github.com/moarvm/moarvm | IRC logs at colabti.org/irclogger/irclogger_logs/moarvm
Set by AlexDaniel on 12 June 2018.
brrt \o 06:30
nwc10 o/ 06:31
brrt wants to refactor the expresion compiler but is unsure just how far to go 06:34
nine .tell brrt what's the goal of the refactor? 07:07
yoleaux nine: I'll pass your message to brrt.
jnthn o/ 08:45
timotimo o/ 08:46
nwc10 that's getting unbalanced... 08:56
AlexDaniel samcv: ETA? I kinda want to get it off my shoulders 09:58
dogbert2 \o (trying to balance things up) 10:19
brrt nine: the main thing is that I want to introduce a new sigil for write-operands 15:07
yoleaux 07:07Z <nine> brrt: what's the goal of the refactor?
timotimo is this to do with exclamation mark templates or something else?
brrt yeah, something like that
timotimo since what expr nodes write isn't really named, right?
brrt I want it to be obvious that the $0 in a write-operand template is not a value but a reference 15:08
correct
timotimo oh, like that you mean
brrt that is the overarching goal
timotimo so we'd have like >0
brrt yeah
I was actually thinking of \$0
timotimo we don't have ops that have more than one write register operand, right?
brrt but whatevs :-)
not yet (maybe never) 15:09
brrt so. I can hack that in, and it'll work.... 15:10
but I'm annoyed by the current level of complexity
timotimo you can just make a \ in front a no-op, and also warn if it's not $0 that follows it :P 15:11
brrt that won't work, the \$0 is parsed as a single token
and more importantly, I want to capture instance of that missing
timotimo ah 15:12
that means the guard ops would become exclamation mark templates, too
since they don't necessarily write
brrt no, no
a template that doesn't yield any value doens't need an exclamation mark 15:13
it is only templates that *overwrite* the register that is used as their output that need exclamation marks
timotimo well, guard ops now sometimes write to a different register than they read from
brrt ah, really?
guard ops write? 15:14
hmmm
timotimo yep, if the register differs
part of the set elimination optimization i think?
brrt example?
oh
timotimo [Annotation: INS Deopt One (idx 9 -> pc 464; line 2706)]
sp_guard r18(1), r6(9), sslot(1), litui32(464)
jnthn Guard ops at the point they are inserted will always be two verions of hte same register, but yes, a future set elimination opt might rewrite the instruction 15:15
And the reason guards do this now is so we have before/after, which lets us do guard elimination
brrt i see 15:16
that makes a bunch of sense
jnthn It's also important that the write does not take place if the guard isn't met :) 15:18
(Yes, I did that mistake. Yes, it was fun to debug. :P)
MasterDuke btw, while people are active. did it make sense in github.com/MoarVM/MoarVM/pull/949 to merge my branch into post-release-opts? or should i have just done master?
timotimo post-release-opts is safe in any case
jnthn It's safe, though its merge schedule is unknown 15:21
MasterDuke well, no hurry for my PR 15:23
timotimo has been writing a blog post about optimizing the performance of a given piece of code, based on the FASTA parsing task off stackoverflow 16:56
but now i'm AFK for a while
i'd say it's about 1/3rd done, at 822 words so far
samcv AlexDaniel: hopefully soon. i finally have actual time. was a family thing for a few days and had no time, but now i do 17:01
AlexDaniel cool
Geth MoarVM: bdf961d5fc | (Samantha McVey)++ | tools/update-changelog.p6
Add many more features to update-changelog.p6

Allow viewing all changes done so far, allow changing view options. Add a few tests and fix a bug in auto categorizing some commits based on directory name.
Use rlwrap if it exists to allow easily editing commit subject's with preentered text of what it was previously.		 18:48
MoarVM: 0b34ce5e23 | (Samantha McVey)++ | tools/update-changelog.p6
Add goto specific commit and print uncomplete entries to update-changelog
samcv i have 14 options but i still don't have enough 18:49
MODIFY (e)dit/(d)rop/(c)ategory/d(o)ne/(T)itlecase; PRINT (b)ody/d(i)ff/num-(l)eft/(C)omplete/(U)uncomplete; GOTO (n)ext/(G)oto commit ID; (Q)uit and save or (s)ave:
travis-ci MoarVM build failed. Samantha McVey 'Add goto specific commit and print uncomplete entries to update-changelog' 19:08
travis-ci.org/MoarVM/MoarVM/builds/421708094 github.com/MoarVM/MoarVM/compare/5...34ce5e23e7
samcv ok at 176 more changelog entries. time for a short break 20:57
this tool is making it much faster to go through the changelog :) and i can merge one commit's subject into another one. though that needs a bit more refinement. and i probably will also add ability to tag items that i need to ask people about 20:58
japhb samcv++ # Making tools to make the job easier 23:04

Powered by Cro and the Raku® Programming Language.

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

Thank you!