Newest at the top
| 2026-01-19 08:39:53 +0100 | tromp | (~textual@2001:1c00:3487:1b00:f96f:f7c1:9b58:4be8) |
| 2026-01-19 08:39:22 +0100 | driib3180 | (~driib@vmi931078.contaboserver.net) (Ping timeout: 255 seconds) |
| 2026-01-19 08:34:58 +0100 | merijn | (~merijn@host-cl.cgnat-g.v4.dfn.nl) (Ping timeout: 256 seconds) |
| 2026-01-19 08:30:14 +0100 | FANTOM | (~fantom@87.75.185.177) |
| 2026-01-19 08:29:57 +0100 | merijn | (~merijn@host-cl.cgnat-g.v4.dfn.nl) merijn |
| 2026-01-19 08:24:43 +0100 | olivial | (~benjaminl@user/benjaminl) benjaminl |
| 2026-01-19 08:24:26 +0100 | olivial | (~benjaminl@user/benjaminl) (Read error: Connection reset by peer) |
| 2026-01-19 08:21:51 +0100 | Square2 | (~Square4@user/square) Square |
| 2026-01-19 08:21:37 +0100 | Square | (~Square@user/square) (Ping timeout: 246 seconds) |
| 2026-01-19 08:20:12 +0100 | driib3180 | (~driib@vmi931078.contaboserver.net) driib |
| 2026-01-19 08:19:13 +0100 | merijn | (~merijn@host-cl.cgnat-g.v4.dfn.nl) (Ping timeout: 250 seconds) |
| 2026-01-19 08:17:21 +0100 | driib3180 | (~driib@vmi931078.contaboserver.net) (Quit: Ping timeout (120 seconds)) |
| 2026-01-19 08:14:09 +0100 | merijn | (~merijn@host-cl.cgnat-g.v4.dfn.nl) merijn |
| 2026-01-19 08:11:58 +0100 | notzmv | (~umar@user/notzmv) notzmv |
| 2026-01-19 08:03:15 +0100 | merijn | (~merijn@host-cl.cgnat-g.v4.dfn.nl) (Ping timeout: 245 seconds) |
| 2026-01-19 07:59:16 +0100 | vanishingideal | (~vanishing@user/vanishingideal) (Ping timeout: 256 seconds) |
| 2026-01-19 07:58:22 +0100 | merijn | (~merijn@host-cl.cgnat-g.v4.dfn.nl) merijn |
| 2026-01-19 07:48:51 +0100 | <Guest70> | Anybody used Accelerate lately, and had good/bad experiences? |
| 2026-01-19 07:47:23 +0100 | merijn | (~merijn@host-cl.cgnat-g.v4.dfn.nl) (Ping timeout: 256 seconds) |
| 2026-01-19 07:45:53 +0100 | haritz | (~hrtz@user/haritz) (Quit: ZNC 1.8.2+deb3.1+deb12u1 - https://znc.in) |
| 2026-01-19 07:42:08 +0100 | <Guest70> | Yes I think so |
| 2026-01-19 07:42:04 +0100 | <Guest70> | I.e. you wouldn't write a "whole program" in Futhark. So "embedded" in the sense that you might have OpenGL code in your Haskell program |
| 2026-01-19 07:41:46 +0100 | <int-e> | Sure. That makes it like OpenCL. Or web assembly. |
| 2026-01-19 07:41:44 +0100 | <Axman6> | Leary: Thank you, I'll take a look |
| 2026-01-19 07:41:21 +0100 | <Guest70> | int-e: I understand, but I mean that Futhark code is designed to exist within a larger program, written in another language |
| 2026-01-19 07:40:56 +0100 | <int-e> | And at least at a glance, Futhark is not 'embedded' in this sense. |
| 2026-01-19 07:40:33 +0100 | <Guest70> | I should say that I don't want truly "magic" (take a normal haskell program and run it on the GPU), but I'd like to be able to hand off big pieces of work to the GPU |
| 2026-01-19 07:40:27 +0100 | <int-e> | Guest70: the 'embedded' in 'EDSL' means it's embedded into a programming language (like Haskell), rather than having its own syntax, with lexer and parser. |
| 2026-01-19 07:40:20 +0100 | merijn | (~merijn@host-cl.cgnat-g.v4.dfn.nl) merijn |
| 2026-01-19 07:39:30 +0100 | <Guest70> | It hands off a lot to OpenCL though |
| 2026-01-19 07:39:30 +0100 | <int-e> | AFAIK and AFAICS graph reduction and the dynamic allocation that it entails is just not GPU friendly; the computations aren't nearly uniform enough. |
| 2026-01-19 07:39:25 +0100 | <Axman6> | I wouldn't call Accelerate low level |
| 2026-01-19 07:39:16 +0100 | <Guest70> | Yes, I think Futhark is still meant to be embedded into your program but not the language |
| 2026-01-19 07:38:49 +0100 | <Guest70> | Maybe they're similar levels of abstraction? |
| 2026-01-19 07:38:42 +0100 | <int-e> | Futhark, hmm. Similar but it drops the 'E' from 'EDSL'? |
| 2026-01-19 07:37:40 +0100 | <Guest70> | Accelerate is very low-level, right? Was hoping for a bit of magic parallelism like futhark |
| 2026-01-19 07:36:34 +0100 | <int-e> | I mean there are things like https://hackage.haskell.org/package/accelerate but they're not really targeting Haskell as much as that they provide an EDSL for vectorizable computations that can be mapped to GPUs. |
| 2026-01-19 07:36:28 +0100 | <Axman6> | Probably Accelerate still |
| 2026-01-19 07:34:54 +0100 | vanishingideal | (~vanishing@user/vanishingideal) vanishingideal |
| 2026-01-19 07:33:24 +0100 | <int-e> | don't? |
| 2026-01-19 07:31:21 +0100 | merijn | (~merijn@host-cl.cgnat-g.v4.dfn.nl) (Ping timeout: 265 seconds) |
| 2026-01-19 07:30:45 +0100 | <Guest70> | What's the state of the art compiling Haskell to GPU these days? |
| 2026-01-19 07:26:16 +0100 | merijn | (~merijn@host-cl.cgnat-g.v4.dfn.nl) merijn |
| 2026-01-19 07:22:19 +0100 | poscat | (~poscat@user/poscat) poscat |
| 2026-01-19 07:19:29 +0100 | poscat | (~poscat@user/poscat) (Remote host closed the connection) |
| 2026-01-19 07:18:18 +0100 | <Leary> | Axman6: I haven't used them, but based on Matthew Pickering's talk 'What we have learned about memory profiling in the last 5 years' <https://www.youtube.com/watch?v=8i8HJiBI0lo> you want to try eventlog-live or ghc-debug. |
| 2026-01-19 07:17:11 +0100 | takuan_dozo | (~takuan@d8D86B9E9.access.telenet.be) (Read error: Connection reset by peer) |
| 2026-01-19 07:17:11 +0100 | takuan | (~takuan@d8D86B9E9.access.telenet.be) |
| 2026-01-19 07:15:40 +0100 | merijn | (~merijn@host-cl.cgnat-g.v4.dfn.nl) (Ping timeout: 255 seconds) |
| 2026-01-19 07:10:47 +0100 | Guest70 | (~Guest70@syn-024-240-142-083.biz.spectrum.com) |