Try this.

#!/usr/bin/perl -w

use strict;
use warnings;

use Data::Dumper;

my $data = join '', 
       "U|(Memory is incorrectly balanced between the NUMA ",
       "nodes of this system, which will lead to poor performance.  ",
       "See),_NUMA_CRIT_,;(/proc/vmware/NUMA/hardware),_NUMA_CRIT_",
       ",__REFERAL__file:__SELF__;( for details on your current ",
       "memory configuration),_NUMA_CRIT_,;";

print "DATA\n$data\n\n";

#(my $m_type = $data) =~ s/^((?:U|M))\|.*$/$1/;  # gets U
#(my $p_list = $data) =~ s/^(?:U|M)\|(.*)$/$1/;  # gets U, i think thi
+s is the reason you weren't getting anything

# This approach does what I think you are trying for without as much w
+ork for you or the computer.
my ($m_type, $p_list) = split /\|/, $data, 2;

my @sub_p_tmp = split /;/, $p_list;

print "MTYPE: $m_type\n";
print "PLIST: $p_list\n";

my @args;

# Try using
# for my $p_tmp ( @sub_p_tmp ) 
# Instead of the c style for loop, unless you are processing HUGE arra
+ys.
for ( my $i = 0; $i <= $#sub_p_tmp; $i++ )
{
    print "$i:  $sub_p_tmp[$i]\n";
# Here's your regex based method fixed to work
#    if ( $sub_p_tmp[$i] =~ m/^\(([^)]*)\),([^,]*),(.*)$/ )
#    {
#        @args = ($1, $2, $3); 
#    print "\nArgs:\n"; 
#    print Dumper(\@args)."\n";
#    }

# Here's a split based approach. 
# in both approaches @args is overwritten each pass through the loop. 
    my @temp = split /,/, $sub_p_tmp[$i]; 
    if ( @temp )
    {
        # Get rid of the extra parens.
        $temp[0] =~ s/^\(|\)$//g;

    # Overwriting the value of @args each time.
    # Do you mean 
    #    push @args, @temp[0..2];
    #     or
    #    push @args, [ @temp[0..2]];
        $temp[0] =~ s/^\(|\)$//g;

        @args = @temp[0..2];
    print "\nArgs:\n"; 
    print Dumper(\@args)."\n";
    }
}
[download]

Thanks TGI, I'll definitely play around with this!