github.com/moarvm/moarvm | IRC logs at colabti.org/irclogger/irclogger_logs/moarvm
Set by AlexDaniel on 12 June 2018.
Geth MoarVM/master: 12 commits pushed by (Timo Paulssen)++, (Bart Wiegmans)++
review: github.com/MoarVM/MoarVM/compare/9...ef0806464d		 11:14
nine Looks like I've come across the first case where I have to change the actual compilation result. The result register kind of an nqp::if depends on the kind of the then and else branches. But I only know those after actually compiling them. So I may have to introduce intermediary registers. 12:24
timotimo nine: alternatively, use istrue_o and then if_i 12:48
spesh does that anyway, but of course if_o results in smaller .moarvm files 12:49
nine How would I use them there? 12:54
timotimo actually, i'm not entirely sure what problem you're actually edscribing 13:14
i think i might have gotten it wrong
sorry about that
nine The result type of nqp::if($cond, $then) is either the type of $cond if $cond and $then have the same type or just obj. The result type of nqp::if($cond, $then, $else) is either the type of $then if $then and $else have the same type or again just obj. In cases where $cond and $then or $then and $else have different types we also need to coerce the result to obj. 13:18
My problem is that I only know the result type of $else after compiling it. So I don't know whether I need a coercer or not.
timotimo oh! 13:19
because the blocks return their result
i was thinking of the underlying bytecode %)
jnthn Also I think there's a void handling where the return values don't matter 13:28
nine There is, yes. Which doesn't exactly make the code easier to digest ;) 13:29
nine So I need an intermediary register for $then and coercion code which I emit after the $else and a goto in $else to jump over the $then coercer 13:30
jnthn Yeah, but the void context things have quite some nice performance impacts :) 13:32
nine This whole if/then/else thing sucks enough to make me consider compiling into a separate bytecode buffer again 16:42
nine Buuut....since I'm thinking of fixing up things afterwards already, would it be possible to fix up the register kind? That's really all that's dependent on the op's children. 16:58
Of course that's not enough, since I also would need the coercing code 17:02
Geth MoarVM: jstuder-gh++ created pull request #976:
Exprjit Templates (09/20/18)		 17:35
MasterDuke .ask brrt for function calls in templates, would there ever be anything in the arglist other than cargs? if not, would it makes to try and remove some of that boilerplate? 17:49
yoleaux MasterDuke: I'll pass your message to brrt.
Geth MoarVM: e21836856f | (Samantha McVey)++ | 2 files
Fix some printf format type specifiers

Fix two in main.c that were causing warnings on clang. Fix some others that were not giving warnings, but should have been specified as unsigned, though they should never become big enough for that to actually matter.		 20:12
Geth MoarVM: jstuder-gh++ created pull request #977:
Implement shift ops for exprjit		 20:26
samcv what is this? JIT ERROR: Negative offset for dynamic label 8 21:59
it occured when i tried to call a method that doesn't exist for that type 22:00
No such method 'base' for invocant of type 'Str'. Did you mean any of these?
timotimo well, that's pretty bad 22:04
jnthn Looks like a "the impossible happened" kind of error... 22:05
Can check if it's expr-JIT sensitive with MVM_JIT_EXPR_DISABLE=1 22:06
samcv sec 22:16
jnthn: yeah that gets rid of the error
code in question:
m: for <PRIMARY SECONDARY TERTIARY> { say 0x385.uniprop("MVM_COLLATION$_".base(16)) }
camelia No such method 'base' for invocant of type 'Str'. Did you mean any of these?
Hash
asec
hash
take

in block <unit> at <tmp> line 1
samcv working on updating us to unicode 11 22:17
jnthn samcv++ 22:20
timotimo cool 22:22
samcv i'm glad there was nothing wrong with the collation, it just had a minor issue regenerating the file. because diving into that would be a huge pain
jnthn Looks like some grapheme clustering changes, alas. 22:24
samcv i'm still amazed at how well it works. and there's only one function which looks a bit ugly
jnthn: yep
i have that fixed afaik already
jnthn Though from the glimpse I took at it, it looked like it might just be a simpliciation.
timotimo imgur.com/sW017ei - super crappy early look at my current task
jnthn But maybe I'm just being an optimist :)
samcv yeah i had to change some things
jnthn NaN% :D 22:25
samcv it *did* make it simpler. my hours of time wasted! now i have to get rid of my code to workaround "should not break within emoji (you may have to modify these rules)"
timotimo yeah, the columns will be replaced by allocation-related columns
samcv something to that effect. i remember i thought the language used was humorous
Further customization of this rule may be necessary for best behavior of emoji zwj sequences [UTS51] 22:26
that's it
jnthn hah, I loved that...
Might as well have said "Good luck with this!"
timotimo can zwj become the new ffs ...
samcv ffs?
timotimo for f's sake
samcv ah that's what i guessed but
wasn't 100% sure 22:27
timotimo just make zwj an expletive
doesn't even matter what you invent for the expanded version
jnthn: you don't happen to have another blog post just about ready to release? :D 22:33
jnthn No, 'fraid not :) 22:37
Didn't decide what the next one will be about yet :)
timotimo that's fair 22:38
jnthn sleep & 22:57
evalable6 Use of uninitialized value of type Callable in numeric context
in block <unit> at /tmp/9…
jnthn, Full output: gist.github.com/7dc88ddf9678d1d4fb...bd97ca4afb
geekosaur .oO { bottle it, stupid bot } 22:58
timotimo %) 22:58

Powered by Cro and the Raku® Programming Language.

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

Thank you!