timotimo good post indeed 00:27
jnthn++
02:56 ilbot3 joined 06:59 domidumont joined 07:04 domidumont joined 07:12 geospeck joined 08:06 Ven joined 08:10 brrt joined 09:07 brrt joined 09:08 zakharyas joined, brrt joined 09:38 scovit joined 10:09 zakharyas joined 10:20 robertle joined 11:00 Ven joined 12:50 Ven joined
dogbert2 jnthn: should I report the CLOSE phaser test problem (valgrind screams) against MoarVM or Rakudo ? 13:00
dogbert2 leans towards MoarVM
jnthn Probably MoarVM 13:21
Though nothing new was added to MoarVM at all in the process of fixing the issue
So it's not a new problem, just uncovering an existing one
dogbert2 agreed, timotimo suspected a missing MVMROOT 13:29
perhaps some part of the MoarVM codebase is exercised a bit more after your recent Rakudo fixes 13:31
nine: did you ever create a PR for the list manipulation bug you found a few days ago? 13:32
13:42 AlexDaniel joined
dogbert2 jnthn: github.com/MoarVM/MoarVM/issues/749 14:06
14:24 zakharyas joined
MasterDuke jnthn: do you have any further thoughts/comments on github.com/MoarVM/MoarVM/pull/689 ? 14:29
14:29 zakharyas joined 14:45 brrt joined
jnthn MasterDuke: Will check it when I get a spare moment :) 14:54
MasterDuke thanks, no particular hurry, just following up on stuff i have open 14:55
14:58 committable6 joined 15:18 domidumont joined
MasterDuke any particular reason MVM_STRING_MAX_STRANDS is only 64? 15:35
MVMStringBody.num_strands is an MVMuint16 15:36
jnthn Yes, it's so we collapse them after a while 15:43
Otherwise, a repeated concat will end up copying an every-lengthening array of strands
*ever-lengthening
MasterDuke heh, yep, just nearly locked up my computer after bumping that up to 32k 15:45
16:10 brrt joined
MasterDuke `if (a->body.storage_type == MVM_STRING_STRAND) fprintf(stderr, "last strand of a = |%s|\n", MVM_string_utf8_encode_C_string(tc, (a->body.storage.strands[a->body.num_strands - 1]).blob_string));` 17:25
^^^ should print just the last strand of `a`, right?
timotimo yeah; don't forget to MVM_free, though 17:26
MasterDuke even when debugging?
17:39 domidumont joined
timotimo what's a memory leak or two between friends … 17:51
MasterDuke huh. string a = "xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx" 18:12
but the last strand of a = 'for (1, 10, 100) -> $limit {my $x; my $y = "x" x 100; $x ~= $y for (1..$limit); say "$limit: ", now - ENTER now}'
guess i'm not taking the strand's start and end into account 18:14
m: my $x; my $y = "x" x 10; $x ~= $y for (1..100_000); say now - INIT now 18:30
camelia 1.55922485
MasterDuke m: my $x; my $y = "xxxxxxxxxx"; $x ~= $y for (1..100_000); say now - INIT now 18:31
camelia 0.05816356
MasterDuke heh. concat checks if the last strand of the lhs == the rhs, in which case it just bumps the repetitions. but in the first case the last strand is just "x", not "xxxxxxxxxx" 18:35
timotimo because different repetition count?
ugexe The good news is the getppid() fix for windows landed in libuv for 1.16.1; the bad news is there is a regression on windows for shell I just noticed that is caused by the last bump to libuv v1.15.0
regression caused by github.com/libuv/libuv/commit/b21c...b8850ffd18
perl6 -e "say shell(q|echo 42|).exitcode" - causes this to only print 0 (the 42 never gets printed) 18:36
MasterDuke timotimo: well, because it's checking a strand vs the whole string
ugexe but seems like since its just a flag that it can be tweaked in moarvm possibly
MasterDuke so i don't think we ever concat strings by just += repetition, e.g., "a" x 3 ~ "a" x 4 -> "a" x 7 18:44
but i wonder if that would be too much special casing to implement? `if (rhs->num_strands == 1 && last strand of lhs == first strand of rhs) result->body.storage.strands[lhs->body.num_strands - 1]).repetitions += rhs->body.storage.strands[0]).repetitions` 18:51
18:57 geospeck joined
MasterDuke can there be a string of one strand where the strands start and end don't encompass the whole blob_string? 19:11
timotimo that sounds just like a substr of another string? 19:12
MasterDuke yeah, guess so
timotimo oh, i get why you're getting the source code as the strand there
it's convinced the best way to represent "x" is to have it be a substring of the whole source code 19:13
could that be it?
MasterDuke maybe. that would explain why when i took start and end into account it was fine 19:14
timotimo right 19:15
one single character substr'd out of a huge string is kind of an edge case where i say we should build a buffer instead 19:16
MasterDuke huh. looks like there's no convenience function to extract a strand as a standalone string
timotimo additionally, at some point we'll want to store very short strings inside the pointer or something :)
aye, strands don't normally have interesting semantic value
MasterDuke right, but it is annoying if i want to see if the "base" of one strand is equal to the "base" of another 19:18
timotimo true that 19:19
MasterDuke i've got to manually substring both strands
timotimo do we have an eqat that has a start point for both strings? 19:20
MasterDuke don't think so
there's MVM_string_equal_at, but it just takes two strings and a length
s/length/offset/ 19:21
afk for a bit 19:22
19:36 dogbert17 joined
dogbert17 ZOFFLOP: t/spec/S10-packages/basic.t (test #83) 19:37
heh ww 19:46
22:26 releasable6 joined
Geth MoarVM: MasterDuke17++ created pull request #750:
Optimize MVM_string_concat when the RHS is a...		 23:25

Powered by Cro and the Raku® Programming Language.

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

Thank you!