Cunningham ECM efforts - Page 7

henryzz · 2022-11-28, 10:41

What proportion of users would be lost due to raising memory limits? Maybe it is time to create a new queue with larger limits. I suspect you could afford to lose a sizeable proportion as higher limits would increase speed.

Would bsquared's recent improvements adding ecm help compensate for lower fb bounds? I realise this is AVX512 only currently be that should be fixable with the correct motivation.

pinhodecarlos · 2022-12-03, 08:29

There are plenty of machines with 2GB per thread. I'm with Henry, new app, make sure you are very clear it will use at least 1.5GB going up to 1.8GB, double or triple the boinc points per Wu, explain the situation in clear mode what we will run on it, set this new app as off by default (client needs to check it).

yoyo · 2022-12-03, 08:51

In yoyo@home I run ecm workunits in Boinc with up to 10 GB RAM.

pinhodecarlos · 2022-12-06, 19:26

I can already see first Wu's out for S2L2194 (2,2194L). Suspect within a week's time 2,1097+ first attempt to post-process will be carried out by Greg.

swellman · 2022-12-10, 03:31

2,2398M is now in active ECM on Yoyo@Home. This is the last of the batch of base-2 Cunninghams (1987) to undergo 9900 curves @B1=2.9e9. It will take a month or so to finish Phase 1 and maybe 2+ months to complete Phase 2. But it is now processing. Only took 22 months to chew through the list!

Both 2,1109+ and 2,2222L are now enqueued with Yoyo, set for 9900 curves @B1=7.6e9. Planning to enqueue the two remaining SNFS 334 composites, namely 2,1108+ and 2,2222M, sometime this weekend. Trying to stay ahead of NFS@Home.

Beyond that, I’ll likely grab the three SNFS 338 composites: 2,2246L, 2,1124+ and 2,1123+ which should take us into May-June(?)

swellman · 2022-12-19, 13:06

2,1109+ has just gone live on Yoyo@Home. Expecting it to take until late January to chew up all the curves in phase 1, much longer to complete phase 2.

Three other SNFS 334 jobs are waiting in Yoyo’s queue.

frmky · 2022-12-19, 23:12

Quote:

Originally Posted by pinhodecarlos

Suspect within a week's time 2,1097+ first attempt to post-process will be carried out by Greg.

2,1097+ is done. The 126M matrix took 98 hours to solve using 4x nVidia A100 gpus.

xilman · 2022-12-20, 11:37

Quote:

Originally Posted by frmky

2,1097+ is done. The 126M matrix took 98 hours to solve using 4x nVidia A100 gpus.

Impressive.

I remember the days when factoring a 111 digit general composite was hard. Now they are showing up as the smaller prime factor.

frmky · 2022-12-20, 19:06

Quote:

Originally Posted by xilman

I remember the days when factoring a 111 digit general composite was hard. Now they are showing up as the smaller prime factor.

Or a 151 digit number as the penultimate prime factor of 7,889M!

swellman · 2023-01-03, 03:19

Quote:

Originally Posted by swellman

2,1109+ has just gone live on Yoyo@Home. Expecting it to take until late January to chew up all the curves in phase 1, much longer to complete phase 2.

Three other SNFS 334 jobs are waiting in Yoyo’s queue.

As a mark on the wall, 2,2222L just started ECM tonight. Only two weeks between consecutive job starts within the recent batch of B1=7.6e9! Unknown if Yoyo can keep up this pace but it is encouraging.

swellman · 2023-01-08, 12:02

Yoyo found a p67 in 2,2222L, thereby fully factoring it!

2022-12-03, 08:29	#68
pinhodecarlos "Carlos Pinho" Oct 2011 Milton Keynes, UK 3×7×13×19 Posts	There are plenty of machines with 2GB per thread. I'm with Henry, new app, make sure you are very clear it will use at least 1.5GB going up to 1.8GB, double or triple the boinc points per Wu, explain the situation in clear mode what we will run on it, set this new app as off by default (client needs to check it). Last fiddled with by pinhodecarlos on 2022-12-03 at 08:30

2022-12-06, 19:26	#70
pinhodecarlos "Carlos Pinho" Oct 2011 Milton Keynes, UK 3×7×13×19 Posts	I can already see first Wu's out for S2L2194 (2,2194L). Suspect within a week's time 2,1097+ first attempt to post-process will be carried out by Greg. Last fiddled with by pinhodecarlos on 2022-12-06 at 19:28

2023-01-08, 12:02	#77
swellman Jun 2012 7×11×53 Posts	p67 Yoyo found a p67 in 2,2222L, thereby fully factoring it!

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Recommended bases and efforts	gd_barnes	Conjectures 'R Us	190	2023-02-26 09:01
Doublecheck efforts; S66/S79 to start with	gd_barnes	Conjectures 'R Us	16	2014-08-07 02:11
Cunningham ECM Now Futile?	R.D. Silverman	GMP-ECM	4	2012-04-25 02:45
ECM efforts mistake?	10metreh	mersennewiki	1	2008-12-28 13:31
ECM Efforts	R.D. Silverman	Factoring	63	2005-06-24 13:41

2022-11-28, 10:41	#67
henryzz Just call me Henry "David" Sep 2007 Liverpool (GMT/BST) 2³×19×41 Posts	What proportion of users would be lost due to raising memory limits? Maybe it is time to create a new queue with larger limits. I suspect you could afford to lose a sizeable proportion as higher limits would increase speed. Would bsquared's recent improvements adding ecm help compensate for lower fb bounds? I realise this is AVX512 only currently be that should be fixable with the correct motivation.

2022-12-03, 08:51	#69
yoyo Oct 2006 Berlin, Germany 2²×3×5×11 Posts	In yoyo@home I run ecm workunits in Boinc with up to 10 GB RAM.

2022-12-10, 03:31	#71
swellman Jun 2012 7·11·53 Posts	2,2398M is now in active ECM on Yoyo@Home. This is the last of the batch of base-2 Cunninghams (1987) to undergo 9900 curves @B1=2.9e9. It will take a month or so to finish Phase 1 and maybe 2+ months to complete Phase 2. But it is now processing. Only took 22 months to chew through the list! Both 2,1109+ and 2,2222L are now enqueued with Yoyo, set for 9900 curves @B1=7.6e9. Planning to enqueue the two remaining SNFS 334 composites, namely 2,1108+ and 2,2222M, sometime this weekend. Trying to stay ahead of NFS@Home. Beyond that, I’ll likely grab the three SNFS 338 composites: 2,2246L, 2,1124+ and 2,1123+ which should take us into May-June(?)

2022-12-19, 13:06	#72
swellman Jun 2012 7·11·53 Posts	2,1109+ has just gone live on Yoyo@Home. Expecting it to take until late January to chew up all the curves in phase 1, much longer to complete phase 2. Three other SNFS 334 jobs are waiting in Yoyo’s queue.