mersenneforum.org 2 Memory Allocation Errors
 Register FAQ Search Today's Posts Mark Forums Read

 2018-01-12, 21:53 #1 petrw1 1976 Toyota Corona years forever!     "Wayne" Nov 2006 Saskatchewan, Canada 460110 Posts 2 Memory Allocation Errors I have a PC doing P-1 in the 5xM range on all 4 cores (It has been doing so for 4 months) It also has a GTX-980 GPU running TF and MisFit. (This was added about 2 months ago) .. AND NOTHING ELSE ... It is running Prime95 28.5B1 First: Over the last few weeks I get a daily message that Windows (7) is running low on memory and it recommends shutting down some programs. When I open Task Manager it tells me that RAM is at about 98% and that I have over 15.5 GB allocated to Prime95. I have 16GB of RAM on this PC; only (thirteen)13,000K allocated in Prime95 client. === To me this looks like a memory leak === Second: I quite often get a message in one of the worker windows to the effect of "Memory Allocation Error - Trying with lower memory allocation" Then I see it restarting that worker with a lower amount or RAM; sometimes this is repeated several times before it finally succeeds and carries on. However, when I add up the memory usage in all 4 workers BEFORE the first Error it is still less than 13,000K. === To me this looks like a bad chunk of RAM === I may be 0/2 in my assumptions. Any thoughts? Thanks
 2018-01-12, 22:37 #2 Dubslow Basketry That Evening!     "Bunslow the Bold" Jun 2011 40
2018-01-12, 22:57   #3
VictordeHolland

"Victor de Hollander"
Aug 2011
the Netherlands

23·3·72 Posts

Quote:
 Originally Posted by petrw1 I have a PC doing P-1 in the 5xM range on all 4 cores (It has been doing so for 4 months) First: Over the last few weeks I get a daily message that Windows (7) is running low on memory and it recommends shutting down some programs. When I open Task Manager it tells me that RAM is at about 98% and that I have over 15.5 GB allocated to Prime95. I have 16GB of RAM on this PC; only (thirteen)13,000K allocated in Prime95 client.
When you allocate so much memory to a single P-1 assignment you'll probably invoke the Brent-Suyama extension. That can take up a lot of memory. What is the B-S exponent in your assignments/results? For instance E=6 or E=12?

Quote:
 Second: I quite often get a message in one of the worker windows to the effect of "Memory Allocation Error - Trying with lower memory allocation" Then I see it restarting that worker with a lower amount or RAM; sometimes this is repeated several times before it finally succeeds and carries on. However, when I add up the memory usage in all 4 workers BEFORE the first Error it is still less than 13,000K.
Even if you set Prime95 to 8MB (it will not run stage 2 then), it will use more than 8MB. The program needs some memory to operate normally, think this amounts to 50-100MB. I don't think it counts those MBs to the limit you set. Windows needs some memory too, else it gets annoyed. Leave some for Mfaktc and the graphic card drivers. Why not set P95 to 10GB? I understand your desire to use everything to the max, but in this case you might hurt performance because of the restarts.

2018-01-12, 23:43   #4
ATH
Einyen

Dec 2003
Denmark

C0716 Posts

Quote:
 Originally Posted by petrw1 It is running Prime95 28.5B1
The newest 29.4b7 has a small speedup for P-1 stage1, might be worth trying it instead? Might also have fixed your issue.

2018-01-13, 00:02   #5
heliosh

Oct 2017
++41

11111012 Posts

Quote:
 Originally Posted by petrw1 When I open Task Manager it tells me that RAM is at about 98% and that I have over 15.5 GB allocated to Prime95.
AFAIK is the memory limit per worker and not per computer. Since you're running multiple workers it tries to get what it could.

2018-01-13, 03:10   #6
Serpentine Vermin Jar

Jul 2014

37·89 Posts

Quote:
 Originally Posted by heliosh AFAIK is the memory limit per worker and not per computer. Since you're running multiple workers it tries to get what it could.
I think that's the case too. It's been a long time since I ran multiple P-1 at the same time, but I vaguely recall it was using XX amount of memory per worker, not total. I could be wrong.

And yeah, definitely consider updating to the latest Prime95 version. P-1 updates...stronger, better, faster...

2018-01-13, 03:59   #7
petrw1
1976 Toyota Corona years forever!

"Wayne"
Nov 2006

460110 Posts

Quote:
 Originally Posted by VictordeHolland When you allocate so much memory to a single P-1 assignment you'll probably invoke the Brent-Suyama extension. That can take up a lot of memory. What is the B-S exponent in your assignments/results? For instance E=6 or E=12?
This particular PC is getting E=12.
Some are getting E=6
Some PCs are neither.

I don't pick. I am running work type of PMINUS1

Thanks though...that is good to know.

And I am going to try 29.4B7

2018-01-13, 04:15   #8
petrw1
1976 Toyota Corona years forever!

"Wayne"
Nov 2006

43×107 Posts

Quote:
 Originally Posted by petrw1 This particular PC is getting E=12. Some are getting E=6 Some PCs are neither. I don't pick. I am running work type of PMINUS1 Thanks though...that is good to know. And I am going to try 29.4B7
Is this expected:

Code:
[Jan 12 22:11] Waiting 15 seconds to stagger worker starts.
[Jan 12 22:11] Worker starting
[Jan 12 22:11] Setting affinity to run worker on CPU core #4
[Jan 12 22:11] P-1 on M54245251 with B1=1000000, B2=20000000
[Jan 12 22:11] Chance of finding a factor is an estimated 4.41%
[Jan 12 22:11] Using AVX FFT length 2880K, Pass1=384, Pass2=7680, clm=1
[Jan 12 22:11] P-1 save file incompatible with this program version.  Restarting stage 1 from the beginning.
[Jan 12 22:11] Error reading intermediate file: m2T45251
[Jan 12 22:11] Renaming m2T45251 to m2T45251.bad1
[Jan 12 22:11] Trying backup intermediate file: m2T45251.bu
[Jan 12 22:11] P-1 save file incompatible with this program version.  Restarting stage 1 from the beginning.
[Jan 12 22:11] Error reading intermediate file: m2T45251.bu
[Jan 12 22:11] Renaming m2T45251.bu to m2T45251.bad2
[Jan 12 22:11] All intermediate files bad.  Temporarily abandoning work unit.
[Jan 12 22:11] P-1 on M54246481 with B1=1000000, B2=20000000
[Jan 12 22:11] Chance of finding a factor is an estimated 4.41%
[Jan 12 22:11] Using AVX FFT length 2880K, Pass1=384, Pass2=7680, clm=1
Ah....

6) Minor performance tweaks were made to stage 1 of P-1. Save files are incompatible
in stage 1. Wait for your P-1 test to reach stage 2 before upgrading.

Last fiddled with by petrw1 on 2018-01-13 at 04:18

 2018-01-13, 09:00 #9 S485122     Sep 2006 Brussels, Belgium 11·151 Posts I would run P-1 factoring on one worker with 4 threads. There is a little bit of performance loss if compared to running 4 workers with one thread each but this is limited to phase 2 and if I remember well the GCD phases are single threaded. The advantage is that you do not have the competition about memory between the workers and you are sure your system will not try to use your disks to swap. Jacob Last fiddled with by S485122 on 2018-01-13 at 09:02 Reason: 15.5 MiB of memory out of 16 ! Added bit about swap.

 Similar Threads Thread Thread Starter Forum Replies Last Post pjaj Information & Answers 7 2011-02-12 02:57 JuanTutors Software 25 2009-02-01 21:07 joel893 Software 0 2008-06-03 14:54 A_Pryme Software 7 2005-02-12 02:28 JuanTutors Software 15 2004-08-30 01:15

All times are UTC. The time now is 10:13.

Wed Apr 21 10:13:00 UTC 2021 up 13 days, 4:53, 0 users, load averages: 1.95, 1.95, 1.97