This function causes an immediate core dump. See also the -u command-line switch in perlrun, which does the same thing. Primarily this is so that you can use the undump program (not supplied) to turn your core dump into an executable binary after having initialized all your variables at the beginning of the program. When the new binary is executed it will begin by executing a goto LABEL (with all the restrictions that goto suffers). Think of it as a goto with an intervening core dump and reincarnation. If LABEL is omitted, restarts the program from the top.
WARNING: Any files opened at the time of the dump will not be open any more when the program is reincarnated, with possible resulting confusion on the part of Perl.
This function is now largely obsolete, partly because it's very hard to convert a core file into an executable, and because the real compiler backends for generating portable bytecode and compilable C code have superseded it. That's why you should now invoke it as CORE::dump(), if you don't want to be warned against a possible typo.
If you're looking to use dump to speed up your program, consider generating bytecode or native C code as described in perlcc. If you're just trying to accelerate a CGI script, consider using the mod_perl extension to Apache, or the CPAN module, CGI::Fast. You might also consider autoloading or selfloading, which at least make your program appear to run faster.