|
Given a machine of this kind:
boinc@localhost:~> lscpu
Architecture: x86_64
CPU op-mode(s): 32-bit, 64-bit
Byte Order: Little Endian
Address sizes: 43 bits physical, 48 bits virtual
CPU(s): 256
On-line CPU(s) list: 0-255
Thread(s) per core: 2
Core(s) per socket: 64
Socket(s): 2
NUMA node(s): 2
Vendor ID: AuthenticAMD
CPU family: 23
Model: 49
Model name: AMD EPYC 7742 64-Core Processor
Stepping: 0
...
NUMA node0 CPU(s): 0-63,128-191
NUMA node1 CPU(s): 64-127,192-255
running SOB (4x64T):
boinc@localhost:~/slots/3> cat stderr.txt
BOINC llr wrapper (version 8.04)
Using Jean Penne's llr (64 bit)
LLR2 Program - Version 0.9.4, using Gwnum Library Version 29.8
LLR command line: primegrid_llr -d -oDiskWriteTime=10 -oThreadsPerTest=64 llr.in
Using all-complex FMA3 FFT length 2880K, Pass1=768, Pass2=3840, clm=1, 64 threads, a = 3, L2 = 2015*1027
and I observe in top(1) a lackluster utiliziation of only max. 41/64ths by each llr process on average (total system: 101/256ths).
Tasks: 2248 total, 5 running, 2243 sleeping, 0 stopped, 0 zombie
%Cpu(s): 0.0 us, 23.5 sy, 28.1 ni, 48.3 id, 0.0 wa, 0.0 hi, 0.1 si, 0.0 st
MiB Mem : 515770.0+total, 189199.3+free, 4230.004 used, 322340.6+buff/cache
MiB Swap: 0.000 total, 0.000 free, 0.000 used. 508124.8+avail Mem
PID USER PR NI VIRT RES SHR S %CPU %MEM TIME+ COMMAND
10503 boinc 39 19 749860 200352 3192 R 4173 0.038 123:36.99 primegrid_llr
10301 boinc 39 19 749864 213016 11768 S 3877 0.040 84:50.40 primegrid_llr
10373 boinc 39 19 749864 200432 3192 R 3492 0.038 115:59.75 primegrid_llr
10369 boinc 39 19 768060 218548 3192 R 2169 0.041 108:03.26 primegrid_llr
Is this a known issue? The picture looks similar when forcibly reducing it to run 16threaded, where it takes about 14/16ths:
top - 21:12:58 up 41 days, 5:18, 5 users, load average: 14.33, 14.68, 14.88
307 boinc 39 19 426664 282868 15536 S 1425 0.430 9424:35 primegrid_llr
[/code] | 
