sub dos2unix { system( qq{perl -i -pe 's/\\cM*//g' $_} ) for @_ }
[download]

(No more inefficient than opening a pipe (?!?!) to mv rather than using unlink())

Well, it's quite inefficient. You're starting a process for each file, and do a substitution for every empty string.

sub dos2unix {
    system perl => '-i', '-pe' => 'tr/\x0D//d' @_ 
}
[download]

Abigail

Well then, you could create a perl script file as follows:

#!/usr/bin/perl

# in-place dos-to-unix conversion (remove ^M's):

my @file_list = <>;  # update: adding these two lines to
chomp @file_list;    #    read a list of file names

for my $file ( @file_list ) {    # not "@ARGV"
    open( I, $file ) or die "$file: $!";
    open( O, ">$file.unx" ) or die "$file.unx: $!";
    while (<I>) {
        s/\cM*//g;
        print O;
    }
    close I;
    close O;
    rename "$file.unx", $file or die "renaming $file.unx to $file: $!"
+;
}
[download]

Note that this differs slightly from the code in your original post: your strange and less efficient approach would cause each "CRLF" sequence in a dos file to be converted to "LFLF", which means your output files would have twice as many "lines" as the originals, and half of those output lines would be empty. The version here simply removes the "CR". (Or maybe it was your intention that dos files should show up in unix as "double-spaced"?)