00:02 b2gills joined 00:19 chansen_ joined 00:35 Kaeipi left, Kaeipi joined 00:43 dogbert17 left 01:29 kvw_5_ joined 01:32 kvw_5 left 01:33 dogbert17 joined 05:04 Geth joined 06:31 squashable6 left 06:32 squashable6 joined
Geth nqp: e4c8a5c30e | (Patrick Böker)++ | README.pod
Remove Travis badge

Travis has been disabled since the beginning of January already.
06:39
nqp: d88a3e4c25 | (Patrick Böker)++ (committed using GitHub Web editor) | README.pod
Merge pull request #707 from patrickbkr/disable-obsolete-CIs

Remove Travis badge
06:43 domidumont joined
[Tux] Rakudo v2021.03-47-g8fc4cb56c (v6.d) on MoarVM 2021.03-16-gd7220f95a
csv-ip5xs0.828 - 0.832
csv-ip5xs-208.053 - 8.446
csv-parser25.002 - 25.687
csv-test-xs-200.367 - 0.370
test7.727 - 7.729
test-t1.892 - 1.954
test-t --race0.863 - 0.983
test-t-2033.498 - 34.265
test-t-20 --race9.041 - 9.414
06:51
06:56 HarmtH left 06:57 HarmtH joined 06:59 patrickb joined
patrickb o/ 06:59
07:09 frost-lab joined 07:40 linkable6 left 07:41 linkable6 joined 07:56 Kaeipi left 07:58 Kaiepi joined 08:54 linkable6 left 08:55 linkable6 joined
MasterDuke very interesting. github.com/Raku/nqp/blob/master/sr...qp#L44-L53 has the same `#?if moar ... ` business with `@ARGS` vs `*@ARGS` that github.com/rakudo/rakudo/blob/mast...c/main.nqp and github.com/rakudo/rakudo/blob/mast...-debug.nqp do 09:09
and easier to experiment with since i don't get that circular module error when trying to run multiple times 09:10
and it's been that way for 8 years 09:11
if i change the `#?if moar` version to match the others, then the nqp build seems to drop into a repl, where ^D starts the next step in the build 09:12
there were mentions of parrot with an 'XXX FIX ME' in previous versions, so i have no idea what's the correct behavior 09:14
09:35 linkable6 left 09:36 linkable6 joined
dogbert17 m: my $a = "»ö«".uniname for ^10_000_000; say now - INIT now 09:38
camelia 1.2926607
dogbert17 m: my $a = uniname("»ö«") for ^10_000_000; say now - INIT now
camelia 1.4239268
dogbert17 m: my $a = uniname("»ö«") for ^10_000_000; say now - INIT now
camelia 1.36861578
dogbert17 heh, the sub version is a lot faster on my home machine 09:39
how old is camelia? 09:40
MasterDuke feb 24 09:43
lizmat dogbert17: the sub version is currently faster as it doesn't dispatch to the method version
this will be corrected in the coming days
dogbert17 ++lizmat, I had a feeling it was on your radar :)
so what can we do about camelia then?
lizmat I'm not sure who runs camelia at the moment 09:47
could be nine, could be moritz_ ?
Geth rakudo: 8f4db8b6b8 | (Elizabeth Mattijsen)++ | src/core.c/List.pm6
Make List.sum again 40% faster

By skipping the special handling of the first element. This also reduces the bytecode size, but still too large to get inlined. However all of the method calls inside "sum" *are* inlined into "sum".
09:50
10:21 gfldex_ is now known as gfldex
Geth rakudo: 07aa4aea9f | (Elizabeth Mattijsen)++ | 2 files
Make Any.sum about 40% faster and check for laziness

  - move it to Any-iterable, as it is more a place to be for such a function
  - use binding instead of assignment
  - add private method !lazy for handling of lazy iterators
  - reduce bytecode size by handling laziness outside in !lazy
  - Any.sum now small enough in bytesize to inline
10:29
10:40 linkable6 left 10:41 linkable6 joined
Geth rakudo: e191266f88 | (Elizabeth Mattijsen)++ | 6 files
Make .sum a proper multi

  - put the "is nodal" trait on the multi
  - all candidates now are "nodal" because of that
10:54
11:01 frost-lab left, frost-lab joined 11:32 frost-lab left
Geth rakudo: 8e56c8c89f | (Elizabeth Mattijsen)++ | 2 files
Introduce (throw|fail)-iterator-cannot-be-lazy methods

  - marked as implementation-detail
  - utility method to fail / throw X::Cannot::Lazy exception
  - to be used by methods not accepting lazy iterators
  - intended to reduce bytecode size of calles to help inlineability
  - removed the !lazy private methods that were used previously
11:37
12:17 sena_kun left, sena_kun joined 14:19 patrickb left
Geth ¦ rakudo: tbrowder self-assigned A DateTime object should output a real number for its posix method github.com/rakudo/rakudo/issues/4286 15:07
16:39
rakudo: ac9f40d871 | (Elizabeth Mattijsen)++ | 8 files
Introduce (fail|throw)-cannot-be-empty helper method

To again reduce bytecode size in the caller's for better inlineabiolity
17:14
17:14 domidumont left
Geth 17:17
rakudo: 71960b162d | (Elizabeth Mattijsen)++ | 20 files
Set more sensible defaults to helper methods

Further reduces bytecode footprint at the callsites
17:37
17:40
17:40 MasterDuke left
tbrowder does anyone know how to find the name or alias of a github emoji submitter on a comment in PRs or issues? 17:57
18:47 HarmtH left 18:50 HarmtH joined
Geth rakudo: b0910d5bd1 | (Christian Bartolomäus)++ | lib/NativeCall.rakumod
Remove unused variable

It was added with 80d6b425ce and it looks like a leftover from developing that patch. (Maybe it was meant to be passed as the value for :deserialize_ast -- which used $fixups in the end.)
18:58
19:18 HarmtH_ joined, HarmtH left
Geth rakudo: 23036f1d7a | (Elizabeth Mattijsen)++ | src/core.c/Any-iterable-methods.pm6
Save one scalar allocation for each non-simple for loop

19:23
19:48 HarmtH_ left 19:50 HarmtH joined
Geth rakudo: f75957dd8e | (Elizabeth Mattijsen)++ | 3 files
Make .chrs a proper multi

Also add a native int candidate, which is about 30% faster.
20:28
20:45 MasterDuke joined, dogbert11 joined 20:49 dogbert17 left
Geth 21:05
moon-child tbrowder: hover over the emoji 21:08
Geth rakudo: 31331a2861 | (Elizabeth Mattijsen)++ | src/core.c/unicodey.pm6
Make uniname a proper multi method

  - move the main logic to the Int / Str classes
  - make the sub version just pass on to the appropriate method
  - add support for other Cool values such as Num and Match
21:24
nqp-configure: e6ddc07aa9 | (Patrick Spek)++ | lib/NQP/Config.pm
Consider /dragonfly/ a BSD as well

In Perl, $^O is dragonfly when running DragonflyBSD. As this doesn't match the /bsd/ that is done to determine bsd-ness, this screws up the
  `make` check, which results in NQP and Rakudo failing to build on
DragonflyBSD.
21:25
nqp-configure: e068508a94 | (Vadim Belman)++ (committed using GitHub Web editor) | lib/NQP/Config.pm
Merge pull request #26 from Tyil/dragonflybsd

Consider /dragonfly/ a BSD as well
tyil \o/
can anyone tell me when I can expect NQP/Rakudo to use this latest update to nqp-configure? 21:26
(it's currently breaking rakudo-star for a user, and I want to make them happy with a nightly that they can use)
Geth rakudo: 80d2034e13 | (Vadim Belman)++ | 3rdparty/nqp-configure
nqp-configure submodule update

21:28
vrurg tyil: was it the answer? ;) 21:29
tyil :D 21:30
that's way faster than I expected, not gonna lie
tbrowder moon-child: thnx, but that doesn't work on an ipad so far (unless there's some finger magic i don't know about) 21:36
moon-child oh, heh, don't know then 21:43
tyil vrurg: I clones latest moarvm, nqp and rakudo for a quick rakudo-star testing drive, but nqp doesn't want to build: `Found a MoarVM binary but was not able to get its version number.` 21:52
I'm going to guess it's an easy thing to fix, but I wouldn't know where to look :<
it says I *should* check out tags, but then I won't have the fix I want
(if it's a chore to do it like this I'll just patiently wait for 2021.4, though) 21:53
Geth nqp: cb008e907a | (Vadim Belman)++ | 3rdparty/nqp-configure
nqp-configure submodule update

21:55
vrurg tyil: ./Config --gen-nqp=master
Config.pl
tyil ah, thanks :> 21:56
vrurg I have forgotten to update the submodule in nqp, but now it's fixed. Not sure if MoarVM is ready for dragonfly though. 21:57
tyil moarvm builds fine on its own, at least 22:09
--gen-nqp is not a valid flag for NQP's Configure.pl it looks like 22:10
(fwiw, I have each moarvm,nqp and rakudo checked out individually, and I build each part from its own repo in rstar) 22:11
vrurg tyil: It's for rakudo build to get the latest NQP.
tyil hmm
vrurg tyil: Since you build separately, you don't need --gen-nqp then. Just update NQP report to the latest master. 22:12
s/report/repo/
tyil I already pulled latest master on nqp 22:13
it is late, perhaps I'm doing something dumb, I should retry in the morning 22:15
vrurg: thanks for your help thus far :>
vrurg tyil: np and good luck! One more platform is good to have. 22:16
Geth rakudo: d14dda68da | (Elizabeth Mattijsen)++ | src/core.c/unicodey.pm6
Make uninames a proper multi method

  - move the main logic to the Str class
  - make the sub version just pass on to the appropriate method
  - makes uninames about 2x as fast
22:19