March 08, 2004

Checkoffs for the Pie-thon

Just in case anyone's keeping track, here are a few things that are working:

  • Bytecode loading
  • Objects
  • Method calls
  • Operator Overloading

oh, yeah, and all that pesky math stuff, subroutines, and strings. They work too, though Unicode needs some help. And variables. We can do variables.

Still need some work on nested namespaces, actual bytecode translation, and object performance, but...

Posted by Dan at March 8, 2004 02:04 PM | TrackBack (0)
Comments

Are you keeping these files in CVS online somewhere? I know I wouldn't mind keeping an eye on the competition. =)

Posted by: Brett at March 8, 2004 06:37 PM

Parrot? Sure, the whole thing's in CVS for anyone to grab. You can get it via anon CVS, rsync, or a web-accessible thrice-daily snapshot, depending on your needs, preferences, and firewall issues. The official point (well, OK, sub-point :) releases are all also on CPAN so if you've got a local CPAN mirror handy you may well already have it.

Details are on the (new, improved, and no longer way out-of-date) Parrot website.

We're pretty easy with checkin access too, so if you feel the need or desire it's yours for the asking... :)

Posted by: Dan at March 8, 2004 06:44 PM

I actually meant your code specific for Pie-thon. I already have an anonymous Parrot checkout.

And as for helping out, I probably should get caught up on my Python work first. Don't want to be cast as a traitor if I end up causing Guido to take a pie in the face. =)

But I do plan on taking a look at the code at some point. Your journal entries have been really cool and made me curious about Parrot.

Posted by: Brett at March 8, 2004 06:51 PM

At the moment there isn't any pie-thon specific code--everything I've done towards it has gone into the repository. The only thing specific to the contest will be the bytecode translator, which I admit I've not started in on. (I ought to start soon, though. All the bytecode disassemblers I've seen for python only do through 2.2, not 2.3)

This'll all be on a bog-standard parrot interpreter--while I do want to win the contest, I'm not going to pull any special clever games for it and, while tempting, pie (mmm, pie!) isn't quite sufficiently tempting to alter parrot's design. Gotta do the right thing in the long run, no matter how much merengue and whipped cream's on the line.

Posted by: Dan at March 8, 2004 06:58 PM

2.3 simplifed the bytecode; FOR_LOOP and SET_LINENO were removed. The cmp_op enumerators were renamed, though in 2.3 .

Python 2.2 forked at rev. 2.37 for Include/opcode.h while Python 2.3 forked at 2.43 in case you care to look at the CVS log to see what changed between the releases for the opcodes.

Posted by: Brett at March 8, 2004 09:32 PM

You mean you're not gonna replace the GC with a leaky malloc system for the benchmark? Jeez, you'd think you wanted a fair comparison or something...

Posted by: Brent at March 9, 2004 02:19 AM

Teh just-in-case list: dropcloth, towel, clean shirt...

Posted by: Clinton at March 9, 2004 08:08 AM

Dan,
Not sure if my recent post to perl6.internals went through as I am not subscribed to the list. I also apologize for replying to this non-related blog entry.

I wanted to let you know I had partial success installing parrot 0.1.0 on Cygwin. It hung on test 12 of extend as previously noted on the list, but I was able to $ kill from another window to allow it to finish. Here are the results:

Cygwin Results

Posted by: Joshua Gatcomb at March 9, 2004 10:16 AM