Re: [PATCH, PR 49089] Don't split AVX256 unaligned loads by default on bdver1 and generic

["H.J. Lu" <hjl.tools@gmail.com>]

On Tue, Jun 14, 2011 at 3:41 PM, Fang, Changpeng <Changpeng.Fang@amd.com> wrote:
>>It probably should go to the 4.6 branch as well.
>
> H.J. Lu's original patch that splits unaligned load and store was checked in gcc 4.7
> trunk. We found that,  splitting unaligned store is beneficial to bdver1, splitting unaligned
> load degrades cfp2006 by 1.3% in geomean on Bulldozer. As a result, we suggest putting
> unaligned store splitting (H.J. original patch + this one) back to 4.6 branch.
>
>  >Note that I find the
>>X86_TUNE_AVX256_SPLIT_UNALIGNED_LOAD_OPTIMAL odd,
>>why not call it simply X86_TUNE_AVX256_SPLIT_UNALIGNED_LOAD?
>
> AVX256_SPLIT_UNALIGNED_LOAD has already  been used for the flag:
> -mavx256-split-unaligned-load, and we intend to keep that flag for performance tuning.
> As a result, we put _OPTIMAL (or _BENEFICAL) at the end for default setting.
>
>
>>I'll defer to x86 maintainers for approval.
>
> So, is it OK to commit this patch to trunk, and H.J's original patch + this to 4.6 branch?
>
>

I have no problems on -mtune=Bulldozer.  But I object -mtune=generic
change and did suggest a different approach for -mtune=generic.

-- 
H.J.