Moved CalculatedSNRs outside the calculate_SNRs definition

This should lead to a 4-7% speed-up for some GW runs. The creation of a named_tuple instance is surprisingly inefficient.

Edit: As @colm.talbot pointed out: This can speed up some runs by 17 %

Good catch, I actually find it's even higher running the fast tutorial, I'm spending 17.4% of time in NamedTuple.

One minor suggestion, could you make the CalculatedSNRs an attribute of the likelihood? So that each likelihood object has its own version. I'm wondering whether this might break with multiprocessing.

I don't think it should break since the CalculatedSNRs is basically like a class definition. I'll move it inside the class anyway.

added 1 commit

• a963c0dd - Made it a class attribute

Compare with previous version

added 9 commits

• a963c0dd...bb5269a7 - 7 commits from branch master
• eba52470 - Merge branch 'master' into minor_speed_ups
• f5ec8269 - Fixed the CalculatedSNRs call

Compare with previous version

added 1 commit

• 10c97891 - Reverted some formatting changes

Compare with previous version

changed the description

added Refactoring label

added Efficiency label

approved this merge request

changed milestone to %0.4.5

approved this merge request

merged

mentioned in commit 3265b9be