mersenneforum.org Planning & Coordination for 2,2330L
 Register FAQ Search Today's Posts Mark Forums Read

 2019-06-22, 14:14 #254 pinhodecarlos     "Carlos Pinho" Oct 2011 Milton Keynes, UK 24·3·101 Posts Any option on server side to send a few wus to clients so to have some buffer or the only option would be to increase sieve range once again to 4k or even 8k.
 2019-06-22, 17:01 #255 fivemack (loop (#_fork))     Feb 2006 Cambridge, England 6,379 Posts Is there any way that the http://factoring.cloudygo.com could record 'relations in last 24 hours' as well as 'relations since the project started' - it would make it rather easier to see what happens when configurations change.
2019-06-22, 20:01   #256
VBCurtis

"Curtis"
Feb 2005
Riverside, CA

35·19 Posts

Quote:
 Originally Posted by pinhodecarlos Any option on server side to send a few wus to clients so to have some buffer or the only option would be to increase sieve range once again to 4k or even 8k.
None that I know of for buffering. What does "only option" refer to?

As for workunit size, we'll be changing that to 8k or 10k when we move to I=15, as yield drops from 6.x on I=16 at ~200M to 2.2 on I=15 at ~300M.
So far, we have sieved Q from 8M to 68M for 533M raw relations, for an average yield of 8.9. The yield on current WUs is in the mid-8's.

I'll run remdups shortly to see what our duplicate rate looks like.
EDIT: For Q from 8-60M, 349.5M unique, 118.5 duplicates, 0 bad relations, total 468M relations. Duplicate rate 25%, better than the C206 that Greg ran on 16e/33 for us (that one was 792M uniq, 373M dup).
I'll run it again when Q reaches 70M.

Last fiddled with by VBCurtis on 2019-06-22 at 21:39

2019-06-23, 02:21   #257
VBCurtis

"Curtis"
Feb 2005
Riverside, CA

35×19 Posts

Quote:
 Originally Posted by fivemack Is there any way that the http://factoring.cloudygo.com could record 'relations in last 24 hours' as well as 'relations since the project started' - it would make it rather easier to see what happens when configurations change.
I'd like this too. Perhaps replace the timestamp of last workunit with this stat? We can see the project-difficulty trend as our numbers gently slide, see who is new and rising fast, and as fivemack said easily quantify changes in client configuration.

2019-06-23, 04:07   #258
SethTro

"Seth"
Apr 2019

111010112 Posts

Quote:
 Originally Posted by VBCurtis I'd like this too. Perhaps replace the timestamp of last workunit with this stat? We can see the project-difficulty trend as our numbers gently slide, see who is new and rising fast, and as fivemack said easily quantify changes in client configuration.
I'll try to add it to the top line and graph page tonight.

2019-06-23, 06:57   #259
SethTro

"Seth"
Apr 2019

3538 Posts

Quote:
 Originally Posted by SethTro I'll try to add it to the top line and graph page tonight.
Done

http://factoring.cloudygo.com/progre...L.c207/daily_r
(which is under Charts tab) also a number on the first page.
http://factoring.cloudygo.com/

Last fiddled with by SethTro on 2019-06-23 at 06:57

2019-06-23, 07:27   #260
fivemack
(loop (#_fork))

Feb 2006
Cambridge, England

6,379 Posts

Quote:
 Originally Posted by SethTro Done http://factoring.cloudygo.com/progre...L.c207/daily_r (which is under Charts tab) also a number on the first page. http://factoring.cloudygo.com/
The daily_r curve is useful, because differentiating the total-work-done curve by eye isn't really possible; thank you

I think it would be even more useful to have the last-24-hours figure at a per-client level.

Last fiddled with by fivemack on 2019-06-23 at 07:27

 2019-06-23, 09:14 #261 pinhodecarlos     "Carlos Pinho" Oct 2011 Milton Keynes, UK 113608 Posts What about associating the clients to teams?
 2019-06-23, 09:37 #262 fivemack (loop (#_fork))     Feb 2006 Cambridge, England 6,379 Posts Some thoughts on hyper-threading and -t lots Median runtimes in various configurations, on the same hardware (fortunately I have three identical computers) Code: One job -t32 1090s = 2180s for two Two jobs -t8 2132s/2 Two jobs -t16 taskset 0-15; 16-31 1742s/2 Two jobs -t16 taskset 0-7,16-23; 8-15,24-31 1915s/2 So, on these dual-socket eight-core machines, the right answer is to run two jobs, one across both sockets and the other on the other hyperthread across both sockets ; I think I'd expected two jobs to be better than one but am a bit surprised that having both jobs use both sockets is significantly better. Last fiddled with by fivemack on 2019-06-23 at 09:37
 2019-06-23, 14:32 #263 VBCurtis     "Curtis" Feb 2005 Riverside, CA 35·19 Posts We've reached Q=70M, so I re-ran remdups: Code: Q Unique Dup Total 8-60M 349.5 118.5 468.0 8-70M 402.4 147.7 550.1 Q=60-70M added 82M relations, but only 53M unique.
2019-06-23, 15:26   #264
lukerichards

"Luke Richards"
Jan 2018
Birmingham, UK

12016 Posts

Quote:
 Originally Posted by SethTro Vebis has ~10 clients (vebis.1, vebis.2, ...) so I joined all of those to vebis. I'm going to join all your clients into lukerichards. for the main tab and then add a new tab for all clients
Is there any way to add instance-1, localhost and lrichards-pre2core to this collection as well, so I can see my complete combined stats without mental arithmetic?

Thanks.

 Similar Threads Thread Thread Starter Forum Replies Last Post ixfd64 Lone Mersenne Hunters 80 2021-01-19 17:17 henryzz Cunningham Tables 16 2010-08-07 05:08 gd_barnes No Prime Left Behind 2 2008-02-16 03:28 gd_barnes Conjectures 'R Us 32 2008-01-22 03:09 wblipp ElevenSmooth 2 2004-02-19 05:25

All times are UTC. The time now is 01:27.

Fri Jan 22 01:27:32 UTC 2021 up 49 days, 21:38, 0 users, load averages: 4.13, 4.21, 3.85