00:56 jnap joined
timotimo i forgot to change the temporary thing in bigint_div, i only did it in bigint_div_num >_> 01:45
dalek arVM: 2e5deb8 | (Timo Paulssen)++ | src/ (2 files):
try to get integer division with negative a behave.
 01:55
arVM: 8b259f7 | (Timo Paulssen)++ | src/core/interp.c:
make sure to abs arguments to gcd_i
arVM: 9cf4c8b | (Timo Paulssen)++ | src/math/bigintops.c:
use temporary mp_ints for subtracting 1 in bigint_div*
arVM: 23c14d7 | (Timo Paulssen)++ | src/ (2 files):
Merge branch 'negative_integer_division'

eab5033 | (Timo Paulssen)++ | src/ (2 files): replace completely wrong appreach with better approach
02:03 benabik joined
diakopter timotimo++ 02:19
anyone with a rakudo push bit? 02:20
timotimo no, it's not timotimo++ 02:21
i did it the wrong way around :)
diakopter timotimo: can you patch something in rakudo for me
timotimo gotta run for a light rail now
diakopter replace all the occurences of perl6-m$(BAT) with $(M_RUNNER) in the template for moar's Makefile 02:22
er 02:24
well, that, and the $(BAT) definition needs added back in at the top, but it seems someone removed it
when on windows.
gah 02:25
[Coke] sounds complicated.
timotimo it would seem to me that if div rounds towards zero but should floor instead, i need to add 1 to the right hand side to force a floor if the lhs is lt 0 02:29
div referring to C div now
but doing that did not seem to help 02:32
diakopter timotimo: you mean subtract 1? 02:33
timotimo yes i did mean that 02:34
since i want the value to be bigger absolutely
what do i do if a is positive and b negative, though? 02:35
that case is not handled by the rat code
diakopter wait
timotimo because there the denominator is forced to be positive
diakopter no you wouldn't subtract 1, you would subtract b..?
timotimo i would subtract b??
diakopter er
you would subtract 1 from the result if a % b != 0 02:36
timotimo i don't think so 02:37
i should be able to maybe i should subtract one from the *result* if that condition holds
not backspaces enough 02:38
i think you are right though. but i should do it completely differently 02:39
i should divide the absolute values and negate if necessary
that should give the correct result and not need to calculate the modulus 02:40
i thinkbthat is what i will do 02:42
if you see random b where words should be separated you know I'm on my phone keyboard 02:44
03:01 jnap joined
timotimo hm. there is no implementation of integer division in libtommath that behaves like java does 03:05
hehe, the latest change to libtommath is merging a pull request made by moritz to fix mod on negative integers :D 03:09
i'm actually not sure about MVM_bigint_div_num 03:20
the / operator there does floating point division, doesn't it?
03:21 FROGGS[mobile] joined
timotimo because there's mp_get_double on both sides of the / 03:21
ah, of course, the result is a num.
so no magic needed
timotimo should probably have tested before pushing 03:22
yup.
of course i made several mistakes 03:23
i shouldn't be up coding at this hour.
03:23 FROGGS[mobile] joined
dalek arVM: 71fab40 | (Timo Paulssen)++ | src/ (2 files):
replace completely wrong appreach with better approach
 03:23
timotimo nobody saw that.
now things are back to 20 failing tests :| 03:31
you know what. forget about all of this. 03:36
03:36 FROGGS[mobile] joined
timotimo this never happened. until tomorrow. when i'll retry and not be as sleep deprived. 03:36
04:02 jnap joined
diakopter timotimo: :) it's okay; we've all been there 04:31
diakopter especially, apparently 05:15
timotimo diakopter: you were right about the remiander check 05:23
i should have listened to you immediately.
dalek arVM: 17effc2 | (Timo Paulssen)++ | src/ (2 files):
fix div_i and div_I to floor rather than round towards 0
 05:24
arVM: f2ce6f8 | (Timo Paulssen)++ | src/core/interp.c:
gcd needs to abs both arguments before calculating
timotimo this makes rounders.t pass completely 05:25
maybe now i can fall asleep
diakopter timotimo: well, I'm quite certain that's not anywhere near the "right" way to do it, even if it works
timotimo are you sure?
diakopter for those tests
yea, I'm sure there's a branchless version
timotimo at least for the bigint stuff we get the remainder for free
well, since it's - 1 in the true case and - 0 in the not-true case, yeah
trivial branchless version
er ... except for the outer branch 05:26
but that can be done with a multiplication.
diakopter confounded again! ;)
timotimo the compiler (or the cpu, even) might be able to do this.
diakopter yeah
timotimo we're in the "make it work" stage right now anyway
diakopter the sufficiently smart compiler knows and sees all
well, enough, anyway 05:27
timotimo it certainly does
i wonder how the rest of rakudo copes with the div difference between jvm/moarvm and parrot
diakopter extremely sparse test coverage?
timotimo that must be it, yeah
timotimo opens a quest on questhub 05:28
questhub.io/realm/perl/quest/52d0d7...7c32000061 ← 05:30
This type cannot unbox to a native string in method Str at src/gen/m-CORE.setting:1020 05:40
in method abs2rel at src/gen/m-CORE.setting:12607
in block frame_name_17 at t/spec/S32-io/io-spec-win.t:266
in block frame_name_1 at t/spec/S32-io/io-spec-win.t:257
i may look into this tomorrow :)
hmm: not ok 823 - Rat.Rat yields correct Rat 05:41
# got: '301281685344656669 1250'
# expected: '301281685344656640 1250'
somehow a num got used instead of a big integer or something? 05:42
that difference is surprisingly big
Files=767, Tests=29473, 912 wallclock secs ( 8.16 usr 1.09 sys + 3192.27 cusr 69.95 csys = 3271.47 CPU) 05:43
the last entry in perl6_pass_rates.csv has 26962; is that Tests= number the number of succeeded tests or run tests? 05:44
anyway. bed. 05:45
jnthn Whoa...timotimo++ hacked all night... 08:06
timotimo: yay, rounders.t now passes fully for me too 08:16
08:19 ssutch joined 08:23 eternaleye joined
FROGGS[mobile] morning o/ 08:46
08:46 Woodi joined 11:09 cognominal joined
jnthn Epic win! The lounge at Frankfurt has a drink-all-you-want weissbier tap 11:10
FROGGS \o/ 11:15
jnthn: what are you in frankfurt for?
jnthn FROGGS: Oh, just connecting here 11:17
FROGGS good :o)
jnthn It's miserable weather. :) 11:18
FROGGS it is very nice in berlin though
11:21 FROGGS[mobile] joined
masak I can see Weissbier being very drink-all-you-want-able. 11:33
11:34 FROGGS[mobile] joined
JimmyZ E:\OpenSource\MoarVM\install\bin\moar --libpath=src\vm\moar\stage0 src\v 11:34
m\moar\stage0\nqp.moarvm --bootstrap --setting=NULL --no-regex-lib --target=mbc
--output=gen\moar\stage1\nqpmo.moarvm gen\moar\stage1\nqpmo.nqp
MAST::Local index out of range
jnthn JimmyZ: You'll almsot certainly have an out of date something or other... 11:35
OK, got a patch here that puts the serialization blob into the bytecode file as is rather than base-64-ing it. 11:36
Shaves a whole megabyte off the build output of NQP
nwc10 cool 11:39
when does your plane leave?
jnthn Boarding starts in 20 ins 11:40
*mins
nwc10 Boarding started here. (Into the car. It takes ages these days :-/) 11:41
masak O(@family) 11:44
jnthn make test in NQP used to be 23s on my laptop. Now it's 21s.
r: say 100 * 21 / 23
camelia rakudo-parrot 263aa0, rakudo-jvm 263aa0: OUTPUT«91.304348␤»
masak \o/ 11:45
jnthn Shaved about 5MB off the memory used by nqp-m -e "while 1 { }" too 11:48
dalek arVM: cf39b8f | jonathan++ | src/ (5 files):
Enable storing serialized data in bytecode section

This avoids the whole base-64 of an MVMString and so forth, speeding both startup and compilation, cutting memory footprint and decreasing the disk size of NQP and Rakudo compilation output.		 11:53
jnthn flight & 11:56
FROGGS yay!! 12:43
jnthn++
12:44 cognominal joined
tadzik woot 12:50
timotimo yay! 12:56
a nice little gift before becoming tuitless for ... what was it, 6 days?
gist.github.com/timo/4b8885415dbef0928f02 \o/ 12:59
diakopter I don't understand why I can't pull or --rebase pull the latest commits (since timotimo's merge) without making merge conflicts 13:00
timotimo it could very well be the case that you pulled one of my commits that i then hastily force-pushed over 13:01
diakopter seems so
how to fix
some kind of force pull? 13:02
timotimo git fetch
git reset --hard origin/master
nwc10 who becomes tuitless? :-( 13:04
diakopter yeah had to do that
timotimo jnthn said so
"22:52 < jnthn> btw, I probably won't have many tuits at all during the next 5-6 days."
diakopter: sorry :(
nwc10 aha 13:05
timotimo but i think you were the only unfortunate soul who got to download these completely botched commits
diakopter I think he will have some p6 tuits ;)
timotimo mayhaps 13:07
diakopter I like how Git Extensions now automatically fetches in teh background every few minutes 13:10
there's a bunch of nom commits in rakudo (back to 9 days) that aren't in moar-support 13:11
(wouldn't that skew the test comparisons?) 13:12
timotimo my change for chained abverbs would, but only 2 tests in total i believe
not sure if the "not set an encoding" commit changes test pass rate 13:13
diakopter oh
13:24 cognominal joined
JimmyZ jnthn: I still got 'MAST::Local index out of range' after updating nqp && moarvm 13:47
timotimo :o 13:48
but how? :(
diakopter JimmyZ: you almost certainly need to make realclean or hard reset
JimmyZ I did do `git clean -xdf' and `git reset HEAD --hard` 13:51
diakopter did you update submodules? 14:18
(maybe something there?)
all three, rakudo, nqp, and moarvm trees
(clean/hard I mean)
14:35 FROGGS[mobile] joined 14:41 lue joined 15:54 lue joined 16:51 FROGGS[mobile] joined 18:04 FROGGS[mobile] joined 18:07 ssutch joined
dalek arVM: 3293aaf | (Tobias Leich)++ | / (6 files):
add op uniisblock to look up unicode blocks directly
 19:04
arVM: 4e2fe17 | (Tobias Leich)++ | / (3 files):
add block lookup, and (evil) hack for property value lookups
FROGGS creates his own rakudo.moar_summary.out 19:21
diakopter :) 19:23
[Coke] oh, we probably have a new one.. 19:44
as of a few hours ago: 19:46
19:46 FROGGS[mobile] joined
[Coke] +"total", 27135, 1060, 641, 1384, 29729, 28495 19:46
r: say 27135/28454 # moar today
camelia rakudo-parrot 263aa0, rakudo-jvm 263aa0: OUTPUT«0.953644␤»
19:58 FROGGS[mobile] joined
FROGGS mine: 20:07
"Synopsis", "pass","fail","todo","skip","plan","spec"
"total", 27460, 686, 639, 1383, 29652, 28495
r: say 27460/28454 20:08
camelia rakudo-parrot 263aa0, rakudo-jvm 263aa0: OUTPUT«0.965066␤»
grondilu 96.5% ? nice 20:12
20:18 FROGGS[mobile] joined
nwc10 slow thing is, well, slow: 7977 nicholas 30 10 382m 359m 14m R 99.4 0.4 3600:00 moar 21:34
FROGGS :o)
timotimo but still awesome :D 23:17

Powered by Cro and the Raku® Programming Language.

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

Thank you!