cleaning and update to the timing
Tests of G. Irrmann shown that:
- the excessive use of
MPI_Wtimestrongly perturb the performance. This is particularly true in the subroutine
lib_mpp/F90) which calls
MPI_Wtimeat very high frequency which strongly affect the performances we are trying to measure.
MPI_Wtimeis much more costly than
SYSTEM_CLOCK(about 20 times more).
SYSTEM_CLOCKused with INTEGER-8 is supposed to have a precision of at least the micro second.
CPUis not really appropriate to check the time spent in communications as use of the MPI library (including time to send/receive) can be seen as a "computing" time.
- AGRIF is not officially supported by the timing
- the estimation of the time spend to perform the timing is inaccurate and can be misleading (showing negative time spent in routines)
- we could easily to value timing.output outputs by providing a few simple analyses and plots based on gnuplot (that is usually available on the machines)
Général cleaning and rewritting of timing.F90
To upload designs, you'll need to enable LFS and have an admin enable hashed storage. More information