Script to test memory consumption for opensim with python API
To run the test, run python test.py n from within the folder where n is the number of loops you
want to test over. This script will save interim files with results from the processes in the Results
folder.
The script runs two loops for each of the main opensim processing pipelines (IK, ID, SO, and JR). The
first loop simply constructs everything to do the processing, but doesnt actually run the
processing. The second loop does the same thing, but it actually runs it - the difference
being whether the .run() command is entered for each of the pipelines. The memory is recorded at the
end of each of these loops. Memory is recorded seperately for the .run()/no-.run() loops.
At the end of the script, the results of the average memory accumulated for the RUN vs. NO-RUN loops for
each of the processes are printed to the terminal.
The raw results of the memory accumulation for each of the processes are also saved to .csv files in the
Results folder. These files can be used to plot and see how the memory increases for each of the processes.
From preliminary testing, the main culprit seems to be the Static Optimization which accumulates 37MB of
memory for every loop (the results are printed below) - this was the average over 20 loops. It should be
noted that the SO only has 10 time-steps SO script - longer files may (likely?) will accumulate more
memory.
Below is the average accumulated memory over 20 loops.
IK - Average memory accumulation per loop WITHOUT run: 0.294 MB
IK - Average memory accumulation per loop WITH run: 0.263 MB
ID - Average memory accumulation per loop WITHOUT run: 0.055 MB
ID - Average memory accumulation per loop WITH run: 0.781 MB
SO - Average memory accumulation per loop WITHOUT run: 0.003 MB
SO - Average memory accumulation per loop WITH run: 37.513 MB
JR - Average memory accumulation per loop WITHOUT run: 1.644 MB
JR - Average memory accumulation per loop WITH run: 2.164 MB