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: Correct vect_cmdline_needed for x86 -m64


On Wed, Oct 04, 2006 at 08:25:18PM +0000, Joseph S. Myers wrote:
> The FAILs
> 
> FAIL: gcc.dg/tree-ssa/gen-vect-11b.c scan-tree-dump-times vectorized 0 loops 1
> FAIL: gcc.dg/tree-ssa/gen-vect-11c.c scan-tree-dump-times vectorized 0 loops 1
> 
> appear on i?86-*-* targets using -m64, but not on x86_64-*-* targets,
> because the tests use { target vect_cmdline_needed } and
> check_effective_target_vect_cmdline_needed fails to handle those two
> cases the same.  This patch fixes this, by handling both targets the
> same and checking for lp64.  OK to commit (to mainline and 4.1
> branch)?
> 
> 2006-10-04  Joseph Myers  <joseph@codesourcery.com>
> 
> 	* lib/target-supports.exp
> 	(check_effective_target_vect_cmdline_needed): Handle i?86-*-* the
> 	same as x86_64-*-*; check for LP64.

No, more is needed here because the result is cached.  The comment above
the proc says that the result won't change for different subtargets, but
you're saying that the result is different for -m32 than for -m64.  See,
for example, check_effective_target_fortran_large_real, which caches a
result for a particular target name.

Janis


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