Testing (see discussion) revealed that constructing sparse arrays with a series of arr[i,j] = k expressions is faster than using a monolithic np.array([[...], [...], ...]) expression.
Ultimately, this proved to be a purely academic distinction. On Python 3.11, python runscripts/manual/runSim.py completes a cell cycle ~1% faster (~15 sec) with this change LOL.
This also fixes the long JIT compilation time bug documented in #1400. Unfortunately, running with --jit was still 1 second slower than with --no-jit.
Testing (see discussion) revealed that constructing sparse arrays with a series of
arr[i,j] = k
expressions is faster than using a monolithicnp.array([[...], [...], ...])
expression.Ultimately, this proved to be a purely academic distinction. On Python 3.11,
python runscripts/manual/runSim.py
completes a cell cycle ~1% faster (~15 sec) with this change LOL.This also fixes the long JIT compilation time bug documented in #1400. Unfortunately, running with
--jit
was still 1 second slower than with--no-jit
.