in reply to Performance and Behavioral changes in Perl 5.8.8 and Perl 5.10.1

Most likely it is that the profiler on 5.10 can instrument BEGIN blocks, which it could not do on 5.8.8.

  • Comment on Re: Performance and Behavioral changes in Perl 5.8.8 and Perl 5.10.1

Replies are listed 'Best First'.
Re^2: Performance and Behavioral changes in Perl 5.8.8 and Perl 5.10.1
by LanX (Saint) on Aug 21, 2017 at 11:57 UTC
    I'm confused and checked several online dicts for the term "to instrument" ...and still not sure what you're saying ...

    Do you mean that the profiler before 5.10 wasn't able to hook into BEGIN blocks but only ordinary subs and that the performance impact is a side effect of this measurement itself?

    Cheers Rolf
    (addicted to the Perl Programming Language and ☆☆☆☆ :)
    Je suis Charlie!

      "To instrument" means "to measure", but to me at least with a connotation of having points where to hook things to.

      I didn't even recognize the original comment as mine, and I'm not really sure what I meant back then. I guess that my reply stemmed from a reading of perl510delta or the changelog of the profiler.

      But yes, I think I suggested back then that the extra granularity/detail resulted in attributing 17 ms to a BEGIN call that was not separately accounted for in 5.8.8. I guess the time spent in the BEGIN block was added to the time spent in the main program.

        Thanks!

        (my first association for "instrumenting" is to provide a bunch of kindergarten kids with vuvuzelas... ;)

        Cheers Rolf
        (addicted to the Perl Programming Language and ☆☆☆☆ :)
        Je suis Charlie!