Gentoo Archives: gentoo-amd64

From: Matt Randolph <mattr@×××××.com>
To: gentoo-amd64@l.g.o
Subject: Re: [gentoo-amd64] Re: x86_64 optimization patches for glibc.
Date: Wed, 27 Jul 2005 02:54:25
Message-Id: 42E6F684.6040009@erols.com
In Reply to: [gentoo-amd64] Re: x86_64 optimization patches for glibc. by Duncan <1i5t5.duncan@cox.net>
1 This is interesting. Guess what effect cpu frequency scaling has on
2 memcpy. I wonder if the difference has to do purely with the cpu
3 frequency, or if it's because the memory controller in on the die.
4
5 At 1800 MHz:
6 $ ./memcpy 1808 1000 1048576
7 Memory to memory copy rate = 1060.210815 MBytes / sec. Block size = 1048576.
8
9 At 1000 MHz:
10 $ ./memcpy 1005 1000 1048576
11 Memory to memory copy rate = 808.985352 MBytes / sec. Block size = 1048576.
12
13 This is an unpatched glibc, of course.
14
15 I've got a single 512MB stick of premium Corsair ram in it (the sister
16 stick lives in another computer until prices drop, at which time I will
17 buy another matched pair of the good stuff).
18
19 The fact that I'm using only one stick and I'm getting speeds similar to
20 those using two, suggests to me that the strangely low numbers some
21 people are getting are not explained away by their using only a single
22 channel.
23
24 --
25 "Pluralitas non est ponenda sine necessitate" - W. of O.
26
27 --
28 gentoo-amd64@g.o mailing list