Reserved for MF - Sequence 3408 - Page 18

swellman · 2015-10-24, 11:39

Just over 1200 additional curves so far with no factor found. Full t50 should finish by Sunday night. Or I can shift to B1=110M for 1-2k additional curves.

Or maybe it's time to abandon ECM? Your call. For now I'll just carry on towards t50 unless you say otherwise.

- Sean

VBCurtis · 2015-10-24, 17:22

How about a couple hundred at 110M, and call it good tonight? I'll start NFS Sun morning USA time.
Edit- anything above half a T50 is sufficient. I didn't calculate how much has been done; if we're past 0.5*T50, no more is needed.

swellman · 2015-10-24, 20:25

So far, ECM has run ~2300 additional curves @B1=43M. In a couple of hours I will kick it up to B1=110M and let it run overnight. I'm on the east coast and have an early start tomorrow, so I can post yea or nay on any factors being found before you start NFS.

Does this work for you? I believe that will push it to well past 0.5t50 and provide sufficient due diligence to avoid a blatant ECM miss.

VBCurtis · 2015-10-24, 20:28

That's terrific, thanks for the extra effort. I'll test-sieve today, and will be ready to fire NFS when I wake up tomorrow.

I haven't previously tried running my full desktop on a single NFS task. Going to try 12-threaded sieving on a 6-core i7. Hopefully, I'll have factors in under a week!

swellman · 2015-10-24, 20:39

Quote:

Originally Posted by VBCurtis

I haven't previously tried running my full desktop on a single NFS task. Going to try 12-threaded sieving on a 6-core i7. Hopefully, I'll have factors in under a week!

Want!

VBCurtis · 2015-10-24, 22:07

Rich's 2nd poly sieves about 15% faster than the first, with yield also higher.
I used alim=rlim=33M, 31-bit LP, 2 large primes per side.
I test-sieved at alim/6, alim/3, alim/2, alim, a range of 1500q each (long enough for sec/rel to stabilize to two sigfigs)

Yield was bad at alim/6, so I will start the run at alim/4 (and with alim 35M instead of 33).
Yields at the other test points were roughly the same, around 4.5. For 140M raw relations, that's a range of about 32Mq, roughly q = 9M to 41M.

swellman · 2015-10-24, 23:36

Ran ECM up to 2680 additional curves @B1=43M, plus the 1550 curves run earlier = 4230 curves, or ~55% of t50.

Now running more curves with B1=110M. Getting throughput of ~66 curves/hr. Should have >700 curves by morning. I'll post the results about 6 AM EDT.

swellman · 2015-10-25, 10:23

Ran another 760 curves @B1=110M with nothing found. Terminated ECM.

It belongs to NFS now. Happy factoring!

swellman · 2015-10-28, 12:35

Quote:

Originally Posted by VBCurtis

That's terrific, thanks for the extra effort. I'll test-sieve today, and will be ready to fire NFS when I wake up tomorrow.

I haven't previously tried running my full desktop on a single NFS task. Going to try 12-threaded sieving on a 6-core i7. Hopefully, I'll have factors in under a week!

How is the speed looking? Just curious about that "12-headed monster".

VBCurtis · 2015-10-28, 22:01

79 hrs wallclock time, 78.3M relations of estimated 143M raw rels needed. If a matrix builds the first try, that'll make nearly exactly 6 days sieve time.

There is a single thread of srsieve running, so NFS has 12/13ths of the CPU. I doubt the matrix will solve in under 24 hr, so it will take me more than a week to factor. :-(

swellman · 2015-10-29, 02:13

8-9 days for a 157 GNFS? Not too shabby for a single machine! Let's hope this sequence starts moving downward.

I'm willing to throw some more ECM into the mix if needed later. Just call for help here. Good luck!

2015-10-24, 17:22	#189
VBCurtis "Curtis" Feb 2005 Riverside, CA 3³×173 Posts	How about a couple hundred at 110M, and call it good tonight? I'll start NFS Sun morning USA time. Edit- anything above half a T50 is sufficient. I didn't calculate how much has been done; if we're past 0.5T50, no more is needed. Last fiddled with by VBCurtis on 2015-10-24 at 17:28*

2015-10-24, 20:25	#190
swellman Jun 2012 101110010000₂ Posts	So far, ECM has run ~2300 additional curves @B1=43M. In a couple of hours I will kick it up to B1=110M and let it run overnight. I'm on the east coast and have an early start tomorrow, so I can post yea or nay on any factors being found before you start NFS. Does this work for you? I believe that will push it to well past 0.5t50 and provide sufficient due diligence to avoid a blatant ECM miss. Last fiddled with by swellman on 2015-10-24 at 20:31

2015-10-28, 22:01	#197
VBCurtis "Curtis" Feb 2005 Riverside, CA 4671₁₀ Posts	79 hrs wallclock time, 78.3M relations of estimated 143M raw rels needed. If a matrix builds the first try, that'll make nearly exactly 6 days sieve time. There is a single thread of srsieve running, so NFS has 12/13ths of the CPU. I doubt the matrix will solve in under 24 hr, so it will take me more than a week to factor. :-( Last fiddled with by VBCurtis on 2015-10-28 at 22:02

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Reserved for MF - Sequence 3366	RichD	Aliquot Sequences	468	2021-01-27 01:16
Reserved for MF - Sequence 4788	schickel	Aliquot Sequences	2934	2021-01-07 18:52
Reserved for MF - Sequence 276	kar_bon	Aliquot Sequences	127	2020-12-17 10:05
Team Sieve #37: 3408:i1287	RichD	Aliquot Sequences	14	2013-08-02 17:02
80M to 64 bits ... but not really reserved	petrw1	Lone Mersenne Hunters	82	2010-01-11 01:57

2015-10-24, 11:39	#188
swellman Jun 2012 2960₁₀ Posts	Just over 1200 additional curves so far with no factor found. Full t50 should finish by Sunday night. Or I can shift to B1=110M for 1-2k additional curves. Or maybe it's time to abandon ECM? Your call. For now I'll just carry on towards t50 unless you say otherwise. - Sean

2015-10-24, 20:28	#191
VBCurtis "Curtis" Feb 2005 Riverside, CA 3³·173 Posts	That's terrific, thanks for the extra effort. I'll test-sieve today, and will be ready to fire NFS when I wake up tomorrow. I haven't previously tried running my full desktop on a single NFS task. Going to try 12-threaded sieving on a 6-core i7. Hopefully, I'll have factors in under a week!

2015-10-24, 22:07	#193
VBCurtis "Curtis" Feb 2005 Riverside, CA 1001000111111₂ Posts	Rich's 2nd poly sieves about 15% faster than the first, with yield also higher. I used alim=rlim=33M, 31-bit LP, 2 large primes per side. I test-sieved at alim/6, alim/3, alim/2, alim, a range of 1500q each (long enough for sec/rel to stabilize to two sigfigs) Yield was bad at alim/6, so I will start the run at alim/4 (and with alim 35M instead of 33). Yields at the other test points were roughly the same, around 4.5. For 140M raw relations, that's a range of about 32Mq, roughly q = 9M to 41M.

2015-10-24, 23:36	#194
swellman Jun 2012 B90₁₆ Posts	Ran ECM up to 2680 additional curves @B1=43M, plus the 1550 curves run earlier = 4230 curves, or ~55% of t50. Now running more curves with B1=110M. Getting throughput of ~66 curves/hr. Should have >700 curves by morning. I'll post the results about 6 AM EDT.

2015-10-25, 10:23	#195
swellman Jun 2012 2⁴·5·37 Posts	Ran another 760 curves @B1=110M with nothing found. Terminated ECM. It belongs to NFS now. Happy factoring!

2015-10-29, 02:13	#198
swellman Jun 2012 2⁴·5·37 Posts	8-9 days for a 157 GNFS? Not too shabby for a single machine! Let's hope this sequence starts moving downward. I'm willing to throw some more ECM into the mix if needed later. Just call for help here. Good luck!