($tag)=split (/\|/,$data,2);
[download]

$tag=substr ($data,0,index($data,'\|',));
[download]

Benchmark: timing 1000000 iterations of split, substr...
     split:  8 wallclock secs ( 6.93 usr +  0.00 sys =  6.93 CPU) @ 14
+4279.32/s(n=1000000)
    substr:  1 wallclock secs ( 1.54 usr +  0.00 sys =  1.54 CPU) @ 64
+8088.14/s(n=1000000)
[download]

If the line you are splitting is very long, then yes, split will be slower than substr. To answer your question, how about something like:

use strict;
use warnings;

while (<DATA>) {
   my $index = index($_,'|');
   next unless $index;               # add error checking!!
   my $tag = substr($_,0,$index);

   expensive_function($tag,$_) if $tag =~ / /;
}

sub expensive_function {
   my ($tag,$line) = @_;
   my ($label,$number)  = split(' ',$tag,2);
   my ($digit) = $label =~ /_(\d+)$/;
   my @field = split(/\|/,$line);

   # now do something with all of this ...
}

__DATA__
START_1 123|
FILE 2222| XXXX| AAAA|
NEW |
END_1|
[download]

This is, of course, complete guess-work. You really should take the time to explain what it is you are trying to accompish - we know that you want to speed your application up, but what does your application do? Take the time to carefully explain yourself and get better answers. I can tell you are in a hurry because you have a plethora of typos. Take a look at How (Not) To Ask A Question.

jeffa

L-LL-L--L-LL-L--L-LL-L--
-R--R-RR-R--R-RR-R--R-RR
B--B--B--B--B--B--B--B--
H---H---H---H---H---H---
(the triplet paradiddle with high-hat)