Reserved for MF - Sequence 4788 - Page 196

EdH · 2013-10-20, 03:44

Quote:

Originally Posted by RichD

I can start the download on Saturday night and begin processing Sunday AM, unless someone else would like to try it. It should be an easy 31-bit with 225M+ rels - as far as 31-bits are easy.

I could attempt this, but I'm not sure whether completion would be timely. Where would I find the relations? Do I have to join NFS@HOME to retrieve them? I could not start until at least tomorrow...

edit: Which polynomial was used?

RichD · 2013-10-20, 05:57

Quote:

Originally Posted by EdH

I could attempt this, but I'm not sure whether completion would be timely. Where would I find the relations? Do I have to join NFS@HOME to retrieve them? I could not start until at least tomorrow...

edit: Which polynomial was used?

I had problems tonight and only got a partial download. You do not need to join NFS@Home. Send a PM to debrouxl and request access to the server. All the .poly, .fb, .ini and other files are in the folder.

I'll keep trying to get the .dat file but it could be another day or two. Feel free to jump in and take over. I'll let you know if I happen to get started but I'm thinking it might be another 48 hours for me.

debrouxl · 2013-10-20, 06:14

Better asking frmky instead of myself

RichD · 2013-10-20, 07:03

Oh yea, frmky is the administrator of the server and debrouxl is the sub-administrator for the lasieved (a.k.a. 14e) jobs but is very active in the queue management. My mistake.

I don't have DSL at home so that is my bottleneck.

This job should take about a week on a Core-i5 system - FWIW.

EdH · 2013-10-20, 13:29

PM sent. I'll see how it goes. I will be attempting to run this across two core2 quad machines using my "new" mpi setup. It will be a good test. I'll post some updates.

EdH · 2013-10-21, 19:29

I'm in the process of retrieving the relations, but I somehow interrupted the download and had to restart. It looks like it will be this evening before I can start. I'll give an estimate once the LA settles in...

EdH · 2013-10-21, 21:24

Quote:

Originally Posted by debrouxl

"C166_4788_5159" sieved by NFS@Home up to 224-225M raw relations.

I am assuming "raw relations" means no duplicate removal or other processing? I know msieve takes care of bad relations and duplicates, but I normally run them against remdups4 whenever I'm manually stepping through. Is this a good practice, or am I wasting that time?

I have had a couple runs where some relations from a previous sieve got mixed into a current one and msieve gave several million error -11 notices, but kept running fine. remdups4, of course, can't find those fumbles...

RichD · 2013-10-22, 00:29

Raw relations are just that, totally raw. They include the good, the bad and the ugly.

The only time I run remdups4 before post-processing is when I want to try to push for a higher target_density, which reduces BL time. Saves a little bit on multiple filtering runs if that is the case.

This is nicely over-sieved. You might be able to get close to 110 TD.

P.S. I haven't had a chance to finish the download, I was going to be your backup. I would like it if you proved I was not needed.

EdH · 2013-10-22, 02:05

Quote:

Originally Posted by RichD

Raw relations are just that, totally raw. They include the good, the bad and the ugly.

The only time I run remdups4 before post-processing is when I want to try to push for a higher target_density, which reduces BL time. Saves a little bit on multiple filtering runs if that is the case.

This is nicely over-sieved. You might be able to get close to 110 TD.

P.S. I haven't had a chance to finish the download, I was going to be your backup. I would like it if you proved I was not needed.

I would like to prove that, too!

The d/l just finished. I'm making a copy prior to decompression and turning it loose...

EdH · 2013-10-22, 17:07

I'm afraid I might be up against a memory problem. I should have asked if 4G in the host machine was enough. It has not returned from "building initial matrix" after two hours. The machine is sluggish over remote vnc. top, via ssh, shows I have less than 90k free memory and swap is using over 2.5G of 5 total.

I had hoped that mpi across two machines would take care of the memory need, but the matrix file construction happens only on the host machine. It is still "trying" to succeed, but it might be a prohibitively slow process.

I can do some memory juggling and perhaps I will in a while, if I don't see any real progress. I think using 2.5G of swap is a "telling" signal.

I am still going to try to solve this, but I will definitely be slower than expected...

EdH · 2013-10-22, 18:39

Update: I have doubled the RAM to 8G in the host machine and moved to a secondary with 4G RAM. The host is using >7.8G, but at least the swap is showing 0!

more later...

Edit: It's later! Does it seem right that this should take a month?

Code:

linear algebra at 0.0%, ETA 708h57m

I'm not sure my attention can span that long, let alone the occasional power bump. Hopefully the power bumps will be smoothed by the UPSs I installed, but still...

2013-10-22, 18:39	#2156
EdH "Ed Hall" Dec 2009 Adirondack Mtns 111011111110₂ Posts	Update: I have doubled the RAM to 8G in the host machine and moved to a secondary with 4G RAM. The host is using >7.8G, but at least the swap is showing 0! more later... Edit: It's later! Does it seem right that this should take a month? Code: linear algebra at 0.0%, ETA 708h57m I'm not sure my attention can span that long, let alone the occasional power bump. Hopefully the power bumps will be smoothed by the UPSs I installed, but still... Last fiddled with by EdH on 2013-10-22 at 19:17

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Reserved for MF - Sequence 3366	RichD	Aliquot Sequences	470	2021-04-22 02:17
Reserved for MF - Sequence 3408	RichD	Aliquot Sequences	474	2021-03-07 20:28
Reserved for MF - Sequence 276	kar_bon	Aliquot Sequences	127	2020-12-17 10:05
Assignments are reserved but not showing up	prism019	GPU to 72	6	2020-09-21 22:11
80M to 64 bits ... but not really reserved	petrw1	Lone Mersenne Hunters	82	2010-01-11 01:57

2013-10-20, 06:14	#2148
debrouxl Sep 2009 977 Posts	Better asking frmky instead of myself

2013-10-20, 07:03	#2149
RichD Sep 2008 Kansas 3,389 Posts	Oh yea, frmky is the administrator of the server and debrouxl is the sub-administrator for the lasieved (a.k.a. 14e) jobs but is very active in the queue management. My mistake. I don't have DSL at home so that is my bottleneck. This job should take about a week on a Core-i5 system - FWIW.

2013-10-20, 13:29	#2150
EdH "Ed Hall" Dec 2009 Adirondack Mtns 3838₁₀ Posts	PM sent. I'll see how it goes. I will be attempting to run this across two core2 quad machines using my "new" mpi setup. It will be a good test. I'll post some updates.

2013-10-21, 19:29	#2151
EdH "Ed Hall" Dec 2009 Adirondack Mtns 2×19×101 Posts	I'm in the process of retrieving the relations, but I somehow interrupted the download and had to restart. It looks like it will be this evening before I can start. I'll give an estimate once the LA settles in...

2013-10-22, 00:29	#2153
RichD Sep 2008 Kansas D3D₁₆ Posts	Raw relations are just that, totally raw. They include the good, the bad and the ugly. The only time I run remdups4 before post-processing is when I want to try to push for a higher target_density, which reduces BL time. Saves a little bit on multiple filtering runs if that is the case. This is nicely over-sieved. You might be able to get close to 110 TD. P.S. I haven't had a chance to finish the download, I was going to be your backup. I would like it if you proved I was not needed.

2013-10-22, 17:07	#2155
EdH "Ed Hall" Dec 2009 Adirondack Mtns 2×19×101 Posts	I'm afraid I might be up against a memory problem. I should have asked if 4G in the host machine was enough. It has not returned from "building initial matrix" after two hours. The machine is sluggish over remote vnc. top, via ssh, shows I have less than 90k free memory and swap is using over 2.5G of 5 total. I had hoped that mpi across two machines would take care of the memory need, but the matrix file construction happens only on the host machine. It is still "trying" to succeed, but it might be a prohibitively slow process. I can do some memory juggling and perhaps I will in a while, if I don't see any real progress. I think using 2.5G of swap is a "telling" signal. I am still going to try to solve this, but I will definitely be slower than expected...