Welcome to the main channel on the development of MoarVM, a virtual machine for NQP and Rakudo (moarvm.org). This channel is being logged for historical purposes.
Set by lizmat on 24 May 2021.
timo a small grep would already get us very close, i think 00:01
like grepping for "Inlining frame" with before-context of one line out of a grep of the whole file with lines "Before:" and the line before it as well as "Inlining frame"
japhb Meanwhile I got the multithreaded attacks test working; my hunch was right and precaching all of the possible pixel combinations worked. 00:19
Current: 00:20
Min: 45.7 ms (21.9 fps) - Ave: 104.0 ms (9.6 fps) - Max: 324.4 ms (3.1 fps)
50%: 97.8 ms - 75%: 131.3 ms - 90%: 151.3 ms - 95%: 161.8 ms - 99%: 286.7 ms
2021.09: 00:22
Min: 60.2 ms (16.6 fps) - Ave: 135.0 ms (7.4 fps) - Max: 363.7 ms (2.7 fps)
50%: 126.0 ms - 75%: 166.9 ms - 90%: 191.1 ms - 95%: 203.6 ms - 99%: 327.1 ms
So the improvement holds there as well, though perhaps not as starkly.
timo all improvement is appreciated :) 00:34
dogbert17 it's very silent here today 10:35
lizmat makes some noise as well :-) 10:36
dogbert17 anyway, there seems to be a flapper hiding in t/spec/S12-methods/lastcall.t, more specifically test #3
hello lizmat :)
ok 2 - sanity test for clearing
Can only use manipulate a capture known in this dispatch 10:37
lizmat scare it out of hiding, I'd say!
dogbert17 in method doit at t/spec/S12-methods/lastcall.t line 21
to repro, run 'MVM_SPESH_NODELAY=1 MVM_SPESH_BLOCKING=1 ./rakudo-m -Ilib t/spec/S12-methods/lastcall.t' a few times 10:38
sena_kun finished with a breakfast and looks at the release 10:40
dogbert17 ++sena_kun
dogbert17 some gdb output can be found here, gist.github.com/dogbert17/d3fee0f0...854faa2f71 11:05
MasterDuke what about an MVM_dump_backtrace(tc)? 11:06
dogbert17 MasterDuke: let's see ...
MasterDuke oh, nm, it's already the first thing there
dogbert17 yeah, added it anyway :) 11:07
MasterDuke what's `record->rec.resume_kind`? 11:08
dogbert17 MVMDispProgramRecordingResumeCaller 11:09
MasterDuke so i guess the find_capture(...) returns false 11:12
dogbert17 so it would seem 11:14
btw, I'm running with a 20k nursery 11:16
now all we need is someone recommending rr :) 11:24
lizmat maybe use rr ? 11:30
.oO( always happy to oblige :-)
sena_kun ouch, there is no merge commit for new-disp in moarvm? 11:45
hmm, I see there is not
sena_kun [-Wunused-variable] 12:05
6721 | MVMCallStackRecord *csrecord = MVM_callstack_allocate_nested_runloop(tc);
Geth MoarVM/2021.10: 48d35da340 | Altai-man++ | docs/ChangeLog
Update ChangeLog for 2021.10 release
12:14
MoarVM/2021.10: 23cc724d16 | Altai-man++ | VERSION
Bump VERSION for release
MoarVM: Altai-man++ created pull request #1576:
2021.10 release
MoarVM: 48d35da340 | Altai-man++ | docs/ChangeLog
Update ChangeLog for 2021.10 release
MoarVM: 23cc724d16 | Altai-man++ | VERSION
Bump VERSION for release
MoarVM: 07060faf14 | Altai-man++ (committed using GitHub Web editor) | 2 files
Merge pull request #1576 from MoarVM/2021.10

2021.10 release
jnthnwrthngtn sena_kun: There was a merge commit for new-disp, it's 962e835dc1 12:31
sena_kun a bit too late. :P 12:34
MoarVM release is done.
jnthnwrthngtn :) 12:38
For future reference, `git log --merges --oneline` is an easy way to see all the merge commits 12:39
timo: For just seeing what gets inlined, MVM_SPESH_INLINE_LOG=1 in the environment will spit inlines out on stderr, and is far cheaper (so doesn't disrupt the program so much) as a spesh log. Sadly, though, also doesn't give that much insight into why. 12:40
timo ooh, of course 12:41