Perl 6 language and compiler development | Logs at | For toolchain/installation stuff see #perl6-toolchain | For MoarVM see #moarvm
Set by AlexDaniel on 12 June 2018.
Geth nqp: 042ffa3d3e | MasterDuke17++ (committed using GitHub Web editor) | t/nqp/059-nqpop.t
Correct some test descriptions
00:45 stmuk joined 00:47 stmuk_ left 00:57 lizmat left
MasterDuke huh, changing infix:<+> in nqp to call nqp::add_i breaks a couple of nqp's tests, but rakdudo builds ok and passes its m-test 01:37
but changing all the math ops in nqp breaks way more of its tests 01:38
01:40 stmuk_ joined
MasterDuke rakudo does still build though 01:41
01:42 stmuk__ joined, stmuk left 01:45 stmuk_ left 02:09 fake_space_whale joined 02:18 skids left 02:28 AlexDaniel left 02:29 AlexDaniel joined 02:30 AlexDaniel left, AlexDaniel joined 02:35 stmuk joined 02:37 stmuk__ left 06:11 ufobat_ joined 06:12 fake_space_whale left
[Tux] This line in the build process: /pro/bin/perl5.26.2 build/ pkgconfig/moar.pc 06:23
causes extreme system load
Rakudo version 2018.06-256-g08b449e1a - MoarVM version 2018.06-322-g41108cdaa
csv-ip5xs1.249 - 1.255
csv-ip5xs-208.331 - 8.437
csv-parser24.904 - 24.922
csv-test-xs-200.453 - 0.456
test10.060 - 10.353
test-t2.313 - 2.354
test-t --race1.028 - 1.099
test-t-2041.239 - 41.680
test-t-20 --race13.441 - 13.936
06:54 lizmat joined
lizmat AlexDaniel: ok, won't do that in the future 08:27
weekly: 08:29
notable6 lizmat, Noted!
lizmat Files=1246, Tests=76432, 384 wallclock secs (16.17 usr 5.31 sys + 2706.94 cusr 259.54 csys = 2987.96 CPU)
08:34 Ven`` joined 09:50 japhb left, japhb joined 10:24 Ven`` left 10:34 robertle joined 11:11 lizmat left 11:20 Zoffix joined
Zoffix AlexDaniel: what's the verdict on the release? Are we skipping it? (Just wondering if post-release branches can be merged in) 11:20
11:21 ChanServ sets mode: +o Zoffix
Zoffix Perl 6 language and compiler development | Logs at | For toolchain/installation stuff see #perl6-toolchain | For MoarVM see #moarvm
Zoffix sets mode: -o Zoffix
stmuk I'm investigating when windows stopped passing 100% 11:32
I'm favouring a star release based on 2016.06 or (2016.06+stuff) 11:33
11:34 robertle left 11:48 pmurias joined
AlexDaniel Zoffix: you can merge post-release branches regardless of what's going to be with 2018.07 :) 11:52
I just woke up so a verdict will be a bit later I think
but either we skip it completely or we will work with some commit in the past, and you can merge post-release stuff now 11:54
stmuk I was amazed to see that windows even had 100% PASS :) 12:04
Geth roast: zoffixznet++ created pull request #455:
Post release 2018.07
roast/master: 19 commits pushed by (Zoffix Znet)++
Zoffix stmuk: of what, make test? 12:06
stmuk yes of roast .. I didn't try the nqp tests
Zoffix ?
jnthn make test is not roast 12:07
Zoffix There's a ton of spectest failures, including a hang:
timotimo hey Zoffix, any word on when recordings of last night's talk will appear online?
yoleaux 03:45Z <MasterDuke> timotimo: how come there's a smrt_numify and smrt_strify, but not smrt_intify?
Zoffix with near-06 release
timotimo: no idea. There was a hangout turned on, so I imagine there'll be a video of it. When, I don't know. This is the meeting organizer, you could ask him: 12:09
stmuk sorry I meant nqp tests not roast
Zoffix Ah
timotimo: but I don't think there's anything interesting in my talks for people who know Perl 6 already 12:10
stmuk I tried git bisect but it seemed useless since most test points didn't compile under windows anyway
Zoffix Other than maybe having a chuckle at me failing miserable to pronounce "Parametarized array ":)
stmuk actually I probably should have just tried nqp or moar bisect in hindsight 12:11
12:20 pmurias left
Zoffix timotimo: I have the slides BTW: and 12:21
timotimo thanks :) 12:22
jnthn Zoffix++ 12:25
stmuk actually what is the proper name for "make test" tests ... maybe "core tests"? 12:28
timotimo hm, sanity tests is maybe too "small"
12:32 pmurias joined, diakopter left
timotimo woo, my grant got a little shout-out \o/ 12:35
Zoffix :) 12:37
12:40 Zoffix left 12:42 diakopter joined
stmuk uses vmware seamless to pretend its not actually windows 12:47
does t\04-nativecall\12-sizeof.t flap? I'd guess not 12:54
maybe it does 12:57 13:18
13:47 robertle joined 14:13 j3nnn1 joined 14:39 Zoffix joined 14:40 nebuchadnezzar joined 15:01 fake_space_whale joined 15:08 lucasb joined
japhb FWIW, Zoffix, I still watch many of the Perl 6 videos after all these years, partly because there's often a different viewpoint or implementation style that I hadn't merged into my own yet, but also partly because I'm trawling for good videos to point Perl 6 skeptics to. 15:10
15:22 AlexDaniel left, AlexDaniel joined 15:36 x[LGWs4x4i]uG2N0 left 15:39 x[LGWs4x4i]uG2N0 joined 16:14 pmurias left 16:16 pmurias joined
Zoffix Join us next week, Saturday, August 4, 2018 for the Monthly SQUASHathon! This month we will be fixing broken ecosystem modules: 16:18
16:25 Zoffix left 16:26 fake_space_whale left 16:31 lizmat joined 17:21 Kaiepi left 17:22 Kaiepi joined 17:56 kiwi_46 joined 17:58 kiwi_46 left 18:00 lucasb left
[Coke] I am happy using a mac for dev work, but am wondering if there's an off the shelf linux laptop anyone would recommend. 18:22
(for p6 dev)
18:36 AlexDaniel left 18:38 scovit left
japhb System76 does Ubuntu-out-of-the-box and decent configs. Mind you, I tend to wipe and install Linux Mint these days, but it's all Debian-alikes anyway. 18:49
They're not exactly cheap though, because not enough economy of scale to get the big discounts. 18:50
[Coke] speaking of AC, my thermostat emailed me today to ask if I left a door open. 18:56
The future is kind of weird. 18:57
stmuk dell xps 13 has good linux support (nice kb too) 19:01
Geth rakudo: lizmat self-assigned Constants slower than literals in range creation
c603340120 | (Elizabeth Mattijsen)++ | src/Perl6/Optimizer.nqp

Part one of fixing R#2138. The case for "my constant $foo = 42" either needs to be fixed differently (as it arrives as a QAST::Var, rather than a QAST::Want or a QAST::WVal). Or maybe "my constant $foo = 42" should be codegenned as a QAST::WVal rather than as a QAST::Var.
19:12 gfldex joined 19:14 robertle_ joined
gfldex is there a hard limit on RAKUDO_MAX_THREADS ? 19:14
lizmat well, there's a default limit in the ThreadPoolScheduler 19:15
you could try raising that with
my $*SCHEDULER = max_threads => 666 ); 19:16
timotimo the number of open file descriptors is often what limits you first
maximum number*
gfldex changing $*SCHEDULER is confusing Telemetry. It doesn't show # of threads anymore. 19:20
lizmat hmmm.....interesting :-) 19:21
do you have some code ? 19:22
gfldex neither setting $*SCHEDULER or RAKUDO_MAX_THREADS, nor both gets Rakudo beyond 11 hardware threads
19:22 dct joined
gfldex i have a wee bit of code :) 19:22
lizmat gfldex: weird, I've seen it close to 64 on my machine
gfldex RAKUDO_MAX_THREADS=32 perl6 -Msnapper -e 'my $*SCHEDULER = max_threads => 18 ); .say for (1..1_000_000).race(:batch(10000)).grep(*.is-prime)'
timotimo it won't add more threads unless it thinks it's worth it 19:23
gfldex i got myself a Threadripper and wanted to rip some threads with Rakudo :(
lizmat: any simple code that works on your end? 19:24
timotimo try running "sleep 1" on a bunch of threads :P 19:27
also, i think you'd need to give .race a degree parameter, too, to get more threads
gfldex :degree(16) did the trick 19:30
wallclock util% max-rss gw gtc
--------- ------ -------- --- --------
32689668 90.88 123128 16 305
i'm not sure it I agree with the default to 4 19:31
gfldex goes to rip loads of threads
before adding :degree(16): 19:37
processed in 6.2727464s with 3455.1053820624265 stations/s
19:37 Zoffix joined
gfldex processed in 3.58713334s with 6065.567666727576 stations/s 19:37
Zoffix gfldex: eventually those parameters will be smarter: R#1741
synopsebot R#1741 [open]: [ASYNC][RFC][perf] RFC: make default $degree in hyper/race depend on number of cores
19:39 Zoffix left
MasterDuke [Coke]: +1 for the dell xps 13 (using it now with kubuntu 18.04) 19:44
gfldex: nice! i still haven't made a ryzen purchase yet... 19:45
lizmat gfldex: sorry was otherwise occupied 20:00
gfldex lizmat: don't worry, I got sorted :)
lizmat if you run -Msnapper, it has already started the Telemetry thread
with the default $*SCHEDULER 20:01
20:51 x[LGWs4x4i]uG2N0 left 20:56 x[LGWs4x4i]uG2N0 joined 21:19 AlexDaniel joined 22:11 Kaiepi left 22:12 j3nnn1 left
synopsebot R#2138 [open]: Constants slower than literals in range creation
lizmat sleep& 22:13
22:21 Kaiepi joined 22:27 skids joined
MasterDuke heh, i now have an nqp where 'say(100000000000000009)' prints '100000000000000009', but it won't build rakudo (and fails a lot of the nqp tests) 22:42
japhb MasterDuke: Are you attempting the num -> int default numeric type conversion? 22:43
MasterDuke japhb: yeah 22:44
timotimo it'd be great to have that 22:47
MasterDuke it's probably beyond me, but maybe i can get started enough to inspire some help 22:48
nqp: say(nqp::inf() + 123) 22:55
camelia Inf
MasterDuke for instance, that now says -9223372036854775685
23:22 PufferBot joined 23:25 skids left
MasterDuke nqp: say(೫e-೩೨೪) 23:27
camelia 5e-324
MasterDuke and that complains: `Error encoding ASCII string: could not encode codepoint 3307` 23:28
heh, 23:32
23:42 dct left