01:48 ilbot3 joined 02:23 lizmat joined 05:58 domidumont joined 06:03 domidumont joined 08:13 zakharyas joined 09:14 domidumont joined 09:27 brrt joined
brrt hi #moarvm 09:27
i figured out that negative synthetic nodes may not be as awesome an idea
... i have a somewhat, slightly better idea 09:28
i add a 'SYNTH' as the zeroth node for any jit tree, and refer to that instead
that is a no-special-branch solution 09:29
nwc10++ interesting talk on the non-butterfly effect
nwc10 Lee++ # bringing the tripod, camera and mike 10:14
11:16 ggoebel joined
brrt yeah, having the videos online is really nice 11:18
nwc10 but we didn't promise anything in advance because
1) we were not sure if it would work out
2) we wanted folks who could to actually show up 11:19
(someone had asked "will there be videos?" and the question appeared to be roughly "can I enjoy the conference from the comfort of the sofa and not have to interact with the nasty huuuumins"
and if everyone stays at home, why have a conference?)
brrt yeah, i can see how that works 11:20
for my part i had prefered to have been there, but $work...
brrt has clearly gone over to the dark side 11:21
12:22 TheLemonMan joined
TheLemonMan timotimo, turn MVM_GC_DEBUG_LOG_FLAGS on and enjoy the nice traces in gdb :D 12:23
the STABLE object might get collected somewhere
interesting, by enabling the debug output I get a crash while compiling rakudo, that's in process_worklist 14:34
and it's actually a debug sentinel to catch null stables
oh, in case you didn't know about rr (rr-project.org/), it's a nifty tool for the debugging lovers 14:38
brrt i did not know 14:42
timotimo i've heard of it, but i haevn't used it yet 14:51
15:03 MasterDuke_ joined
MasterDuke_ ahoy #moarvm. just performing a driveby dropping of a bug report if anybody is interested 15:04
timotimo OK
MasterDuke_ RT #128035
synopsebot6 Link: rt.perl.org/rt3//Public/Bug/Displa...?id=128035
MasterDuke_ i submitted a Rakudo PR github.com/rakudo/rakudo/pull/867/, but it turns out to be a MoarVM issue 15:05
just for the heck of it i tried changing the > on this line github.com/MoarVM/MoarVM/blob/94ed...gint.c#L18 to a >= 15:09
and that fixed the problem in the original RT, but not surprisingly broke the nqp and rakudo spectests 15:10
and with that my work here is done, later all... 15:11
timotimo why would we be unable to unbox a 64bit big int?
MasterDuke_ (sucked back in) no idea 15:12
m: use nqp; say nqp::unbox_i(2**64); say nqp::unbox_i(2**64 - 1) 15:13
camelia rakudo-moar c9b18c: OUTPUT«Cannot unbox 65 bit wide bigint into native integer␤ in block <unit> at <tmp> line 1␤␤»
MasterDuke_ m: use nqp; say nqp::unbox_i(2**64 - 1);
camelia rakudo-moar c9b18c: OUTPUT«-1␤»
timotimo that seems legit 15:14
MasterDuke_ m: use nqp; say nqp::unbox_i(2**63);
camelia rakudo-moar c9b18c: OUTPUT«-9223372036854775808␤»
MasterDuke_ m: use nqp; say nqp::unbox_i(2**63 - 1);
camelia rakudo-moar c9b18c: OUTPUT«9223372036854775807␤»
MasterDuke_ m: say int.Range.max 15:15
camelia rakudo-moar c9b18c: OUTPUT«9223372036854775807␤»
MasterDuke_ m: say uint.Range.max
camelia rakudo-moar c9b18c: OUTPUT«18446744073709551615␤»
timotimo hmm
MasterDuke_ i'm a little unclear why the funtion's name is mp_get_int64, but it returns a MVMuint64 instead of a MVMint64 15:20
TheLemonMan you should just check github.com/MoarVM/MoarVM/blob/mast...int.c#L102 in the if arm whether the sign-extended version becomes negative
or simply check if you're going to set the MSB when assembling the result 15:21
timotimo MasterDuke_: maybe because the sign is a separate bit in the mp library we use? no clue. 15:23
MasterDuke_ (this reminds me of physics tests where i'd get something like g = -9.8m/s**2, which i always tried to argue was a fine answer)
TheLemonMan you can easily test this by checking out the result of (int64_t)(1ULL << 63) 15:24
MasterDuke_ well, i really do need to run now, but i'll play around with your suggestions this evening if nobody else get a fix in before then 15:26
thanks all
TheLemonMan and MP_LT == mp_cmp_d(i, 0) can be optimized out with SIGN(i) if I understand the code correctly 15:27
timotimo oh, that sounds sensible 15:29
TheLemonMan done o/ 15:50
the mp_neg(i,i) calls can also be thrown away by using SIGN(i) ^= 1 15:54
or SIGN(i) ^= MP_NEG to be fancy heh
timotimo most stuff we do is non-destructive 16:02
16:39 domidumont joined 16:55 domidumont joined 17:29 FROGGS joined 18:02 pyrimidine joined 18:16 lizmat joined 18:22 TheLemonMan joined 22:54 Ven_ joined

Powered by Cro and the Raku® Programming Language.

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

Thank you!