Perl 6 language and compiler development | Logs at colabti.org/irclogger/irclogger_logs/perl6-dev | For toolchain/installation stuff see #perl6-toolchain | For MoarVM see #moarvm
Set by AlexDaniel on 12 June 2018.
AlexDaniel hmmmmmm github.com/rakudo/rakudo/issues/21...-407591812 00:26
[Coke] nqp repo is showing a potential vuln. 00:34
timotimo in the javascripd code, yeah 00:35
surely pmurias has already seen it, too
[Coke] ^^pmurias, maybe?
00:37 fake_space_whale joined
Geth ¦ nqp: AlexDaniel assigned to pmurias Issue Flagged as vulnerable github.com/perl6/nqp/issues/437 00:46
02:08 mst joined
Geth nqp/revert-485-unfudge-test: 1b1ddea13e | 陈梓立++ (committed using GitHub Web editor) | 0 files
Revert "[MoarVM Bump] Brings 4 commits"

This reverts commit 3a4cbe04237a7eae15c045385f43a178b470d74f.		 02:18
nqp/revert-485-unfudge-test: ce177f5cb1 | 陈梓立++ (committed using GitHub Web editor) | t/nqp/019-file-ops.t
Revert "unfudge test, co #274 (#485)"

This reverts commit 15c9774719f98c0c04252a3a1712b35c84b29931.		 02:19
nqp: TisonKun++ created pull request #490:
Revert "unfudge test, co #274"
synopsebot NQP#274 [open]: github.com/perl6/nqp/issues/274 t/nqp/19-file-ops.t failure in pre-2016.01
02:22 ufobat_ joined 02:26 ufobat left
Geth nqp: aff38355c1 | 陈梓立++ (committed using GitHub Web editor) | t/nqp/019-file-ops.t
Fudge test 019-file-ops.t

Fails when MoarVM/MoarVM test with option
or `MVM_OPTIONS="--no-jit" CC='gcc'` or `MVM_OPTIONS="" CC='gcc'` or `MVM_OPTIONS="--no-jit" CC='clang'` or `MVM_OPTIONS="" CC='clang'`
would not unfudge it before a certainly fix.		 02:27
02:30 dct left 03:15 stmuk joined 03:17 stmuk_ left 04:40 squashable6 joined 05:59 fake_space_whale left 06:05 benjikun left, MasterDuke left
[Tux] Rakudo version 2018.06-247-ga71e370bb - MoarVM version 2018.06-318-gb8f9f85f2
csv-ip5xs1.189 - 1.271
csv-ip5xs-208.066 - 8.892
csv-parser25.600 - 25.602
csv-test-xs-200.457 - 0.465
test10.300 - 11.018
test-t2.302 - 2.397
test-t --race1.018 - 1.048
test-t-2040.866 - 44.630
test-t-20 --race14.084 - 14.187
 06:22
06:51 robertle joined 08:10 lizmat left 08:23 lizmat joined
lizmat Files=1246, Tests=76406, 382 wallclock secs (16.32 usr 5.78 sys + 2682.04 cusr 264.13 csys = 2968.27 CPU) 08:25
Geth nqp/nqp-in-the-browser: 870ad0844c | (Paweł Murias)++ | 3 files
[js] Add .npmignore, bump versions in package.json after releases
 08:48
nqp/master: 51 commits pushed by pmurias++, (Paweł Murias)++
review: github.com/perl6/nqp/compare/aff38...34978a639c		 09:01
09:11 pmurias joined 09:50 timotimo left, timotimo joined 11:38 robertle left 11:40 robertle joined 11:59 robertle left 12:10 robertle joined 12:24 diakopter left
Geth nqp/truffle: 131819e902 | (Paweł Murias)++ | src/vm/jvm/runtime/org/perl6/nqp/truffle/nodes/variables/NQPBindLocalVariableNode.java
[truffle] Cleanup by removing unneeded code
 13:15
nqp/truffle: 9453e7fe5b | (Paweł Murias)++ | 11 files
[truffle] Implement native types on variables
nqp/truffle: e300cac81e | (Paweł Murias)++ | src/vm/jvm/runtime/org/perl6/nqp/truffle/nodes/expression/NQPSmartIntifyNode.java
[truffle] Use @Specialization in NQPSmartIntify
13:41 [Tux] joined 13:48 ufobat_ left 14:31 AlexDaniel joined, AlexDaniel` joined 14:33 ZofBot joined, ChanServ sets mode: +v ZofBot 14:35 AlexDaniel left 14:36 AlexDaniel joined 14:38 AlexDaniel left, AlexDaniel joined 14:51 fake_space_whale joined
Geth nqp/truffle: d25e35c901 | (Paweł Murias)++ | 12 files
[truffle] Use the double to string conversion from graal.js
 14:53
pmurias .tell MasterDuke I replace our num->str conversion with a faster one from the truffle js implementation, it would be interesting how our performance now compares 15:14
yoleaux pmurias: I'll pass your message to MasterDuke.
15:24 robertle left 15:34 skids joined 15:53 ufobat_ joined, robertle joined 16:26 fake_space_whale left
[Tux] Rakudo version 2018.06-247-ga71e370bb - MoarVM version 2018.06-318-gb8f9f85f2
csv-ip5xs1.292 - 1.523
csv-ip5xs-208.649 - 9.453
csv-parser25.525 - 26.417
csv-test-xs-200.449 - 0.523
test9.949 - 10.215
test-t2.274 - 2.284
test-t --race0.974 - 1.080
test-t-2040.647 - 43.801
test-t-20 --race13.471 - 14.361
 16:45
main reason to re-post is a reboot after some system updates 16:46
lizmat [Tux]++ 16:51
17:50 Ven`` joined 17:55 Ven`` left 17:57 Ven`` joined 18:14 Ven`` left 18:18 Ven`` joined 18:22 Ven`` left 19:25 Ven`` joined 19:35 Ven`` left 20:11 MasterDuke joined
MasterDuke jnthn, pmurias, timotimo: that code i was timing earlier with moar/jmv/truffle is now fastest (excluding startup) on truffle after pmurias++ recent commits 20:22
yoleaux 15:14Z <pmurias> MasterDuke: I replace our num->str conversion with a faster one from the truffle js implementation, it would be interesting how our performance now compares
timotimo nice 20:23
MasterDuke it is more variable though (the fast runs are faster, but the slow runs are slower) 20:27
timotimo oh, mhm 20:28
MasterDuke oh, that was for untyped, let me re-run it with typed variables
20:31 robertle left
MasterDuke using typed variables makes moar faster (from 3.1s down to 1.9s), but doesn't change jmv or truffle 20:33
timotimo probably still NYI on truffle? 20:35
but it surprises me a bit that it doesn't make our jvm code faster 20:36
i have got to go take a shower, the temperature & humidity is getting unbearable
20:37 literal left 20:39 literal joined
MasterDuke his most recent commits did implement them for truffle 20:42
pmurias timotimo: it's a bit suprising too for me 20:44
MasterDuke pmurias: i guess it was the faster num stringification that made the difference 20:45
pmurias timotimo: I'll have to brush up on assembly and see how the JVM optimizes it
MasterDuke: the stringification was the slow part for sure
MasterDuke same for moarvm. perf shows dtoa_grisu3 the most expensive routine at 30%, then MVM_coerce_smart_numify at 9.3%, and then MVM_coerce_n_s at 5% 20:48
MasterDuke wonders if that java profiling tool can actually give useful info for truffle/nqp code 20:53
Geth rakudo: 434ede491d | (Elizabeth Mattijsen)++ | src/core/Routine.pm6
Make sure that wrapped onlystar routines are not optimized away

If a proto with an onlystar is wrapped, it still kept the "onlystar" flag, which means optimizers can optimize it away. This commit always resets the flag, regardless of whether it was set or not.
20:57 skids left
pmurias MasterDuke: you can see how the code is optimized using the ideal graph visualizer: ssw.jku.at/General/Staff/TW/igv.html 20:57
although my laptop screen is too small to read the graphs comfortably 20:58
github.com/graalvm/simplelanguage/...ter/sl#L62 too pass stuff to the igv 20:59
MasterDuke: I skimmed through the assembly that's generated for our benchmark
MasterDuke i'm trying to remember what the tool was i was experimenting with a little while ago (trying to profile the jvm rakudo build) 21:00
pmurias and it seems the only call we have is to the coercion
MasterDuke ah, visualvm i think... 21:01
pmurias spending time on optimizing the num to str conversion at this point doesn't seem to be the best use of time :) 21:02
MasterDuke pmurias: for sure
pmurias I think it will be easier to beat moar on complex stuff rather then a simple loop where it passes stuff to a C routine :) 21:03
lizmat .tell moritz en.wikipedia.org/wiki/Perl_6 points to leanpub.com/perl6regex/ which shows work in progress ?? 21:22
yoleaux lizmat: I'll pass your message to moritz.
moritz lizmat: both should link to www.apress.com/us/book/9781484232279 instead 21:28
yoleaux 21:22Z <lizmat> moritz: en.wikipedia.org/wiki/Perl_6 points to leanpub.com/perl6regex/ which shows work in progress ??
lizmat ok, will fix
MasterDuke pmurias: btw, i just repeated the commands to build a native image for substratevm and it does create a lowercase org.perl6.nqp.truffle.svm.truffle6 21:30
lizmat moritz: fixed 21:32
moritz lizmat: thanks 21:33
lizmat: I fixed the publication year (it was published yesteryear, by a slim margin :D) 21:34
lizmat ah, ok :-)
moritz like, around Dec 22 or so 21:35
MasterDuke pmurias: interesting, the image (for that bench code) has a longer total running time, but has virtually no startup time 21:36
22:32 benjikun joined 22:35 pmurias left 23:20 ufobat_ left

Powered by Cro and the Raku® Programming Language.

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

Thank you!