<< 28-03-2025 >>
02:08:11*derpydoo joined #nim
02:43:30*ensyde joined #nim
02:48:04*alexdaguy joined #nim
04:46:50*alexdaguy quit (Ping timeout: 260 seconds)
05:08:31*ntat joined #nim
06:56:25*ntat quit (Quit: leaving)
07:07:49*ntat joined #nim
07:15:59*ensyde quit (Ping timeout: 260 seconds)
07:17:48*ensyde joined #nim
08:21:34*ntat quit (Quit: leaving)
10:05:56*derpydoo quit (Quit: derpydoo)
11:15:01*ensyde quit (Ping timeout: 248 seconds)
11:36:05*ntat joined #nim
12:50:06FromDiscord<madonuko> sent a code paste, see https://paste.rs/1VltD
12:52:04FromDiscord<madonuko> wait, forgot to -d:release
12:53:15FromDiscord<madonuko> uh, still pretty slow even with -d:danger
12:53:16*ntat quit (Quit: leaving)
13:11:34FromDiscord<madonuko> so uhh I'm not sure if this is a good idea but I made it run multithreaded
13:11:55FromDiscord<madonuko> sent a code paste, see https://paste.rs/MvSyW
13:18:34FromDiscord<z1r343l2> What's syncio
13:18:44FromDiscord<janakali> try `seq[string]` instead of HashSet and then `lines.deduplicate()`, hashing every line is probably a lot slower
13:19:24FromDiscord<madonuko> In reply to @janakali "try `seq[string]` instead of": need to also sort
13:19:45FromDiscord<lainlaylie> what does the profiler say
13:20:09FromDiscord<madonuko> uh, I don't actually know how to use a profiler
13:21:02FromDiscord<lainlaylie> good opportunity to learn
13:21:15FromDiscord<lainlaylie> In reply to @z1r343l2 "What's syncio": https://nim-lang.github.io/Nim/syncio.html
13:21:42FromDiscord<lainlaylie> automatically imported by system module unless nimPreviewSlimSystem
13:22:06FromDiscord<madonuko> well my program kinda doesn't finish right now https://media.discordapp.net/attachments/371759389889003532/1355170082036121620/image.png?ex=67e7f3fd&is=67e6a27d&hm=19fee71a9bd1d45408e1215805b4733fede9a246c53082b79a8282945b1fae33&
13:52:40FromDiscord<lainlaylie> basically compile with --debugger:native and then use your typical c profiling tools
13:55:01FromDiscord<tapeda> Git Treeform/hotty is a convenient way too
14:15:37*coldfeet joined #nim
14:30:08*ntat joined #nim
15:16:48*xutaxkamay_ joined #nim
15:17:29*xutaxkamay quit (Read error: Connection reset by peer)
15:17:45*xutaxkamay_ is now known as xutaxkamay
15:46:59FromDiscord<__nycto__> I was doing some profiling and found that iterating over a `seq` in a `Table` was forcing the hash of the key to get double calculated, and creating a copy of the `seq` to do the iteration. I had to do some privateAccess and import shenanigans to fix it. But I'm hoping there is a better way that I just didn't see. Any ideas about how to improve this? https://github.com/Nycto/AHGrid/blob/main/src/private/util.nim
15:58:13FromDiscord<lainlaylie> sent a code paste, see https://paste.rs/Jnl4B
16:20:43FromDiscord<__nycto__> sent a code paste, see https://paste.rs/BkXMj
16:21:53FromDiscord<__nycto__> sent a code paste, see https://paste.rs/MhNAF
16:46:37FromDiscord<maxb0tbeep> is there a good way to store data in a similar way to a java 2d array?
16:47:42FromDiscord<maxb0tbeep> like `int[][] myArray = new int[3][3];` to make a 3x3 array
16:47:47FromDiscord<maxb0tbeep> is there an equivalent?
16:48:30FromDiscord<__nycto__> seq[seq[int]] is one way
16:48:59FromDiscord<__nycto__> array[3; array[3; int]] is another
16:49:39FromDiscord<__nycto__> If you’re doing matrix operations, though, you should use a lib like vmath
16:50:59FromDiscord<maxb0tbeep> alright, thanks
17:37:39*coldfeet quit (Quit: Lost terminal)
17:46:32FromDiscord<Elegantbeef> sent a code paste, see https://paste.rs/FRz22
18:51:13FromDiscord<threefour> I know there is an `{.inline.}` calling convention, but is there a way to inline at the call location, instead of for every call of a function?
19:00:57Amun-Ra"inline at call location"?
19:28:25*GnuYawk quit (Quit: The Lounge - https://thelounge.chat)
19:28:45*GnuYawk joined #nim
19:41:39FromDiscord<heysokam> In reply to @threefour "I know there is": nope. the syntax wouldn't have room to have it either
19:42:06FromDiscord<heysokam> you'd need to create a template or macro for it
19:59:31FromDiscord<.bobbbob> that's basically what a template is isnt it?
20:00:32FromDiscord<.bobbbob> or you could have the "function" be a template that gets inlined but then have a normal function thats a wrapper for the template if you want that
20:00:38FromDiscord<.bobbbob> sounds like premature optimization
20:01:55FromDiscord<threefour> Nah, just a theoretical question because Zig has that. But on second thought, Zig does comptime differently to where that's actually useful.
20:22:23FromDiscord<heysokam> In reply to @.bobbbob "that's basically what a": A template -always- inlines. he said he wants to inline only sometimes
20:23:03FromDiscord<Elegantbeef> The default convention inlines sometime
20:23:03FromDiscord<Elegantbeef> What does that even mean
20:23:13FromDiscord<heysokam> In reply to @threefour "Nah, just a theoretical": zig has `inline for`, but I'm not sure if it has inline function calls
20:23:45FromDiscord<heysokam> (edit) "In reply to @threefour "Nah, just a theoretical": zig has `inline for`, but I'm not sure if it has ... inline" added "in-place"
20:23:53FromDiscord<Elegantbeef> Isn't that loop unrolling?
20:23:57FromDiscord<heysokam> yea
20:52:06*ntat quit (Quit: leaving)
21:00:31FromDiscord<.bobbbob> In reply to @heysokam "A template -always- inlines.": I mean you put the code in the template, and when you want to inline it you "call" the template, but then you have a wrapper proc that just calls the template in it, so when you dont want to inline you call that
21:03:14FromDiscord<.bobbbob> but then gcc might deduplicate your code? 👀
21:05:18FromDiscord<.bobbbob> actually does the nim .inline. pragma use C inline? because im under the impression inline isnt guarateed to inline anyways and you have to use attribute((always_inline)) to do that
21:27:31*zgasma joined #nim
22:04:00FromDiscord<leorize> it does use C's inline
22:04:37FromDiscord<leorize> it does, see\: https://ziglang.org/documentation/master/#toc-call↵(@heysokam)
22:14:05FromDiscord<heysokam> oh, didn't know that `@call` can take an inline parameter. ty leorize, good to know
22:30:24FromDiscord<__nycto__> sent a code paste, see https://paste.rs/ws4Ea
22:33:16FromDiscord<Elegantbeef> sent a code paste, see https://paste.rs/I03nE
22:33:43FromDiscord<Elegantbeef> That should be identical to your solution
22:33:51FromDiscord<Elegantbeef> Though you need to do `entry[]` I guess
22:37:09FromDiscord<__nycto__> sent a code paste, see https://paste.rs/Vaq23
22:37:35FromDiscord<Elegantbeef> `.addr[]` 😛
22:37:47FromDiscord<Elegantbeef> What's a little cheekiness between friends
22:38:51FromDiscord<__nycto__> Ding ding, we have a winner 😄
22:39:02FromDiscord<__nycto__> `==22902== Collected : 843149179`
22:39:27FromDiscord<__nycto__> Just gotta go take a quick shower
22:39:41FromDiscord<__nycto__> (edit) "Just gotta go take a quick shower ... " added "after checking that one in"
22:41:11FromDiscord<Elegantbeef> The ones that really win are the ones that don't have to see the code
22:41:35FromDiscord<Elegantbeef> I argue it's less sick then using `importutils`
22:41:41FromDiscord<__nycto__> Totally agreed
22:41:46FromDiscord<Elegantbeef> That module is a silly billy that should arguably never be used
22:42:03FromDiscord<__nycto__> If I get the time, I may submit a PR to add a new version of `withValue`
22:42:41FromDiscord<Elegantbeef> I think the thing is that it returns `ptr T` so it's never wise use it implicitly with non mutable
22:44:08FromDiscord<__nycto__> Interestingly, it's also just a _tooooch_ slower than the `importutils` version -- a difference of about 10,000,000 instructions in my benchmark code. Which is still pretty good
22:44:09FromDiscord<Elegantbeef> Nim needs borrowing and `let T` types pronto 😄
22:44:46FromDiscord<Elegantbeef> Does your compiler have LTO?
22:45:24FromDiscord<Elegantbeef> Oh your code is mildly different
22:45:42FromDiscord<Elegantbeef> You did `>` and it does `>=` so perhaps that's the culprit, or just the fact it's taking an address
22:45:59FromDiscord<__nycto__> I'm using GCC, so it should
23:33:56*pbsds355 joined #nim
23:35:24*pbsds35 quit (Ping timeout: 252 seconds)
23:35:25*pbsds355 is now known as pbsds35