2026/02/14

Newest at the top

2026-02-14 22:00:22 +0100 <[exa]> probie: you have unboxed or primitive vectors?
2026-02-14 22:00:09 +0100L29Ah(~L29Ah@wikipedia/L29Ah) (Ping timeout: 260 seconds)
2026-02-14 21:59:26 +0100 <probie> https://paste.tomsmeding.com/NRYKh5Fj
2026-02-14 21:59:21 +0100 <[exa]> that looks like a lot of indirection
2026-02-14 21:58:23 +0100caubert(~caubert@user/caubert) (Ping timeout: 252 seconds)
2026-02-14 21:58:11 +0100 <tomsmeding> why are there so many loads for only two stores? I assume this is different code than you posted originally?
2026-02-14 21:57:03 +0100 <probie> I don't think it's LLVM's problem here; GHC is just not generating good code https://paste.tomsmeding.com/8ZYY5Pka
2026-02-14 21:57:02 +0100merijn(~merijn@host-cl.cgnat-g.v4.dfn.nl) (Ping timeout: 256 seconds)
2026-02-14 21:56:55 +0100 <[exa]> probie: btw try to unroll the loop manually, that might give llvm enough decisive force
2026-02-14 21:56:50 +0100 <tomsmeding> very recently IIRC some of them started working on NCG too
2026-02-14 21:56:37 +0100 <tomsmeding> GHC has SIMD primops, but they only work with LLVM
2026-02-14 21:56:12 +0100 <[exa]> probie: I find it better than relying on the compiler accidentaly noticing that I want SIMD (but yeah it's still :( )
2026-02-14 21:54:32 +0100machinedgod(~machinedg@d75-159-126-101.abhsia.telus.net) (Ping timeout: 252 seconds)
2026-02-14 21:53:21 +0100caubert(~caubert@user/caubert) caubert
2026-02-14 21:52:13 +0100 <probie> [exa]: Because I shouldn't need to
2026-02-14 21:49:55 +0100merijn(~merijn@host-cl.cgnat-g.v4.dfn.nl) merijn
2026-02-14 21:45:43 +0100 <[exa]> ok not bad :)
2026-02-14 21:45:36 +0100peterbecich(~Thunderbi@71.84.33.135) peterbecich
2026-02-14 21:44:27 +0100 <tomsmeding> (never used it)
2026-02-14 21:44:08 +0100 <tomsmeding> [exa]: https://gitlab.com/Gusten_Isfeldt/futhask
2026-02-14 21:43:46 +0100Pixi(~Pixi@user/pixi) Pixi
2026-02-14 21:43:41 +0100 <[exa]> (man, can we FFI to futhark?)
2026-02-14 21:43:03 +0100 <[exa]> probie: btw why not make a small FFI to a relatively portable C?
2026-02-14 21:38:35 +0100Pixi(~Pixi@user/pixi) (Quit: Leaving)
2026-02-14 21:38:17 +0100caubert(~caubert@user/caubert) (Ping timeout: 250 seconds)
2026-02-14 21:37:41 +0100merijn(~merijn@host-cl.cgnat-g.v4.dfn.nl) (Ping timeout: 244 seconds)
2026-02-14 21:34:44 +0100 <tomsmeding> possibly, yes
2026-02-14 21:33:37 +0100peterbecich(~Thunderbi@71.84.33.135) (Ping timeout: 264 seconds)
2026-02-14 21:32:46 +0100merijn(~merijn@host-cl.cgnat-g.v4.dfn.nl) merijn
2026-02-14 21:29:31 +0100pavonia(~user@user/siracusa) siracusa
2026-02-14 21:29:02 +0100 <probie> I wonder if it's the use of `read` instead of `unsafeRead`?
2026-02-14 21:28:52 +0100 <probie> It's not giving me SIMD instruction :'(
2026-02-14 21:26:38 +0100ouilemur(~jgmerritt@user/ouilemur) ouilemur
2026-02-14 21:25:54 +0100 <geekosaur> also I mentioned the settings file bvecause I saw a claim in backscroll that the correct llvm version wasn't on their PATH, which means a settings file edit to point to the correct one
2026-02-14 21:23:31 +0100 <geekosaur> there was a point where `opt` parameters changed and ghc didn't know how tgo call newer versions correctly
2026-02-14 21:23:07 +0100 <geekosaur> they can be pretty critical but I don't think it's critical for recent versions
2026-02-14 21:22:08 +0100 <int-e> Hmm. For me, the baked in commands (from the settings file) have a version, e.g. llc-14 for ghc-9.10.3 and llc-19 for ghc-9.12.2. I wonder what the binary distributions put there... ghc --info | grep LLVM will show that info without you having to go looking for the settings file.
2026-02-14 21:21:49 +0100merijn(~merijn@host-cl.cgnat-g.v4.dfn.nl) (Ping timeout: 250 seconds)
2026-02-14 21:18:38 +0100 <int-e> I don't
2026-02-14 21:18:17 +0100 <tomsmeding> do you happen to know how critical those version upper bounds are, in GHC's LLVM support?
2026-02-14 21:18:07 +0100s3np41(~s3np41@078088254000.unknown.vectranet.pl)
2026-02-14 21:17:56 +0100 <int-e> (Rather than digging into the history I'll just assume this hasn't changed recently except for bumping versions.)
2026-02-14 21:17:44 +0100 <tomsmeding> I see
2026-02-14 21:16:55 +0100merijn(~merijn@host-cl.cgnat-g.v4.dfn.nl) merijn
2026-02-14 21:16:26 +0100 <int-e> tomsmeding: https://gitlab.haskell.org/ghc/ghc/-/blob/master/compiler/GHC/SysTools/Tasks.hs#L341-350 (and beyond)
2026-02-14 21:15:32 +0100 <geekosaur> "LLVM opt command" (and similar for llc)
2026-02-14 21:15:14 +0100 <geekosaur> $(ghc --print-libdir)/settings
2026-02-14 21:15:13 +0100caubert(~caubert@user/caubert) caubert
2026-02-14 21:14:58 +0100 <geekosaur> I think you can specify a path in the settings file?
2026-02-14 21:14:18 +0100 <tomsmeding> lame