This is the mail archive of the
mailing list for the GCC project.
Re: recognize x86 CPU variants take II
- From: Jan Hubicka <jh at suse dot cz>
- To: Janis Johnson <janis187 at us dot ibm dot com>
- Cc: Jan Hubicka <jh at suse dot cz>, gcc-patches at gcc dot gnu dot org, rth at cygnus dot com, patches at x86-64 dot org
- Date: Wed, 12 Dec 2001 09:22:31 +0100
- Subject: Re: recognize x86 CPU variants take II
- References: <20011211124354.A29171@atrey.karlin.mff.cuni.cz> <20011211134945.A2758@us.ibm.com>
> On Tue, Dec 11, 2001 at 12:43:54PM +0100, Jan Hubicka wrote:
> > Tue Dec 11 11:23:36 CET 2001 Jan Hubicka <email@example.com>
> > * i386.c (override_options): Recognize various CPU variants and set
> > SSE/MMX/3dNOW flags accordingly.
> > * i386.h (MASK_NOMMX, MASK_NOSSE, MASK_NOSSE2, MASK_NO3DNOW,
> > MASK_NO3DNOW_A): New.
> > (MASK_*): Renumber.
> > (TARGET_FLAGS): Use new masks.
> > (CPP_CPU_SPECS): Recognize new CPU variants.
> > * invoke.texi (-mcpu): Update documentation.
> This change could use some tests so we'll know when it breaks and to
> ensure that you get what you expect to begin with.
> See http://gcc.gnu.org/ml/gcc-patches/2001-11/msg02128.html for an
> example of what I plan for testing prefetch support for x86 targets;
> similar tests could check that the expected type of fp instructions
> are being used, or that instructions from extension sets are only used
> for the appropriate sets of options.
I will try to prepare something next week. Actually the combinations are
quite tested, as we do use heavilly SSE for x86_64 work and opositely if
some SSE leaks to normal code, most of current machines will get testsuite
failures, but I agree that it makes sense to write some specific test.