github.com/moarvm/moarvm | IRC logs at colabti.org/irclogger/irclogger_logs/moarvm
Set by AlexDaniel on 12 June 2018.
MasterDuke might find this interesting www.mono-project.com/news/2018/09/...harp-jit/, "writing a jit compiler in c#" 00:35
Geth MoarVM: jstuder-gh++ created pull request #964:
Fix wrong field in exprjit cast template and add macro		 03:35
Geth MoarVM: 3388d1229b | (Jeremy Studer)++ | 2 files
Fix wrong field in exprjit cast and add macro

In the 'objprimbits' template, accidently used the wrong field name for cast's from_size value ('boxed_primitive' instead of 'bits').
In this case, 'boxed_primitive' and 'bits' are both the same size, but that doesn't make it less wrong just less painfully wrong 😉.
Added a macro that makes this idiom more DRY and less prone to this type of error.		 07:36
MoarVM: da2072b57e | (Bart Wiegmans)++ | 2 files
Merge remote-tracking branch 'jstuder/exprjit_sm_template_err'
nine Compiling labels + fixup was surprisingly easy after a good night's sleep. Now on to calls which are...much more involved. 09:08
yoleaux 13 Sep 2018 21:42Z <japhb> nine: Inline::Python won't build for me at Rakudo HEAD
AlexDaniel M#2280 interesting ticke 14:56
t
R#2280
synopsebot R#2280 [open]: github.com/rakudo/rakudo/issues/2280 [math][regression] Core dump in sequence when iterator is dividing by a negative number
timotimo have we tried jit and such yet? 15:03
i.e. turning off the jit
AlexDaniel c: MVM_JIT_DISABLE=1 HEAD say (11**5, */-2 … 0)[31]
committable6 AlexDaniel, ¦HEAD(ed9b48d): « «exit signal = SIGFPE (8)»»
AlexDaniel c: MVM_SPESH_DISABLE=1 HEAD say (11**5, */-2 … 0)[31] 15:04
committable6 AlexDaniel, ¦HEAD(ed9b48d): « «exit signal = SIGFPE (8)»»
timotimo OK 15:05
so at least the implementations of interpreter and jit are the same
c: HEAD signal(SIGFPE).tap({ say "oh my" }); say (11**5, */-2 … 0)[31]
committable6 timotimo, ¦HEAD(ed9b48d): «oh my␤«timed out after 10 seconds» «exit signal = SIGHUP (1)»»
Geth MoarVM/fork-safety: 4 commits pushed by (Bart Wiegmans)++ 18:01
nine Unhandled exception: Bytecode validation error at offset 96, instruction 15: callsite index 0 out of range 0..4294967295 18:05
I accept the bytecode validation error but I dare to disagree with the reason ;) 18:06
diakopter nine: :> 18:44
nine Ok, now that callsites are written, too, it's just annotations that are missing before the first real frame is compiled fully 18:45
diakopter does MoarVM have a stack depth limit these days 20:01
timotimo it does not
not that i know, at least
i think if you just sub a { a }; a; it'll malloc until the OS dies (more likely that it'll hang and stall than that it'll reliably kill the process, in my experience at least) 20:02
diakopter yeah. at one point it did, I think 20:05
timotimo i know parrot had something like that 20:05
which i believe was not more than just having a counter for how many stack frames have been used yet, and you could configure it with an op 20:06
Geth MoarVM: 578c2f7623 | (Samantha McVey)++ | 2 files
Add support for utf-16 decodestream

Adds support for utf-16. At the moment only handles the same endianess as the processor. We may want to have utf16le and utf16be as options.
When we decode a utf16 string (standard, not decodestream) we check for a BOM, byte order marker. But if we are using decodestream we may be starting decode partway through the file, so we will probably have to save this data.
For now this gets utf16 working for those who need it.		 20:29
diakopter timotimo: can NativeCall stuff be used from NQP, or does it need Perl 6 20:31
timotimo it can, but it's a bit of manual work that's involved :) 20:31
actually 20:32
there's t/nativecall/01-basic.t inside the nqp repository
diakopter oh yeah.
lizmat samcv: that is making Tux quite happy 20:53
samcv yay
lizmat he would be even happier with utf16le and utf16be
:-) 20:54
samcv heh
ok :)
lizmat i think it's one of the blockers for him to really take Perl 6 into production
samcv where does he get these utf16 files :P
ah
lizmat customers
on Windows
samcv well i'll do that next then
lizmat ++samcv
samcv after i get some food :) 20:55
timotimo oh, huh. using exit seems to cause the profiler to get very wrong numbers (though that seems to just be a negative number being put into an unsigned integer 22:57

Powered by Cro and the Raku® Programming Language.

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

Thank you!