Strange output for timing results

[donnaaboise]

I just did an AMRClaw run and noticed some strange timing results for the regridding time. I don't have a lot of insight into what might have happened, other than that I had set the regrid interval to 1 (regrid every time step) and the buffer region to 0. I've saved the data files and setrun.py file, as well as the example.

============================== Timing Data ==============================

Integration Time (stepgrid + BC + overhead)
Level           Wall Time (seconds)    CPU Time (seconds)   Total Cell Updates
 clock_rate         1000
  1                    62.787                 62.664            0.492E+07
  2                   130.457                130.201            0.443E+08
  3                   107.296                107.045            0.379E+09
total                 300.540                299.910            0.428E+09

All levels:
stepgrid               64.123                 63.988    
BC/ghost cells        154.562                154.231
Regridding       -2146807.188                675.064  
Output (valout)        28.806                 28.371  

Total time:      -2146472.862               1008.295  
Using  1 thread(s)

Note: The CPU times are summed over all threads.
      Total time includes more than the subroutines listed above
Note: timings are also recorded for each output step
      in the file timing.csv.

[mjberger]

That' s a hoot. Is it repeatable? I think we recently read that the timer starts at some arbitrary point in time, and can have weird behavior like that. (RAndy ?) More upsetting is huge cost of BC/ghost cells: 154 out of 300 seconds. Do you have aux arrays? If so, try out the new "fasterfilpatch" branch. If not, what is causing this? Are you running linear advection with no interior work per patch? — Marsha

…

On Jul 17, 2019, at 3:15 PM, Donna Calhoun ***@***.***> wrote: I just did an AMRClaw run and noticed some strange timing results for the regridding time. I don't have a lot of insight into what might have happened, other than that I had set the regrid interval to 1 (regrid every time step) and the buffer region to 0. I've saved the data files and setrun.py file, as well as the example. ============================== Timing Data ============================== Integration Time (stepgrid + BC + overhead) Level Wall Time (seconds) CPU Time (seconds) Total Cell Updates clock_rate 1000 1 62.787 62.664 0.492E+07 2 130.457 130.201 0.443E+08 3 107.296 107.045 0.379E+09 total 300.540 299.910 0.428E+09 All levels: stepgrid 64.123 63.988 BC/ghost cells 154.562 154.231 Regridding -2146807.188 675.064 Output (valout) 28.806 28.371 Total time: -2146472.862 1008.295 Using 1 thread(s) Note: The CPU times are summed over all threads. Total time includes more than the subroutines listed above Note: timings are also recorded for each output step in the file timing.csv. — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub <#248?email_source=notifications&email_token=AAGUGC3L3MDDFAA23WXPPN3P74LPFA5CNFSM4IEQHORKYY3PNVWWK3TUL52HS4DFUVEXG43VMWVGG33NNVSW45C7NFSM4G7XHBMQ>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AAGUGC2ZMFCVR5TPD6OCVRLP74LPFANCNFSM4IEQHORA>.

[mandli]

Looks like the same issue @rjleveque had. I am not sure that we ever came up with a way to fix it reliably.

[donnaaboise]

Well, I have 9 aux array variables, and yes the problem i just a scalar transport problem.

[mjberger]

fasterFilpatch will make a huge difference then. Kyle - is it merged yet? — Marsha

…

On Jul 17, 2019, at 4:17 PM, Donna Calhoun ***@***.***> wrote: Well, I have 9 aux array variables, and yes the problem i just a scalar transport problem. — You are receiving this because you commented. Reply to this email directly, view it on GitHub <#248?email_source=notifications&email_token=AAGUGC7RGXWWKUOYJVGHU43P74SXFA5CNFSM4IEQHORKYY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOD2ELCWY#issuecomment-512274779>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AAGUGCZNL7I4XZSLXSV6P5LP74SXFANCNFSM4IEQHORA>.

[mandli]

For AMRClaw it is.