[quote=bsquared;181135]Is it faster than QS even at 90 digits?!?[/quote]

Depends on your processor.

[quote]Also, does it use degree 4 automatically depending on input size or does it need a flag? And would I need a new factMsieve.pl?[/quote]

Beta 3 automatically does degree 4 up to 102 digits.

[quote](I guess this should be moved to the msieve discussion thread...)[/quote]

I don't think so, so I won't.

[QUOTE=bsquared;181135]Is it faster than QS even at 90 digits?!?

Also, does it use degree 4 automatically depending on input size or does it need a flag? And would I need a new factMsieve.pl?

(I guess this should be moved to the msieve discussion thread...)[/QUOTE]

There are a few benchmark posts over there in the "msieve 1.42 beta feedback" thread. According to a benchmarks, it is significantly faster than QS (Yafu, which is faster than msieve QS on a C2D / WinXP 32) for a c92. Another benchmark showed that quartic polynomials are ~40% faster than quintic ones for c95's. The crossover between quartics and quintics is around 100-102 digits.

Msieve 1.42 beta 3 will automatically use degree 4 in a range from 85 to 102 digits.

[quote=bsquared;181115]Also, I'll reserve 15-17M.
[/quote]

Done: 2062521 relations.

Reserving 13-15M

[QUOTE=bsquared;181115]We have 38594402 unique relations (45813049 raw), and msieve isn't happy with that. So if you could post the 49-52 range that would help. Also, I'll reserve 15-17M.[/QUOTE]

049-052.rels.gz is up at the obvious address; reserved 14-15 which should be done inside twelve hours (oh, bother, you reserved it while I was waiting for 049-052 to finish uploading; it's running on machines at work that I can't log into; hey, it's only one evening's compute wasted)

Tom

[quote=fivemack;181156]049-052.rels.gz is up at the obvious address; reserved 14-15 which should be done inside twelve hours (oh, bother, you reserved it while I was waiting for 049-052 to finish uploading; it's running on machines at work that I can't log into; hey, it's only one evening's compute wasted)

Tom[/quote]

Since it's not your electricity and you don't seem that upset about it, I'll let mine keep running. It should be done in the next hour anyway, since I'm running that range on 32 fast CPUs. After that finishes and the download of your range completes, I'll start another filtering run. We'll have about 7.5M more raw relations, so hopefully this will give us a matrix.

2nd filtering run
 

We are now up to 51317468 raw relations (42221599 unique) and no matrix. Log attached. I have all the relations now from 15M-61M, but my 13-15 range is on a different system that has apparently crashed. I'm still trying to figure out from the sysadmin what happened, and I'm not sure when I'll have access again. So consider 13-15 lost (this also effects 2200 CPU hours worth of 2.877- sieving, so I hope the data is not permanently gone). Tom, your range will be useful after all!

I'll try again once we acquire another 5M relations. I'll take 13-14M.

014-015.rels.gz in the obvious place. Taking 12-13.

[quote=fivemack;181229]014-015.rels.gz in the obvious place. Taking 12-13.[/quote]

Downloaded 1138428 rels from 14-15.

Finished 13-14: 1128944 relations.

[B]Reserving 10-12.[/B]

With numbers like this one, I usually had a matrix at about 46M unique relns. The end is near! :smile:

I'll take 9-10 as well. That'll make 6M+ relations by morning if fivemack finishes his results by then. I'll hopefully be able to start the post-processing before I leave for a long weekend camping with the family. With any luck, we'll have factors when I get back on sunday.

[quote=bsquared;181336]I'll take 9-10 as well. That'll make 6M+ relations by morning if fivemack finishes his results by then. I'll hopefully be able to start the post-processing before I leave for a long weekend camping with the family. With any luck, we'll have factors when I get back on sunday.[/quote]

Could you make a perl script that submits factors to the DB automatically?