- Currently DSECS spends 70% of time in [numpy.linalg.svd](https://numpy.org/doc/stable/reference/generated/numpy.linalg.svd.html). How can this be improved? - I have tried [Scalene](https://github.com/plasma-umass/scalene) as a code profiler. I put the output of a basic DSECS benchmark at https://swarmpal-benchmarks.netlify.app/profile.html - One run of DSECS takes ~3mins on a modern laptop (8 cores), but took ~20mins on VRE (1 core). `numpy.linalg.svd` uses multiple threads
numpy.linalg.svduses multiple threads