Thank you for providing sample input, but unfortunately, when I run the code on this sample input, the output is empty. Could you provide sample input that produces some output, and to play it safe also provide that output, each inside <code> tags? See also Short, Self-Contained, Correct Example.

In general, as has already been suggested, a hash table provides for much faster lookups than a linear scan with nested loops.

use warnings;
use strict;
use List::Util qw/ shuffle /;
use Time::HiRes qw/ gettimeofday tv_interval /;
use Test::More tests=>2;

my @looking_for = qw/ foo bar quz baz /;
my @looking_in = shuffle qw/ foo bar quz baz / x 100_000,
    qw/ some other stuff we're not looking for / x 2_000_000;

{
    my $t0 = [gettimeofday];
    my $found_count;
    for my $haystack (@looking_in) {
        for my $needle (@looking_for) {
            if ( $needle eq $haystack ) {
                $found_count++;
            }
        }
    }
    is $found_count, 400_000, 'linear scan';
    diag sprintf "that took %.3fs", tv_interval($t0);
}

{
    my $t0 = [gettimeofday];
    my $found_count;
    my %needles_hash = map { ($_=>1) } @looking_for;
    diag 'needles_hash: ', explain \%needles_hash;
    for my $haystack (@looking_in) {
        if ( $needles_hash{$haystack} ) {
            $found_count++;
        }
    }
    is $found_count, 400_000, 'hash lookup';
    diag sprintf "that took %.3fs", tv_interval($t0);
}

__END__

1..2
ok 1 - linear scan
# that took 2.274s
# needles_hash: {
#   'bar' => 1,
#   'baz' => 1,
#   'foo' => 1,
#   'quz' => 1
# }
ok 2 - hash lookup
# that took 0.599s
[download]

When you're asking the question "do the strings in the haystack contain any of the needles", or in general when what you're looking for is not a fixed string but can be expressed as a regex, an alternative is to build a regex.

use warnings;
use strict;
use List::Util qw/ shuffle /;
use Time::HiRes qw/ gettimeofday tv_interval /;
use Test::More tests=>2;

my @looking_for = qw/ foo bar quz baz /;
my @looking_in = shuffle qw/ xyfooz abcbarx 123quzy abazz / x 10_000,
    qw/ some other stuff we're not looking for / x 100_000;

{
    my $t0 = [gettimeofday];
    my $found_count;
    for my $haystack (@looking_in) {
        for my $needle (@looking_for) {
            if ( $haystack =~ /\Q$needle\E/ ) {
                $found_count++;
            }
        }
    }
    is $found_count, 40_000, 'linear scan';
    diag sprintf "that took %.3fs", tv_interval($t0);
}

{
    my $t0 = [gettimeofday];
    my $found_count;
    my ($needles_regex) = map {qr/$_/} join '|', map {quotemeta}
        sort { length $b <=> length $a or $a cmp $b } @looking_for;
    diag "needles_regex: ", explain $needles_regex;
    for my $haystack (@looking_in) {
        if ( $haystack =~ $needles_regex ) {
            $found_count++;
        }
    }
    is $found_count, 40_000, 'regex';
    diag sprintf "that took %.3fs", tv_interval($t0);
}

__END__

1..2
ok 1 - linear scan
# that took 2.716s
# needles_regex: qr/bar|baz|foo|quz/
ok 2 - regex
# that took 0.212s
[download]

But your basic problem is using two loops resulting in 880*75000 comparisons. And not the full scan of files.

Using lookup hashes %A and %B with would speed this up dramatically.

Just use $look;$sec;$cls as keys.

The fastest way are hash slices then:

@A{keys %B} will give you the intersection.

Edit

But maybe easier to understand is to loop over the keys of the smaller hash and to lookup in the bigger one. This will also preserve order.

Update

Same principle of you want to avoid holding the bigger file in memory, just read it line per line while comparing against a hash of the smaller one.

I got but how can I do this because my knowledge in perl is limited. Which direction I have to understand first?

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]

I got this but the result I receive is not the same from the first script. I was analysing this code and I notice I will not cover the all combinations. My first result was 6382 lines and the result of this script was 928. The lines of second result file is in the first result file but it still missing some lines

Looks like your script creates multiple output records, so if

l100107,bbbbbbb,c_0300,loc,12,6

in FileA matches

l100107,bbbbbbb,c_0300,389

in FileB, the output is 6 lines (the value of $qtd the last column)

l100107,bbbbbbb,loc,12,389
l100107,bbbbbbb,loc,12,389
l100107,bbbbbbb,loc,12,389
l100107,bbbbbbb,loc,12,389
l100107,bbbbbbb,loc,12,389
l100107,bbbbbbb,loc,12,389

Is that what you want ?

Also, can you please explain what this code line does.

last if $count == $max;
[download]

poj