Potential New Team Aliquot Sequence Project (18528) - Page 3

VBCurtis · 2021-02-27, 16:14

Jagotu-
What would help is a summary (most of this info is listed at the end of the log):

Time spent on poly select (you can sum size and root, or report them separately)
Time spent on sieving
Time spent on linear algebra
Q-range sieved (not in summary)
Number of relations needed, if the first filtering pass didn't build a matrix.

The relative timings of the 3 phases help me tune the target relations count- if the matrix step is too long, add relations and vice versa.

The Q-range helps to set Q-initial. We target a ratio of 7 to 9 for Q-final to Q-initial. If that ratio gets too big, we get extra duplicate relations and should increase starting-Q; if that ratio is too small we're missing out on speed by not sieving small Q so we should decrease starting-Q.

jagotu · 2021-02-28, 12:16

Next up is a C140, which I ECM'd to t43.08 (yafu default). I won't have access to the nfs machine until next sunday, so if anyone else wants to try their luck with a bigger-ish factoring job, feel free to. If noone does I'll get to it next week.

VBCurtis:
I'll get you the numbers once I get access to the machine. I remember the initial relations weren't nearly enough, there were like 4-5 relation increases necessary. But I'll know for sure once I get back to my machine.

VBCurtis · 2021-02-28, 16:32

I'll knock out the C140 today.

VBCurtis · 2021-03-01, 06:43

C140 done, next line has zero ECM done. Ready for jagotu to continue, if interested.

jagotu · 2021-03-01, 11:11

Sure, I'll move it ahead a bit :)

jagotu · 2021-03-02, 16:37

Next up is a C133, that I ecm'd to t40 (2557 curves @ 3M). Again, anybody feel free to knock it out now, or I'll get to it with my nfs beast on sunday.

EdH · 2021-03-02, 16:55

Quote:

Originally Posted by jagotu

Sure, I'll move it ahead a bit :)

You advanced the sequence pretty well. Thanks!

Quote:

Originally Posted by jagotu

Next up is a C133, that I ecm'd to t40 (2557 curves @ 3M). Again, anybody feel free to knock it out now, or I'll get to it with my nfs beast on sunday.

I, like charybdis, will wait until larger sizes show up, so if anyone would like to wet their feet with a somewhat smaller composite, please join in. Or, if it is still available, jagotu may feel free to continue.

The only request is that if someone does start NFS on these, post a claim so duplicate work is minimized.

Thanks to all for the effort in this.

jagotu · 2021-03-07, 09:22

As no one else showed interest I'm gonna keep moving the sequence myself for now :)

VBCurtis-
I lost the original polyselect time and used a msieve poly, but here's the rest of the stats.

Code:

PID12252 2021-02-27 09:06:36,892 Info:Square Root: Total cpu/real time for sqrt: 6638.88/605.567
PID12252 2021-02-27 09:06:36,893 Info:Filtering - Singleton removal: Total cpu/real time for purge: 2002.05/1290.05
PID12252 2021-02-27 09:06:36,893 Info:Quadratic Characters: Total cpu/real time for characters: 177.73/36.9895
PID12252 2021-02-27 09:06:36,893 Info:Linear Algebra: Total cpu/real time for bwc: 104006/13794.1
PID12252 2021-02-27 09:06:36,893 Info:Linear Algebra: Aggregate statistics:
PID12252 2021-02-27 09:06:36,893 Info:Linear Algebra: Krylov: CPU time 63639.9, WCT time 8464.69, iteration CPU time 0.07, COMM 0.01, cpu-wait 0.0, comm-wait 0.0 (114000 iterations)
PID12252 2021-02-27 09:06:36,893 Info:Linear Algebra: Lingen CPU time 1553.49, WCT time 99.41
PID12252 2021-02-27 09:06:36,893 Info:Linear Algebra: Mksol: CPU time 35393.31,  WCT time 4685.03, iteration CPU time 0.08, COMM 0.01, cpu-wait 0.0, comm-wait 0.0 (57000 iterations)
PID12252 2021-02-27 09:06:36,893 Info:Filtering - Duplicate Removal, removal pass: Total cpu/real time for dup2: 3720.93/2629.5
PID12252 2021-02-27 09:06:36,893 Info:Filtering - Duplicate Removal, removal pass: Aggregate statistics:
PID12252 2021-02-27 09:06:36,893 Info:Filtering - Duplicate Removal, removal pass: CPU time for dup2: 1862.1s
PID12252 2021-02-27 09:06:36,893 Info:Lattice Sieving: Aggregate statistics:
PID12252 2021-02-27 09:06:36,894 Info:Lattice Sieving: Total number of relations: 112300802
PID12252 2021-02-27 09:06:36,894 Info:Lattice Sieving: Average J: 8190.44 for 826803 special-q, max bucket fill -bkmult 1.0,1s:1.081120
PID12252 2021-02-27 09:06:36,894 Info:Lattice Sieving: Total time: 1.44759e+06s
PID12252 2021-02-27 09:06:36,894 Info:Filtering - Merging: Total cpu/real time for merge: 460.13/60.5791
PID12252 2021-02-27 09:06:36,894 Info:Filtering - Merging: Total cpu/real time for replay: 142.89/122.54
PID12252 2021-02-27 09:06:36,894 Info:Generate Free Relations: Total cpu/real time for freerel: 2345.43/148.544
PID12252 2021-02-27 09:06:36,894 Info:Generate Factor Base: Total cpu/real time for makefb: 26.13/3.23953
PID12252 2021-02-27 09:06:36,894 Info:Filtering - Duplicate Removal, splitting pass: Total cpu/real time for dup1: 708.25/473.076
PID12252 2021-02-27 09:06:36,894 Info:Filtering - Duplicate Removal, splitting pass: Aggregate statistics:
PID12252 2021-02-27 09:06:36,894 Info:Filtering - Duplicate Removal, splitting pass: CPU time for dup1: 471.9s
PID12252 2021-02-27 09:06:36,894 Info:Square Root: Total cpu/real time for sqrt: 6638.88/605.567

Q range sieved: 10000000-10000000

jagotu · 2021-03-11, 23:32

The sequence reached a C157 that didn't want to give under ECM.

I did a full t45 (4480@11e6) and an additional 1200@43e6.
Yafu wanted to reach t48.31 (total 3821@43e6, so 2621@43e6 more than I did).

As the number is out of my ballpark, I'm gonna leave it to someone's bigger powerhouse.

EdH · 2021-03-11, 23:46

Quote:

Originally Posted by jagotu

The sequence reached a C157 that didn't want to give under ECM.

I did a full t45 (4480@11e6) and an additional 1200@43e6.
Yafu wanted to reach t48.31 (total 3821@43e6, so 2621@43e6 more than I did).

As the number is out of my ballpark, I'm gonna leave it to someone's bigger powerhouse.

Thank you for what you did. Let's see if someone else might want to run it, before I look it over. Maybe another smaller composite will make an appearance for you to tackle.

VBCurtis · 2021-03-12, 00:29

Crack it, Ed.

2021-02-28, 12:16	#24
jagotu Oct 2019 5·7 Posts	i1416 C140 Next up is a C140, which I ECM'd to t43.08 (yafu default). I won't have access to the nfs machine until next sunday, so if anyone else wants to try their luck with a bigger-ish factoring job, feel free to. If noone does I'll get to it next week. VBCurtis: I'll get you the numbers once I get access to the machine. I remember the initial relations weren't nearly enough, there were like 4-5 relation increases necessary. But I'll know for sure once I get back to my machine.

2021-03-02, 16:37	#28
jagotu Oct 2019 5·7 Posts	C133 i1423 Next up is a C133, that I ecm'd to t40 (2557 curves @ 3M). Again, anybody feel free to knock it out now, or I'll get to it with my nfs beast on sunday.

2021-03-11, 23:32	#31
jagotu Oct 2019 5×7 Posts	C157 i1436 The sequence reached a C157 that didn't want to give under ECM. I did a full t45 (4480@11e6) and an additional 1200@43e6. Yafu wanted to reach t48.31 (total 3821@43e6, so 2621@43e6 more than I did). As the number is out of my ballpark, I'm gonna leave it to someone's bigger powerhouse.

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Aliquot sequence reservations	schickel	Aliquot Sequences	3439	2021-04-16 16:41
Another Aliquot Sequence site	schickel	Aliquot Sequences	67	2012-01-20 17:53
Aliquot sequence worker for factordb	yoyo	FactorDB	6	2012-01-12 20:58
YA aliquot-sequence-chasing script	fivemack	Aliquot Sequences	5	2009-09-28 16:40
Useful aliquot-sequence links	10metreh	Aliquot Sequences	2	2009-07-31 17:43

2021-02-27, 16:14	#23
VBCurtis "Curtis" Feb 2005 Riverside, CA 4737₁₀ Posts	Jagotu- What would help is a summary (most of this info is listed at the end of the log): Time spent on poly select (you can sum size and root, or report them separately) Time spent on sieving Time spent on linear algebra Q-range sieved (not in summary) Number of relations needed, if the first filtering pass didn't build a matrix. The relative timings of the 3 phases help me tune the target relations count- if the matrix step is too long, add relations and vice versa. The Q-range helps to set Q-initial. We target a ratio of 7 to 9 for Q-final to Q-initial. If that ratio gets too big, we get extra duplicate relations and should increase starting-Q; if that ratio is too small we're missing out on speed by not sieving small Q so we should decrease starting-Q.

2021-02-28, 16:32	#25
VBCurtis "Curtis" Feb 2005 Riverside, CA 4737₁₀ Posts	I'll knock out the C140 today.

2021-03-01, 06:43	#26
VBCurtis "Curtis" Feb 2005 Riverside, CA 3×1,579 Posts	C140 done, next line has zero ECM done. Ready for jagotu to continue, if interested.

2021-03-01, 11:11	#27
jagotu Oct 2019 5·7 Posts	Sure, I'll move it ahead a bit :)

2021-03-12, 00:29	#33
VBCurtis "Curtis" Feb 2005 Riverside, CA 1001010000001₂ Posts	Crack it, Ed.