diff options
| author | Paul E. McKenney <paulmck@kernel.org> | 2024-01-22 12:15:55 -0800 |
|---|---|---|
| committer | Paul E. McKenney <paulmck@kernel.org> | 2024-01-22 12:20:09 -0800 |
| commit | d1a5d020205ba69caefd59dcc514f425c566159e (patch) | |
| tree | 874c1479f95e204fecc03dc3a615fd6ce7afa431 | |
| parent | 4c316f71b4bae1610e8753184767d0301e7914d5 (diff) | |
| download | perfbook-d1a5d020205ba69caefd59dcc514f425c566159e.tar.gz | |
CodeSamples/cpu: Take nthreads times the samples for coe
For fre and rfe, there are nthreads samples per run, but only one sample
per run for coe. Therefore, make perftemporal.sh compensate by running
coe.sh nthreads times more often. All in pursuit of pretty graphs!
Signed-off-by: Paul E. McKenney <paulmck@kernel.org>
| -rw-r--r-- | CodeSamples/cpu/perftemporal.sh | 11 |
1 files changed, 10 insertions, 1 deletions
diff --git a/CodeSamples/cpu/perftemporal.sh b/CodeSamples/cpu/perftemporal.sh index f1a157da..d965cc2f 100644 --- a/CodeSamples/cpu/perftemporal.sh +++ b/CodeSamples/cpu/perftemporal.sh @@ -67,7 +67,16 @@ lscpu > $tag/lscpu.out bash coereduce.sh < $tag/coe-nvals.out > $tag/coe-nvals.dat for pgm in coe fre rfe do - for ((i=0;i<reps;i++)) + # Allow for coe getting only one data point per run, + # compared to fre and rfe getting $nthreads per run. + if test "$pgm" = coe + then + nr="$((reps*nthreads))" + else + nr="$reps" + fi + echo $pgm nr: $nr + for ((i=0;i<nr;i++)) do sh $pgm.sh $nthreads done > $tag/$pgm.out |