🦋 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.
00:06 reportable6 left 00:09 reportable6 joined 01:09 linkable6 left, evalable6 left, linkable6 joined 01:10 linkable6 joined 01:11 evalable6 joined 03:29 MasterDuke joined 03:45 Kaipei joined 04:33 Xliff joined
Xliff Has anyone considered the possibility of building an in-memory cashe for ..precomp blobs? 04:34
04:49 codesections left 04:57 frost joined 05:44 jdv left 06:04 frost left 06:05 frost joined 06:07 reportable6 left 06:08 reportable6 joined 06:30 Xliff left 08:09 frost left
lizmat I thought "use" would actually mmap precomps? probably should ask on #moarvm 08:13
nine What would be the benefit of such a cache? We load (mmap) each precomp file once at most. 09:24
[Tux] Friday: 09:41
Rakudo v2022.07-87-gbc3b58241 (v6.d) on MoarVM 2022.07-20-g757524899
csv-ip5xs0.764 - 0.812
csv-ip5xs-205.011 - 5.066
csv-parser3.345 - 3.486
csv-test-xs-200.421 - 0.429
test6.554 - 6.667
test-t1.428 - 1.472
test-t --race0.837 - 0.876
test-t-2019.989 - 20.294
test-t-20 --race6.351 - 7.501
09:47 Kaipei left 09:50 sena_kun joined
[Tux] Rakudo v2022.07-90-g5e1f092f5 (v6.d) on MoarVM 2022.07-20-g757524899 09:51
10:16 frost joined
lizmat [Tux]: fwiw, I cannot reproduce that performance regression since Friday 10:25
10:53 Kaipei joined 12:06 reportable6 left 12:09 reportable6 joined 12:29 frost left 12:58 frost joined 13:23 frost left 13:31 sena_kun left 13:33 sena_kun joined 13:43 sena_kun left 13:45 sena_kun joined 14:09 frost joined 17:04 Xliff joined
Xliff P6-GLib suite timing statistics for Raku 2022.07.90.g.5.e.1.f.092.f.5 17:08
Total number of projects: 33 (502148 loc)
Total non-parallel compile times: 8672.000 (262.788 avg)
Total parallel compile times: 1307.643 (39.626 avg) 6.632x speedup
nine: I was thinking the cache would be persistent across executions of raku. 17:20
So mmap once, keep for as long as the daemon stays resident
However, now that I think about it, that might require a non-insignificant, and non-significant amount of change to moarvm. 17:21
Is it possible to check how long it takes to load a set of precomp files from disk, and then the times it takes to mmap them? 17:22
Should be several orders of magnitude. Multiply that by hours of raku use.
The time saved should be significant.
Xliff & 17:23
17:23 Xliff left 17:26 Kaipei left
nine Such a cache already exists. It's called "your operating system's page cache" 17:26
17:28 Kaipei joined 18:06 reportable6 left, reportable6 joined
Geth rakudo/rakuast: 88d3702379 | (Stefan Seifert)++ | src/Raku/ast/name.rakumod
RakuAST: support GLOBAL::{} lookups
rakudo/rakuast: 06b7d398b0 | (Stefan Seifert)++ | src/Raku/Actions.nqp
RakuAST: fix lookup of GLOBAL when using the new frontend

Need to bind the GLOBAL hll sym at the start of compilation as that's what GLOBAL is looked up from.
20:21 nebuchadnezzar joined
Geth ¦ problem-solving: rba assigned to lizmat Issue Wanted: new raku infrastructure jack-of-all-trades github.com/Raku/problem-solving/issues/352 20:48
¦ problem-solving: rba assigned to codesections Issue Wanted: new raku infrastructure jack-of-all-trades github.com/Raku/problem-solving/issues/352
¦ problem-solving: coke assigned to codesections Issue artistic license not a default on github github.com/Raku/problem-solving/issues/353 21:00
21:16 epony left 21:56 sena_kun left 22:03 MasterDuke left 22:04 epony joined 23:37 linkable6 left, linkable6 joined 23:49 Kaipei left 23:50 linkable6 left 23:51 linkable6 joined 23:56 linkable6 left 23:57 linkable6 joined