samcv [TuxCM], have you ever tried running it with the FIFO priority and higher IO priority to see how much difference it makes? If the timings are pretty close, then it's likely other processes on the computer are not affecting it much. 01:08
You can live dangerously and just do `sudo chrt -f 1 ionice -c 2 -n 0 MYCOMMAND.raku` or if you want to do it safer, and run as your own user with only the capabilities needed (cap_sys_nice): 01:31
`_USER=$USER _HOME=$HOME sudo -E capsh --keep=1 --user=$USER --inh=cap_sys_nice --addamb=cap_sys_nice -- -c 'export USER=$_USER HOME=$_HOME; chrt -f 1 ionice -c 2 -n 0 whoami'` The variable exporting and setting is not needed for security reasons. `whoami` confirms it is the correct user, it's just sudo command tries to be helpful by setting environment variables, while capsh changes the user, but doesn't touch the environment 01:32
This sets the lowest FIFO (realtime) mode (on a scale of 1 to 100, one is the lowest). While it sets the highest io niceness within the normal io priority system. There is a realtime io priority as well, but i didn't use it, since cap_sys_nice does not seem to grant that right 01:34
02:19 lucasb left, frost-lab joined 03:19 linkable6 left, evalable6 left 03:22 linkable6 joined, evalable6 joined 03:59 leont left
Geth rakudo: vrurg++ created pull request #4096:
Implement invocation protocol for roles
04:03
05:53 b2gills left
Geth nqp: 779fc8e66c | Coke++ | 5 files
match 3-digits of test file
05:57
nqp: 70fac46e3d | Coke++ | t/nqp/117-strs.t
Add tests for nqp::x
[Coke] ok, I did the easy one. :) 05:58
zzz
06:28 squashable6 left 06:30 squashable6 joined 06:35 [Coke] left 06:37 [Coke] joined, [Coke] left, [Coke] joined 06:59 maggotbrain joined 07:00 sena_kun joined
[TuxCM] Rakudo v2020.11-109-g1ae04788c (v6.d) on MoarVM 2020.11-82-g6294b666f
csv-ip5xs0.853 - 0.953
csv-ip5xs-208.709 - 8.900
csv-parser25.044 - 25.078
csv-test-xs-200.395 - 0.438
test8.151 - 8.493
test-t1.919 - 2.190
test-t --race0.860 - 0.881
test-t-2034.077 - 35.231
test-t-20 --race9.561 - 9.662
07:36
samv I run "cpupower frequency-set -g performance", other suggestions welcome 07:38
sena_kun releasable6, status 07:49
releasable6 sena_kun, Next release in ā‰ˆ11 days and ā‰ˆ11 hours. 1 blocker. 0 out of 109 commits logged
sena_kun, Details: gist.github.com/3d94dbfeb403805b17...b90f075827
08:16 domidumont joined 08:58 domidumont left 09:00 domidumont joined 09:04 domidumont left 09:07 Altai-man joined 09:10 sena_kun left
lizmat Files=1346, Tests=117144, 223 wallclock secs (30.22 usr 8.56 sys + 3109.99 cusr 294.15 csys = 3442.92 CPU) 09:18
09:18 domidumont joined 09:26 patrickb joined
[TuxCM] Rakudo v2020.11-109-g1ae04788c (v6.d) on MoarVM 2020.11-82-g6294b666f
csv-ip5xs0.805 - 0.811
csv-ip5xs-208.007 - 8.272
csv-parser24.029 - 26.047
csv-test-xs-200.381 - 0.382
test7.405 - 7.752
test-t1.916 - 1.922
test-t --race0.811 - 0.815
test-t-2032.731 - 32.872
test-t-20 --race9.153 - 9.382
09:32
09:40 patrickb left 10:12 Xliff joined 11:16 Xliff left 11:17 [Tux] left
Geth rakudo: bd944e7cfe | (Elizabeth Mattijsen)++ | 2 files
Add support for native shaped1 slice assignment from native

This makes @a[@i] = @j on a shaped 1-dimensional native array about 15% faster, and a few % faster than on an unshaped native array. Difference caused by not having to check indices for < 0 values, as that is done internally for shaped arrays.
11:29
11:30 [Tux] joined 11:47 frost-lab left, [Tux] left 12:05 [Tux] joined 13:08 sena_kun joined 13:09 Altai-man left 13:12 MasterDuke left 13:25 leont joined 13:27 MasterDuke joined 13:40 b2gills joined 13:58 finsternis left 13:59 finsternis joined 14:10 cog joined
Geth nqp/fix-multithreaded-compilation-issues: 2af4e1cff0 | (Stefan Seifert)++ | src/QAST/Block.nqp
Fix parallel compilation occasionally losing frames

CUIDs must as their name suggest be unique within a computation unit. We rely on this when we use CUIDs as keys for lookup hashes for the static frames into which we compile blocks.
When multiple compilations (like EVAL) are run in parallel, it was possible ... (7 more lines)
15:45
nqp/fix-multithreaded-compilation-issues: 2c79b30814 | (Stefan Seifert)++ | src/how/NQPClassHOW.nqp
Remove long obsolete code

8 years ago in commit b6c88ffae3de22ead132d5751ab11517ebacd4dc the %caches lexical was turned into an attribute. Since now every instance of the meta class has its own caches hash, there was actually no longer a need to retain the nested hash keyed on the meta object's memory address. The cache method was adapted accordingly, but the obsolete initialization in add_method was kept needlessly. Lay it to rest, finally.
rakudo/rakuast: 11f489d814 | (Jonathan Worthington)++ | 8 files
Implement contextualizers in RakuAST
nqp/fix-multithreaded-compilation-issues: 4454625470 | (Stefan Seifert)++ | src/how/NQPClassHOW.nqp
Remove superfluous setting of debugtypename on mixins

The debug type name is alreay set in new_type which is called from generate_mixin. Cached mixins will keep their debug type name. There's no need to set it again. I just didn't realize this when trying to find all places where we were still missing a debug type name.
15:53
nqp/fix-multithreaded-compilation-issues: bf6bdf5538 | (Stefan Seifert)++ | src/how/NQPParametricRoleHOW.nqp
Fix "no such attribute" errors on mixin created by concurrent code

When multiple thread create the same mixin type at the same time, they will all create types, but only the winner's will make it into the mixin cache and will be used by all competing threads. However, NQPParametricRoleHOW's specialize method was not thread safe.
... (6 more lines)
nqp: niner++ created pull request #687:
Fix multithreaded compilation issues
16:00
nqp/fix-multithreaded-compilation-issues: abef8549c2 | (Stefan Seifert)++ | src/how/NQPParametricRoleHOW.nqp
Fix "no such attribute" errors on mixin created by concurrent code

When multiple thread create the same mixin type at the same time, they will all create types, but only the winner's will make it into the mixin cache and will be used by all competing threads. However, NQPParametricRoleHOW's specialize method was not thread safe.
... (6 more lines)
16:03
16:14 morayj joined
lizmat nine++ 16:16
jdv79 here i thought raku make multithreading easier;)
nine++ # maybe that fixed some familar sounding errors i've randomly seen 16:17
lizmat this is part of the torturing core devs :-)
leont nine++ # I know how difficult these issues are 16:29
Geth rakudo/rakuast: 104677bc78 | (Jonathan Worthington)++ | 5 files
Implement quoted method calls in RakuAST
16:35
16:53 patrickb joined 17:07 Altai-man joined 17:09 sena_kun left
Geth rakudo/rakuast: dd2ff3b270 | (Jonathan Worthington)++ | 5 files
Implement meta method calls in RakuAST
17:12
17:35 maggotbrain left 18:11 morayj left 18:13 morayj joined
Geth rakudo/rakuast: 225b07792a | (Jonathan Worthington)++ | 5 files
Start on RakuAST statement modifier support

This adds plumbing for them, along with if/unless condition modifiers and while/until loop modifiers. No AST tests or deparse support yet. The other modifiers are trickier, needing setting the topic or, potentially, a thunk.
18:20
18:47 domidumont left 18:57 sivoais left
Geth rakudo: f6c7ddb286 | (Elizabeth Mattijsen)++ | src/core.c/Rakudo/Iterator.pm6
Introduce Rakudo::Iterator.TailWith

This basically takes an iterator and a value, and will start producing the value indefinitely after the given iterator is exhausted.
19:41
19:42 morayj left
Geth rakudo: 2274aa85a0 | (Elizabeth Mattijsen)++ | 2 files
Fix issue with 1dim native shaped array initialization

Before this, @a = 1,2,3 would not initialize any elements that were not consistently set. So it would actually be an implicit splice into the array. This is now rectified.
19:44
roast: d21cb937c4 | (Elizabeth Mattijsen)++ | 3 files
Untodo now passing tests
19:48
19:52 leont left, leont joined 20:50 patrickb left 21:08 sena_kun joined 21:10 Altai-man left 22:09 sena_kun left
Geth roast: vrurg++ created pull request #709:
Make test conform to the new coercion semantics and role invocation
22:14
22:29 HarmtH joined 23:25 dogbert17 left 23:35 cog_ left 23:36 cog_ joined