mersenneforum.org  

Go Back   mersenneforum.org > Great Internet Mersenne Prime Search > Software

Reply
 
Thread Tools
Old 2007-03-28, 21:17   #1
E_tron
 
E_tron's Avatar
 
Sep 2002
Austin, TX

3·11·17 Posts
Default Any room for optimization on Core 2 processors?

Will there be room for improved assembly code on Intel's Core 2 architecture? Prime95 v24.14 was release well ahead of the Core 2 extravaganza. Without code specifically optimized for Core 2, it crushes its netburst predecessor and leaves everything else in the dust:

4096K FFT length:
126.041 ms on Core 2 Duo @ 2.0ghz
126.115 ms on a P4(90nm) @ 3.4ghz

2048k FFT length:
48.920 ms. on Core 2 Duo @ 2.4ghz
52.848 ms on P4(90nm) @ 3.8ghz
50.557 ms on P4(130nm) @ 4.0ghz

Will we see optimizations that will leave netburst in the dust everything else in the dust’s dust?
E_tron is offline   Reply With Quote
Old 2007-03-28, 22:54   #2
retina
Undefined
 
retina's Avatar
 
"The unspeakable one"
Jun 2006
My evil lair

33·233 Posts
Default

Quote:
Originally Posted by E_tron
Will there be room for improved assembly code on Intel's Core 2 architecture? Prime95 v24.14 was release well ahead of the Core 2 extravaganza. Without code specifically optimized for Core 2, it crushes its netburst predecessor and leaves everything else in the dust
I'm no expert but: the original code was optimised for SSE2 on the P4. The C2 has much superior SSE2 performance over the P4. So I think that means, with some tweaks to make best use of the new FP pipelines, that we can see even more improvement over the iteration times you posted above. As to the actual amount of improvement I would speculate it would not be very high, perhaps not more than 10%. But that is only a guess so everyone please feel free to prove me wrong and get a better than 10% improvement.
retina is offline   Reply With Quote
Old 2007-03-29, 14:52   #3
dsouza123
 
dsouza123's Avatar
 
Sep 2002

2·331 Posts
Default

Unlikely, the gain has already been realized by switching to the new hardware !

As was shown a 4096K FFT run on a Core 2 Duo @ 2.0 ghz
equals a P4 @ 3.4 ghz.

Run the Core 2 Duo @ 3.4 ghz or the P4 @ 2.0 ghz
and the performance increase will be quite evident.
dsouza123 is offline   Reply With Quote
Old 2007-03-30, 06:06   #4
Ender
 
Ender's Avatar
 
Feb 2007
Romania

3 Posts
Default Both cores

Maybe the ability for mprime to use both cores on a C2D, running the same exponent, will be a nice feature in a future version?
Ender is offline   Reply With Quote
Old 2007-03-30, 06:59   #5
Andi47
 
Andi47's Avatar
 
Oct 2004
Austria

2·17·73 Posts
Default

Quote:
Originally Posted by Ender View Post
Maybe the ability for mprime to use both cores on a C2D, running the same exponent, will be a nice feature in a future version?
This would be nice if someone wants to have a certain exponent tested quickly, but testing two exponents (one per core) is more efficient.

Edit: See also this thread about hyperthreading and Dual Core with upcoming version 25.2

Last fiddled with by Andi47 on 2007-03-30 at 07:02
Andi47 is offline   Reply With Quote
Reply

Thread Tools


Similar Threads
Thread Thread Starter Forum Replies Last Post
Room Temperature Alloy a1call Lounge 0 2017-05-20 14:34
IRC room? kracker Lounge 3 2012-06-17 20:59
GMP-Fermat and multi-core processors ixfd64 Factoring 1 2011-03-16 16:07
Intel Quad Core optimization joblack Hardware 1 2009-06-11 02:42
Importance of dual channel memory for dual core processors patrik Hardware 3 2007-01-07 09:26

All times are UTC. The time now is 19:36.


Tue Oct 26 19:36:11 UTC 2021 up 95 days, 14:05, 0 users, load averages: 1.55, 2.26, 2.49

Powered by vBulletin® Version 3.8.11
Copyright ©2000 - 2021, Jelsoft Enterprises Ltd.

This forum has received and complied with 0 (zero) government requests for information.

Permission is granted to copy, distribute and/or modify this document under the terms of the GNU Free Documentation License, Version 1.2 or any later version published by the Free Software Foundation.
A copy of the license is included in the FAQ.