Kent wrote: >Important question: > >How did you prove this was a performance issue. > >Because unless you've > >a) profiled your code >b) benchmarked your code >c) seen that the performance is significantly impeded by that function call >d) seen that the performance is in the function overhead itself in that >function call
I used a nytprof to check performance, function call reduces it. >It may be simpler and more elegant to just write that line of code in the 5 >places by hand. I think you right. Shlomi Fish >Some ideas: > >1. Implement the subroutine in XS using https://metacpan.org/release/Inline-C >or whatever. > >2. Create a code like: > >sub my_RAND_SUB_PLEASE_EXPAND >{ >return (int ( rand 2 ) ? 1 : -1); >} ... >And then compile the code to a more optimised version by passing it through a >filter such as: >perl -lapE 's/\Qmy_RAND_SUB_PLEASE_EXPAND()\E/(int ( rand 2 ) ? 1 : -1)/g' Too little reason to add new dependencies from CPAN and use such a powerful magic :) But source filters seems interesting to me. I think I'll find use for them in the future. Kent, Shlomi Fish, thanks for your replies!
