mersenneforum.org

mersenneforum.org (https://www.mersenneforum.org/index.php)
-   Software (https://www.mersenneforum.org/forumdisplay.php?f=10)
-   -   pfgw factorial testing too slow? (https://www.mersenneforum.org/showthread.php?t=27287)

bur 2021-10-30 06:11

pfgw factorial testing too slow?
 
I had done some factorial prime testing with [URL="http://prpnet.primegrid.com:12002/server_stats.html"]PRPnet[/URL] with i5 and i3 CPUs before and a test took about 25-30 hours.

Now recently I began testing using an i10-10900k and a test in the 282000! range takes more than 40 hours. I have no idea why.

FFT is about 450k, it's a 10 core with 20 MB L3 cache, I'm running 10 tasks simultaneously. It's not ideal, but better L3/task ratio than on the other computers.

It's running Ubuntu. I tried different pfgw versions, the one supplied with the PRPnet client (3.7.10) and also 4.0.1 from sourceforge. No difference.

Each pfgw instance is using 99-100% of a core. Total RAM usage is 1GB/16GB. I used that same computer for yafu and cado and never had any issues with slow speeds.

axn 2021-10-30 06:24

[QUOTE=bur;592036]a test took about 25-30 hours.[/QUOTE]

It would be helpful if there was some indication of the sizes of the numbers being tested when you were getting the above timings.

For example, n=280k ought to take appr. 35% more time compared to n=250k (based on naive scaling calculations) on the same CPU. Of course, it would be worse, since the impact of large L3 will be less when you have bigger FFT.


All times are UTC. The time now is 23:06.

Powered by vBulletin® Version 3.8.11
Copyright ©2000 - 2022, Jelsoft Enterprises Ltd.