Optimizing vectorized array operations

Help
10

The reason in this MRE is actually obvious :slight_smile: ā€¦

At each iteration of the outer loop on i:

  • In the fast version, the inner loop on j is executed only N_samples=1000 times, which means that only 1000 elements of pack_tot are updated.
  • In the slow version all the N_grid_pack=800000 elements of pack_tot are updated (and the whole line_pack is set to zero, although you really need only 1000 elements).
2 Likes