Closed glwagner closed 5 years ago
This is the result of running sandbox/benchmark.py
on my MacBook Pro with 4 cores:
Benchmark results for 100 iterations:
** Build time **
================
Model Time
----- ----
DNS | 14.45 s
Constant Smagorinsky | 14.65 s
** Run time **
==============
Model Time Slowdown vs DNS
----- ---- ---------------
DNS | 46.51 s | 1
Constant Smagorinsky | 76.17 s | 1.637622
There's significant variation between runs, so it'd be nice to have a better/more reproducible bencharking --- @kburns, what's the best way to benchmark long-running code?
PS see dedaLES/benchmarks.py
for key code and sandbox/sample_benchmark.py
for how to use it.
Results for my desktop at MIT:
Benchmark results for 100 iterations:
** Build time **
================
Model Time
----- ----
DNS | 18.88 s
Constant Smagorinsky | 19.66 s
** Run time **
==============
Model Time Slowdown vs DNS
----- ---- ---------------
DNS | 40.79 s | 1
Constant Smagorinsky | 89.29 s | 2.188969
Adds some primitive benchmarking tools. Minor updates to API.