mersenneforum.org Reserved for MF - Sequence 4788
 Register FAQ Search Today's Posts Mark Forums Read

2009-03-24, 07:37   #23
10metreh

Nov 2008

2×33×43 Posts

Quote:
 Originally Posted by mdettweiler The next line (2358) has a C92 that survived full ECM. My resources are completely tied up for tonight, so if anyone else wants to do QS on this, go ahead. If nobody's grabbed it by the time I get on tomorrow (probably late morning) I'll do it myself since things should be freed up a bit by then. The number is: Code: 23815769543009537132187278752799013868767385674615392411879275450950012601757447373080004859
I'd like to see this done by this afternoon/early evening (GMT), because this gives me most time to compute more.

2009-03-24, 07:38   #24
mdettweiler
A Sunny Moo

Aug 2007
USA (GMT-5)

11000011000012 Posts

Quote:
 Originally Posted by 10metreh I'd like to see this done by this afternoon/early evening (GMT), because this gives me most time to compute more.
Okay, no problem. I presume, thus, that you'll do it? (considering that it's 7AM GMT right now)

2009-03-24, 07:47   #25
10metreh

Nov 2008

1001000100102 Posts

Quote:
 Originally Posted by mdettweiler Okay, no problem. I presume, thus, that you'll do it? (considering that it's 7AM GMT right now)
Nope - no computer access for a while. I won't have much time until the weekend to do jobs much larger than C85.

Last fiddled with by 10metreh on 2009-03-24 at 07:49

 2009-03-24, 08:21 #26 axn     Jun 2003 126C16 Posts Msieve is cranking away on the C92 even as we speak! ETA: 2 hours.
2009-03-24, 11:22   #27
mklasson

Feb 2004

2·3·43 Posts

Quote:
 Originally Posted by mdettweiler Oh, wait! I just realized what's going on. henryzz, are you by chance using the 64-bit version of gnfs-lasieve4I12e? I'm using the 32-bit version, which of course is about half as fast--so that would explain the discrepancy perfectly.
I'm not so sure about that. The 64-bit version is significantly slower than the 32-bit version on my system... I don't know why, but I've seen other people report the same problem (which was why I checked and noticed it in the first place).

Maybe comparing yield over a certain range is a more portable way of measuring the poly quality?

2009-03-24, 12:30   #28
axn

Jun 2003

22·32·131 Posts

Quote:
 Originally Posted by mklasson Maybe comparing yield over a certain range is a more portable way of measuring the poly quality?
Without knowing the actual factor base sizes, large prime bounds, etc.., it is an apples to oranges comparison, innit?

2009-03-24, 12:39   #29
mklasson

Feb 2004

2·3·43 Posts

Quote:
 Originally Posted by axn Without knowing the actual factor base sizes, large prime bounds, etc.., it is an apples to oranges comparison, innit?
I'm not suggesting it should be used for anything serious... It just seemed a possibly better idea for them than comparing the sec/rel with different processors.

2009-03-24, 12:47   #30
10metreh

Nov 2008

2×33×43 Posts

Quote:
 Originally Posted by axn Msieve is cranking away on the C92 even as we speak! ETA: 2 hours.
And... (I'm not at home, so I won't be able to do much work on the next iteraton(s))

Last fiddled with by 10metreh on 2009-03-24 at 12:49

2009-03-24, 12:50   #31
axn

Jun 2003

22·32·131 Posts

Quote:
 Originally Posted by mklasson I'm not suggesting it should be used for anything serious... It just seemed a possibly better idea for them than comparing the sec/rel with different processors.
Even though I quoted your post specifically, it was intended at the general discussion. The two different speeds posted here for the two different polys didn't have enough supporting information to tell if it is an anomaly or par for the course (quite possible that both are using the default values for factMsieve, but we don't know). I was just trying to highlight that fact

Anyway, the difference between the two CPUs (3GHz Q6600 (4MB L2) vs 2.2GHz E4500 (1MB L2)) is enough to explain the bulk of the discrepancy.

For proper comparison of two polys, the SOP is to trial sieve at different q-values with the same set of parameters, and look at the sieve speed (which is IMO, better than yield count) for the _same_ processor.

EDIT:- The discussion is getting pretty offtopic. So...

Last fiddled with by axn on 2009-03-24 at 12:52 Reason: offtopic

2009-03-24, 13:08   #32
bsquared

"Ben"
Feb 2007

63428 Posts

Quote:
 Originally Posted by mklasson I'm not so sure about that. The 64-bit version is significantly slower than the 32-bit version on my system... I don't know why, but I've seen other people report the same problem (which was why I checked and noticed it in the first place). Maybe comparing yield over a certain range is a more portable way of measuring the poly quality?
There are two 64 bit gnfs-lasieve*Ie versions, one with and one without assembly optimizations. The one with is approximatly twice as fast as the one without (and twice as fast as the 32bit version), but AFAIK, is only available on linux.

2009-03-24, 13:16   #33
mklasson

Feb 2004

2·3·43 Posts

Quote:
 Originally Posted by bsquared There are two 64 bit gnfs-lasieve*Ie versions, one with and one without assembly optimizations. The one with is approximatly twice as fast as the one without (and twice as fast as the 32bit version), but AFAIK, is only available on linux.
Ah, what horrible bad luck for me to be running windows then... As usual, inline asm seems to be the culprit.

 Similar Threads Thread Thread Starter Forum Replies Last Post RichD Aliquot Sequences 431 2020-10-27 23:51 sweety439 And now for something completely different 17 2017-06-13 03:49 frmky Aliquot Sequences 36 2011-04-28 06:27 schickel Aliquot Sequences 51 2011-01-05 02:32 petrw1 Lone Mersenne Hunters 82 2010-01-11 01:57

All times are UTC. The time now is 05:28.

Wed Oct 28 05:28:57 UTC 2020 up 48 days, 2:39, 2 users, load averages: 2.24, 1.94, 1.71