I find YAML is usually easier to read for debugging structures than Data::Dumper

#!/usr/bin/perl

# http://perlmonks.org/?node_id=1179628

use strict;
use warnings;

my $rec;
my %HoH;

while(<DATA>)
  {
  if( /^\[(.*?)\]/ )
    {
    $rec = $1;
    }
  elsif( defined $rec and /(\S+)\s*=\s*(".*"|\S+)/ )
    {
    $HoH{$rec}{$1} = $2;
    }
  }
use YAML; print Dump \%HoH;

__DATA__
[CELL_NAME1]
COMMENT           = "Perl parsing"
FIRST             = "TEST1"
SECOND            = "ID1"
THIRD             = 123
FOURTH            = "THREE"
FIFTH             = 12345
SIXTH             = 6789
SEVENTH           = QWERTY

[CELL_NAME2]
COMMENT           = "Tester"
FIRST             = "TEST2"
SECOND            = "ID2"
THIRD             = 1234
FOURTH            = "FOUR"
FIFTH             = 12345
SIXTH             = BOARD
SEVENTH           = MOUSE

[CELL_NAME3]
COMMENT           = "Parser"
FIRST             = "TEST3"
SECOND            = "ID3"
THIRD             = 12345
FOURTH            = "FIVE"
FIFTH             = 12345
SIXTH             = PAD
SEVENTH           = KEY
[download]