View Single Post
Old 2021-01-15, 12:34   #58
phillipsjk
 
Nov 2019

1068 Posts
Default

Is stage 1 P-1 factoring now supposed to fit completely in the L3 cache? Is that where the speedup came from?

Code:
top - 05:19:52 up  6:14,  1 user,  load average: 63.25, 59.63, 50.47
Tasks: 588 total,   1 running, 587 sleeping,   0 stopped,   0 zombie
%Cpu(s):  0.0 us,  0.3 sy, 96.6 ni,  3.1 id,  0.0 wa,  0.0 hi,  0.0 si,  0.0 st
MiB Mem : 128925.9 total, 127437.0 free,    803.2 used,    685.7 buff/cache
MiB Swap:      0.0 total,      0.0 free,      0.0 used. 127236.8 avail Mem 

  PID USER      PR  NI    VIRT    RES    SHR S  %CPU  %MEM     TIME+ COMMAND    
13527 james     30  10 4483612 134276   7880 S  1556   0.1 186:08.48 mprime     
13680 james     30  10 4483616 134428   7988 S  1553   0.1 180:02.59 mprime     
13834 james     30  10 4483480 133496   7496 S  1551   0.1 174:31.16 mprime     
13377 james     30  10 4483480 134888   7636 S  1542   0.1 189:46.74 mprime
The prime.log makes no reference to the new version yet.


This is my 4 CPU Opteron system running 4 separate instances (because mprime is not NUMA aware).


Having trouble verifying but I think the L3 cache on that chip is actually 2 8MB caches.


Edit:
Because I was only using half of my L3 cache, I doubled the number of workers per instance.
Got some reassuring messages in prime.log:


Code:
[Fri Jan 15 06:27:57 2021 - ver 30.4]
Exchanging program options with server
Getting assignment from server
PrimeNet success code with additional info:
Server assigned P-1 factoring work.
Got assignment 6914F714740D2AF13BE05BCE07ABA42D: P-1 M102220457
 Sending expected completion date for M102220457: Jan 21 2021

Edit: or is the resource usage 135MB, not 135kB?

Last fiddled with by phillipsjk on 2021-01-15 at 13:32 Reason: Update, typos
phillipsjk is offline   Reply With Quote