This is the mail archive of the
libc-alpha@sourceware.org
mailing list for the glibc project.
Re: [PATCH] PowerPC: memmove default implementation cleanup
- From: Adhemerval Zanella <azanella at linux dot vnet dot ibm dot com>
- To: OndÅej BÃlka <neleai at seznam dot cz>
- Cc: "GNU C. Library" <libc-alpha at sourceware dot org>
- Date: Tue, 24 Jun 2014 15:35:06 -0300
- Subject: Re: [PATCH] PowerPC: memmove default implementation cleanup
- Authentication-results: sourceware.org; auth=none
- References: <53A9896A dot 3090503 at linux dot vnet dot ibm dot com> <20140624161141 dot GA25996 at domone dot podge>
On 24-06-2014 13:11, OndÅej BÃlka wrote:
> On Tue, Jun 24, 2014 at 11:21:30AM -0300, Adhemerval Zanella wrote:
>> This patch removes the powerpc specific logic in memmove and instead
>> include default implementation with MEMCPY_OK_FOR_FWD_MEMMOVE defined.
>> This lead in a increase performance, since the constraints to use
>> memcpy in powerpc code are too restrictive and memcpy can be used for
>> any forward memmove (and the rest to PPC implementation is just a copy
>> of default one).
>>
>> Checked on powerpc32-fpu and powerpc64be. OK to apply?
>>
> That looks ok, I would go step more and say that aliasing memcpy to
> memmove will differ only by few cycles and could improve performance
> due of cache locality but I did not do test yet.
>
Thanks for review. Now with optimized power7, I think I can check if alias
shows a performance boost (although such few cycles will be hard to evaluate
due os jitter). However, I don't think we can enable now for all power
implementations (since for power4, power6, etc. that have optimized
memcpy, this will mean another branch).