brrt good * 06:09
dalek arVM: 8f82955 | (Jimmy Zhuo)++ | src/6model/containers.c:
fixed signal SEGV from RT #125371, now "my $a := $a; say $a" outputs (Mu)
 06:14
synbot6 Link: rt.perl.org/rt3/Public/Bug/Display...?id=125371
JimmyZ m: my multi bar ($) { }; my sub foo () { my $q := bar($q); }; foo(); 06:17
camelia rakudo-moar 375435: OUTPUT«===SORRY!=== Error while compiling /tmp/ZpY4XuWPwY␤Cannot use variable $q in declaration to initialize itself␤at /tmp/ZpY4XuWPwY:1␤------> ($) { }; my sub foo () { my $q := bar($⏏q); }; foo();␤ expecting any of:␤ argumen…»
brrt JimmyZ: shouldn't that null container be a vmnull? 07:22
like, in the first place :-) 07:25
JimmyZ brrt, I think it's not 09:47
brrt no, it probably isnt
JimmyZ it don't SEGGV if it is vmnull
brrt but i think it ought to be?
JimmyZ I think it is not since regs in moarvm is not vmnull too. 09:48
by default, it is NULL 09:49
timotimo i think we use null vs vmnull to differentiate stuff that needs to be autoviv'd 09:53
at least in some places
dalek arVM/even-moar-jit: 66f2bf8 | brrt++ | src/jit/ (6 files):
Define argument registers
 10:14
brrt i need to move use information calculation to the tiling step.... 11:24
because we often address nodes deeply 11:25
brrt wonders how to manage the complexity of it all
brrt how... to fix it 11:50
dalek arVM/even-moar-jit: efe423e | brrt++ | / (7 files):
WIP on compiling tiles
 11:53
dalek arVM/even-moar-jit: aedf261 | brrt++ | / (4 files):
Calculate usage information after tiling

During tiling we change the tree, and only after tiling do we know which values are really used during which instructions. Hence usage information should really be calculated post-tiling (except if ever used during optimization, which is not currently the case).
Also fixed a nasty bug I overlooked in tiling DO		 19:09

Powered by Cro and the Raku® Programming Language.

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

Thank you!