This is the mail archive of the libc-alpha@sourceware.org mailing list for the glibc project.

Index Nav:	[Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav:	[Date Prev] [Date Next]	[Thread Prev] [Thread Next]
Other format:	[Raw text]

Re: [PATCH] Optimize SSE 4.1 x86_64 memcmp

From: Florian Weimer <fweimer at redhat dot com>
To: "Joseph S. Myers" <joseph at codesourcery dot com>
Cc: GNU C Library <libc-alpha at sourceware dot org>
Date: Fri, 31 Jan 2014 22:24:14 +0100
Subject: Re: [PATCH] Optimize SSE 4.1 x86_64 memcmp
Authentication-results: sourceware.org; auth=none
References: <52EBBCC2 dot 7090807 at redhat dot com> <Pine dot LNX dot 4 dot 64 dot 1401311728250 dot 26476 at digraph dot polyomino dot org dot uk>

On 01/31/2014 06:34 PM, Joseph S. Myers wrote:

On Fri, 31 Jan 2014, Florian Weimer wrote:

In case BSWAP turns out to be too slow for some CPUs, it could be replaced
with BSF and a shift (both constant time in current processors), like we
already do in some of the existing memcmp implementations.


I wonder how the performance of MOVBE compares with the other options
(though availability of MOVBE is sufficiently limited it may not be worth
having another version of memcmp using it).

Interesting. It might allow to recover the one-cycle overhead in somecases, although the number of µops after decoding should remain thesame. But as you say, it's probably not worth maintaining yet anotherversion just for that.


--
Florian Weimer / Red Hat Product Security Team

References:
- [PATCH] Optimize SSE 4.1 x86_64 memcmp
  - From: Florian Weimer
- Re: [PATCH] Optimize SSE 4.1 x86_64 memcmp
  - From: Joseph S. Myers

Index Nav:	[Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav:	[Date Prev] [Date Next]	[Thread Prev] [Thread Next]