Mathew: reserved to you. I had downloaded the raw dataset yesterday, so Greg can remove the file from the server once you have downloaded it as well :smile:

[QUOTE=fivemack;346610]May I reserve L1217?[/QUOTE]
Done.

May I reserve L2910?

Reserved for you :smile:

L2910 should be done by tomorrow night - msieve-svn923 running on 48 cores is a beast. Reserving L3335B.

I definitely agree, msieve-svn927 (Brian's windows compile hosted by Jeff) is definitely a beast even on my 16-core machine. I was able to finish an 8.83M^2 matrix in 21.5 hours.

Also, probably of more interest here, the L1911 c167 is done!
[CODE]Sun Jul 21 22:18:25 2013 prp70 factor: 1804884259792467690673278289080097940019161105656487784369968446751059
Sun Jul 21 22:18:25 2013 prp98 factor: 10890740644097947260228498261961007481694342816974640553214282223161855561117576608098921466911279
Sun Jul 21 22:18:25 2013 elapsed time 24:31:29[/CODE]

Here's the matrix snippet from the log:
[CODE]
Sat Jul 20 23:28:04 2013 matrix is 8830757 x 8830983 (2587.2 MB) with weight 664939915 (75.30/col)
Sat Jul 20 23:28:04 2013 sparse part has weight 589918881 (66.80/col)
Sat Jul 20 23:28:04 2013 using block size 8192 and superblock size 1966080 for processor cache size 20480 kB
Sat Jul 20 23:28:38 2013 commencing Lanczos iteration (16 threads)
Sat Jul 20 23:28:38 2013 memory use: 2110.1 MB
Sat Jul 20 23:28:52 2013 linear algebra at 0.0%, ETA 21h 3m
Sat Jul 20 23:28:57 2013 checkpointing every 410000 dimensions
Sun Jul 21 21:03:43 2013 lanczos halted after 139654 iterations (dim = 8830756)
Sun Jul 21 21:03:58 2013 recovered 28 nontrivial dependencies
[/CODE]
And here's the log in case anyone is interested:

Does your machine seriously have 20MB of cache?!

[QUOTE=jasonp;346975]Does your machine seriously have 20MB of cache?![/QUOTE]

:smile: Yes. The machine has two Xeon E5-2687W processors. According to the specs, each one is:
3.1 GHz
8 physical cores
8 x 256KB L2 cache
20MB L3 cache
2 × 8.0 GT/s QPI

Each processor has a quad-channel memory controller and I have 8 sticks of DDR3-1600 8GB ECC registered ram. So, in all I have 16 cores and 64GB of ram in this machine. I can upgrade this to 128GB of ram, but the heat sinks sit so low to the board it is difficult to get in a couple of the sticks. So far I haven't needed to go above 64, so I'm waiting it out to see if I actually need to upgrade.

L2910 done
 

[code]
Sun Jul 21 14:58:22 2013 matrix is 10574608 x 10574789 (3219.4 MB) with weight 1008116182 (95.33/col)
Mon Jul 22 22:40:08 2013 BLanczosTime: 117054
Tue Jul 23 01:28:03 2013 prp69 factor: 170482021489151537357699348890742814346693197489876187260812758554841
Tue Jul 23 01:28:03 2013 prp100 factor: 3623148469119597902267795436100577680557086912329700399770621334437924863121580494990506310144533121
[/code]

L3335B job didn't start properly at first (I think it helps to do

echo 3 | sudo tee /proc/sys/vm/drop_caches

to clear disc caches and arrange a chunk of free memory on each NUMA node before running jobs with numactl --cpunodebind -l) but I restarted and it should be done by tomorrow morning.

L3335B done
 

[code]
Tue Jul 23 08:06:10 2013 matrix is 7776937 x 7777137 (2353.1 MB) with weight 729728721 (93.83/col)
Wed Jul 24 00:04:17 2013 BLanczosTime: 60119
Wed Jul 24 01:55:22 2013 prp61 factor: 3797410774154194497864115807319596644680039582655941653929281
Wed Jul 24 01:55:22 2013 prp103 factor: 7392747394671201897269984299282206285729796825963649282602156827514331604710773398485841113481594513721
[/code]

L1217 done
 

Nice split:
[code]
Wed Jul 24 18:44:50 2013 prp99 factor: 213259382328802880095976186759993372537516136049359453960430400784759051997987651012758130372705369
Wed Jul 24 18:44:50 2013 prp100 factor: 1536227261299077816800753932335420524137049379474377174657167162598668533635625901254408246761846299
[/code]