nebuchadnezzar hello 12:07
timotimo o/ 12:15
nebuchadnezzar if I rebuild moarvm from dyncall to libffi, should I rebuild nqp and rakudo ? 12:37
timotimo no need 12:38
nebuchadnezzar great, thanks
dalek arVM/even-moar-jit: 7da475f | brrt++ | src/jit/linear_scan.c:
Treat the different copy/phi nodes separately

They were treated as the same, but because they are subtly different, they are best treated differently		 17:24
arVM/even-moar-jit: f24555a | brrt++ | src/jit/ (2 files):
Change some types and format strings to make clang happy

Contrary to my expectations, MVMint8 != char
MoarVM/even-moar-jit: b308641 | brrt++ | / (3 files):
MoarVM/even-moar-jit: Implement MVM_JIT_IF and MVM_JIT_EITHER via tiling
MoarVM/even-moar-jit:
timotimo is that a rebase or new commits? 17:29
brrt that's new commits 17:39
i finally had a seat on the train :-)
timotimo cool! 17:40
yay for trains
with seats on them
but not actually *on*
brrt haha 17:43
brrt is going to get icecream 17:45
timotimo yay!
Zoffix How would you go about this: we have a function that recurses and there's a case to trigger a segfault by recursing to deep? 22:49
m: 7 ~ "\x[308]" x 110_000
camelia rakudo-moar 626a22: OUTPUT«(signal SEGV)»
Zoffix This is recursing in twiddle_trie_node() that eventually gets a stackoverflow
timotimo if we used the C stack, you could run your stack into your heap, but that's not a SEGV usually 22:50
oh
that is the C stack
so maybe that's it
Zoffix So to fix it we need to "run stack into heap"? 22:51
timotimo oh, no, that's how it explodes 22:52
Zoffix oh
timotimo there's one big area of memory. at one end lives the heap, at the other lives the stack
both grow towards each other
well, in the realm of virtual memory it's all a bit more interesting
but when your stack reaches your heap and you try to put more stuff on the stack ... boom
Zoffix K, I think I'll leave this for now and try something easier :) 22:53
That was this ticket: rt.perl.org/Ticket/Display.html?id...et-history
timotimo since the algorithm to change this trie is based on traversing it down and then up, and the nodes of the trie don't have an "up" pointer (i haven't looked, but i guess this is the case), we'd have to introduce our own data structure to facilitate "going up" 22:55
essentially, we'd make the same thing the stack was, but hopefully a bit more memory efficient, and swap-out-able 22:56
i need to go to bed
toodles 22:57
Zoffix night 22:59
konobi madvise? 23:04
mmm... not what i was thinking of 23:05

Powered by Cro and the Raku® Programming Language.

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

Thank you!