Dear Weiyu,
The CPU percentage has no meaning in linux and using top (please google
"python cpu usage high").
Normally, there should be no reason why MWCS is slower or faster
depending of mwcs_wlen (window length), except maybe some edge case
where the wlen x sampling_rate is a prime number.
Best regards,
Thomas
Le 16/06/2019 à 06:51, weiyu a écrit :
Hi MSNoise group,
I was faced with a new problem in computing MWCS with multiprocess operation. I defined
different Mwcs_Wlen for different filters. Then I started 10 instances of the code using
command “msnoise -t 10 compute-mwcs”. When the Mwcs_Wlen was 10s, the code was working
properly. But when the Mwcs_Wlen was more than 10s such as 20s, 30s, the 10 processes
occupied all my 40 physical cores. And the computational speed became very slow.
I don’t know what happened, so I stoped the code and reseted the DTT jobs. And I
recalculated them in one process using command “msnoise compute-mwcs”.I found when
computed 10s window length, the process occupied 100% of one core. But when computed
others, the process occupied 4000% of one core.
The cpu utilization percentage was obtained from the top command in Linux,and I computed
with msnoise 1.5.1.
Looking forward to know the reason of this problem.
Thanks!
Weiyu
_______________________________________________
MSNoise mailing list
MSNoise(a)mailman-as.oma.be
http://mailman-as.oma.be/mailman/listinfo/msnoise