|
This channel is intended for people just starting with the Raku Programming Language (raku.org). Logs are available at irclogs.raku.org/raku-beginner/live.html Set by lizmat on 8 June 2022. |
|||
|
00:49
jgaz joined
00:56
jgaz left
01:29
jgaz joined
01:30
jgaz left
01:31
kjp left
01:32
jgaz joined,
jgaz left
01:33
jgaz joined
01:34
jgaz left
01:47
jgaz joined,
jgaz left
01:49
jgaz joined,
jgaz left
04:26
kjp joined
05:05
Tirifto left
05:06
Tirifto joined
05:09
Tirifto left
05:10
Tirifto joined
05:17
hythm joined
05:22
hythm left
|
|||
| .vushu | Can somebody explain that some times this code unit module Benchmark; sub test is export { my int $n; for 0..^1000000000 { $n +=1; } say $n; } use Benchmark; sub MAIN { test; } Running it more times: 1st run: real 1m57.742s user 2m47.524s sys 0m0.586s second run: real 0m11.996s user 0m11.798s sys 0m0.049s third run: ` | 05:51 | |
| @librasteve Ireal 0m12.376s user 0m12.186s sys 0m0.037s | 06:02 | ||
| @librasteve I'm guessing that whenever the jit decides to kicks in it's getting slower as you explained yesterday? | 06:09 | ||
| ab5tract_ | I don’t see a speed difference with `my int $n; for ^100_000_000 { $n +=1 }`. Takes *50 seconds* on an M2 Pro… some optimization must be missing | 08:08 | |
| *don’t see a speed difference between have the `say $n` at the end | |||
|
08:46
dakkar joined
08:54
lizmat joined
11:44
teatwo joined,
teatwo left
11:45
teatwo joined
11:47
tea3po left
|
|||
| lizmat | And yet another Rakudo Weekly News hits the Net: rakudoweekly.blog/2023/11/06/2023-45-confvids/ | 13:05 | |
| rcmlz | time raku -e 'my Int $n; for ^100_000_000 {$n += 1}; say $n' 8,94s user time raku -e 'my int $n; for ^100_000_000 {$n += 1}; say $n' 2,14s user Interesting, "int" and "Int" realy does make a difference in this example (Rakudo™ v2023.10) | 13:16 | |
| Just for a laugh I also tried C++ and Rust with max. optimization switched on (c++ -O3 count_to.cpp and rustc -O count_to.rs) and booth seem to unroll and hardcode the result as it is always 0 seconds, even for very large upper bounds. It is just an artificial example .... | 13:21 | ||
| using raku --optimize=3 ... to trigger optimizations does not seem to have an impact in this specific example. Interesting. | 13:27 | ||
| librasteve | o/ | 13:40 | |
| I also timed raku with int (2.76s) vs perl (2.01s) so raku is 37% slower than perl when you compare apples vs apples - nice! | 13:47 | ||
| (raku still has a bunch of optimisation opportunities going forward + concurrent processing whereas perl is prtty much maxed out by now) | 13:49 | ||
| rcmlz: ah yes, static compilation and loop unrolling ... | 13:52 | ||
| ab5stract: I suspect (hope!) that your 50s on an M2 is for 1_000_000_000 and not 100_000_000 that you said ;-) | 13:53 | ||
| otherwise my M1 is 5x faster than your M2 | 13:54 | ||
|
17:05
jgaz joined
|
|||
| ab5tract_ | librasteve: Indeed, I was running it with RAKUDO_RAKUAST=1 | 17:29 | |
| 1.98s | |||
| And I also see the speed difference between having the `say $n` and not. Weird! | 17:35 | ||
| Actually, scratch that. Further testing revealed a similar baseline for both | 17:39 | ||
|
17:40
dakkar left
18:24
lizmat joined
19:13
ab5tract__ joined
19:18
ab5tract_ left,
ab5tract__ is now known as ab5tract_,
KOTP left
19:19
KOTP joined
23:45
tea3po joined
23:47
tea3po left,
tea3po joined
23:48
teatwo left
23:53
tea3po left
23:54
tea3po joined
|
|||