2018 15e post-processing reservations and results

[pinhodecarlos]

Originally Posted by pinhodecarlos

L2785B Failed with td=115, giving a shot with 100. Let’s see what comes up!

Failed with 100. Unreserving composite, please add more relations. TIA

[fivemack]

Taking F1971

[VBCurtis]

Originally Posted by VBCurtis

May I have another 15-20M relations? Density 128 does not build, and density 124 yields a ~32M matrix.
L2995B is GNFS188 and has 23M matrix; I hope I can do better than 31M for GNFS191!

Fivemack is already solving this matrix, but since more relations were queued I thought I'd find out how much help they would have provided.
First try: 452.6M raw relations, 324M unique. TD 128 failed, TD 124 matrix 31.3M.
Second try: 461.5 raw relations, 329.4M unique. TD 128 failed, TD 124 matrix 30.7M.
So, 9M extra relations in this case were not a meaningful difference; a 2% improvement in matrix size might improve LA time by 4%.
I retract my previous claim that 450M relations is always enough for 32LP jobs; however, I think that any job that needs 450M relations for 32LP ought to be done with 33LP instead.

[fivemack]

Thu Aug  2 23:50:50 2018  p79 factor: 4852482693399566031309100342466502203821581948525327672263255869963531563887371
Thu Aug  2 23:50:50 2018  p155 factor: 25601695898174599365362426477703552041646980275314877660484633243074989602867424075941978673943728332378716879550916402098860939133559296665461200154894043

68.4 hours on 10 cores Xeon Silver 4114

Log attached and at https://pastebin.com/eMf58gNh

[swellman]

p63 factor: 232555578644599927086757417642069159548766630657783545861246961
p75 factor: 118853751876715560420209556615837399854608332350671977978213659035577629141
p86 factor: 33164063418443223168062297507538876395929474531708325019581611805139958383009502113931

TD=102, 246M raw, 184M unique. Truncated log attached

[pinhodecarlos]

L2785A: https://pastebin.com/hk1zjP0D
130^101+101^130 C242 cofactor: https://pastebin.com/1aEiJUjn

[VBCurtis]

Taking 145^89_89^145 (GNFS 180).

[fivemack]

I've finished L1328, though the combination of doing parallel square-root on four processes and of having done the SNFS on a number with two known factors means the end of the log looks quite messy.

Second-largest factor is 42029642098058070120505155457595135814424295841322041625673209628515064207327

About 388.7 hours on 6 cores i7/4930K for a 26.80M matrix at density 126 (128 didn't work). Mangled log attached and at https://pastebin.com/L14VAFNq

[VBCurtis]

I just finished the matrix for l2995b, a GNFS-188.
The square root phase crashes after 2-3 minutes with:

Tue Aug  7 19:46:55 2018  commencing square root phase
Tue Aug  7 19:46:55 2018  reading relations for dependency 1
Tue Aug  7 19:46:58 2018  read 11497681 cycles
Tue Aug  7 19:47:27 2018  cycles contain 39663842 unique relations
Tue Aug  7 19:49:28 2018  error: relation 25830062 corrupt

What's my plan?

[axn]

Originally Posted by VBCurtis

I just finished the matrix for l2995b, a GNFS-188.
The square root phase crashes after 2-3 minutes with:

Tue Aug  7 19:46:55 2018  commencing square root phase
Tue Aug  7 19:46:55 2018  reading relations for dependency 1
Tue Aug  7 19:46:58 2018  read 11497681 cycles
Tue Aug  7 19:47:27 2018  cycles contain 39663842 unique relations
Tue Aug  7 19:49:28 2018  error: relation 25830062 corrupt

What's my plan?

Have you tried running the sqrt with the explicit dependency # (from 2 onwards)? Maybe you'll get lucky and the corruption won't affect others?

[VBCurtis]

Originally Posted by axn

Have you tried running the sqrt with the explicit dependency # (from 2 onwards)? Maybe you'll get lucky and the corruption won't affect others?

Thanks! #2 failed, but #3 is multiplying. Maybe I'll get lucky, or maybe I'll be manually running sqrts for a day or so.