cpu vs. wall clock times
2 posters
Page 1 of 1
cpu vs. wall clock times
In comparing cpu and wall clock times in a job information report I see three cases:
a) cpu time just slightly smaller than wall clock time - OK
b) cpu time a fair bit more than wall clock time - presumably multi-threaded solutions
but also
c) very small cpu time, but a substantial wall clock time, e.g.
0.002999 0.870433
0.002998 0.231191
0.002998 0.70621
0.002998 0.460266
Why this last case? and are both numbers accurate?
- David
a) cpu time just slightly smaller than wall clock time - OK
b) cpu time a fair bit more than wall clock time - presumably multi-threaded solutions
but also
c) very small cpu time, but a substantial wall clock time, e.g.
0.002999 0.870433
0.002998 0.231191
0.002998 0.70621
0.002998 0.460266
Why this last case? and are both numbers accurate?
- David
davidcok- Posts : 32
Join date : 2014-04-29
Re: cpu vs. wall clock times
David,
I am sorry, but I have no special insight into this issue. I would suggest running examples of those particular three situations on your local computer, to see if you observe something similar. This may be a characteristic of the benchmarks, rather than an issue with StarExec. Wallclock and CPU times for sequential solvers can differ greatly if a lot of system time is required, for example. This could be the case for reading very large benchmarks into memory. This is just speculation, of course.
I am sorry, but I have no special insight into this issue. I would suggest running examples of those particular three situations on your local computer, to see if you observe something similar. This may be a characteristic of the benchmarks, rather than an issue with StarExec. Wallclock and CPU times for sequential solvers can differ greatly if a lot of system time is required, for example. This could be the case for reading very large benchmarks into memory. This is just speculation, of course.
Page 1 of 1
Permissions in this forum:
You cannot reply to topics in this forum
|
|