Fil: A Python Memory Profiler for Scientific Computing
TimeFriday, 13 November 20205:10pm - 5:15pm EST
DescriptionFil is a new, open source memory profiler for Python aimed specifically at the needs of scientific computing and other batch data processing applications. It finds the peak memory usage, and the location in your Python code from which the relevant allocations came, which is exactly what you need to optimize. It can track memory allocations from Python code, but also in Python extensions written in C, C++, Cython, and (at least when gfortran is used) Fortran 90 allocations as well. It is easy to use, with a simple yet visually expressive UX. For applications where most of the computation is delegated to C libraries and where most allocations are large, the performance overhead is low, allowing profiling with complete datasets. It attempts to report sources of memory usage when out-of-memory crashes occur.