MPC5200B memory performance

Daniel Schnell daniel.schnell at marel.com
Fri May 18 20:27:43 EST 2007


Hi,

I found the following link to get more insights about the memcpy performance of my board:

http://www.greyhound-data.com/gunnar/glibc/index.htm
 
One can download a testsuite for memory performance and let it run. Not only memcpy, but also memcmp and memset is tested.

This gave quite interesting results. Most of the time I get better values from optimized functions than from the glibc functions. As a rule of thumb: depending on the buffer size used, 15-20 % speedup, sometimes > 100% !

I will definetly integrate some of these functions in our software and then use 
-Wl,--wrap,memcpy on the linker line to replace all occurences of memcpy, memcmp, etc. also for external libraries.


Best regards,

Daniel Schnell

-- 
Daniel Schnell                   | daniel.schnell at marel.com
Hugbúnaðargerð                   | www.marel.com

-----Original Message-----
From: linuxppc-embedded-bounces+daniel.schnell=marel.com at ozlabs.org [mailto:linuxppc-embedded-bounces+daniel.schnell=marel.com at ozlabs.org] On Behalf Of Matthias Fechner
Sent: 16. maí 2007 18:03
To: linuxppc-embedded at ozlabs.org
Subject: Re: MPC5200B memory performance

Hello Daniel,

* Daniel Schnell <daniel.schnell at marel.com> [15-05-07 11:22]:
> I get the following values on my board:

my result is:
Test (10000) memcpy of sizes (1024) ....
10000 memcpy. Time per memcpy: 1814 [nsec] (564 MB/sec)  finished.
Test (10000) memcpy of sizes (2048) ....
10000 memcpy. Time per memcpy: 3433 [nsec] (596 MB/sec)  finished.
Test (10000) memcpy of sizes (4096) ....
10000 memcpy. Time per memcpy: 6687 [nsec] (612 MB/sec)  finished.
Test (10000) memcpy of sizes (8192) ....
10000 memcpy. Time per memcpy: 21454 [nsec] (381 MB/sec)  finished.
Test (10000) memcpy of sizes (16384) ....
10000 memcpy. Time per memcpy: 205551 [nsec] (79 MB/sec)  finished.
Test (1000) memcpy of sizes (32768) ....
1000 memcpy. Time per memcpy: 379875 [nsec] (86 MB/sec)  finished.
Test (1000) memcpy of sizes (51200) ....
1000 memcpy. Time per memcpy: 588792 [nsec] (86 MB/sec)  finished.
Test (1000) memcpy of sizes (102400) ....
1000 memcpy. Time per memcpy: 1126511 [nsec] (90 MB/sec)  finished.
Test (100) memcpy of sizes (1048576) ....
100 memcpy. Time per memcpy: 11307890 [nsec] (92 MB/sec)  finished.
Test (10) memcpy of sizes (10485760) ....
10 memcpy. Time per memcpy: 120783600 [nsec] (86 MB/sec)  finished.
Test (5) memcpy of sizes (52428800) ....
5 memcpy. Time per memcpy: 673867800 [nsec] (77 MB/sec)  finished.
	   


Best regards,
Matthias

-- 

"Programming today is a race between software engineers striving to build bigger and better idiot-proof programs, and the universe trying to produce bigger and better idiots. So far, the universe is winning." -- Rich Cook _______________________________________________
Linuxppc-embedded mailing list
Linuxppc-embedded at ozlabs.org
https://ozlabs.org/mailman/listinfo/linuxppc-embedded



More information about the Linuxppc-embedded mailing list