ATTN Tetsuji Maverick- RE: SSE3 for AMD Dual Core

Message boards : Number crunching : ATTN Tetsuji Maverick- RE: SSE3 for AMD Dual Core
Message board moderation

To post messages, you must log in.

1 · 2 · Next

AuthorMessage
Daniel Schaalma
Volunteer tester
Avatar

Send message
Joined: 28 May 99
Posts: 297
Credit: 16,953,703
RAC: 0
United States
Message 173061 - Posted: 30 Sep 2005, 14:32:03 UTC

Tetsuji, I will soon be putting a couple of new machines online, the AMD Athlon64 X2 Dual Core 4800+ units, and perhaps a Dual Core Opteron machine. All these CPU's have the SSE3 Instruction set, and I was wondering if you have any plans to compile a new version of the SETI science app to make use of SSE3 on these new AMD CPU's. The current SSE3 science app is evidently Intel specific. Thanks for your time, and BIG thanks for all the current optimized apps!
Happy Crunching!

Regards, Daniel.
ID: 173061 · Report as offensive
Tetsuji Maverick Rai
Volunteer tester
Avatar

Send message
Joined: 25 Apr 99
Posts: 518
Credit: 90,863
RAC: 0
Japan
Message 173064 - Posted: 30 Sep 2005, 14:48:53 UTC
Last modified: 30 Sep 2005, 15:22:43 UTC

No I don't have any plan to build further applications. Sorry... And yes, SSE3 optimization code of ICC always checks the processor. So SSE2 version (W version) is the best way. There's little difference between SSE3 (P version) and SSE2 (W or N version) even on P4. Look http://www.marisan.nl/seti/reference.htm or other benchmark pages.

SSE3 instruction set is just a small addition to SSE2.

Maybe I can hack Intel's library technically and legally, but it won't benefit so much. Since I have no AMD chips, I will need your help for testing (just run small testing programs and give me the results.) It may be an interesting challenge.

BTW marisan's English phpBB board looks hackable (I haven't tried.)
Luckiest in the world. WMD = Weapon of Mass Distraction.
Click this table.
ID: 173064 · Report as offensive
Profile -= Vyper =-
Volunteer tester
Avatar

Send message
Joined: 5 Sep 99
Posts: 1652
Credit: 1,065,191,981
RAC: 2,537
Sweden
Message 173072 - Posted: 30 Sep 2005, 15:38:15 UTC

Hey Tetsuji, i'm on it also.. Here is my email so i can participate in your testing also..

laracroft@telia.com

_________________________________________________________________________
Addicted to SETI crunching!
Founder of GPU Users Group
ID: 173072 · Report as offensive
Ned Slider

Send message
Joined: 12 Oct 01
Posts: 668
Credit: 4,375,315
RAC: 0
United Kingdom
Message 173090 - Posted: 30 Sep 2005, 16:31:39 UTC

There isn't really any real advantage to using SSE3. I built otherwise identical SSE2 and SSE3 clients for AMD Athlon64 on Linux and they benchmarked within a few seconds of each other (the SSE2 client was actually a few seconds faster!). This was on an Athlon64, not an X2, and on Linux, not Windows, but I see no reason for it to be any different.

Ned

*** My Guide to Compiling Optimised BOINC and SETI Clients ***
*** Download Optimised BOINC and SETI Clients for Linux Here ***
ID: 173090 · Report as offensive
Tetsuji Maverick Rai
Volunteer tester
Avatar

Send message
Joined: 25 Apr 99
Posts: 518
Credit: 90,863
RAC: 0
Japan
Message 173092 - Posted: 30 Sep 2005, 16:38:07 UTC
Last modified: 30 Sep 2005, 16:57:05 UTC

Ned is correct. SO I stop hacking Intel's library, though it looks easier than I expected. sse3 only adds very few instructions, and that the speed advantage is very small.

BTW my RAC is zero :)
Luckiest in the world. WMD = Weapon of Mass Distraction.
Click this table.
ID: 173092 · Report as offensive
Hans Dorn
Volunteer developer
Volunteer tester
Avatar

Send message
Joined: 3 Apr 99
Posts: 2262
Credit: 26,448,570
RAC: 0
Germany
Message 173097 - Posted: 30 Sep 2005, 17:36:11 UTC - in response to Message 173092.  

BTW my RAC is zero :)


Strange... Everyone except you is crunching away with your client :o)


Regards Hans

ID: 173097 · Report as offensive
Tetsuji Maverick Rai
Volunteer tester
Avatar

Send message
Joined: 25 Apr 99
Posts: 518
Credit: 90,863
RAC: 0
Japan
Message 173098 - Posted: 30 Sep 2005, 17:41:01 UTC - in response to Message 173097.  
Last modified: 30 Sep 2005, 18:07:35 UTC

BTW my RAC is zero :)


Strange... Everyone except you is crunching away with your client :o)


Regards Hans


It's because now I'm up to the ears in beta test project. As you see in my computers, only two results are present in the database now.

All my 3 machines are running setiathome_enhanced for beta test project. Sometimes (for ex. during the long outage, or while I'm compiling new version of crunchers) I attach to other projects. Beta is more efficient, but takes longer time. And now I hope major bugs are fixed. Now optimizatin is for testing/debugging in beta project to me. That's one of the reasons I'm reluctant on aggressive optimization now.
Luckiest in the world. WMD = Weapon of Mass Distraction.
Click this table.
ID: 173098 · Report as offensive
Tetsuji Maverick Rai
Volunteer tester
Avatar

Send message
Joined: 25 Apr 99
Posts: 518
Credit: 90,863
RAC: 0
Japan
Message 173135 - Posted: 30 Sep 2005, 19:54:19 UTC

PS:

And I am not concerned with credits so much (though if I told I'm not interested in credits at all, it would be a lie.....) So I am pouring full computer power to beta project whose credits don't appear in the stats.
ID: 173135 · Report as offensive
Hans Dorn
Volunteer developer
Volunteer tester
Avatar

Send message
Joined: 3 Apr 99
Posts: 2262
Credit: 26,448,570
RAC: 0
Germany
Message 173138 - Posted: 30 Sep 2005, 19:59:30 UTC - in response to Message 173135.  
Last modified: 30 Sep 2005, 20:00:13 UTC

PS:

And I am not concerned with credits so much (though if I told I'm not interested in credits at all, it would be a lie.....) So I am pouring full computer power to beta project whose credits don't appear in the stats.


Fine.

I'll be leeching .... erm, downloading your patches as soon as seti_enhanced gets out :o)

Regards Hans
ID: 173138 · Report as offensive
Profile Speedy67 & Friends
Volunteer tester
Avatar

Send message
Joined: 14 Jul 99
Posts: 335
Credit: 1,178,138
RAC: 0
Netherlands
Message 173139 - Posted: 30 Sep 2005, 20:00:55 UTC - in response to Message 173064.  

Hi TMR,


BTW marisan's English phpBB board looks hackable (I haven't tried.)


Of course I'm interested in what you mean.. is it that you think phpBB2 boards are hackable in general, or just the english one at marisan.nl....? :)

Greetings,
Sander



ID: 173139 · Report as offensive
Tetsuji Maverick Rai
Volunteer tester
Avatar

Send message
Joined: 25 Apr 99
Posts: 518
Credit: 90,863
RAC: 0
Japan
Message 173142 - Posted: 30 Sep 2005, 20:11:23 UTC - in response to Message 173139.  
Last modified: 30 Sep 2005, 20:24:03 UTC

Hi TMR,


BTW marisan's English phpBB board looks hackable (I haven't tried.)


Of course I'm interested in what you mean.. is it that you think phpBB2 boards are hackable in general, or just the english one at marisan.nl....? :)

Greetings,
Sander


To tell the truth, I tried just after I registered :) It's not hackable to me. I wrote "It looks hackable" because it looked older. In the bottom, it says "Powered by phpBB © 2001, 2002 phpBB Group ;" Now I think it means the skin (or template?) of the forums is old. As you may know, earlier versions of phpBB (2.0.12 or earlier) are very easy to hack. By changing cookie, any member can be the admin, and hijack the forums. For further details see this. BOINC synergy or another stats forum (I forgot) had this problem before, and I emailed to the admin.

Much earlier versions (2.0.9? or earlier) had more serious problem.

Anyway when I find a vulnerability I won't post in the public, instead email to the admin ASAP :)
Luckiest in the world. WMD = Weapon of Mass Distraction.
Click this table.
ID: 173142 · Report as offensive
Profile Chilean
Volunteer tester
Avatar

Send message
Joined: 6 Apr 03
Posts: 498
Credit: 3,200,504
RAC: 0
Chile
Message 173146 - Posted: 30 Sep 2005, 20:16:50 UTC - in response to Message 173064.  

No I don't have any plan to build further applications. Sorry... And yes, SSE3 optimization code of ICC always checks the processor. So SSE2 version (W version) is the best way. There's little difference between SSE3 (P version) and SSE2 (W or N version) even on P4. Look http://www.marisan.nl/seti/reference.htm or other benchmark pages.

SSE3 instruction set is just a small addition to SSE2.

Maybe I can hack Intel's library technically and legally, but it won't benefit so much. Since I have no AMD chips, I will need your help for testing (just run small testing programs and give me the results.) It may be an interesting challenge.

BTW marisan's English phpBB board looks hackable (I haven't tried.)


Odd.... on the benchamarks... your version 7 is faster than your 8.1
ID: 173146 · Report as offensive
Profile Speedy67 & Friends
Volunteer tester
Avatar

Send message
Joined: 14 Jul 99
Posts: 335
Credit: 1,178,138
RAC: 0
Netherlands
Message 173148 - Posted: 30 Sep 2005, 20:25:20 UTC - in response to Message 173142.  

Hi TMR,

BTW marisan's English phpBB board looks hackable (I haven't tried.)


Of course I'm interested in what you mean.. is it that you think phpBB2 boards are hackable in general, or just the english one at marisan.nl....? :)


To tell the truth, I tried just after I registered :) It's not hackable to me. I wrote "It looks hackable" because it looked older. In the bottom, it says "Powered by phpBB © 2001, 2002 phpBB Group ;" As you know, earlier versions of phpBB (2.0.12 or earlier) are very easy to hack.


I know, I try to keep things up-to-date. I was running 2.0.16, updated to 2.0.17 today just to be sure. Thanks for the concern anyway. :)
Why it says © 2001, 2002 I don't know... I'm looking into it. Maybe something's not as is should be, and you turn out to be right after all.. :)


Anyway when I find a vulnerability I won't post in the public, instead email to the admin ASAP :)


Thanks. :)

Greetings,
Sander




ID: 173148 · Report as offensive
Tetsuji Maverick Rai
Volunteer tester
Avatar

Send message
Joined: 25 Apr 99
Posts: 518
Credit: 90,863
RAC: 0
Japan
Message 173164 - Posted: 30 Sep 2005, 21:26:45 UTC - in response to Message 173146.  


Odd.... on the benchamarks... your version 7 is faster than your 8.1


yes, on my p4 boxes, 8.1 was always faster... Several people reported 7 was faster on this forum. I don't know why. So you'd better try both versions yourself on your boxes to your content.
ID: 173164 · Report as offensive
[BOINCstats] Willy
Volunteer tester

Send message
Joined: 4 Mar 01
Posts: 202
Credit: 152,243
RAC: 0
Netherlands
Message 173174 - Posted: 30 Sep 2005, 21:59:07 UTC - in response to Message 173142.  

BOINC synergy or another stats forum (I forgot) had this problem before, and I emailed to the admin.


I guess that was me.

I don't trust the security of PHPBB anymore. I updated after the warning, and kept up-to-date as soon as an update was available. Still my server was hacked by using a PHPBB exploit even though I was running the lastest (4 days old) version op PHPBB.
But this is all way off topic.


Join team BOINCstats
ID: 173174 · Report as offensive
Tetsuji Maverick Rai
Volunteer tester
Avatar

Send message
Joined: 25 Apr 99
Posts: 518
Credit: 90,863
RAC: 0
Japan
Message 173239 - Posted: 1 Oct 2005, 3:43:56 UTC - in response to Message 173174.  

BOINC synergy or another stats forum (I forgot) had this problem before, and I emailed to the admin.


I guess that was me.

I don't trust the security of PHPBB anymore. I updated after the warning, and kept up-to-date as soon as an update was available. Still my server was hacked by using a PHPBB exploit even though I was running the lastest (4 days old) version op PHPBB.
But this is all way off topic.


sorry to mention it here :( But in reality there are still a lot of forums using older versions of phpBB.
Luckiest in the world. WMD = Weapon of Mass Distraction.
Click this table.
ID: 173239 · Report as offensive
Profile Karsten Vinding
Volunteer tester

Send message
Joined: 18 May 99
Posts: 239
Credit: 25,201,931
RAC: 11
Denmark
Message 173694 - Posted: 2 Oct 2005, 21:38:34 UTC
Last modified: 2 Oct 2005, 21:56:55 UTC

Regarding making a specialized version with SSE3 for AMD 64 with these instructions.

I can't help you with the programming, but I found a handy little tool called iccpatch.

It can be found on this page, together with some documentation: http://www.swallowtail.org/naughty-intel.html#patches

This checks binaries compiled with ICC9 for specific processor checks, and if it finds any it replaces them with noop commands.

The only thing it leaves behind is the check for the processors abilities, and if the processor reports ability to run SSE3, it will work on the processor, regardless if its an AMD or Intel processor.

I have only checked the program on my AMD Sempron, with the SSE versions of TMR's clients, and it identifies and removes some commands. I don't see any difference in crunch time on my machine, and didn't expect to either.

But I'm positive the program could be used to make the SSE3 compiles work on AMD64 processors with SSE3 as well.
ID: 173694 · Report as offensive
Tetsuji Maverick Rai
Volunteer tester
Avatar

Send message
Joined: 25 Apr 99
Posts: 518
Credit: 90,863
RAC: 0
Japan
Message 173750 - Posted: 3 Oct 2005, 0:57:20 UTC - in response to Message 173061.  
Last modified: 3 Oct 2005, 1:14:20 UTC

Tetsuji, I will soon be putting a couple of new machines online, the AMD Athlon64 X2 Dual Core 4800+ units, and perhaps a Dual Core Opteron machine. All these CPU's have the SSE3 Instruction set, and I was wondering if you have any plans to compile a new version of the SETI science app to make use of SSE3 on these new AMD CPU's. The current SSE3 science app is evidently Intel specific. Thanks for your time, and BIG thanks for all the current optimized apps!
Happy Crunching!

Regards, Daniel.


I found it how to cheat ICC (not IPP yet and forever), but I also found it illegal (see this post for details) and so I won't tell how to or build it. Sorry.....
Luckiest in the world. WMD = Weapon of Mass Distraction.
Click this table.
ID: 173750 · Report as offensive
Daniel Schaalma
Volunteer tester
Avatar

Send message
Joined: 28 May 99
Posts: 297
Credit: 16,953,703
RAC: 0
United States
Message 174130 - Posted: 4 Oct 2005, 12:21:44 UTC - in response to Message 173064.  

No I don't have any plan to build further applications. Sorry... And yes, SSE3 optimization code of ICC always checks the processor. So SSE2 version (W version) is the best way. There's little difference between SSE3 (P version) and SSE2 (W or N version) even on P4. Look http://www.marisan.nl/seti/reference.htm or other benchmark pages.

SSE3 instruction set is just a small addition to SSE2.

Maybe I can hack Intel's library technically and legally, but it won't benefit so much. Since I have no AMD chips, I will need your help for testing (just run small testing programs and give me the results.) It may be an interesting challenge.

BTW marisan's English phpBB board looks hackable (I haven't tried.)


Tetsuji, thanks much for the info. I was not aware that there was so little difference between the SSE2 and SSE3 instruction set. You'll have to forgive my ignorance, as I have not done any programming since 1983, and even then, I was only working with BASIC, and it didn't require a compiler. I guess I should break down and buy a copy of Visual C++ and start learning it. I will let you know when I get the new dual core chip. I should be in a position to buy it in about two weeks. I would be happy to do some test crunching on it for you, if there is anything that you plan on coding for it. Sorry it took so long to reply, but I have been sick with pneumonia, and I have not spent much time here at the keyboard.

Regards, Daniel.
ID: 174130 · Report as offensive
Profile James Von Schmittou

Send message
Joined: 4 Jun 02
Posts: 5
Credit: 983,878
RAC: 0
United States
Message 179450 - Posted: 18 Oct 2005, 1:23:40 UTC

i have a dual core on the way, and would be willing to test the SSE3 version "if you make it".
ID: 179450 · Report as offensive
1 · 2 · Next

Message boards : Number crunching : ATTN Tetsuji Maverick- RE: SSE3 for AMD Dual Core


 
©2024 University of California
 
SETI@home and Astropulse are funded by grants from the National Science Foundation, NASA, and donations from SETI@home volunteers. AstroPulse is funded in part by the NSF through grant AST-0307956.