This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: Patch to refine PowerPC testsuite conditions
- From: Janis Johnson <janis187 at us dot ibm dot com>
- To: "Joseph S. Myers" <joseph at codesourcery dot com>
- Cc: gcc-patches at gcc dot gnu dot org
- Date: Thu, 11 May 2006 11:26:05 -0700
- Subject: Re: Patch to refine PowerPC testsuite conditions
- References: <Pine.LNX.4.64.0605111235240.2453@digraph.polyomino.org.uk>
On Thu, May 11, 2006 at 12:41:29PM +0000, Joseph S. Myers wrote:
> This patch cleans up the conditions under which various PowerPC tests
> are run.
>
> There is one test, gcc.target/powerpc/980827-1.c, which has an asm
> using floating-point registers and so cannot work with -msoft-float or
> on E500 targets. An effective-target keyword powerpc_fprs is added
> for this test.
>
> Many tests use -maltivec but vary in whether and how they restrict
> this to targets supporting -maltivec. (E500 targets cannot support
> AltiVec because the instructions conflict, and Aix does not support
> AltiVec.) Some tests are XFAILed on powerpc-*-eabispe*,
> powerpc-*-aix* or both, but this is not done consistently, the tests
> should be skipped rather than XFAILed, and targets can have both E500
> and non-E500 multilibs (so the skipping should depend on the
> properties of the multilib under test, not just on the target name
> being *-spe). Thus, an effective-target keyword powerpc_altivec_ok is
> added, and used for all such tests not explicitly specifying a
> particular CPU.
>
> The test gcc.dg/20020103-1.c fails on various PowerPC targets (bug
> 11594). As a known failure with a bug open, it should be XFAILed in
> order to reduce the number of unexpected FAILs appearing in test
> results. (Conversely, all XFAILs should have corresponding open bugs
> in Bugzilla.)
>
> Tested powerpc-none-eabispe and powerpc64-unknown-linux-gnu. OK to
> commit?
This is OK, although if it's not too painful, or if you disagree, there
are a couple of things that I would prefer to change. The first is that
powerpc_altivec_ok seems more verbose than necessary; just "altivec"
would be enough. The second is that instead of
/* { dg-do compile { target powerpc*-*-* } } */
/* { dg-require-effective-target powerpc_altivec_ok } */
you could use
/* { dg-do compile { target { powerpc*-*-* && powerpc_altivec_ok } } */
I don't feel strongly about these, so you can check it in as-is if you
prefer. Thanks for this big cleanup.
As for the ICEs for powerpc-none-eabispe, see if the patch in
http://gcc.gnu.org/ml/gcc-patches/2006-05/msg00429.html would be useful.
Janis