SIFT Profiling
SIFT can be started with a custom Heap Profiler, which is using the Python
tracemalloc module underneath. When done so it creates a directory
<YYYYddmm_HH-MM>_uwsift_heap_profile
in the current working directory and stores
memory usage snapshot data into it. The command to take a snapshot every 2.0
seconds is as follows:
python -m uwsift --profile-heap 2.0
Afterwards the snapshorts must be combined:
python ./uwsift/util/heap_analyzer.py --combine ./combined_stats.prof --snapshot-dir <<YYYYddmm_HH-MM>_uwsift_heap_profile>
This creates the file combined_stats.prof
, which can be visualized with
matplotlib for analysis:
python ./uwsift/util/heap_analyzer.py --load ./combined_stats.prof --plot