github.com/moarvm/moarvm | IRC logs at colabti.org/irclogger/irclogger_logs/moarvm
Set by AlexDaniel on 12 June 2018.
nwc10 good *, #moarvm 10:05
jnthn o/ 10:07
nwc10 \o 10:08
MasterDuke jnthn: if you've got a couple spare minutes sometime (hopefully today) i'd appreciate a look at github.com/MoarVM/MoarVM/pull/1308 10:09
nwc10 could someone who knows how to test the debugserver have a look at the first commit of my pull request (#1307) and if that one is good, merge that one? The other two are (I feel) "contraversial" but the first removes a bunch of inefficiencies
MasterDuke the nqp and rakudo builds are currently a bit panicky and that fixes it, but confirmation that it's correct and complete would be good 10:10
nwc10 jnthn: you're popular :-) 10:11
MasterDuke that's the problem with being good at your job
nwc10 is deep somewhere in $ork invovling images
brrt \o 11:14
tellable6 2020-05-20T08:01:13Z #moarvm <nwc10> brrt \o
brrt (yes, I still exist)
lizmat brrt o/
brrt ohai lizmat 11:15
nwc10 good *, brrt
brrt ohai nwc10, too :-)
MasterDuke brrt: ahoy. btw, there have been a couple questions recently about what it would take to support the jit on arm. is it tricky work? or relatively straightforward grunt work? any difference if it's just AArch64? 11:24
brrt MasterDuke: I'd think there would be a good part messy work involved 11:25
I think it ought to be possible but not necessarily easy
especially the 'legacy' JIT is very heavy on x86-isms
MasterDuke ok, good to know
lizmat brrt so would it make sense to start a different JIT effort for ARM > 11:36
?
brrt not sure I'd say that either 11:39
lizmat well, with the next gen of Apple hardware being ARM again... I'd hate to have to run Raku on that without JIT :-) 11:39
brrt lizmat: on a pure technical ground, I think somebody (me?) should probably fix the legacy JIT not to suck, or to be removed altogether 11:40
On the other hand, that someone should probably have possession of a good ARM hackmachine, and since I'm not buying apple... :-P
timotimo lazy initialization of lookup hashes in static frames sounds like something we'd want to either lock or make access atomic for
lizmat hehe... well, I don't have an ARM either 11:41
and for me personally, I doubt that my next notebook will be an Apple, as Catalina managed to completely mess up just about all of my private media workflows
brrt I'm far too satisfied with my linux thinkpad 11:43
Ironically I got myself 32gb to be able to use gradle+intellij... now I still have to use gradle but rarely use intellij, so I hardly ever use that much memory 11:45
timotimo any additional ram makes the disk faster, though i assume you have an SSD which is already pretty fast even with cold caches 11:50
brrt It came with an NVMe which has windows on it, and I placed an SSD into the expansion slot 11:52
on which I installed fedora
xiaomiao pro tip: swap on nvme is fast enough 12:01
brrt yeah, but it still has windows on there
which is handy in a pinch 12:02
xiaomiao put the windows in a VM ;)
brrt could, but won't :-) 12:03
xiaomiao me neither, waste of diskspace 12:05
the "windows 10" tech demo my thinkpad came with still needs a few years of development before it's useful
jnthn brrt: I figure, so far as the JIT goes, transitioning completely to using the expr JIT would make things easier so far as porting goes? 12:06
brrt jnthn: yes; which, unfortunately, means that someone needs to hacking work to make the expression JIT more complete still 12:07
of which I have a bunch of half-finished branches
jnthn brrt: I guess the good news is that dispatch is going to get a bit more regular... :) 12:10
brrt I think I saw bits and pieces there 12:12
:-)
nine MasterDuke: I think with timotimo++'s insight into spesh facts, we can merge your GC fixing PR in good conscience. 12:38
MasterDuke cool. then i'll do that now and bump nqp+rakudo 12:39
Geth MoarVM: 1a46b0ae7c | (Daniel Green)++ | src/spesh/optimize.c
Maybe fix MVM_panic from entering GC during spesh

The `MVM_gc_mark_thread_(un)blocked(tc)` calls were incorrect and caused
  `MoarVM panic: Must not GC when in the specializer/JIT` at random times.
However, just removing them changed to causing `MoarVM panic: Invalid owner in item added to GC worklist`. Setting `MVM_GC_DEBUG` to `3` would reliably trigger the panic when building NQP. Removing overwriting the spesh facts fixed this, but I'm not 100% that the rest of the code is 100% correct.		 12:41
MoarVM: af1475ec15 | (Daniel Green)++ | src/spesh/optimize.c
Simplify optimize_getcurhllsym

Because `ins->info` is what's used to determine how many operands to access, we don't need to reallocate new operands in the gethllsym case.
MoarVM: a5df9ead10 | MasterDuke17++ (committed using GitHub Web editor) | src/spesh/optimize.c
Merge pull request #1308 from MasterDuke17/maybe_fix_gethllsym_spesh_optimization
nwc10 brrt: I have not (yet) got one of these: geizhals.eu/raspberry-pi-4-model-b...01761.html 12:43
travis-ci MoarVM build passed. MasterDuke17 'Merge pull request #1308 from MasterDuke17/maybe_fix_gethllsym_spesh_optimization' 12:55
travis-ci.org/MoarVM/MoarVM/builds/696441324 github.com/MoarVM/MoarVM/compare/6...df9ead10f7
brrt nwc10: cortex a72... is that an aarch64 chip? 13:17
google says it is
interesting
nwc10 yes. and there's a beta of "Rasbperry Pi OS" for 64 bit 13:27
IIRC 13:29
brrt dubitable new achievement in my motoring carreer; I had to jump-start my car 13:51
note to anyone trying this, make sure that the cables you use are rated for a decent amount of amperes
without this it doesn't work as well 13:52
MasterDuke yeah. i helped a friend jump his car about a year ago. absolutely nothing happened at all with his cables. tried mine, which were a bit more substantial, and it started right up on the first attempt 13:53
brrt hehe... I used the cables from my solar installations... it ultimately did work 13:55
MasterDuke ha
brrt with the help of a helpful neighbor

Powered by Cro and the Raku® Programming Language.

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

Thank you!