github.com/moarvm/moarvm | IRC logs at colabti.org/irclogger/irclogger_logs/moarvm
Set by AlexDaniel on 12 June 2018.
05:28 robertle left 06:49 domidumont joined 07:02 brrt joined 07:10 patrickb joined
brrt \o 07:17
timotimo o/ 07:18
nwc10 \o
brrt I'm debugging the expression JIT code now.... 07:33
timotimo what it spits out, or the implementation? 07:40
07:41 zakharyas joined 07:45 sena_kun joined 09:03 brrt` joined 09:04 brrt left
brrt` what it spits out 09:06
09:07 brrt` is now known as brrt
timotimo what does it spit? 09:08
brrt well, it should spit out correct code for lt_n, but it does not 09:10
timotimo at least now we're actually putting in lt_n, not lt_i 09:12
09:22 japhb left 09:40 brrt left
timotimo ISTR somewhere jnthn said that *ref -> * is handled "in code gen", that probably means after the perl6 static optimizer is already done with it 09:49
so ... perhaps the optimization is a waste anyway? 09:50
jnthn timotimo: Well, maybe not, in so far as I think we need to try and do some analysis of these things in the Perl 6 optimizer 09:58
Otherwise we can never do lexical => local lowering for natives in any non-trivial case 09:59
timotimo OK, but i don't think the optimization i just put in will do a lot for that
i'll push to master, since there's already a release branch
jnthn OK 10:00
Well, yeah, we need to do the general analysis there, also to look at the arguments of subs we call with natives to see if there's any `is rw` params (or multis with `is rw` candidates)
And if not, we know we can pass the value there
10:01 domidumont left
timotimo this one is only for assign_i and assign_n 10:01
though actually no reason not to expand this one it to assign_s 10:02
10:23 sena_kun left, sena_kun joined
timotimo doesn't appear even once in the core setting :) 10:24
assign_s or even assign_n
10:58 brrt joined
lizmat .seen pamplemousse 11:06
yoleaux I saw pamplemousse 3 Jul 2019 16:51Z in #moarvm: <pamplemousse> Thank you all, sorry I didn't catch that sooner
lizmat misses progress reports :-) 11:07
brrt yeah, pamplemousse said she was going to have a 4th-of-july celebration
lizmat aha... ! 11:08
:-)
11:38 domidumont joined 11:39 zakharyas left
brrt long story short, C compilers are cleverer than NQP compilers... 12:12
12:42 brrt left 13:01 brrt joined 13:06 brrt left 13:15 zakharyas joined 13:25 brrt joined 13:37 pamplemousse joined
brrt \o pamplemousse 14:30
pamplemousse o/ brrt
brrt pamplemousse: lizmat asked if you had a progress report ready perchance 14:33
lizmat pamplemousse o.
indeed :-)
pamplemousse I sadly don't have one ready to go quite yet, but I'm actually writing it up now (I had a success this morning and want to update it to include that). 14:36
I'll post a link as soon as I have it done, sorry that it's been a bit. 14:37
lizmat pamplemousse: no problem, better make a good post than a hasty one with errors... 14:38
14:44 AlexDaniel left, AlexDaniel joined
brrt :-) 14:48
lizmat and yet another Perl 6 Weekly hits the Net: p6weekly.wordpress.com/2019/07/08/...ythifying/ 15:30
15:59 zakharyas left 16:05 robertle joined 16:06 patrickb left 17:22 domidumont left 17:31 patrickb joined 17:58 brrt left
samcv timotimo, can you summarize the confproc? 19:07
*prog
pamplemousse Progress update is finished! yakshavingcream.blogspot.com/2019/...etely.html 19:18
19:38 pamplemousse left
timotimo oh 19:40
i mean i wrote a blog post about it :D
but essentially it's a very limited subset (not a true subset actually, since some of the ops behave differently) of moarvm bytecode, and there's a couple of points in moarvm itself that can be configured (at run time) to run specific parts of that code to make decisions 19:41
normally, the profiler will start recording the call graph immediately when mvmprofilestart (or what its name is) is called, but with a confprog you can ask it to only start recording a call graph when the confprog says "yes" 19:42
so you can have your regular program but only profile everything happening inside of "super-interesting-subroutine"
19:45 brrt joined
brrt pamplemousse++ 19:46
and, lizmat++ ofc :-)
jnthn lizmat++ # weekly 20:04
20:12 robertle left
Geth MoarVM/jit-expr-float: 66d72541dd | (Bart Wiegmans)++ | 2 files
[JIT] Fix floating point comparisons

  - cmp_num tile uses the correct input registers
  - LT, LE must check that the Parity Flag (set if unordered) isn't set		 20:55
MasterDuke any opinions on whether github.com/MoarVM/MoarVM/pull/1133 should be merged ahead of the release? 21:01
oh, nm. only really necessary after i merged the default-int nqp and rakudo prs, and those won't be in the imminent release anyway 21:02
Geth MoarVM: bdw++ created pull request #1134:
JIT - Floating Point Support for Expression JIT		 21:32
timotimo yeeeaaaahh 21:33
jnthn Wow, brrt++ 21:36
Will take a look at that :)
I figured we'd have had a relase by now, but since we haven't...I guess we can do a release branch so we can get back to breaking stuff. 21:37
brrt thank you :-) I'm going to sleep now, so I'll read your reviews later
jnthn Rest well o/ 21:38
jnthn also afk for a while
21:41 patrickb left 21:42 brrt left 21:43 patrickb joined
timotimo (just a rebase) 21:51
Geth MoarVM/dump_all_heapsnapshot_data: 16 commits pushed by (Timo Paulssen)++
review: github.com/MoarVM/MoarVM/compare/0...c94bffcd4b
22:44 sena_kun left 22:46 japhb joined 22:51 patrickb left 22:52 AlexDaniel left 22:53 AlexDaniel joined 23:17 jnthn left 23:18 jnthn joined
Kaiepi P6str's storage spec uses `sizeof(MVMString *) * 8` as its number of bits, but uses `ALIGNOF(void *)` for its alignment 23:25
shouldn't it use ALIGNOF(MVMString *)?
or is there a reason why void * is used instead
timotimo i can't think of anything other than "it's usually the same anyway" 23:41

Powered by Cro and the Raku® Programming Language.

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

Thank you!