brrt \o #moarvm 07:48
jnthn brrt: Reasons for malloc/memcpy'ing it make some sense. 08:39
brrt ugh, segfaults 10:32
damn segfaults 10:52
oh.... 10:54
ret
*ahum* 10:55
oh, now i'm over to bus errors 10:57
much amaze
FROGGS ó.ò 10:58
brrt *ahem* 11:00
turns out storing floating points in the place of your former stack base pointer isn't a good idea :-D
FROGGS ohhh
well, even I can agree on that point :o) 11:01
brrt thats just one of them, though
hmm 11:02
ok, now it works for the example code
by the way, i think i should make a test library of jittable fragments in moar-jit
FROGGS nice :o) 11:03
brrt: that would be nqp code, right?
then nqp/t/jit/* would be the right place for it
brrt yes, although i could probably compile them to moarvm bytecode
hmm, i could do that, yes 11:04
my original plan was to make a directory under moarvm and hard-compile the moarvm bytecode fragments
FROGGS nah
brrt much like the dynasm fragments are precompiled
well, if you say so :-) 11:05
FROGGS we already have feature/backend specific subdirs in nqp/t :o)
brrt i'm fine with every solution 11:08
as long as its black
FROGGS sure 11:09
FROGGS goes off to paint everything black
brrt hmm
ls
wrong window :-) 11:10
FROGGS np :o)
brrt oh, i've seen where my segfault comes from 11:11
movabs r10, 0x0; call r10
doesn't seem right, does it
brrt (its equivalent to int (*fp)(int, int) = NULL; fp();) 11:13
brrt and.. now it works 11:18
dalek arVM/moar-jit: 3df5f80 | (Bart Wiegmans)++ | / (6 files):
Support floating point passing and returning
11:24
brrt wonders if there is a way to have git ignore the dynasm output files in the diff log 11:25
anyway, gotta run, see you in a bit
timotimo oh yay, floats can be passed and returned 14:55
that's a good step
brrt :-) 14:56
now working on further arithmetic
jnthn :) 15:08
timotimo and then some future arithmetic?
jnthn hotel &
dalek arVM/moar-jit: ac1d715 | (Bart Wiegmans)++ | / (5 files):
More arithmetic, including floating-point ops.
15:41
timotimo those diffs are teh huge >_< 15:43
jnthn It's brrt's secret plan to overtake me for lines committed in Moar :P 15:57
timotimo oooh clever :) 16:02
brrt :-D 16:10
timotimo - i think that's because dynasm rebuilds the whole file and this file is committed with the rest
timotimo yes, of course
brrt its not 'fair' to count those lines :-) 16:11
timotimo hah.
of course not
lee__ brrt: with js files github will collapse .min.js diffs 16:17
brrt hmmm.... 16:21
but i can hardly call my c file output .min.js just to please github
my real secret plan is to take all credit for spesh-related speedups 16:22
muhahaah
jnthn :P 16:46
.oO( does he get all the blame for the speshbugs too? :P )
brrt well, can't have one without the other 16:49
i think.. the next thing is the sp_guard* ops 16:50
also
what does wval do?
(bbiab) 16:53
timotimo it grabs an object from a SC iirc
jnthn Yup. 17:03
And if the object is in gen2, can just reference it directly.
dalek arVM: 722c847 | (Tobias Leich)++ | / (3 files):
auto detect computed goto support
19:07
jnthn Cute :) 19:11
FROGGS :o) 19:12
timotimo: I'm sorry but I could not resist
jnthn is glad somebody did it :) 19:13
FROGGS you don't like build toolchains, do you :o) 19:14
hmmm, stage parse took 57s with cgoto
but my battery is low so it is perhaps just that
last time I checked it was a bit faster, and that was like three weeks ago 19:15
timotimo yays 19:16
FROGGS timotimo: would be nice when you could test the impact on stage parse... I have no power cable where atm :/ 19:17
here* 19:18
timotimo last time i looked it was an improvement
my computer is kinda busy showing SGDQ :)
FROGGS *g* 19:21
np, I didn't meant *now* btw
timotimo okay 19:22
hoelzro MVM_exception_throw_adhoc(tc, "Named argument '%s' already used", MVM_string_utf8_encode_C_string(tc, name)); // <- won't this cause a memory leak? 20:48