Perl 6 - the future is here, just unevenly distributed

IRC log for #parrot, 2014-07-15

Parrot | source cross referenced

| Channels | #parrot index | Today | | Search | Google Search | Plain-Text | summary

All times shown according to UTC.

Time Nick Message
01:30 FROGGS_ joined #parrot
02:34 kid51_ joined #parrot
02:36 kid51__ joined #parrot
05:45 davidfetter joined #parrot
06:16 bighugedog joined #parrot
08:16 basiliscos joined #parrot
09:47 Psyche^ joined #parrot
10:19 Tene joined #parrot
11:02 kid51 joined #parrot
11:50 bighugedog joined #parrot
16:11 rurban1 joined #parrot
16:42 Chirag joined #parrot
17:07 Coke looks like one more parrot test is aborting, have a few failures that need to be ticketed & fudged.
17:07 Coke er, rakudo-parrot
17:19 Chirag rurban: If I am understanding the problem correctly, maybe we can check the type for the incoming "PMC *signature"  at the beginning of the call itself and then maybe create type-specific "*call_object"s ??
17:29 rurban which call?
17:29 rurban the call_obj is always a CallContext PMC
17:29 rurban which holds various sig arrays (args and arg flags)
17:35 Chirag yes it is always a PMC.. so to overcome boxing of signatures, we will detect the type carried by PMC *signature, right?
17:37 Chirag and I mean do we try to stop using call_object altogether? and instead come up with independent logic for each type?
17:56 rurban hmm, I'm not sure
17:58 rurban A union type as with Pcc_cell sounds good
18:00 rurban Hmm, we already do that: ATTR struct Pcc_cell *positionals
18:02 Chirag hmm.. so the methods that send PMC *call_object might need a change in definition to send a union-type? or we expand them in our call itself?
18:03 rurban we cannot touch the CallContext itself, we want to improve the sigs
18:05 Chirag ok.. so some sort of a patch-work, I guess
18:15 rurban mostly like inlining access to CallContext attrs
18:16 rurban getting rid of the autoboxing would be nice, yes
18:17 rurban the call context already was prepared to hold arbitrary type (PSNI) args, which was thrown away to support a PMC array to hold it
18:17 rurban until the 1.6 release or so
18:18 rurban http://wknight8111.blogspot.com/2009/10/optimizing-parrot.html
18:18 Chirag I can look at that commit and try to get it back..
18:19 rurban it's not just a simple commit. It was whole refactoring. look for bacek and Context
18:20 rurban I'm looking now at the kill_pccinvoke branch
18:21 rurban 863cfa471593f0342b45f1d973e9bfc49291c9b1
18:25 rurban e.g. e5dead2ed00fdcfc383ad82fe92c74636f8f2a45 changed the sig array to a PMC
18:25 rurban tig src/inter_call.c is nice to use
19:06 Chirag just finished going through the code... I guess all this work is now non-existent..
19:11 rurban1 yep
19:13 Chirag the challenge/task will be to merge this into our current version
19:50 Coke what's the "unstable" note regarding for rakudo/parrot?
19:51 Coke ah, sounds like folks are on it. nevermind.
19:51 rurban1 unstable? last month all the parrot spectest ran okay, this month not. it's getting closer though
20:33 rurban updated http://wiki.enlightenedperl.org/gsoc2014/ideas/improve_performance_of_method_signatures a bit with the done work
20:34 Coke ah. that's mostly due to fudging issues in the suite now that we're trying to maintain 3 backends that all have varying states of todos. ideally, you should only have a particular failure for a day or so until someone can fudge the test properly for that backend. (but if you have something fudged on all, then unfudge it because it works in one place, but it fails on the other two, that will fail until someone fudges it just for the failing backend).
20:34 Coke it's kind of a
20:35 Coke whoops. It's kind of annoying, but all we have to do to fix it is finish rakudo. :)
20:45 rurban documented now all of the current ideas in the wiki
22:13 kid51 joined #parrot
23:00 basiliscos joined #parrot
23:58 khisanth__ joined #parrot

| Channels | #parrot index | Today | | Search | Google Search | Plain-Text | summary

Parrot | source cross referenced