GNU/Linux S5R3 "power users" App 4.35 available

log in

Advanced search

Message boards : Cruncher's Corner : GNU/Linux S5R3 "power users" App 4.35 available

1 · 2 · 3 · 4 · Next
Author Message
Profile Bernd Machenschalk
Volunteer moderator
Project administrator
Project developer
Avatar
Send message
Joined: 15 Oct 04
Posts: 3612
Credit: 128,728,497
RAC: 53,026
Message 81482 - Posted: 22 Feb 2008, 10:30:59 UTC
Last modified: 22 Feb 2008, 14:16:06 UTC

Built from the same codebase as the 4.33/4.34 Apps and includes assembler-coded prefetching instructions in the "Hough" code.

It also serves as a compatibility test for a new (virtual) build machine I set up. So if you find it crashing instantly, it's probably a bad library reference.

With the larger variety of Apps it had gotten harder to write an app_info.xml that satisfies everyone's needs. You may want to take a look into the file before installing the App (i.e. restarting the client) to see if your previous App version is covered.

Only run this if you're sure of what you're doing, in particular your CPU supports SSE instructions.

Find it on the Power User Apps page.

Happy crunching!

BM

Profile M. Schmitt
Avatar
Send message
Joined: 27 Jun 05
Posts: 480
Credit: 15,872,262
RAC: 0
Message 81485 - Posted: 22 Feb 2008, 11:05:09 UTC

For those who used the 64Bit app like me, I added the following to the app_info.xml right before th last </app_info> tag:

<app_version>
<app_name>einstein_S5R3</app_name>
<version_num>428</version_num>
<api_version>6.1.0</api_version>
<file_ref>
<file_name>einstein_S5R3_4.35_i686-pc-linux-gnu</file_name>
<main_program/>
</file_ref>
<file_ref>
<file_name>einstein_S5R3_4.35_graphics_i686-pc-linux-gnu</file_name>
<open_name>graphics_app</open_name>
</file_ref>
</app_version>

cu,
Michael
____________

Profile M. Schmitt
Avatar
Send message
Joined: 27 Jun 05
Posts: 480
Credit: 15,872,262
RAC: 0
Message 81486 - Posted: 22 Feb 2008, 11:18:50 UTC

Same thing with app 4.31!

I lost an already finished WU, because I forgot to add this at one host. :(

cu,
Michael
____________

rroonnaalldd
Send message
Joined: 12 Dec 05
Posts: 116
Credit: 537,221
RAC: 0
Message 81489 - Posted: 22 Feb 2008, 12:51:19 UTC - in response to Message 81486.

Same thing with app 4.31!

I lost an already finished WU, because I forgot to add this at one host. :(

cu,
Michael


...but that's no problem because Einstein resent WU's. :-)

____________
Profile M. Schmitt
Avatar
Send message
Joined: 27 Jun 05
Posts: 480
Credit: 15,872,262
RAC: 0
Message 81491 - Posted: 22 Feb 2008, 13:07:48 UTC - in response to Message 81489.

Same thing with app 4.31!

I lost an already finished WU, because I forgot to add this at one host. :(

cu,
Michael


...but that's no problem because Einstein resent WU's. :-)


Just almost 16h crunching for nothing...
And what's the difference between new and resent WUs? ;-)

This app runs like hell, really good work! It's very fast even with the shorter WUs.

cu,
Michael

____________
Profile Bernd Machenschalk
Volunteer moderator
Project administrator
Project developer
Avatar
Send message
Joined: 15 Oct 04
Posts: 3612
Credit: 128,728,497
RAC: 53,026
Message 81493 - Posted: 22 Feb 2008, 13:33:30 UTC
Last modified: 22 Feb 2008, 13:33:52 UTC

I updated the app_info.xml in the package (and the md5sum on the page), and also added the warning I already wrote at the 4.32 thread in the initial post.

The app_info.xml now has antries for:

$ grep version_num power_apps/app_info.xml
420
421
424
427
428
431
435


HTH

BM
rroonnaalldd
Send message
Joined: 12 Dec 05
Posts: 116
Credit: 537,221
RAC: 0
Message 81495 - Posted: 22 Feb 2008, 13:54:04 UTC - in response to Message 81491.


Just almost 16h crunching for nothing...
And what's the difference between new and resent WUs? ;-)

This app runs like hell, really good work! It's very fast even with the shorter WUs.

cu,
Michael


No difference. You will get the same WU's to crunch and past a few hours later, credits would be granted...

____________
Wedge009
Send message
Joined: 5 Mar 05
Posts: 37
Credit: 26,911,577
RAC: 80,788
Message 81496 - Posted: 22 Feb 2008, 14:03:58 UTC - in response to Message 81493.

I think ziegenmelker was making a rhetorical question there, rroonnaalldd. :)

No apparent problems so far, switching from 4.27 to 4.35 at 19.9%. Too early to make any judgement on speed improvements.
____________
Soli Deo Gloria

Annika
Avatar
Send message
Joined: 8 Aug 06
Posts: 720
Credit: 494,410
RAC: 0
Message 81497 - Posted: 22 Feb 2008, 14:35:25 UTC

No instant crashes on my box. We'll see about performance. Since my two current Einstein WUs are only at 23 respectively 6 percent finished one should be able to see a difference in crunch times if the app is much faster (than my previous 4.27).

Profile M. Schmitt
Avatar
Send message
Joined: 27 Jun 05
Posts: 480
Credit: 15,872,262
RAC: 0
Message 81502 - Posted: 22 Feb 2008, 17:44:37 UTC
Last modified: 22 Feb 2008, 17:48:59 UTC

The first WU crunched pure by 4.35 is finished.

- 24.8 % runtime

compared to what it should have taken with 4.27 following my formulas. Host is an AMD X2 5000@3,3 GHz, OpenSuse 10.3, 64 Bit.

The WU took 18,632 sec instead of ~24,760 sec.
The relative seq. no. is 23, where 0 would be the trough.

This is a great improvement! Congrats to everybody who takes part in this enhancement!

My AMD XP 3000+ will take some more time to finish its first WU, but its also running very noticeable faster.

cu,
Michael
____________

Profile M. Schmitt
Avatar
Send message
Joined: 27 Jun 05
Posts: 480
Credit: 15,872,262
RAC: 0
Message 81514 - Posted: 23 Feb 2008, 1:36:42 UTC

I have 4 finished results on my X2 now with an average speed increase of 34.6 %.

Speed means [credits/s] or better credits per hour. ;-)

My Athlon XP 3000+ also finished a WU and was 38.6 % faster, but in this case the calculated values are not as precise because of lacking reference WUs.

One result pending, the others are valid.

Great work!

cu,
Michael
____________

Profile Donald A. Tevault
Avatar
Send message
Joined: 17 Feb 06
Posts: 443
Credit: 73,513,404
RAC: 0
Message 81529 - Posted: 23 Feb 2008, 4:43:42 UTC
Last modified: 23 Feb 2008, 4:44:10 UTC

I've switched all of my SSE capable machines over to this, and it's making a huge difference. (And, that's when compared to the 4.21 and 4.27 power-users apps.)

My 6000+ machine will now be completing workunits in about five hours time. (That's down from about eight hours with the 4.21 power-users app.)
____________

Profile Zxian
Avatar
Send message
Joined: 23 Oct 06
Posts: 40
Credit: 5,121,474
RAC: 0
Message 81534 - Posted: 23 Feb 2008, 7:29:04 UTC

I've switched over my one E2160 machine to this app, and things seem to be working so far.

rroonnaalldd
Send message
Joined: 12 Dec 05
Posts: 116
Credit: 537,221
RAC: 0
Message 81536 - Posted: 23 Feb 2008, 8:45:12 UTC
Last modified: 23 Feb 2008, 9:40:05 UTC

...switched from 4.21 and get message in log:
file projects/einstein/einstein_4.21 not found
Restarting task using einstein_4.21

It is crunching now from last checkpoint. So far no lib-problems seen on my little suse64-box. This WU will be the first unit crunched only with 4.35.

edit link and wu

____________

Profile Donald A. Tevault
Avatar
Send message
Joined: 17 Feb 06
Posts: 443
Credit: 73,513,404
RAC: 0
Message 81540 - Posted: 23 Feb 2008, 9:37:14 UTC

Only one problem so far. . .

This is a signal 11 problem from a 4.27 workunit that I switched over to 4.35. (The computer is a P-III 450 running Ubuntu Gutsy Gibbon.)

Signal 11 problem

This happened while I was in bed, so I don't know whether there was a communication problem.
____________

rroonnaalldd
Send message
Joined: 12 Dec 05
Posts: 116
Credit: 537,221
RAC: 0
Message 81541 - Posted: 23 Feb 2008, 9:42:35 UTC - in response to Message 81540.

Only one problem so far. . .

This is a signal 11 problem from a 4.27 workunit that I switched over to 4.35. (The computer is a P-III 450 running Ubuntu Gutsy Gibbon.)

Signal 11 problem

This happened while I was in bed, so I don't know whether there was a communication problem.


intresting is your stderr_txt:

800, libgcc_s.so.1 must be installed for pthread_cancel to work

____________
rroonnaalldd
Send message
Joined: 12 Dec 05
Posts: 116
Credit: 537,221
RAC: 0
Message 81557 - Posted: 23 Feb 2008, 18:24:21 UTC - in response to Message 81536.

...switched from 4.21 and get message in log:
file projects/einstein/einstein_4.21 not found
Restarting task using einstein_4.21

It is crunching now from last checkpoint. So far no lib-problems seen on my little suse64-box. This WU will be the first unit crunched only with 4.35.

edit link and wu


*jingle* WU is crunched and need only ~8hours on a E6320 with virtualized Suse64-10.2. Seems to be the fastest linux-app ever.
Great job!

____________
Profile Zxian
Avatar
Send message
Joined: 23 Oct 06
Posts: 40
Credit: 5,121,474
RAC: 0
Message 81576 - Posted: 23 Feb 2008, 23:35:16 UTC

App 4.35 result
App 4.27 result


All I can say is wow. Great job at improving the speed!

Profile tullio
Send message
Joined: 22 Jan 05
Posts: 1953
Credit: 3,100,647
RAC: 31,787
Message 81586 - Posted: 24 Feb 2008, 5:08:09 UTC
Last modified: 24 Feb 2008, 5:08:42 UTC

I had a 35% reduction in time from 4.27 to 4.35 and a 30 credits/h rate (pending), about three times what I get in QMC@home. SuSE Linux 10.3 on AMD Opteron 1210.
Tullio
____________

Profile Jos van Wolput
Send message
Joined: 11 Feb 05
Posts: 47
Credit: 800,840
RAC: 0
Message 81589 - Posted: 24 Feb 2008, 7:35:50 UTC

I get a 33% speed improvement after upgrading from 4.27 to 4.35 and no problems.
Really great!
Thanks!

____________

1 · 2 · 3 · 4 · Next

Message boards : Cruncher's Corner : GNU/Linux S5R3 "power users" App 4.35 available


Home · Your account · Message boards

This material is based upon work supported by the National Science Foundation (NSF) under Grants PHY-1104902, PHY-1104617 and PHY-1105572 and by the Max Planck Gesellschaft (MPG). Any opinions, findings, and conclusions or recommendations expressed in this material are those of the investigators and do not necessarily reflect the views of the NSF or the MPG.

Copyright © 2016 Bruce Allen