I tried to get some of what you need but didn't get the use of $i, $j, $k or $modified.

This isn't a complete solution, but it uses the suggestion by LanX to create a hash of the smaller B file and loop through the A file just once. This should speed up your program considerably.

I used 2 pseudo files to stand in for your real files. I hope this gives you some direction for your problem.

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

$|=1;

my $fileA =<<EOF;
l100101,aaaaaaa,a_0100,loc,10,1
l100101,aaaaaaa,a_0100,loc,11,1
l100101,aaaaaaa,a_0100,loc,12,6
EOF

my $fileB =<<EOF;
l103709,bbbbbbb,c_0200,929
l100109,bbbbbbb,b_0100,442
l100107,bbbbbbb,c_0300,389
EOF

my $filea = $ARGV[0];
my $fileb = $ARGV[1];
my $FileC = "result.csv";


open ( FA, '<', \$fileA) || die ( "File $filea Not Found!" );
open ( FB, '<', \$fileB) || die ( "File $fileb Not Found!" );
#open ( FC, ">", $FileC) || die ( "File $FileC Not Found!" );

my %B;
while ( <FB> ) {
    chomp;
    my($look, $sec, $cls, $max) = split ",";
    $B{"$look,$sec,$cls"} = $max;
}

my @A;
while ( <FA> ) {
    chomp;
    my($look, $sec, $cls, $att, $idx, $qtd) = split ",";
    my $keyA = "$look,$sec,$cls";
    
    if (exists $B{$keyA}) {
         my $max = $B{$keyA};
         my $tot = $qtd - 1;
         if ($tot >= 0) {
             print join(",", $look, $sec, $cls, $att, $idx, $max), "\n
+";    
         }   
    }
}
[download]