From: Tom Metro <[EMAIL PROTECTED]>
   Date: Thu, 03 Mar 2005 18:45:21 -0500

   Sean Quinlan wrote:
   > ...Parrot will be able to save and reuse it's bytecode,
   > which might give you something close to a platform specific executable...

   Wasn't the objective a non-platform specific way to install modules?

Parrot bytecode is intended to be platform independent, even to machine
endian independence.  The theory (last I looked) is that it should be
possible to compile to bytecode on a little-endian machine (e.g.) and
run it on a big-endian machine, albeit possibly not as efficiently.  But
at present, this doesn't seem to be in the test suite -- at least that I
can see.

   In any case, portable bytecode is helpful, but I think we'll still
   see a certain percentage of modules that need to be compiled.

As machines get faster and ease of cross-platform installation gets more
important, I expect the need for C-level hackery will go down.  I
suspect this is overused even at present.  Several years ago, I wound up
rewriting a mini-app with an XS hack in pure Perl, partly because I had
trouble compiling it on my config and partly because an experiment
showed that the increase in runtime would be less than 15%.

   Unless someone deems it worthwhile to rewrite, say the MySQL network
   protocol libraries, in Parrot.

That wouldn't be necessary; you'd just need to write it in something
that compiles to PBC.  Like Perl.  ;-}

   (Although on platforms with dynamically loadable libraries, if Parrot
   can talk to them without needing compiled glue code, like an XS, then
   the need for compiled code should go way down.)

That is also the theory, and I think it currently works in practice
(though personally I've had problems with some of the example code).

   The easiest thing for the end user, of course, would be something
like the MacOS "fat binary" that supported both 680x0 and PPC in the
same install . . .

                                        -- Bob Rogers
                                           http://rgrjr.dyndns.org/
 
_______________________________________________
Boston-pm mailing list
Boston-pm@mail.pm.org
http://mail.pm.org/mailman/listinfo/boston-pm

Reply via email to