diff options
| author | Eugeniy Mikhailov <evgmik@gmail.com> | 2012-09-06 19:15:22 -0400 |
|---|---|---|
| committer | Eugeniy Mikhailov <evgmik@gmail.com> | 2012-09-06 19:15:22 -0400 |
| commit | fea99e35f681f2c536a48f41383c7408b158fd31 (patch) | |
| tree | 4245f096a67b181a9a1761fa32c7afb67f96320f | |
| parent | 1b4a582eff7d083e747cd3c1f2548743875c97ac (diff) | |
| download | Nresonances-fea99e35f681f2c536a48f41383c7408b158fd31.tar.gz Nresonances-fea99e35f681f2c536a48f41383c7408b158fd31.zip | |
added some notes aboit optimization performance
| -rw-r--r-- | xmds2/realistic_Rb/OptimizationNotes.txt | 17 |
1 files changed, 17 insertions, 0 deletions
diff --git a/xmds2/realistic_Rb/OptimizationNotes.txt b/xmds2/realistic_Rb/OptimizationNotes.txt new file mode 100644 index 0000000..4aea670 --- /dev/null +++ b/xmds2/realistic_Rb/OptimizationNotes.txt @@ -0,0 +1,17 @@ +I use gcc 4.7.1 + +Enabling OpenMP option makes my 6 core CPU busy on all 6 of them. +Surprisingly on Time x Velocity = 10000 x 10 grid it takes longer them with +disabled OpenMp option and only single core working. I have got 230 seconds +vs 160 seconds. Besides optimizing wisdom of FFT takes forever with OpenMP. + + +on a grid Time x Velocity = 10000 x 100 +with OpenMP execution time = 988 seconds +without = 1708 seconds + +with 4 thread +on a grid Time x Velocity = 10000 x 100 +with OpenMP execution time = 1036 seconds +without = 1603 seconds + |