mersenneforum.org - Linear algebra reservations, progress and results

mersenneforum.org (https://www.mersenneforum.org/index.php)

- NFS@Home (https://www.mersenneforum.org/forumdisplay.php?f=98)

- - Linear algebra reservations, progress and results (https://www.mersenneforum.org/showthread.php?t=20023)

[code]
Wed Feb 3 08:48:00 2016 p95 factor: 12988503186222327646421248730895222403293905784149398502620481281112320157272209614019269690491
Wed Feb 3 08:48:00 2016 p159 factor: 518335114642293378691256234379505566150951390081075444098532223237755512044427487395657159499307225212792813634256606208660579492770434929349443574020113409031
[/code]

119.7 hours on six cores of i7/5820K for 15.6M density-120 matrix. Log attached.

Reserving C161_P201_plus_1 for post processing, once sieving nears completion.

Reserving 146^61+61^146 for post processing.

4051^71-1 completed - 90 hours for 12.99M matrix. Log attached.

[CODE]prp75 factor: 279274616733523689420549341866562257753979070922335553016705017106816974741
prp179 factor: 12118051067257084674115490347233065965201598080089357315147723455637969877312374084872855670357510735066190365056976013646056286675455603648094759669868167197319449527159692130881
[/CODE]

I'll take 12479_61_minus1. Thanks.

128^95+95^128 Factored

[code]
p68 = 23144924940060930504142723222526535829414212140961804786225308913977
p71 = 67007404223119987076182324317164971756140037268878967582148260231233633
p72 = 878501540133886952102725149264266226184054224758023789229451091417720817
[/code]

Log file for 128^95+95^128.

[CODE]p72 factor: 263615508766572376761965157810265545216275677830154922343498587550270229
p115 factor: 7994398119225158258804307836393693167720834296122169067092886389267152091868349999345814661476258564700539077285423[/CODE]

Usually "D" jobs are around 100 to 125 hours on our NUC.

We just started 12269_61_minus1 (TD=100) and the ETA is over 1,100 hours.

Did we do something wrong?

:mike:

[CODE]Wed Feb 10 13:24:55 2016 commencing linear algebra
Wed Feb 10 13:25:10 2016 read 20345397 cycles
Wed Feb 10 13:26:06 2016 cycles contain 66603558 unique relations
Wed Feb 10 13:33:11 2016 read 66603558 relations
Wed Feb 10 13:36:04 2016 using 20 quadratic characters above 4294917295
Wed Feb 10 13:42:09 2016 building initial matrix
Wed Feb 10 13:56:21 2016 memory use: 8343.3 MB
Wed Feb 10 13:56:35 2016 read 20345397 cycles
Wed Feb 10 13:56:38 2016 matrix is 20345220 x 20345397 (8219.0 MB) with weight 2351373112 (115.57/col)
Wed Feb 10 13:56:38 2016 sparse part has weight 1951095079 (95.90/col)
Wed Feb 10 14:05:03 2016 filtering completed in 2 passes
Wed Feb 10 14:05:08 2016 matrix is 20344558 x 20344735 (8218.9 MB) with weight 2351354016 (115.58/col)
Wed Feb 10 14:05:08 2016 sparse part has weight 1951089631 (95.90/col)
Wed Feb 10 14:06:37 2016 matrix starts at (0, 0)
Wed Feb 10 14:06:40 2016 matrix is 20344558 x 20344735 (8218.9 MB) with weight 2351354016 (115.58/col)
Wed Feb 10 14:06:40 2016 sparse part has weight 1951089631 (95.90/col)
Wed Feb 10 14:06:40 2016 saving the first 48 matrix rows for later
Wed Feb 10 14:06:43 2016 matrix includes 64 packed rows
Wed Feb 10 14:06:46 2016 matrix is 20344510 x 20344735 (7917.1 MB) with weight 1985767002 (97.61/col)
Wed Feb 10 14:06:46 2016 sparse part has weight 1871966901 (92.01/col)
Wed Feb 10 14:06:47 2016 using block size 8192 and superblock size 294912 for processor cache size 3072 kB
Wed Feb 10 14:08:43 2016 commencing Lanczos iteration (2 threads)
Wed Feb 10 14:08:43 2016 memory use: 6798.9 MB
Wed Feb 10 14:14:03 2016 linear algebra at 0.0%, ETA 1160h 1m
Wed Feb 10 14:15:48 2016 checkpointing every 20000 dimensions[/CODE]

A 20Mx20M matrix with density 100 is fantastically large for a SNFS difficulty 249 task, so it looks like the task was significantly undersieved. How much RAM does your NUC have ?

If the numbers on the "Crunching" page are to be believed (= if the new numbers are not subject to the recent glitch), all of C164_P207_plus_1, C162_842592_8035, C157_933436_12482 and C174_135_52, which were moved to the "Queued for post-processing" state, are way undersieved, and at least C157_933436_12482 has fewer relations than the minimum count to build a (huge) matrix.
I do intentionally queue numbers with short ranges, and adjust them later - don't mistake my short ranges for proper ranges :smile:

Our NUC has 16GiB memory.

We are trying to eliminate downtime between jobs. We have C164_P207_plus_1 mostly downloaded but it might also need additional sieving so we are not sure what to do.

:mike: