mersenneforum.org

mersenneforum.org (https://www.mersenneforum.org/index.php)
-   NFS@Home (https://www.mersenneforum.org/forumdisplay.php?f=98)
-   -   BOINC NFS sieving - RSALS (https://www.mersenneforum.org/showthread.php?t=12458)

pinhodecarlos 2012-01-08 08:55

Taking 1051_71_minus1.

fivemack 2012-01-08 09:27

[QUOTE=frmky;285349]Yay, that was enough.
linear algebra completed 16186 of 12269808 dimensions (0.1%, ETA 77h10m)
:smile:[/QUOTE]

Just out of interest, what's the hardware for that? That's a very impressive speed for a non-MPI job ...

pinhodecarlos 2012-01-08 09:29

[QUOTE=fivemack;285371]Just out of interest, what's the hardware for that? That's a very impressive speed for a non-MPI job ...[/QUOTE]

I think it is a MPI job.

debrouxl 2012-01-08 10:00

Greg: great :smile:

Carlos: OK, I'll somehow generate more relations. The yield on the 40M-45M range has been miserable, I seriously think I uploaded the wrong poly: I initially trial sieved with 29-bit LPs, then tried 28-bit LPs with lackluster success, but seem to have forgotten to go back to the 29/58 version and uploaded the 28/56 version...

pinhodecarlos 2012-01-08 10:16

[QUOTE=debrouxl;285373]

Carlos: OK, I'll somehow generate more relations. The yield on the 40M-45M range has been miserable, I seriously think I uploaded the wrong poly: I initially trial sieved with 29-bit LPs, then tried 28-bit LPs with lackluster success, but seem to have forgotten to go back to the 29/58 version and uploaded the 28/56 version...[/QUOTE]

Until you get this one fixed I have plenty to play around, 698419_37_minus1 and 293_97_minus1 are ready for post-processing. Tomorrow I will start 1051_71_minus1 and it must be a 2 days task.

jasonp 2012-01-08 13:28

At a guess, Greg is using 8 quad-cores with infiniband.

frmky 2012-01-08 18:15

[QUOTE=jasonp;285387]At a guess, Greg is using 8 quad-cores with infiniband.[/QUOTE]
Exactly. But these are Core 2 quads with DDR2 memory, and I've found it's faster to use only 2 cores on each rather than all four.

pinhodecarlos 2012-01-08 19:46

[QUOTE=frmky;285420]Exactly. But these are Core 2 quads with DDR2 memory, and I've found it's faster to use only 2 cores on each rather than all four.[/QUOTE]

Why? What about only three?

frmky 2012-01-08 21:49

[QUOTE=pinhodecarlos;285426]Why? What about only three?[/QUOTE]
Memory contention. Empirically, two is optimal.

wblipp 2012-01-08 22:16

[QUOTE=frmky;285444]Memory contention. Empirically, two is optimal.[/QUOTE]

Can you do anything else with the other cores? Perhaps some ECM stage 1?

pinhodecarlos 2012-01-09 14:57

1051_71_minus1 will de done in less than 16 hours. Next I'll do 698419_37_minus1 and 293_97_minus1.


All times are UTC. The time now is 22:26.

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