NFS results and reservations - Page 29

fivemack · 2016-10-19, 09:48

64 threads on E5-2699v4 (I'm assuming it must be that because of the reported L3 cache size) seems a peculiar number; has Ryan written about his approach for choosing this?

On a dual E5 v4 box with 64GB per memory controller I would be fairly inclined to run a 2x2 MPI job to keep down the number of cross-ring memory accesses, but I guess this has been tried and found ineffective.

Next week, all being well, I get to see what MPI-over-10Gbit-Ethernet is like; if I remember Greg rightly, it's not amazing, but I can't recall whether Greg was using 10GbaseT or direct-connect SFP+ for his experiments (the latter is what I have, and should have significantly lower latency).

swellman · 2016-10-19, 11:58

Quote:

Originally Posted by fivemack

64 threads on E5-2699v4 (I'm assuming it must be that because of the reported L3 cache size) seems a peculiar number; has Ryan written about his approach for choosing this?

No he has not brought it up in our communications. Maybe around the forum but I doubt it.

Quote:

On a dual E5 v4 box with 64GB per memory controller I would be fairly inclined to run a 2x2 MPI job to keep down the number of cross-ring memory accesses, but I guess this has been tried and found ineffective.

Next week, all being well, I get to see what MPI-over-10Gbit-Ethernet is like; if I remember Greg rightly, it's not amazing, but I can't recall whether Greg was using 10GbaseT or direct-connect SFP+ for his experiments (the latter is what I have, and should have significantly lower latency).

Good luck!

swellman · 2016-10-23, 13:36

Factored C134_121_115 using GNFS.

Code:

prp63 = 543503370355392774592616452370240981438448800614361173067190499
prp71 = 51969817699619942254710352117648136980643520849181482174053385613952689

swellman · 2016-11-05, 23:23

Quote:

Originally Posted by swellman

Ryan P has started working on C307_149_120 via some ECM/SNFS.

Currently in LA, with an ETA of mid April.

swellman · 2016-12-05, 13:11

Quote:

Originally Posted by swellman

Ryan P has started working on C307_149_120 via some ECM/SNFS.

C307_149_120 has been factored through the efforts of Ryan Propper (sieving - 6 days) and Greg Childers (LA etc - 20 days).

p111*p197

Log file is here http://pastebin.com/n9v8Qxur

Nothing more to say.

LaurV · 2016-12-11, 15:47

Wow! Now we SNFS in the 1024 bit range of the RSA huh?
When GNFS?

Congrats!

swellman · 2016-12-12, 16:54

Ryan Propper has factored C194_139_131 via GNFS

Quote:

p62=61900812667083715395439754704611524371883355591925146043705171
p132=556631937303753659395380600942209216747998682905221986126761613827703949542493079107945373211831009670991345314730268938132525692121

swellman · 2016-12-26, 13:46

Ryan has factored C192_139_81 using SNFS. There was some confusion about who was working on this number - see here. The number has been removed from the 15e queue of NFS@Home. Log file attached.

Code:

p84=918913919767906167873884209042152546683335234911092921560013054824865535890411627651
p108=109699092250470932814216854763718751860024893059519189618637561239511608416240934956915893950561035148579189

swellman · 2017-01-28, 13:48

Ryan is currently sieving C258_142_69.

swellman · 2017-02-13, 22:07

Ryan Propper used SNFS to factor C196_135_94.

p93 * p104

swellman · 2017-02-18, 18:57

Quote:

Originally Posted by swellman

Ryan is currently sieving C258_142_69.

Factored by Ryan using SNFS.

p72*p187

2016-10-23, 13:36	#311
swellman Jun 2012 2²×13×59 Posts	Factored C134_121_115 using GNFS. Code: prp63 = 543503370355392774592616452370240981438448800614361173067190499 prp71 = 51969817699619942254710352117648136980643520849181482174053385613952689

2017-01-28, 13:48	#317
swellman Jun 2012 BFC₁₆ Posts	C258_142_69 Ryan is currently sieving C258_142_69.

2017-02-13, 22:07	#318
swellman Jun 2012 2²×13×59 Posts	C196_135_94 Factored Ryan Propper used SNFS to factor C196_135_94. p93 * p104

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Reservations	ET_	FermatSearch	226	2021-04-25 19:47
ECM results and reservations	swellman	XYYXF Project	281	2018-08-27 01:28
Linear algebra reservations, progress and results	wombatman	NFS@Home	2768	2018-06-02 15:06
Reservations	kar_bon	Riesel Prime Data Collecting (k*2^n-1)	129	2016-09-05 09:23
Where are my reservations	fivemack	PrimeNet	3	2016-02-08 17:58

2016-10-19, 09:48	#309
fivemack (loop (#_fork)) Feb 2006 Cambridge, England 7²×131 Posts	64 threads on E5-2699v4 (I'm assuming it must be that because of the reported L3 cache size) seems a peculiar number; has Ryan written about his approach for choosing this? On a dual E5 v4 box with 64GB per memory controller I would be fairly inclined to run a 2x2 MPI job to keep down the number of cross-ring memory accesses, but I guess this has been tried and found ineffective. Next week, all being well, I get to see what MPI-over-10Gbit-Ethernet is like; if I remember Greg rightly, it's not amazing, but I can't recall whether Greg was using 10GbaseT or direct-connect SFP+ for his experiments (the latter is what I have, and should have significantly lower latency).

2016-12-11, 15:47	#314
LaurV Romulan Interpreter Jun 2011 Thailand 7·1,373 Posts	Wow! Now we SNFS in the 1024 bit range of the RSA huh? When GNFS? Congrats!