if ($Conf[7] =~ /NO/i) {
...
    elsif($Conf[7] =~ /YES/i) {
[download]

@Conf appears to be setup outside of this loop. Evaluate this once and save the result instead of parsing a regex every time through the loop.

    $attrID =~ s/\s*//g;
    $port =~ s/\s*//g;
[download]

I think replacing the '*' with '+' might help here.

    $val = join("", $val, $port);
[download]

How about $val .= $port?

    push @Data, [@entry];
[download]

Since @entry is a lexical variable, you should be able to use \@entry instead of [@entry] which needs to make a copy of the list.

Unless I'm missing something, both blocks of code do the same thing. It might be better from a maintenance standpoint to check if either condition is true and only have the block once.

---
print map { my ($m)=1<<hex($_)&11?' ':'';
$m.=substr('AHJPacehklnorstu',hex($_),1) }
split //,'2fde0abe76c36c914586c';
[download]

Just a few things that come to mind:

• if ($port =~ /\-/) {...} can be replaced by $port =~ s/\-//g;
• you can simply push @Data, [$Time, $mth, $val, $value];
• It may be faster to do ($year, $month, $day) = (substr($date, 0, 4), substr($date, 4, 2), substr($date, 6)); to split the date
• if $Conf[7] seems to be set outside the while, so you could go for a simpler if statement rather than do the same regex each 1.5 million times through the loop or even forego the conditional statement in the loop completely
• you might have a look at the various Date and Time modules on CPAN that might do a faster job on date/time conversions

Hope this helps, -gjb-

great ideas...
you wouldn't happen to have a suggestion for some revised conditional statements would you...I've been staring at this code too long....

Thanks again

As I understand ir, using the 'i' modifier in Perl slows down the regexp code. You might benchmark using /YES/ || /yes/ instead.

You appear to be using the s operator to remove any spaces. My guess is that you can replace that with tr/ //d instead, and get a speed improvement.

Finally, it looks like the two blocks of code are identical. Can you instead check for (NO and $value not zero) or YES? to do that block of code?

This is untested off the cuff code

#!/usr/bin/perl
use warnings;
use strict;

# move declarations out of loop
#
my ( $date, $atime, $etime, $mth, $port, $attrID, $value );
my ( $HH, $MM, $SS );
my ( $year, $month, $day );
my @entry ;

while (<SSLOG>) {
  # chomp one item not seven
  chomp;

  ($date, $atime, $etime, $mth, $port, $attrID, $value) =
      split /;/, $_ ;

  # combine conditions into one
  # could be optimized by knowing the data 
  # and short circuiting
  #
  if ( ($Conf[7]=~ /NO/i) && ($value!= /^0/) 
       or ( $Conf[7] =~ /YES/i)     
  ) {
     # Move time computation inside conditional
     #     
     # Convert the log time to UTC time
     ( $HH, $MM, $SS ) = split ( /:/, $etime );
     ( $year, $month, $day ) = 
           $date =~ m/^(.{4})(.{2})(.*)/;
      $month -= 1;
      $Time= timelocal($SS, $MM, $HH, $day, $month, $year);

      # Build a hash of MTypeHandles with their unique ports
      $Hash{$mth}{$port}++;

      $attrID =~ s/\s*//g;
      $port   =~ s/\s*//g;

      # Remove unnecessary condition
      #
      # if ( $port =~ /\-/ ) {
      #    $port =~ s/\-//;
      # }
      $port =~ s/\-//;

      #remove unnecessary variable
      #
      #$key = $attrID;
      #$val = $DS_Info{$key};
      #
      $val = $DS_Info{$attrID};

      #Remove unnecessary join.
      #$val = join ( "", $val, $port );
      #
      $val .= $port;

      # Remove unnecessary variable
      # @entry = ( $Time, $mth, $val, $value );
      #
      # This is a big array you are making
      # What do you do with it? 
      # Is there any other choice
      #
      # Populate the data array
      # push @Data, [@entry];
      #
      push @Data, [ $Time, $mth, $val, $value ];
  }
}
[download]

Thank you rir,

Your suggestions and a combination of the others have improved performance greatly.

Thanks again rir, and others who who posted suggestions

$Hash{$mth}{$port}++;
# and especially:
push @Data, [@entry];
[download]

The easiest thing to do is probably write ($Time, $mth, $val, $value) out to a new file and process that in a second pass just as you now process @Data. Or, possibly, you can do it in one pass and not save the data at all.

-Nitrox

This advice is not very helpful in this case. Devel::DProf collects statistics for subroutine calls but this script doesn't have them. Devel::SmallProf which is line oriented profiler is more useful in such cases.

--
Ilya Martynov, ilya@iponweb.net
CTO IPonWEB (UK) Ltd
Quality Perl Programming and Unix Support UK managed @ offshore prices - http://www.iponweb.net
Personal website - http://martynov.org