00:27 librasteve_ left 00:33 Sgeo__ joined 00:36 Sgeo_ left 01:07 Aedil left
disbot6 <antononcube> @habere-et-disper Interesting question! I might have to research if for the post "Who has the better code highlighter now?" 01:31
02:10 oodani left 02:12 oodani joined 02:48 hulk joined 02:49 kylese left 03:15 hulk left, kylese joined 03:36 sibl joined 05:27 melezhik joined 05:53 Aedil joined, Aedil left 05:54 Aedil joined 06:52 Aedil left 07:45 abraxxa joined 07:52 abraxxa left, abraxxa1 joined 08:04 sourceable6 left, sourceable6 joined, itaipu left, Geth left, notable6 left, sourceable6 left, sourceable6 joined, linkable6 left, oodani left, greppable6 left, oodani joined 08:05 greppable6 joined, linkable6 joined, notable6 joined 08:06 abraxxa1 left 08:08 abraxxa joined 08:16 itaipu joined 08:24 sibl left 08:30 sibl joined 08:43 Sgeo__ left 08:51 dakkar joined 09:28 sibl left 09:35 sibl joined 09:48 sibl left 10:16 melezhik left 10:30 tejr left 10:46 sibl joined 11:02 sibl left 11:03 sibl joined 11:10 sibl left 12:12 Geth joined
Geth Papers/main: 06ed981d01 | (Elizabeth Mattijsen)++ | announcements/20251211.md
Announce John Haltiwanger as Raku Steering Council member
 12:13
CCR/main: 6905391c2c | (Elizabeth Mattijsen)++ | 5 files
Collect / Conserve remaining jnthn articles
 12:40
13:33 MoC joined
Geth CCR/main: 55e244351f | (Elizabeth Mattijsen)++ | 7 files
Remaster jnthn's last 4 blog posts
 13:34
13:36 Aedil joined
jdv arent there texas versions of everything? 14:41
14:43 Sgeo joined
jdv i guess sawyer might be onto something 14:43
[Coke] texas versions of almost everything. not sure it's 100% 15:42
lizmat jdv: context ? 16:15
ah, github.com/rakudo/rakudo/issues/6039 I assume? 16:17
jdv yes 16:28
16:30 tgt joined 16:32 tgt left 16:34 tgt joined, tgt left 17:12 melezhik joined 17:59 abraxxa left 18:15 dakkar left 18:16 melezhik_ joined 18:34 melezhik_ left, melezhik_ joined 18:52 melezhik_ left 18:58 melezhik_ joined 19:05 melezhik_ left 19:09 melezhik_ joined
[Coke] anyone seen jonathanstowe? 19:13
.seen jonathanstowe
seenbot missing?
lizmat perhaps ping them to be sure 19:14
19:33 melezhik_ left 19:42 melezhik left 20:03 tgt joined 20:28 tejr joined 20:31 Aedil left 20:37 tgt left 20:46 MoC left 20:57 Pixi__ left
disbot6 <librasteve> weekly: github.com/Raku-Steering-Council/P...06ed981d01 21:00
21:00 librasteve_ joined
librasteve_ weekly: github.com/Raku-Steering-Council/P...06ed981d01 21:00
notable6 librasteve_, Noted! (weekly)
librasteve_ ab5tract: ++ 21:03
[Coke] working on a remote windows 11 pro box - having trouble getting the powershell to have 'raku --version' emit correctly. 21:13
chcp 65001 not sufficient, trying stackoverflow.com/questions/400987...g-to-utf-8 - anyone else have a known good way to do this?
ah, refound it: $OutputEncoding = [Console]::InputEncoding = [Console]::OutputEncoding = New-Object System.Text.UTF8Encoding 21:15
disbot6 <librasteve> what about texas atomics? (<=== a fun sentence to write btw) 21:22
ugexe those already exist 21:31
disbot6 <librasteve> cant see them in the docs docs.raku.org/type/atomicint#Operators (happy to try and add them in) 21:32
ugexe atomic-fetch-add etc 21:33
disbot6 <librasteve> oh, i see ... was looking for texas infixes and so on 21:34
<librasteve> thanks! 21:35
21:36 Pixi joined 22:15 silug left 22:17 silug joined 22:19 silug left 22:20 silug joined 22:25 silug left 22:27 silug joined 22:30 silug left 22:35 silug joined
disbot6 <shalokshalom> This language performance comparison has a Raku implementation, and its currently the slowest language šŸ™ I am not a Raku developer, so maybe somebody skilled can look into it. github.com/niklas-heer/speed-compa...ibniz.raku 22:53
lizmat so, does this calculate the wallclock including startup? 22:59
if so, that's a battle that cannot be won
korvo I could try to contribute a slower language. They use Docker and Earthly, so I might be able to get a Monte image included; Monte's faster than CPython once loaded but takes a moment to start up. 23:02
Is Raku like E/Monte in running guards at runtime? If so, then removing the `int` and `num` annotations could speed things up at the cost of safety. 23:03
lizmat no, the native variants are faster...
korvo: so does the measurement include startup or not? 23:04
23:07 habere-et-disper joined
korvo lizmat: Yes, it does. They measure exec'ing `raku leibniz.raku` with the hyperfine tool (github.com/sharkdp/hyperfine) which does a few warmup iterations before taking the measurement. 23:07
lizmat but that would only warmup memory, not warmup Raku itself 23:08
korvo For Janet (a Lisp), they have both interpreted and compiled variants. The compiled variant has to set up Janet's package manager and do some bootstrapping, but the speed comparison is only based on final runtime. 23:10
lizmat OOC, what are typical values found in "rounds.txt" ? 23:13
korvo It is a checked-in file at src/rounds.txt. It currently contains "1000000000". 23:15
I'm reading about MoarVM right now. Is it possible to precompile bytecode? For SBCL (another Lisp), the Earthfile first prebuilds a FASL file (FASt Load IIRC, a mix of native code and structs) and benchmarks that file. I think that CPython gets a similar benefit since it would emit a .pyc file with compiled bytecode during warmup. 23:23
lizmat which version of Rakudo is being used ?
korvo They're using the `rakudo-star:latest` image from Docker Hub, I think. Updated 3 days ago. You'd know better than me which version that is. 23:24
lizmat ok, that's pretty decent :-) 23:25
on my machine the script with 1000000000 runs for 51.336165009 seconds excluding startup 23:26
with a small modification: 23:27
$x *= -1e0; -> $x = $x * -1e0;
and $pi += $w; -> $pi = $pi + $w;
the metaops *= and += don't handle natives directly yet, so that would cause upgrading to Num, which for this hot code, would be expensive 23:28
korvo Nice! 23:30
lizmat but yeah, that's about it that can be done to the code 23:31
without resorting to NQP 23:33
if I rewrite the inner loop with nqp, the script with 1000000000 runs for 25.673542532 seconds, so more than 2x as fast 23:35
$x = nqp::mul_n($x,-1e0);
$y = nqp::sub_i(nqp::mul_i(2,$i),1);
$w = nqp::div_n($x,$y);
$pi = nqp::add_n($pi,$w);
that's about the maximum that could possibly be obtained in Rakudo, I'd say... 23:36
I sorta suspect the reason for the difference is actually: $w = $x / $y I only need to replace that one by NQP to get the same results 23:38
I guess there's no candidate for num / int, so it takes the slow path there 23:39
will look at fixing this tomorrow, after some shuteye
&
23:56 habere-et-disper left

Powered by Cro and the RakuĀ® Programming Language.

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

Thank you!