melezhik Hi devs! I've dropped a proposal for RIP - Rakudo Issues Player - www.reddit.com/r/rakulang/comments...es_player/ I'd appreciate any feedback. 04:55
this is what earlier was referred as automation testing system for known Rakudo issues ... cc [Coke]
lizmat Files=1346, Tests=117143, 229 wallclock secs (29.80 usr 8.28 sys + 3096.27 cusr 279.94 csys = 3414.29 CPU) 08:01
MasterDuke why do github.com/rakudo/rakudo/blob/mast...6#L96-L137 and github.com/rakudo/rakudo/blob/mast...#L618-L645 allocate BOOTHashes? 09:32
lizmat to catch any additional named parameters ? Like any method ? 09:39
FWIW, a few months ago I tried to write a pragma that would make the implicit *%_ in method signatures optional
MasterDuke oh yeah, the number of BOOTHash allocations from it exactly matches the number of calls to it 09:40
lizmat I forget why that didn't work out...probably something deep in the setting compilation depended on it
Geth rakudo: 65f24a8dbd | (Elizabeth Mattijsen)++ | 8 files
Change SET-SELF object init to use standard .new

There no longer is a performance reason to do so. However, in many places in the core, this is not as simple to do, as using the
  "is built" attribute trait there causes bootstrap issues.		 10:36
Geth rakudo: c02c9cdb85 | (Elizabeth Mattijsen)++ | src/core.c/native_array.pm6
Make array::(int|num|str)array dispatchable
 11:46
Geth rakudo: 46e9468feb | (Elizabeth Mattijsen)++ | 2 files
Make the native array code generation easier

Just run the file, dammit!		 12:12
rakudo: 82d31e0d43 | (Elizabeth Mattijsen)++ | 2 files
Make the native shaped array code generation easier

Just run the file, dammit!		 12:16
rakudo: 137d49b02b | (Elizabeth Mattijsen)++ | 2 files
Make the native shaped array postcircumfix generation easier

Just run the file, dammit!		 12:23
Geth rakudo: 53ad24af5d | (Elizabeth Mattijsen)++ | 2 files
Native shaped postcircumfixes do not need to be indented
 12:45
Geth rakudo: 0f4970da2f | (Elizabeth Mattijsen)++ | 2 files
Add missing candidats for 1dim shaped native arrays

  :kv :p :k :v were missing. Sadly there qre no tests to signal this
absence (yet, anyway).		 13:30
lizmat afk for a few hours& 13:31
travis-ci Rakudo build failed. Elizabeth Mattijsen 'Make array::(int|num|str)array dispatchable' 14:58
travis-ci.org/rakudo/rakudo/builds/747401818 github.com/rakudo/rakudo/compare/6...2c9cdb8588
lizmat hmmm.. wonder if that was a flapper in repl.t 15:07
Geth rakudo: 7c0956b81b | (Elizabeth Mattijsen)++ | 3 files
Remove postcircumfix[int] candidates

Turns out calling an [int] candidate with an native int **CURRENTLY** does not have any intlexref/intposref elimination, which in turn makes the code un-inlineable. So remove these candidates for now until we
  *do* have proper optimizations for this (hint: if you're looking for
some cool stuff to dig your teeth into, there's it!).
This makes @a[$i] with $i being a native int, about 2x as fast.		 16:35
Geth rakudo: 02e48d8127 | (Elizabeth Mattijsen)++ | 2 files
Add support for > 1 adverb to 1dim native shaped arrays

To make them work just like the unshaped ones. Specifically
  :exists:(kv|p) and :delete:(kv|p)		 17:59
Geth rakudo: 42fceb0855 | (Elizabeth Mattijsen)++ | 2 files
Add basic slice candidate to native 1dim shaped arrays

To make these type of accesses work directly without any intervening AT-POS logic		 19:04
rakudo: 2c5b545760 | (Elizabeth Mattijsen)++ | 2 files
Add basic Callable candidate to native 1dim shaped arrays

To make these type of accesses work directly without any intervening AT-POS logic		 19:11
lizmat well, that should make most accesses to native shaped 1dim arrays about 2x as fast as accessing unshaped native arrays
(before jnthns fixes and these fixes, they were about 8x as slow) 19:12
afk for a few hours&
[Tux] Rakudo v2020.11-91-g2c5b54576 (v6.d) on MoarVM 2020.11-54-ga595d9ddc
csv-ip5xs0.812 - 0.831
csv-ip5xs-208.122 - 8.769
csv-parser24.590 - 25.724
csv-test-xs-200.385 - 0.390
test7.832 - 8.238
test-t1.881 - 1.929
test-t --race0.830 - 0.849
test-t-2031.843 - 31.877
test-t-20 --race8.978 - 9.503
 20:26
timotimo oh wow! 2x as fast that's amazing! 22:52
i did not expect it to pass unshaped in performance this quickly
[Coke] nice. 22:56

Powered by Cro and the Raku® Programming Language.

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

Thank you!