flee@dictionopolis.cs.psu.edu (Felix Lee) (10/30/90)
I've been playing with profiling Perl scripts using the Perl debugger hooks, and I'm wondering if I'm reinventing the wheel. Does someone already have a profiling version of perldb.pl? So far, I've got call graph output a la gprof, without cycle detection. No time sampling yet; that's next. And then maybe a statement frequency counter a la tcov. -- Felix Lee flee@cs.psu.edu
lwall@jpl-devvax.JPL.NASA.GOV (Larry Wall) (10/30/90)
In article <Fl43vd_2@cs.psu.edu> flee@dictionopolis.cs.psu.edu (Felix Lee) writes:
: I've been playing with profiling Perl scripts using the Perl debugger
: hooks, and I'm wondering if I'm reinventing the wheel. Does someone
: already have a profiling version of perldb.pl?
Somebody (don't recall who) used to have one, but the debugger has changed
enough since then that I wouldn't imagine it'd be that useful, unless they've
been tracking it.
: So far, I've got call graph output a la gprof, without cycle
: detection. No time sampling yet; that's next. And then maybe a
: statement frequency counter a la tcov.
Given the great variation in the times taken by various Perl statements,
I would think time sampling on a line-by-line basis would be cool. Of
course, it would help if I made caller work from a signal handler...
If the time it takes for the signal handler to run approaches the sample
interval, however, you could get some sample tick fratricide.
Larry