in reply to RE: big performance gain
in thread Training wheels again

No, I didn't preload CGI.pm, I just said "use CGI qw(:standard);".

I did this because I wanted to measure the performance of a CGI script that's handed over to Apache::Registry.

Replies are listed 'Best First'.
RE: RE: RE: big performance gain
by merlyn (Sage) on Sep 10, 2000 at 20:06 UTC
    There's a huge performance difference between use CGI and use CGI qw(:standard) because the latter has to make 150 or so symbols through an import, where the former simply loads the "lincoln loader" selfloader and compiles subroutines as needed.

    -- Randal L. Schwartz, Perl hacker

[reply]
[d/l]
[select]
      Following up to my own post, this invocation of the debugger shows just how many symbols were added by CGI: 
      halfdome.holdit.com> perl -debug

Loading DB routines from perl5db.pl version 1.0402
Emacs support available.

Enter h or `h h' for help.

main::(-e:1):   bug
  DB<1> x scalar keys %main::
0  73
  DB<2> use CGI qw(:standard);

  DB<3> x scalar keys %main::
0  227
  DB<4>

      [download]
      Note that the main symbol table jumped from 73 stabs to 227 stabs. There's work behind this, and that's gonna slow you down a bit. It also takes a lot of space.

      -- Randal L. Schwartz, Perl hacker

[reply]
[d/l]
        i only recently took the necessary code out of CGI.pm (i practically only ever needed the part for taking params off a POST-form), and packed it into my own code. not very clean, i know, but Benchmark told me the code with use CGI.pm took 500secs for 1_000_000 iterations, the other one only 3 secs. So... I'm sticking with that.
[reply]

In Section Perl Monks Discussion