01:07 vendethiel joined 01:17 geekosaur joined 02:36 vendethiel joined
nwc10 smmn jnthn 06:13
nine_ nwc10 is taking the "no vovel rule" wrt jnthn quite seriously 06:26
moritz did you mean "nwc10 s tkng th "n vwl rl" wrt jnthn qt srsly"? :-) 06:28
nine_ f crs 06:29
nwc10 (gd mrng #mrvm) 06:30
hsnt wrkd - stll n jnthn :-(
06:46 domidumont joined 06:51 domidumont joined
masak .oO( gd mrng vtnm ) 06:59
07:03 zakharyas joined
moritz gd mrning msk 07:24
07:27 leont joined 07:33 leont joined 08:02 brrt joined
jnthn ywwwns 08:08
brrt waves
09:05 domidumont joined
masak jnthn: did you know that in Swedish, "y" is a vowel? 10:10
not half a one like in English
jnthn masak: Ystad! 10:11
I think I did know that :)
masak ...and if you mispronounce "Ystad" with a short "y" sound, it means "curdled" (as in cheese) instead of the name of a city :P 10:12
jnthn fwiw, the MoarVM release process is 1) easy, and 2) documented in docs/relesae_guide
The only step only I can do at present is actually upload the release 10:13
masak jnthn: how many distinct people have cut a MoarVM release?
jnthn But everything else is do-able but people with a commit bit
masak: Not enough! :P
I'm pretty sure some folks have done the bulk of the work and I just uploaded the result, though.
nwc10 cym 10:33
arse
cwm
w is a vowel in Welsh, and Welsh loanwords 10:34
jnthn What is the world cwming to...
nwc10 oh, apparently Welsh doesn't just need ŵ 10:35
but also ẃ ẅ and ẁ 10:36
(says one page, but not another)
jnthn The only Welsh word I know is cwrw :P 10:38
nwc10 I didn't know *that* one, but I guessed
jnthn blogs.perl.org/users/yuki_kimoto/20...cking.html is entertaining in so far as MoarVM has for probably close to 2 years been able to completely optimize away arity checks on most hot-paths... 10:42
...so dynamic languages aren't doomed to have to do that as a runtime check. 10:43
At least, not post-optimization
nine_ Also giving Javascript and PHP as examples for "good" "natural" behavior is...intriguing
moritz also, the arguments in there are pretty weak
I think most users of programming languages that got this one right would agree that argument checking is just essential to getting good feedback from the compiler 10:45
or the runtime, whatever
11:00 domidumont joined 11:02 brrt joined
jnthn Sure, the lang design side of the argument also strikes me as weak, but lang design is always something folks can debate. The perf bit is a decades-out-of-date view though. :) 11:04
11:04 diakopte1 joined, arnsholt_ joined, japhb_ joined 11:05 ggoebel114 joined 11:07 colomon joined 11:30 retupmoc_ joined
psch maybe i'm just getting hung up on bad english, but it seems to me that the blog author doesn't really have any argument except "it should be slower but i don't want to measure"..? 11:37
ah, no, there's "other dynamic language do it too" as well 11:38
12:00 avar joined 13:02 colomon joined 13:06 domidumont joined 13:09 vendethiel joined
[Coke] I could have done the moarvm release, btw, but I have no idea if what's on master was good to release. the actual bits of commiting and pressing buttons, sure, that part's easy. 14:02
Maybe we can plan for someone aside from jnthn to do May's release of Moar, JIC.
14:06 colomon joined
jnthn Aye. I can cut a Moar release this evening. 14:08
[Coke] that'd be great. I'll do the nqp/rakudo by wednesday at the latest. 14:11
14:18 colomon joined
brrt i'm wondering how a linear scan register allocation algorithm can deal with pre-coloring 14:19
and with 'dead ranges', i.e. a range were everything must be spilled... 14:20
14:23 domidumont joined
brrt hmmmm 14:24
14:28 colomon joined
timotimo when you have a range where everything must be spilled, just break up the whole algorithm into two pieces of data? 14:31
14:32 vendethiel joined
timotimo jnthn: how do we feel about reverting the "set up VMNulls everywhere" code until we find something that's more efficient? 14:48
jnthn Well, it fixes quite a few SEGVs? 14:49
And we're talking "several percent"?
timotimo you said it was costly; i didn't do a precise measurement yet, but perf showed those instructions were hot in that function 14:56
i'm not yet very good at reading perf, though
and it only fixes segvs if you run moar on fuzzed bytecodes 14:57
jnthn m: my $x := $x; say $x
camelia rakudo-moar 1aabef: OUTPUT«===SORRY!=== Error while compiling /tmp/KOuXFsHcVu␤Cannot use variable $x in declaration to initialize itself␤at /tmp/KOuXFsHcVu:1␤------> my $x := $⏏x; say $x␤ expecting any of:␤ term␤»
timotimo afl-fuzz is very good at finding instructions, rewiring one of the object operands to point at an object register that hasn't been written to yet, and saying "ha, i crashed it!"
jnthn Hm, there was some case a bit like that in RT that SEGV'd that I figured it'd also fix
But we have had crashes in the past resulting from such things, anyway 14:58
I'd rather keep it and work out how to fix things efficiently later.
timotimo if you say it ain't that bad, we'll leave it in for this month's release
it's likely that people would notice a performance impact, but unlikely they'd stumble upon the segfaults; at least i thought so until you said we actually had this kind of segfault happen in the wild already
jnthn Well, I didn't notice any notable increase in the Tux Test::CSV numbers 14:59
Which is a good canary for real world code
I'd be much more worried if I saw a big jump there
We might check the daily benchmark runs also
timotimo ah, of course! you have those! why did i already forget about them? 15:00
nine_ timotimo: This may come as a shock to you, but there is a possibility, that you are actually human ;) 15:01
timotimo darn, all the effort i put into making sure i'm not human gone to waste :(
jnthn
.oO( botornot )
 15:02
timotimo did someone inject some botox into me while i was sleeping? 15:04
moritz if you have to ask.. probabl yes :-) 15:14
timotimo that'd explain why i'm not a bot, then 15:19
because clearly that's what botox does
15:42 leont joined
jnthn Ugh, tired 16:38
Will rest a little then cut the release
17:05 vendethiel joined
dalek arVM: 46feef2 | jnthn++ | docs/ChangeLog:
ChangeLog for 2016.04.
 17:09
arVM: 0d03bb6 | jnthn++ | VERSION:
Bump VERSION.
 17:10
17:11 leont joined
japhb_ That was a quick rest! 17:15
timotimo powernap, perhaps 17:16
japhb pernaps
On my desktop, that's only different by a couple *pixels* 17:17
jnthn www.moarvm.org/releases/MoarVM-2016.04.tar.gz 17:18
[Coke]: ^^ :)
Website can wait until after more rest :)
leont is still observing weird Proc::Async related ordering issues in TAP::Harness occasionally. Figuring out if it is my code, rakudo or moar that's causing it it tricky :-/ 17:35
Though things have gotten a lot better
timotimo yeah, i bet!
that's good to hear
hoelzro leont: ordering issue? 18:15
I encountered something like that some time ago
leont It seems I'm getting my result in the wrong order, but I don't know why. I just know I sometimes get errors suggesting my output was misordered, or even weirder errors 18:16
hoelzro lemme see if I can dig up that RT I filed
ah, here it is: rt.perl.org/Ticket/Display.html?id=126425 18:17
leont No, my issue is dealing with output from stdout 18:21
[Coke] jnthn: Danke. 18:22
18:22 domidumont joined
hoelzro ah ok 18:24
leont Also, I have no idea how «while ($!grammar.subparse($!buffer)) -> $match { $!buffer.=substr($match.to);» could give me «Start argument to substr out of range. Is: 6, should be in 0..5» errors :-/ 18:38
I'd almost suspect the .act it's in isn't serialized properly, but that doesn't sound likely does it 18:39
18:54 FROGGS joined 18:55 Ven joined 19:25 vendethiel joined
dalek href="https://moarvm.org:">moarvm.org: b6f6332 | jnthn++ | / (4 files):
Update for 2016.04 release.
 19:27
20:56 lizmat joined 22:05 vendethiel joined 22:16 geekosaur joined 22:31 colomon joined

Powered by Cro and the Raku® Programming Language.

Please report any issues / comments / feature requests as an issue on App::Raku::Log.

Thank you!