00:13 sxmx joined 01:22 Xliff left 01:43 sena_kun joined 02:37 kvw_5_ joined 03:10 HarmtH_ joined 03:11 HarmtH left 03:31 djinni` left 03:55 djinni` joined 04:54 vrurg left 04:55 vrurg joined 06:35 frost-lab joined 07:31 ggoebel joined 07:48 domidumont joined 07:49 [Tux] joined 08:02 [Tux] left 08:08 [Tux] joined 08:11 domidumont left 08:16 [Tux] left 08:25 [Tux] joined 08:59 patrickb joined 09:29 frost-lab left, frost-lab joined
dogbert17 .seen lizmat 09:35
tellable6 dogbert17, I saw lizmat 2021-03-25T14:05:30Z in #moarvm: <lizmat> jnthn: so any idea how many people watched ?
dogbert17 m: wordcase("abc def ghi") for ^100_000 09:47
camelia WARNINGS for <tmp>:
Useless use of "wordcase(\"abc def ghi\")" in expression "wordcase(\"abc def ghi\")" in sink context (line 1)
dogbert17 m: wordcase("abc def ghi") for ^100_000; say now - INIT now; 09:48
camelia WARNINGS for <tmp>:
Useless use of "wordcase(\"abc def ghi\")" in expression "wordcase(\"abc def ghi\")" in sink context (line 1)
dogbert17 m: "abc def ghi".wordcase for ^100_000; say now - INIT now;
camelia 3.9773093
dogbert17 lizmat: any theory as to the timing differences?
dogbert17 is probably doing something stupid 09:49
10:04 sena_kun left 10:06 frost-lab left 10:08 sena_kun joined
lizmat looks 10:21
looks like the first case is constant folded, the second isn't ? 10:23
dogbert17 ^^ 10:24
dogbert17 m: my $a = wordcase("abc def ghi") for ^100_000; say now - INIT now; 10:26
camelia 0.03031934
dogbert17 m: my $a = "abc def ghi".wordcase for ^100_000; say now - INIT now;
camelia 3.9287443
dogbert17 lizmat: so this is to be expected then?
lizmat I'm not saying that: I'm merely stating the current fact :-) 10:27
and wordcase is relatively expensive, so the difference is more pronounced 10:29
dogbert17 I'm looking a bit into differences beteween sub and method form of some routines. In many cases the sub form runs in twice the time compared to the method form 10:30
dogbert17 is trying to lure lizmat into doing some cool opts 10:31
lizmat that might work: wordcase has been on my long list for a while
10:32 domidumont joined
dogbert17 aha, I'd better keep investigating. Perhaps something else will turn up :) 10:32
a quick look at the profiler shows zero GC's vs 288 GC's 10:34
among other things
lizmat yeah, the sub version is equivalent to: my $a = 42 for ^100_000 10:35
dogbert17 now that'll be fast :) 10:36
MasterDuke interesting. because the sub just calls the method on the argument
dogbert17 that is indeed interesting 10:37
lizmat "wordcase" is marked "is pure" 10:42
so open to optimization of a given input 10:43
so I think the optimizer then just runs the wordcase("abc") at optimize time, and WVals the result 10:44
MasterDuke huh. rakudo only has a single method marked 'is pure' (`src/core.c/Str.pm6:841:    method !hexify(str $char) is pure {`) 10:46
but 156 subs 10:47
lizmat I'm not sure it even makes sense on methods ?
sena_kun Was there an RSC meeting last week? 11:59
lizmat yes, there was 12:00
sena_kun I see no "minutes" note posted, but glad it was.
13:21 b2gills joined
nine From IO::Handle docs: "Using getc to get a single keypress from a terminal will only work properly if you've set the terminal to "unbuffered"." 13:36
Good hint! No how would one do that exactly?
leont Didn't someone (gfldex?) file an issue recently along the lines of "when should support argumented blocks? Or was that just a suggestion on his blogpost. 13:55
15:30 rba joined 15:31 camelia left 15:50 camelia joined 16:50 ggoebel left
japhb nine: I think I answered this recently, let me go see if it was in a logged channel. 17:20
Yeah, here: colabti.org/irclogger/irclogger_lo...03-17#l557 17:22
lizmat just pushed github.com/rakudo/rakudo/commit/5d...e023d82829 # dogbert17 17:46
"Make Str.wordcase about 5x as fast" 17:50
17:52 domidumont left
japhb Commit descriptions like that just warm my heart. :-) 17:56
20:52 vrurg left 20:54 sena_kun left 21:02 vrurg joined 21:58 patrickb left
gfldex leont: I mentioned it in my blog and jnthn picked up on it in 6guts.wordpress.com/2021/03/15/tow...in-moarvm/ . 22:46
leont: Close to the bottom of his post.
leont That explains why I couldn't find it, that's not where I was looking 22:48
Though it's not entirely what I meant. Even without new dispatching I'd really like to be able to rename $_ to something clearer, based on the knowledge of the match 23:31