|
🦋 Welcome to the IRC channel of the core developers of the Raku Programming Language (raku.org #rakulang). This channel is logged for the purpose of history keeping about its development | evalbot usage: 'm: say 3;' or /msg camelia m: ... | Logs available at irclogs.raku.org/raku-dev/live.html | For MoarVM see #moarvm Set by lizmat on 8 June 2022. |
|||
| [Coke] | why is it .add_phaser? | 00:57 | |
| should it be changed to add-phaser at some point? | |||
|
01:01
Voldenet left
01:06
Voldenet joined
05:02
vrurg_ joined
05:03
vrurg left
05:24
vrurg_ left
05:25
vrurg joined
08:19
sena_kun joined
08:34
sena_kun left
08:43
finanalyst joined
09:01
kjp_ joined,
RakuIRCLogger left,
kjp left
|
|||
| lizmat | [Coke]: that touches on a more general thing: in the Raku grammar I have tried to use kebab-case where possible. However, some of it goes back to NQP | 09:32 | |
| and thus back to snake_case | |||
| also add_phaser is not tested anywhere, and could thus be considered to be an implementation-detail | 09:34 | ||
| so I've run into something that's really puzzling, at least to me: | 10:44 | ||
| m: use Telemetry; say (^Inf).grep(*.is-prime).skip(999999).head; say T<cpu-user>/1000000 | |||
| camelia | 15485863 5.405457 |
||
| lizmat | iow, give me the millionth prime number | 10:45 | |
| in the single threaded case, this takes 5.4 CPU seconds | |||
| if we hyper that: | |||
| m: use Telemetry; say (^Inf).hyper.grep(*.is-prime).skip(999999).head; say T<cpu-user>/1000000 | 10:46 | ||
| camelia | 15485863 21.335601 |
||
| lizmat | it takes quite a few CPU seconds more | 10:47 | |
| however, if we set a bigger batch size (larger than the 64 by default): | |||
| m: use Telemetry; say (^Inf).hyper(batch => 1024).grep(*.is-prime).skip(999999).head; say T<cpu-user>/1000000 | |||
| camelia | 15485863 3.584926 |
||
| lizmat | it takes a lot less CPU! | ||
| but the weird thing (to me at least) is that it takes *less* CPU than the single threaded case! | 10:48 | ||
| m: say 3.584926 / 5.405457 | 10:49 | ||
| camelia | 0.66320498 | ||
| lizmat | on my M1 the difference is even more pronounced 0.52 | 10:50 | |
| so what causes this difference? | 10:55 | ||
| m: use Telemetry; use nqp; my $i = -1; nqp::while(++$i < 15485863, nqp::isprime_I(nqp::unbox_i($i))); say T<cpu-user>/1000000 | 11:07 | ||
| camelia | 1.960859 | ||
| lizmat | so that would be the NQP minimum CPU usage | ||
| Geth | rakudo/main: 097b50cd0d | (Elizabeth Mattijsen)++ | t/12-rakuast/fixed-in-rakuast.rakutest Properly mark failing test as todo Apparently "make test" files are not fudged |
11:34 | |
|
12:50
lizmat left
13:04
lizmat joined
|
|||
| Geth | rakudo/main: 2e4bc48a1b | (Elizabeth Mattijsen)++ | src/core.c/Any-iterable-methods.rakumod Micro-optimize basic iterators .push-all / sink-all specifically, by creating lexical shortcuts to static attributes, so that these can be lowered into locals in the hot code paths. Makes: for ^10000 { (^1000).map(* + 1).iterator.push-all(IterationBuffer.CREATE) } about 1.5% faster |
13:13 | |
| lizmat | looks like the problem is in grep: | 13:47 | |
| m: use Telemetry; say (^Inf).map({ $_ if .is-prime}).skip(999999).head; say T<cpu-user>/1000000 | 13:48 | ||
| camelia | 15485863 2.240862 |
||
| lizmat | m: use Telemetry; say (^Inf).grep(*.is-prime).skip(999999).head; say T<cpu-user>/1000000 | ||
| camelia | 15485863 5.768551 |
||
| lizmat | consequently with the proper hyper: | 13:49 | |
| m: use Telemetry; say (^Inf).hyper(batch => 2048).map({ $_ if .is-prime}).skip(999999).head; say T<cpu-user>/1000000 | 13:50 | ||
| camelia | 15485863 3.76585 |
||
| Geth | rakudo/main: 72b46fe292 | (Elizabeth Mattijsen)++ | src/core.c/Any-iterable-methods.rakumod Some more lexical shortcuts These were missed in 2e4bc48a1b |
14:04 | |
| lizmat refrains from being drawn into the grep vortex for now | |||
| [Coke] | (add_phaser not tested) so not spec, so not part of docs. Excellent :) | 14:10 | |
| lizmat | greppable6: add_phaser | ||
| greppable6 | lizmat, 11 lines, 5 modules: gist.github.com/1eec5b78ac2819b892...943af07936 | ||
| lizmat | mostly the usual suspects :-) | 14:12 | |
| [Coke] | ... weird, I no longer see the add_phaser ticket. :| | 14:15 | |
|
14:27
vrurg_ joined,
vrurg left
15:31
nine left,
nine joined
17:25
sena_kun joined
|
|||
| Geth | rakudo/main: 6846c49ed1 | (Elizabeth Mattijsen)++ | src/core.c/Any-iterable-methods.rakumod Adding some more lexical shortcuts In the grep hierarchy of push-alls |
18:36 | |
| rakudo/r1378: 16981ad9ba | ab5tract++ | 2 files Provide better feedback for ambiguous adverb usage This fixes an LTA situation: > my %h = :1k; say 1 ~ %h<k>:exists ~ 1 ===SORRY!=== MVMArray: Can't pop from an empty array ... (10 more lines) |
19:14 | ||
| nqp/r1378: 25362cf803 | ab5tract++ | src/HLL/Grammar.nqp Provide better feedback for ambiguous adverb usage This fixes an LTA situation: > my %h = :1k; say 1 ~ %h<k>:exists ~ 1 ===SORRY!=== MVMArray: Can't pop from an empty array ... (9 more lines) |
19:20 | ||
| nqp/r1378: 3e1c47538a | ab5tract++ | src/HLL/Grammar.nqp Provide better feedback for ambiguous adverb usage This fixes an LTA situation: > my %h = :1k; say 1 ~ %h<k>:exists ~ 1 ===SORRY!=== MVMArray: Can't pop from an empty array ... (9 more lines) |
19:25 | ||
| nqp/r1378: 8dba195cb8 | ab5tract++ | src/HLL/Grammar.nqp Provide better feedback for ambiguous adverb usage This fixes an LTA situation: > my %h = :1k; say 1 ~ %h<k>:exists ~ 1 ===SORRY!=== MVMArray: Can't pop from an empty array ... (9 more lines) |
19:26 | ||
| [Coke] | ab5tract: "nowehere" | ||
| Geth | rakudo/r1378: 3c5988198a | ab5tract++ | 2 files Provide better feedback for ambiguous adverb usage This fixes an LTA situation: > my %h = :1k; say 1 ~ %h<k>:exists ~ 1 ===SORRY!=== MVMArray: Can't pop from an empty array ... (10 more lines) |
19:28 | |
| ab5tract | [Coke]: thanks! | ||
| [Coke] | ab5tract++ | 19:31 | |
| Geth | nqp/main: fa489fe301 | ab5tract++ (committed using GitHub Web editor) | src/HLL/Grammar.nqp Provide better feedback for ambiguous adverb usage This fixes an LTA situation: > my %h = :1k; say 1 ~ %h<k>:exists ~ 1 ===SORRY!=== MVMArray: Can't pop from an empty array ... (9 more lines) |
19:50 | |
| rakudo/main: 3ce05fa0db | ab5tract++ (committed using GitHub Web editor) | 2 files Provide better feedback for ambiguous adverb usage This fixes an LTA situation: > my %h = :1k; say 1 ~ %h<k>:exists ~ 1 ===SORRY!=== MVMArray: Can't pop from an empty array ... (10 more lines) |
19:51 | ||
| rakudo/main: 3a928a5616 | ab5tract++ | t/12-rakuast/fixed-in-rakuast.rakutest Test for ambiguous adverb exception This is only properly testable in RakuAST, as base dies in NQP's HLL::Grammar without an exception class to verify against (#1378). We could test against the text of the die message, but my experience has led me to consider that bad practice and the situation in base is only temporary anyway. |
20:15 | ||
|
21:24
finanalyst left
|
|||
| Geth | rakudo/main: a26b756a50 | ab5tract++ | t/12-rakuast/fixed-in-rakuast.rakutest Tests indices for shaped arrays automatically coerced to Int This is to keep R#1297 (#1297) from regressing in RakuAST. |
21:43 | |
| linkable6 | R#1297 [open]: github.com/rakudo/rakudo/issues/1297 [shaped][Fixed in RakuAST] Indices for shaped arrays need manually coerced to numbers | ||
| Geth | rakudo: ab5tract++ created pull request #5576: Give META_REDUCE_RIGHT behavior for 1-arg case |
21:59 | |
|
22:00
sena_kun left
|
|||
| leont | I haven't looked into it yet, but I'm guessing this Perl issue is also relevant for Raku quoting constructs github.com/Perl/perl5/issues/22228 | 22:55 | |
| [Coke] | leont++ | 22:58 | |
| ab5tract | not just Perl. macOS won't even successfully highlight the `$weird` example | 23:06 | |