You could maintain a small shift buffer with the last two lines read, which would work even when the data file is large.

use strict;

my $value = 0;
my @linebuf;
open IN, "data.file" or die "$!";
for (0..1) {
   $_ = <IN>;
   push @linebuf,$_;
}
while (<IN>) {
   if (/^--/ and $linebuf[0] =~ /ABC SET/) {
     print "Matched ",$linebuf[0];
     $value++;
   }
   shift @linebuf;
   push @linebuf,$_;
}
close IN or warn "$!";
print "total found is $value\n";
[download]