mersenneforum.org  

Go Back   mersenneforum.org > Great Internet Mersenne Prime Search > Software

Reply
 
Thread Tools
Old 2004-02-09, 08:39   #1
Matthias C. Noc
 
Matthias C. Noc's Avatar
 
Dec 2003

23·3 Posts
Default Improving the RAM allocation for Prime 95

When allocating different amounts of RAM to Prime95 has only a true impact while stage 1 and stage 2 processing (there it is visible in the Range of B1 and B2 values and the speed the calculation is running), then I think the RAM allocation options shouldn’t go by Daytime and Nighttime, but should reflect the process. So one would allocate more memory to stage 1 and stage 2 calculations and is reducing it during the primality test. I know the user could do that by watching the process of prime95 and allocate the RAM "by hand" when the time comes, but I'm normally not watching prime95 work.

Or, if one is sticking with the daytime/nighttime option, one could add an option that prime95 is arranging its calculation by how much RAM is available. So when one gets a new exponent 3-5 days before one is finishing the current one, there is normally enough nighttime with more RAM available to run in this time the stage 1 and stage 2 for the new exponent while the work on the “current” exponent is suspended until daytime or the stage 1 and stage 2 tests are finished.

I think that could be a possible way to improve the process without too much re-programming of a new client. Or am I wrong?

Thanks,

Matthias
Matthias C. Noc is offline   Reply With Quote
Old 2004-02-09, 14:23   #2
Reboot It
 
Reboot It's Avatar
 
Aug 2002
London, UK

7×13 Posts
Default

Have you seen this option from UNDOC.TXT?
Quote:
Since P-1 stage 2 runs faster with more memory available you can have the program only run stage 2 at night when more memory is available. In prime.ini set:
OnlyRunStage2WithMaxMemory=1
This allows you to set a sensible daytime RAM allocation that allows you to work and yet allocate much more RAM at night. If an exponent completes P-1 Stage 1 during the daytime, it skips Stage 2 for the time being and starts on the main LL test itself. When it reaches the start of nighttime (as configured), it pauses the LL test and runs Stage 2 with all of the memory you gave it. If it finds a factor it abandons the LL test, but if not it resumes from where it left off.

You might also find this option useful:
Quote:
SequentialWorkToDo=0
There's a full explanation of that option also in UNDOC.TXT.
Reboot It is offline   Reply With Quote
Old 2004-02-09, 17:02   #3
Matthias C. Noc
 
Matthias C. Noc's Avatar
 
Dec 2003

1816 Posts
Default

Quote:
Originally Posted by Reboot It
Have you seen this option from UNDOC.TXT?
...
There's a full explanation of that option also in UNDOC.TXT.
Thanks! The UNDOC.TXT was looking too difficult for a newcomer like me, so I gave it up after a few paragraphs. But it seems that many good idea are buried there. I think it would be better to add more of them into the easily accessible option/preferences, so people like me can find them.

Best,

Matthias
Matthias C. Noc is offline   Reply With Quote
Old 2004-02-12, 19:34   #4
Reboot It
 
Reboot It's Avatar
 
Aug 2002
London, UK

7·13 Posts
Default

Quote:
Originally Posted by Matthias C. Noc
UNDOC.TXT was looking too difficult for a newcomer like me...
I think that's why Undoc.txt exists, so that there is neither too much to overwhelm somebody starting out and also so that you can learn the basics before you start fiddling with settings that can have undesired consequences if you don't understand them.

Glad to have been of help, though.
Reboot It is offline   Reply With Quote
Reply

Thread Tools


Similar Threads
Thread Thread Starter Forum Replies Last Post
Improving the queue management. debrouxl NFS@Home 10 2018-05-06 21:05
improving factorization method bhelmes Computer Science & Computational Number Theory 7 2017-06-26 02:20
an idea for improving the factor table ixfd64 PrimeNet 5 2013-11-08 05:41
Improving website speed Unregistered Information & Answers 1 2011-04-02 02:17
Improving Sieving by 18%. cipher Prime Sierpinski Project 10 2009-07-01 13:34

All times are UTC. The time now is 23:57.


Wed Oct 27 23:57:59 UTC 2021 up 96 days, 18:26, 0 users, load averages: 1.09, 1.03, 1.00

Powered by vBulletin® Version 3.8.11
Copyright ©2000 - 2021, Jelsoft Enterprises Ltd.

This forum has received and complied with 0 (zero) government requests for information.

Permission is granted to copy, distribute and/or modify this document under the terms of the GNU Free Documentation License, Version 1.2 or any later version published by the Free Software Foundation.
A copy of the license is included in the FAQ.