If I were dealing with a space delimited file of simple characteristics, and I wanted to insert it into a multidimensional array, along with a /start/ .. /end/ conditional, I might do it like this:

while( <$fh> ) {
  chomp;
  next unless length;
  if( /start/ .. /end/ ) {
    push @array, [ split /\s+/, $_ ];
  }
}
[download]

Your definition of the start and end pattern are difficult for me to guess without knowing more about the input data, but you were probably on the right track to begin with there.

If there's any risk that the input file could contain things like spaces embedded within quoted fields, or other more complex constructs that are challenging for split, then you would want to shift to using Text::CSV, specifying space as the delimiter rather than commas.

Oh, one last thought; if the input data has fixed-width fields, then splitting on space becomes less important; you could just unpack based on a fixed unpack specification.