This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: PATCH: Add vzeroupper optimization for AVX
- From: "H.J. Lu" <hjl dot tools at gmail dot com>
- To: Andi Kleen <andi at firstfloor dot org>
- Cc: gcc-patches at gcc dot gnu dot org, Uros Bizjak <ubizjak at gmail dot com>, Richard Henderson <rth at redhat dot com>
- Date: Mon, 25 Oct 2010 10:03:59 -0700
- Subject: Re: PATCH: Add vzeroupper optimization for AVX
- References: <20101025085724.GA17893@intel.com> <87y69m4csn.fsf@basil.nowhere.org>
In Mon, Oct 25, 2010 at 6:33 AM, Andi Kleen <andi@firstfloor.org> wrote:
> "H.J. Lu" <hongjiu.lu@intel.com> writes:
>>
>> At RTL expansion time, the vzeroupper optimization generates a
>> vzeroupper_nop before function call and functin return if 256bit AVX
>> instructions are used. The vzeroupper pass is run before final pass.
>> It scans all reachable blocks:
>>
>> 1. Remove vzeroupper_nop when:
>> ? ? a. The upper 128bits of AVX regiters are known dead.
>
> Is dead the correct term here? It would seem to need "unused"
>
It has be to be cleared to zero, not unused.Maybe I should use zero.
--
H.J.