Closed kenken-neko closed 6 months ago
did you run any benchmarks comparing before and after? reading the diagonal is by far the least time consuming part of the pseudoinverse computation
All modified and coverable lines are covered by tests :white_check_mark:
Project coverage is 85.99%. Comparing base (
b014e06
) to head (ead917e
).
:umbrella: View full report in Codecov by Sentry.
:loudspeaker: Have feedback on the report? Share it here.
@sarah-ek
No comparison was made because the current benchmark test does not have an compute_pseudoinverse
implementation. I made this PR because I believe it is obvious that any short process reduces processing time.
It will take some time to implement the benchmark due to my work. I will submit review request once the benchmark comparison is complete.
I am temporarily closing this PR because I do not know when I will be able to implement the benchmark.
Change
The process of reading singular values twice (the two processes below) in the implementation of
compute_pseudoinverse
was reduced to once.Reason
To reduce the processing time by reducing the number of readings.