This is the mail archive of the
mailing list for the GCC project.
Re: Fwd: x86 issue: unwelcome mmx instruction
- From: Andrew Pinski <pinskia at physics dot uc dot edu>
- To: stuart at apple dot com (Stuart Hastings)
- Cc: gcc-patches at gcc dot gnu dot org (GCC Patches)
- Date: Fri, 18 Aug 2006 17:24:58 -0400 (EDT)
- Subject: Re: Fwd: x86 issue: unwelcome mmx instruction
> This x86-specific testcase provokes an MMX instruction in the absence
> of any MMX builtins (grep the assembly for "%mm"). The trivial patch
> "fixes" the problem, but the original code very deliberately checked
> TARGET_SSE, and I don't understand why.
> In other words, I don't claim I've found the "correct fix" here.
> Apple (and many others, I expect) wants to avoid random MMX
> generation. What should we do?
This is the same fix I came up with after Uros showed this problem in PR 24073 (which I filed).
Can you look at the other bug reports related to vector generation before posting