|
weekly Perl 6 status meetings with phase transitions: Wed 17:30 UTC | IR clogs at irclog.perlgeek.de/phasers/today Set by moderator on 30 September 2011. |
|||
|
06:30
cognominal joined
10:34
sorear joined,
jnthn joined
11:04
cognominal_ joined
11:36
colomon joined
16:15
cognominal___ joined
16:21
cognominal_ joined
16:50
cognominal joined
17:01
pmichaud joined
|
|||
| pmichaud | I will be missing #phasers again today -- have little to report. Tuit shortage continues (and likely continues for at least another 3-4 days). | 17:02 | |
| jnthn | o/ pmichaud | 17:03 | |
| moritz | my prereport: | 17:25 | |
| * fixed most cases of variables needing to be predeclared (which didn't used to error out) | |||
| * started to catch typed exceptions from the setting | 17:26 | ||
| * more thoughts on docs/faqs | |||
| * hacking on bigint in nqp | |||
| * talked with which seems to be the libtommath maintainer. He'll probably implement a conversion from mp_int to float/double | 17:27 | ||
|
17:27
prammer joined
|
|||
| moritz | (which we'll need) | 17:27 | |
| jnthn | \\o/ | 17:28 | |
|
17:28
diakopter joined
|
|||
| moritz | * used native ints in the setting to speed up Str.{succ,pred} | 17:28 | |
| EOR | |||
| sorear | DID: merged /serialize branch, released v11, implemented BEGIN in niecza | ||
| broke constant Foo = CLR::(...) | |||
| discussed my \\x a bit with TimToady | 17:29 | ||
| EOR | |||
| diakopter | did: completed first hack at porting 6model/dotnet to luajit. will do: make it work. eor. | ||
| jnthn | This week... | 17:30 | |
| * Started working on bigint support | |||
| * Realized that really depended on doing other REPR bits | |||
| * Did a large refactor of REPR handling | |||
| * Will help with bigint; also serves as ground work to enable compact structs and packed arrays | |||
| * Various optimizations and bug fixes | |||
| * Worked on my talk for TCPW and LPW | |||
| Over the next week... | |||
| * Going to TCPW | |||
| * Bit more work on optimizer | |||
| * Bit more work on BigInt | |||
| * Join in with diakopter++'s nqplua hacking | |||
| EOR | |||
| sorear | will do: update documentation. Fix constant = CLR. Add some kind of use-from-clr. Continue wondering why niecza-the-compiler is so slow. EOR | 17:32 | |
| [Coke] | did: minor ticket/fudge wrangling. lamenting lack of time to hack on smolder replacement. | ||
| sorear | o/ | 17:33 | |
| moritz | any more prereports? | 17:37 | |
| sorear needs to find a way to make sense of all these flat profiles | 17:39 | ||
| moritz | anything we want/need to discuss? | 17:42 | |
| sorear | is everyone on the same page with bigints? | 17:43 | |
| jnthn is fine with everything moritz++ has been doing :) | |||
| (And very glad he's picked it up.) | |||
| TimToady | do we need to be on the same page, as long as everything hides behind the Int abstraction? | ||
| or does it leak? | 17:44 | ||
| jnthn | I'm aiming towards it being hidden behind the Int abstraction. | ||
| That's a bit part of why I spent the last week doing the repr changes :) | |||
| *big | |||
| diakopter | I encountered some odd discrepancies between the behaviors of luajit when loading RakudoRuntime.lbc (luajit bytecode) and RakudoRuntime.lua (concatenated source of all the *.lua files under lua/runtime). Probably some bugs that need reported to the luajit author. | 17:45 | |
| jnthn | diakopter: Hm, odd. | ||
| TimToady | you do realize that p6 will put more stress on a VM than almost anything out there :) | ||
| diakopter | I might be able to golf one of them down enough to be a reasonable bug report | 17:46 | |
| instead of a 30,000 line lua file | 17:47 | ||
| (NQPSetting.lua) | |||
| still, loading from bytecode saves only 20ms or so | |||
| jnthn | That's some really quite fast parsing :) | 17:48 | |
| ...or slow bytecode loading ;) | 17:49 | ||
| sorear | lua is famed for that | ||
| I am wondering whether I want to keep niecza's compiler in Perl6 | 17:50 | ||
| TimToady | just make Perl 6 run faster ;) | 17:51 | |
| sorear | the problem with that is that the profile is quite flat | 17:52 | |
| I need to make *everything* faster to get a significant improvement, I think | |||
| moritz | sorear: wouldn't it be easier to profile differently than rewriting the whole compiler in another language? | 17:53 | |
|
17:53
cognominal_ joined
|
|||
| sorear | moritz: I've tried three different Mono profilers and one crude P6 profiler and I can't generally get any function to show above 2% on the self-compile task | 17:55 | |
| TimToady | there are likely to be other abstraction levels than functions that one could factor stuff out from | 17:57 | |
| there are also likely to be optimizations for commonly occuring languages | 17:58 | ||
| diakopter | if I could get RedGate's decompiler to work on niecza.exe and its dlls, I could use the vstudio profiler | ||
| TimToady | parse bits with yacc until you get an error, then reparse to get a better error, say | ||
| there are probably things common across many different functions that are potentially factorable | 17:59 | ||
| sorear | diakopter: does vstudio support ilasm as a language? | ||
| diakopter | no | ||
| sorear | the problem with niecza.exe is that, well, it's not made from C# | 18:00 | |
| currently it uses a lot of C#-illegal identifies, but that could be fixed | |||
| diakopter | right; I'd need to make Reflector decompile it | ||
| sorear | TimToady: I do not quite understand what you mean by commonly occurring languages. | 18:02 | |
| TimToady | Standard Perl 6 vs mutated varieties | 18:03 | |
| sorear | ah. | 18:04 | |
| TimToady | depending on how much overhead there is in regen of a language | ||
| std used to cache some of that | |||
| diakopter | on another note, Pm mentioned that his new regex engine is still going to compile patterns to PIR (with gotos), so I won't be able to use it wholesale in nqplua. <sad /> | 18:06 | |
| sorear | I don't think language regen is a huge deal; LTM stuff totals to <10% on the profiles | 18:10 | |
| sorear wonders if it would be useful to just write niecza/c and then use a native profiler :D | 18:15 | ||
| diakopter | :) | ||
| sorear out | 18:24 | ||
|
18:41
cognominal___ joined
19:42
colomon joined
|
|||