* 
> generated tests.
> The first patch adds cl_khr_fp64 requirement to all generated tests, the 
> second one adds necessary infrasttructure to 
> tests/cl/program/program_tester.c.
> 
> I'm not sure if I fully understand the TODO comment that was there.
> According to 
> http://www.khronos.org/registry/cl/sdk/1.2/docs/man/xhtml/clGetDeviceInfo.html
> cl_khr_fp64 must be returned by all OpenCL 1.2 devices. Checking for
> the presence of this extension should be enough, or am I missing something?
>

I think you are correct, the presence of the extension should be enough.

> These patches change 10 store-double* tests from warn to skipped on my amd 
> turks
> (and to passed on intel cpu implementation). Note that the compiler only 
> complains about missing extension for type double, vector types (doubleN) 
> pass.
> 

This seems like a bug to me, are the complaints coming from clang or
from headers in libclc?

Both patches are:

Reviewed-by: Tom Stellard <[email protected]>

-Tom

> regards,
> Jan
> 
> Jan Vesely (2):
>   cl: check for cl_khr_fp64 in generated double store tests
>   program-tester: Allow double type
> 
>  generated_tests/generate-cl-store-tests.py | 10 ++++++++--
>  tests/cl/program/program-tester.c          | 11 +++++++----
>  2 files changed, 15 insertions(+), 6 deletions(-)
> 
> -- 
> 1.8.3.1
> 
> _______________________________________________
> Piglit mailing list
> [email protected]
> http://lists.freedesktop.org/mailman/listinfo/piglit
_______________________________________________
Piglit mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/piglit

Reply via email to