C304_150_149 - Page 2

swellman · 2016-08-26, 20:43

Many thanks to all for the helpful tips. Ryan has implemented them and has made progress. He is currently attempting to sieve additional uniques and incrementing TD to find the "sweet spot" wrt ETA. I'm not sure there's anything sweet about this monster but he's engaged it. I'm just hoping he can get an ETA under 4000 hours but who knows? I'll report his progress as things er progress.

swellman · 2016-10-18, 21:34

Update: LA is in full swing with an ETA of -4800 hours (early May).

swellman · 2017-04-07, 18:05

150^149+149^150 has been factored. The mighty giant lies slain.

Ryan ran the sieving, Greg Childers the LA/postprocessing. It was a great effort and personally I am in awe of their accomplishment.

Code:

p112=4589717414070503932356610139809713246293714200289874762254641434246464550705097961311966128119332304236622320109
p192=452371514527436596669656271361055896200135049572909801471049275612551239658558845331878122140222413088563004287778808937264541956636576326085349433639397620869454531909234048542341368429340431

xilman · 2017-04-07, 18:24

Quote:

Originally Posted by swellman

150^149+149^150 has been factored. The mighty giant lies slain.

Ryan ran the sieving, Greg Childers the LA/postprocessing. It was a great effort and personally I am in awe of their accomplishment.

Code:

p112=4589717414070503932356610139809713246293714200289874762254641434246464550705097961311966128119332304236622320109
p192=452371514527436596669656271361055896200135049572909801471049275612551239658558845331878122140222413088563004287778808937264541956636576326085349433639397620869454531909234048542341368429340431

Very nice!

frmky · 2017-04-07, 19:56

Here's the msieve log file for the number:

Batalov · 2017-04-07, 22:18

Most impressive! Kudos!

Batalov · 2017-04-08, 17:51

... This factorization (a 326-digit SNFS difficulty, a clear record peak for quite some time to stay*) is probably a good reason to get all XYYXF webpages refereshed?

The latest update is date-stamped on Jan.4.
_________________
*and not likely to be beaten before current XYYXF pages are extended some day (from 150 to 200, like they were extended when all numbers y<x<=100 were almost finished).

henryzz · 2017-04-08, 21:17

Am I correct in thinking that that was 16x24 threads?

I assume that raising the lp bound would only make the matrix harder. Are the msieve problems fixed with large datasets?

Nice record. Can't wait for it to be beaten.

frmky · 2017-04-09, 06:36

Yes, the linear algebra was completed on 16 compute nodes, each with a Xeon E5-2650 v4 cpu with DDR4 memory connected by FDR Infiniband. This cpu has 12 cores with hyperthreading, so 24 msieve threads per node.

The actual Infiniband data transfer rate during the computation was about 700 MB/s in each direction on each node as measured by the HCA counters.

XYYXF · 2017-04-09, 06:46

Quote:

Originally Posted by Batalov

... This factorization (a 326-digit SNFS difficulty, a clear record peak for quite some time to stay*) is probably a good reason to get all XYYXF webpages refereshed?

The latest update is date-stamped on Jan.4.

You are right!
Going to update them in a day or two.

2017-04-09, 06:36	#20
frmky Jul 2003 So Cal 2106₁₀ Posts	Yes, the linear algebra was completed on 16 compute nodes, each with a Xeon E5-2650 v4 cpu with DDR4 memory connected by FDR Infiniband. This cpu has 12 cores with hyperthreading, so 24 msieve threads per node. The actual Infiniband data transfer rate during the computation was about 700 MB/s in each direction on each node as measured by the HCA counters. Last fiddled with by frmky on 2017-04-09 at 06:44

2016-08-26, 20:43	#12
swellman Jun 2012 2²·13·59 Posts	Many thanks to all for the helpful tips. Ryan has implemented them and has made progress. He is currently attempting to sieve additional uniques and incrementing TD to find the "sweet spot" wrt ETA. I'm not sure there's anything sweet about this monster but he's engaged it. I'm just hoping he can get an ETA under 4000 hours but who knows? I'll report his progress as things er progress.

2016-10-18, 21:34	#13
swellman Jun 2012 3068₁₀ Posts	Update: LA is in full swing with an ETA of -4800 hours (early May).

2017-04-07, 19:56	#16
frmky Jul 2003 So Cal 2·3⁴·13 Posts	Here's the msieve log file for the number:

2017-04-07, 22:18	#17
Batalov "Serge" Mar 2008 Phi(4,2^7658614+1)/2 3⁶×13 Posts	Most impressive! Kudos!

2017-04-08, 17:51	#18
Batalov "Serge" Mar 2008 Phi(4,2^7658614+1)/2 3⁶·13 Posts	... This factorization (a 326-digit SNFS difficulty, a clear record peak for quite some time to stay) is probably a good reason to get all XYYXF webpages refereshed? The latest update is date-stamped on Jan.4. _________________ and not likely to be beaten before current XYYXF pages are extended some day (from 150 to 200, like they were extended when all numbers y<x<=100 were almost finished).

2017-04-08, 21:17	#19
henryzz Just call me Henry "David" Sep 2007 Cambridge (GMT/BST) 13370₈ Posts	Am I correct in thinking that that was 16x24 threads? I assume that raising the lp bound would only make the matrix harder. Are the msieve problems fixed with large datasets? Nice record. Can't wait for it to be beaten.