Linear algebra reservations, progress and results

[Jarod]

Originally Posted by VBCurtis

Anything from the 14e queue should fit in 10GB and finish in less than 2 weeks. If you dodge the candidates with the largest assigned Q-range, you should be under a week and under 8GB. Matrix difficulty on 14e roughly correlates to Q-range sieved.

For example, I'm doing a GNFS-186 from the 15e queue, which fits in 9GB and will take 400hr running on half a 6-core i7. That's bigger than any job you should encounter on 14e.

Thanks for the insight. How long do you think a 20 – 100 M will take say on 8 cores?

[Jarod]

I would just like to correct my above post it is 20 – 110 M

[VBCurtis]

3 days? Perhaps less. Try it and report back:
1. The number of relations and target density you used.
2. The size of the matrix (you can round to nearest 100k, or not)
3. The number of hours the solution took.

If you look through this thread (e.g. post #2565), you'll get a pretty good idea of what some previous 30-bit or small 31-bit projects took to solve. With an 8-core, you might break 50 hours for the matrix-solve step.

[fivemack]

Wed Apr  4 10:45:54 2018  p58 factor: 1155978240967059981545236794906098102147329427908865568601
Wed Apr  4 10:45:54 2018  p150 factor: 103914941017585432569807404537280209718564853033928931580814904984029336379068716559184041945595873614118964758599832464919670696456300942029189406901

47.4 hours for a 13.44M density-112 (114 didn't work) matrix on 14 cores i9-7940X. Maximum reported memory use 7403.5MB.

Log attached and at https://pastebin.com/e52z8G25

[fivemack]

Originally Posted by fivemack

weight of 64591916 cycles is about 4521521435 (70.00/cycle)

I've thrown it back for another two hundred million raw relations

Wed Apr  4 20:11:00 2018  weight of 47012938 cycles is about 3291067733 (70.00/cycle)

So once I've found the largest working density, that should be a matrix at least reasonably tractable on excessive hardware.

[swellman]

C184 from Lucas(3865) second Aurifeuillian factor

p56 factor: 11650548468013607480636572444551824125879428408390406561
p129 factor: 107421293507746607930099795638622153254964830031896106394841885992088863386544251790675380303933540799807850243915748690952737701

(Releases primal scream)

533M raw / 409M unique, log attached

[swellman]

C233_133_86 factored

prp53 factor: 48123005051671884100596570627622910027934285907981289
prp180 factor: 220961970465109701680862878060671110534178284477988742932023352385881799781006963132654422483371568642550168863830804058261022202461328997384502687247398959971962088052635841973157

(silently stares at screen)

459M raw / 360M unique

[Dubslow]

lol

[VBCurtis]

Originally Posted by swellman

C233_133_86 factored
(silently stares at screen)

Who ECMs these days, anyway?

Thanks to you, I'll finish that t55 on my next 14e submission...

[swellman]

I ran 15k curves @B1=43e6, then another 12k curves @B1=3e8 on the L3865 composite. Unlucky seems a fair assessment, but I’m a bit biased. Reported this result to Marin M.

Yoyo@Home ran a full t55 on C233_133_86, i.e. 18000 curves @B1=11e7. Plus I ran a touch more (~800 curves @B1=3e8). A p53?! Really?

Them’s the breaks...

[Dubslow]

Originally Posted by Dubslow

I'll take C209_96656723_29

matrix is 19883667 x 19883915 (6153.1 MB) with weight 1520507965 (76.47/col)
sparse part has weight 1414151927 (71.12/col)
using block size 8192 and superblock size 589824 for processor cache size 6144 kB
commencing Lanczos iteration (4 threads)
memory use: 5194.7 MB
linear algebra at 0.0%, ETA 400h28m883915 dimensions (0.0%, ETA 400h28m)    
checkpointing every 50000 dimensions83915 dimensions (0.0%, ETA 406h33m)    
linear algebra completed 19883509 of 19883915 dimensions (100.0%, ETA 0h 0m)     
lanczos halted after 314436 iterations (dim = 19883640)
recovered 22 nontrivial dependencies
BLanczosTime: 1433833

commencing square root phase
handling dependencies 1 to 64
reading relations for dependency 1
read 9941909 cycles
cycles contain 27874064 unique relations
read 27874064 relations
multiplying 27874064 relations
multiply complete, coefficients have about 765.64 million bits
initial square root is modulo 7395469
GCD is 1, no factor found
reading relations for dependency 2
read 9940716 cycles
cycles contain 27871564 unique relations
read 27871564 relations
multiplying 27871564 relations
multiply complete, coefficients have about 765.58 million bits
initial square root is modulo 7386667
GCD is N, no factor found
reading relations for dependency 3
read 9941408 cycles
cycles contain 27880866 unique relations
read 27880866 relations
multiplying 27880866 relations
multiply complete, coefficients have about 765.83 million bits
initial square root is modulo 7424887
sqrtTime: 10099
p58 factor: 6772755222280762934306144869239832781600209926161442442841
p151 factor: 5633673172606053749567341300705965212594925381326504811982365282795063601234344646337559626043606810943430235730154306584910699812115366158348985111879
elapsed time 403:50:31

https://pastebin.com/TSR27JYV