There is a common mechanism for all descriptors, to check whether a descriptor is ready for reading/writing, please check out IO::Select.

The main point is to call can_read or can_write before you really go reading/writing. This way, you even don't need to bother blocking/non-blocking mode. More important, you gain a better control.

On the other hand, I attached a simple sample to show pipe:

parn.pl:
use strict;
use warnings;

$| ++;

print "Entering parn.pl...\n";
my $chld;
open($chld, "perl -w chld.pl|") || die "failed to open child process\n
+";
my @ready;
while (1) {
    my $from_chld = <$chld>;
    print $from_chld * 2, "\n";
}
   
chld.pl:
use strict;
use warnings;

$| ++;

while (<>) {
   if ($_ < 100) {
      print "$_";
   }
}
[download]

Update: for getc, although it returns one char a time, but unfortunately, this only happens after return is hit. An alternative way is presented in perlfunc under getc.