This is the mail archive of the gcc-patches@gcc.gnu.org mailing list for the GCC project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: [PATCH] Adjust gcc.target/powerpc/atomic-p{7,8}.c tests for PR59448 (PR testsuite/64930)


On Thu, Feb 12, 2015 at 2:51 AM, Jakub Jelinek <jakub@redhat.com> wrote:
> Hi!
>
> As PR59448 started promoting consume to acquire and as written by Torvald,
> that isn't going to change any time soon unless the standard is clarified,
> this patch adjusts the two PPC testcases for that.
>
> Ok for trunk?
>
> 2015-02-12  Jakub Jelinek  <jakub@redhat.com>
>
>         PR testsuite/64930
>         * gcc.target/powerpc/atomic-p7.c: Adjust expected count of isync
>         instructions for 2015-01-14 get_memmodel changes.
>         * gcc.target/powerpc/atomic-p8.c: Likewise.
>
> --- gcc/testsuite/gcc.target/powerpc/atomic-p7.c.jj     2014-11-11 00:05:43.000000000 +0100
> +++ gcc/testsuite/gcc.target/powerpc/atomic-p7.c        2015-02-11 18:36:48.208282824 +0100
> @@ -14,7 +14,9 @@
>  /* { dg-final { scan-assembler-times "stdcx" 6 } } */
>  /* { dg-final { scan-assembler-not "stqcx" } } */
>  /* { dg-final { scan-assembler-times "bl __atomic" 6 } } */
> -/* { dg-final { scan-assembler-times "isync" 12 } } */
> +/* As since PR59448 GCC promotes consume to acquire, the expected isync count
> +   is 16 rather than 12.  */
> +/* { dg-final { scan-assembler-times "isync" 16 } } */
>  /* { dg-final { scan-assembler-times "lwsync" 8 } } */
>  /* { dg-final { scan-assembler-not "mtvsrd" } } */
>  /* { dg-final { scan-assembler-not "mtvsrwa" } } */
> --- gcc/testsuite/gcc.target/powerpc/atomic-p8.c.jj     2014-11-11 00:05:43.000000000 +0100
> +++ gcc/testsuite/gcc.target/powerpc/atomic-p8.c        2015-02-11 18:37:00.992070778 +0100
> @@ -14,7 +14,9 @@
>  /* { dg-final { scan-assembler-times "stdcx" 7 } } */
>  /* { dg-final { scan-assembler-times "stqcx" 7 } } */
>  /* { dg-final { scan-assembler-not "bl __atomic" } } */
> -/* { dg-final { scan-assembler-times "isync" 20 } } */
> +/* As since PR59448 GCC promotes consume to acquire, the expected isync count
> +   is 25 rather than 20.  */
> +/* { dg-final { scan-assembler-times "isync" 25 } } */
>  /* { dg-final { scan-assembler-times "lwsync" 10 } } */
>  /* { dg-final { scan-assembler-not "mtvsrd" } } */
>  /* { dg-final { scan-assembler-not "mtvsrwa" } } */

Okay.

Thanks, David


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]