github.com/moarvm/moarvm | IRC logs at colabti.org/irclogger/irclogger_logs/moarvm
Set by AlexDaniel on 12 June 2018.
01:13 evalable6 joined 01:23 ggoebel joined 03:08 dogbert17 left 03:19 zakharyas joined 03:36 zakharyas left
MasterDuke can anyone else confirm that reverting github.com/MoarVM/MoarVM/pull/1111 doesn't change profiles, i.e., they still have `<anon> gen/moar/stage2/NQPHLL.nqp:1824` at the top? 06:48
08:16 sena_kun joined
MasterDuke timotimo: the time column in moarperf shows '<exclusive> / <inclusive>', but the buttons up top are inclusive then exclusive 09:39
i think something should be switched
09:58 squashable6 joined 10:18 MasterDuke left 11:48 dogbert17 joined
dogbert17 .seen timotimo 11:53
yoleaux I saw timotimo 22 Jun 2019 23:08Z in #moarvm: <timotimo> what with four sata cables coming out of it
12:09 kanopis joined 12:14 MasterDuke joined, MasterDuke left, MasterDuke joined 12:25 kanopis left
dogbert17 MasterDuke: are you looking onto the gen/moar/stage2/NQPHLL.nqp:1825 problem? 12:31
MasterDuke dogbert17: yeah. at first i assumed it was becuase of something i did(n't) do in github.com/MoarVM/MoarVM/pull/1111 12:38
but i reverted that and nothing changed, so now at least i don't think i'm to blame (probably) 12:39
dogbert17 I think the problems start with github.com/MoarVM/MoarVM/commit/1b...ba8809fe1f
MasterDuke huh. it goes away if you revert that? 12:40
dogbert17 the previous commit, i.e. b8db58753317c19a2220a3148cd168bc3a691585 looks good on my system, the one above is the first one where things starts to look wonky 12:41
with commit 1b76b3456881e6c29fab713c3c7f59ba8809fe1f the most expensive routine is said to have taken 36893488147417.62ms (profiling a program which takes ~1 sec in total) 12:44
MasterDuke ugh, yeah, also trying to figure out why that still happens 12:45
can you repro that reliably?
dogbert17 everything is good, as far as I can tell, on the previous commit b8db5875
MasterDuke but with 1b76b3456881e6c29fab713c3c7f59ba8809fe1f the profiles always have the absurdly expensive routine? 12:46
dogbert17 yes 12:47
MasterDuke what perl6 code are you using to test?
dogbert17 a super program :-) gimme a sec to gist it
gist.github.com/dogbert17/02202b34...c1a2ac387c 12:48
projecteuler.net/problem=3 12:51
MasterDuke arg, doesn't do the absurdly expensive routine for me 12:53
dogbert17 that's odd 12:54
did you sort on exclusive time? 12:55
MasterDuke yep 12:56
dogbert17 that's interesting, we must be doing something different 12:58
MasterDuke well, the times it has happened for me, running the exact same command again results in a normal profile 13:00
so it's something somewhat non-deterministic
dogbert17 interesting 13:04
MasterDuke: awake? 14:11
MasterDuke yep 14:12
dogbert17 I have changed my mind slightly :-) 14:13
look at github.com/MoarVM/MoarVM/commit/69...1cb10e7fba
MasterDuke hm. something changes around that commit?
dogbert17 if I remove the two line change in src/profiler/log.c things work for me on head
at least with my example prg, perhaps I should test a few more 14:15
MasterDuke huh. i don't really know why that's doing it. maybe timotimo has an idea? 14:17
dogbert17 he seems to have disappeared, perhaps his computer is busted
MasterDuke does anything change if you use a confprog? 14:18
dogbert17 have never tried that. I don't know how to use them :( 14:20
14:37 sena_kun left, sena_kun joined
MasterDuke hm, tools/compile_confprog.p6 gives `===SORRY!=== Error while compiling /home/dan/Source/perl6/MoarVM/tools/compile_confprog.p6 Variable '$custom_reg_base' is not declared at /home/dan/Source/perl6/MoarVM/tools/compile_confprog.p6:692` 14:45
dogbert17: if i add an fprintf in that `if` it prints 4 times when just doing a normal --profile 14:48
which seems a little odd when the variable name it's checking is `was_entered_via_confprog` 14:49
dogbert17 true 14:51
MasterDuke moving github.com/MoarVM/MoarVM/blob/mast...log.c#L106 to be inside the block it's directly after made those prints go away, but the profile looks the same 14:52
wait. the `<anon> gen/moar/stage2/NQPHLL.nqp:1824` is still there, but now the percentage of the rest of the inclusive times are different 14:57
dogbert17 does it look better?
MasterDuke at HEAD (plus my fprintf), `<anon> gen/moar/stage2/NQPHLL.nqp:1824` is 100% inclusive, and `<unit-outer>` is at 50% 14:58
14:58 patrickb joined
MasterDuke if i move github.com/MoarVM/MoarVM/blob/mast...log.c#L106 inside the block it's directly after, then `<anon> gen/moar/stage2/NQPHLL.nqp:1824` is still 100% inclusive, but `<unit-outer>` is at 100% also (and the rest go slowly down from there) 14:59
dogbert17 and did you remove lines 116-117? github.com/MoarVM/MoarVM/blob/mast...log.c#L116 15:00
MasterDuke nope, just added a print inside the if
dogbert17 ok, I removed them and things started to look a lot better on my end 15:01
patrickb .tell jnthn I think the install location of perl6.moarvm did not change. It was and is installed to ${RAKUDO_PREFIX}/share/perl6/runtime/perl6.moarvm. Do you still hit a problem with this? 15:04
yoleaux patrickb: I'll pass your message to jnthn.
MasterDuke i don't understand the confprog stuff well enough to know if that's the correct fix, but dogbert17++ for making progress
15:16 patrickb left 15:43 robertle joined 16:26 ggoebel left
Kaiepi can someone review github.com/MoarVM/MoarVM/pull/1126 ? i don't expect it to be merged before the next release, just i have a feeling i may be using the wrong type for families and want to have it ready as soon as it gets released 16:55
if you guys have time, i know people are busy with blockers atm
18:12 MasterDuke left 19:19 MasterDuke joined
MasterDuke huh, webchat.freenode.net looks completely different now 19:21
20:56 ggoebel joined
MasterDuke .tell timotimo bunch of relevant-to-you chat today. starting at colabti.org/irclogger/irclogger_lo...9-06-23#l6 21:10
yoleaux MasterDuke: I'll pass your message to timotimo.
22:56 sena_kun left

Powered by Cro and the Raku® Programming Language.

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

Thank you!