Hi % allison has joined #parrotsketch hi Jesse hey leo, allison hiya % audreyt has joined #parrotsketch greetings hey audrey. we'll give folks a couple more minutes to appear, then get cracking. sure right then. Hi everybody. leo: how's things? - still waiting for some clarification WRT namespaces and classes - and WRT integer division overflow - debug improvements (gdb and pdb), better/nice trace - lot of JIT codegen improvements on ppc and x86 - added libreadline support to ParrotIO - MMD: handle empty :multi() - implement modular pow algorithm for pow_n_n_i - http://shootout.alioth.debian.org/sandbox/benchmark.php?test=partialsums&lang=all is now at 1 second with Parrot JIT, that would be 1.2s on the sandbox machine or faster than all incl. Intel Fortran / C (the current PIR test is running w/o JIT) EOR (ooh re libreadline) so, blocking on those couple items from chip. anything else? nothing else ok audrey: what's new and exciting in pugsland? - learned 8 hours ago that I can make it to Germany and attend 2/3 of GPW and stay at Leo's afterwards to make pugs->parrot object space work - reformulated Perl6's lexical import/pad/etc structure after brief q/a with lwall and extended discussion with gaal, using pdd20 as a starting point -- though we'd require true "state" variable still - shaved pugs compilation from 20min to 5min on my laptop by using YAML as bytecode [sic - really "intermediate parse tree form"] on-disk format applying the same transformation to Test.pm tomorrow would likely shave the "make test" time from 200min to ~30min or (if optimised) from 100min to ~15min oo woot - hackathon is about to start tomorrow; theme for me is compiling the newly formulated object and pad space to perl5 using putter's recent work of a rules engine implemented in p5 and stevan's work of the metaobject protocol in p5 all we lack is prim bindings; with them we should be able to compile p6 to (fast) p5 -- it helps that larry will be there to make rulings onthefly re: semantics - will record the approaches we take with parrot in mind, and re-expressed them to PIR/PASM when I get to Leo's place. - as noted above, all Pugs data structures from the Haskell side can be serialized and deserialized as YAML now, so interop with Parrot compiler tools in the future should become much easier using common YAML schema for things such as ASTs and annotations. - not blocking on anything; enjoying Tel Aviv and Jerusalem massively. ^D cool :) allison? working on Punie this week the first few control structures and mainly getting the operator precedence parser into place I set that up with a translation table on the PAST to POST transformation so that adding the rest of the operators (the ones that near equivalents in PIR/PASM) is a matter of adding 2 entries: one in the OPP and one in the translation table I'll probably to more of those this week also continuing with watching all commits I see some small increases in entropy, but overall not bad EOR ok. blocking on anything? chip, are you here? nope, not blocking And if no chip, I think we're into general question time allison: does demo.p1 represent the state-of-the-world for punie? more or less it has most of the statements I know Punie can run now nod. I see that it passes base_if now but with cominatorial explosions, Punie can do more than what's there glad to see that the test cases from perl1 can still make useful contributions yeah, the perl1 tests have been my target along the way wonderful variables are the biggest thing I need to add next perl1 has no lexical variables so you don't need to touch the padinfo which makes is considerably easier :) (in a good way.) it does have my really? I thought it's a perl5 thing. or are you planning to add "my" to perl1 as part of punie extensions? hmmm... maybe not thought I saw it in the tests, but doesn't seem to be there when I grep for it (and perl5's lexical scoping rules are very baroque, so you probably doesn't want to go there) well, that's a nice simplification I too thought my was new-for-5 indeed -- a global pad is all you need (perl1 didn't have packages, either.) yeah, I noticed that which is why Punie translates the whole thing to one big .sub in PIR nod. one single compilation unit is all that's needed btw, is it still using the space-inefficient AG module though, of course, the next target after Perl 1 is Perl 5, but that's a ways down the road (at least a couple months ;) or did luqui's revamped algorithm make into the tree? allison: well, in that case you can take larry's p5top5 optrees as input :) I looked at the revamped algorithm, and rejected it, though I don't remember why But, TGE does need extensive work for one thing, I plan to convert it so it compiles tree grammars down to PIR (and then bytecode) at the moment, it does everything at runtime k. thanks for the exciting update :) Ok. with no chip and no patrick, I think we're free until next week See you all then thanks obra thanks Jesse % allison has left allison!~chatzilla@sub17-30.member.dsl-only.net % Chas has joined #parrotsketch Hmmm... So who's the Norwegian Blue? % Chas has left #parrotsketch % rafl has left rafl!~rafl@armitage.dreker.org