On Fri, Apr 7, 2017 at 4:54 PM, Jakub Jelinek <ja...@redhat.com> wrote:
> Hi!
>
> I've noticed the avx512f-vgetmantpd-2.c testcase eats lots of
> CPU time on KNL with -m32 and FAILs. The problem is
> excess precision, I've added -mfpmath=sse to it to fix that.
>
> Tested with
> make -j272 -k check-gcc RUNTESTFLAGS='--target_board=unix\{-m32,-m64\}
> i386.exp'
> on KNL, ok for trunk?
>
> 2017-04-07 Jakub Jelinek <ja...@redhat.com>
>
> * gcc.target/i386/avx512f-vgetmantps-2.c: Add -mfpmath=sse to
> dg-options.
> * gcc.target/i386/avx512f-vgetmantpd-2.c: Likewise.
OK.
Thanks,
Uros.
> --- gcc/testsuite/gcc.target/i386/avx512f-vgetmantps-2.c.jj 2017-04-07
> 05:52:04.000000000 -0400
> +++ gcc/testsuite/gcc.target/i386/avx512f-vgetmantps-2.c 2017-04-07
> 09:22:13.051209011 -0400
> @@ -1,5 +1,5 @@
> /* { dg-do run } */
> -/* { dg-options "-O2 -mavx512f -std=c99" } */
> +/* { dg-options "-O2 -mavx512f -std=c99 -mfpmath=sse" } */
> /* { dg-require-effective-target avx512f } */
> /* { dg-require-effective-target c99_runtime } */
>
> --- gcc/testsuite/gcc.target/i386/avx512f-vgetmantpd-2.c.jj 2017-04-07
> 05:51:49.000000000 -0400
> +++ gcc/testsuite/gcc.target/i386/avx512f-vgetmantpd-2.c 2017-04-07
> 09:21:46.648317195 -0400
> @@ -1,5 +1,5 @@
> /* { dg-do run } */
> -/* { dg-options "-O2 -mavx512f -std=c99" } */
> +/* { dg-options "-O2 -mavx512f -std=c99 -mfpmath=sse" } */
> /* { dg-require-effective-target avx512f } */
> /* { dg-require-effective-target c99_runtime } */
>
>
> Jakub
