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

Fang, Changpeng Changpeng.Fang@amd.com
Fri Jun 17 22:49:00 GMT 2011


Hi, 

 I added AVX256_SPLIT_UNALIGNED_STORE to ix86_tune_indices
and put m_COREI7, m_BDVER1 and m_GENERIC as the targets that
enable it.

Is this OK?

Thanks,

Changpeng



________________________________________
From: H.J. Lu [hjl.tools@gmail.com]
Sent: Friday, June 17, 2011 1:08 PM
To: Fang, Changpeng
Cc: Richard Guenther; gcc-patches@gcc.gnu.org
Subject: Re: [PATCH, PR 49089] Don't split AVX256 unaligned loads by default on bdver1 and generic

On Fri, Jun 17, 2011 at 10:45 AM, Fang, Changpeng
<Changpeng.Fang@amd.com> wrote:
>>Why not just move AVX256_SPLIT_UNALIGNED_STORE
>>and AVX256_SPLIT_UNALIGNED_LOAD to ix86_tune_indices?
>
> I would like to keep the -m option so that at least we can explicitly turn
> off the splittings when regressions are found!

I prefer to implement it the same way as:

x86_accumulate_outgoing_args
x86_arch_always_fancy_math_387

> By the way, I can add an index for store splitting, if you want.
>

Yes, please.

--
H.J.

-------------- next part --------------
A non-text attachment was scrubbed...
Name: 0002-pr49089-enable-avx256-splitting-unaligned-load-store.patch
Type: text/x-patch
Size: 3496 bytes
Desc: 0002-pr49089-enable-avx256-splitting-unaligned-load-store.patch
URL: <http://gcc.gnu.org/pipermail/gcc-patches/attachments/20110617/5906a97a/attachment.bin>


More information about the Gcc-patches mailing list