[QUOTE]That rule of only running LA for 3 of 4 cores still stands even for a i7?[/QUOTE]We try to start each job out with four cores and then test three and maybe two to see which is fastest. The "harder" the job is the more likely it is going to go faster with fewer cores. We do the "nc1" part of the job when we have another job running.

[QUOTE]Which machine did you use for 161_115_minus1?[/QUOTE]Our i7. We could have run four cores but three cores was projected to be a few minutes less.

L2575B done
 

[code]
prp88 factor: 3818795127999177924147227181959554946376119456756019077913829275252288527723645287600201
prp103 factor: 4392241986643411109718595536566776892318495252274492224657555414826440503175386707147582587797325789801
[/code]

Runtime: 6hrs download. 5.5hrs cycle-building; 1hr matrix-building; 29.5hrs x 24 CPUs linalg; 70 mins x 8 CPUs square root.

The log is for L2575B, I just used the wrong name for the directory!

Looks like L2605A will be ready for completion soon; shall I do that too?

[QUOTE=fivemack;322365]Looks like L2605A will be ready for completion soon; shall I do that too?[/QUOTE]
It's all yours. When the Est. Pending Rels hits 0, it's ready for download.

[QUOTE]…29.5hrs x 24 CPUs linalg…[/QUOTE]We really must be doing something wrong because we usually can use only three cores before performance drops off dramatically.

:down:

I'm only using three cores on each of the eight processors (two per socket) in that machine, because otherwise performance drops off dramatically :)

(see gosh3 in the .zip file for the exact combination of taskset, numactl and mpirun commands that I use)

L2605A will be even more oversieved than L2575B was, because the range was extended even further (but I don't know precisely when). Filtering it at target density 90 or 100 will probably be possible.

L2605 post-processing has started; the download took nine hours because my ADSL hasn't been running very fast. With luck and a following wind should expect factors by the end of Christmas Day.

Anyone postprocessing 853_83_minus1? I would do it.

[QUOTE=Batalov;322437]Anyone postprocessing 853_83_minus1? I would do it.[/QUOTE]

You are. :smile:


[COLOR=green]SB EDIT 2:10pm Dec-25: ftping now.[/COLOR]

[code]prp88 factor: 1516335665261748709137973391705243492332466590177299610265665349298098187499264788222251
prp96 factor: 739205173227316594615577377145357991779736202583720051830228260674218223073381288289689117365717[/code]