| 24 Nov 2022 |
 pie_ | perfect time to retry /dev/shm | 02:56:04 |
| pie_ | something I mighthave messed up last time is not setting the temp dir variables to be on shm too | 02:59:48 |
| pie_ | hm, I got steam-run to eval. I probably broke it though? | 03:37:18 |
| pie_ | time to build spidermonkey and rust because of polkit... | 03:37:59 |
|  Nick Cao joined the room. | 05:03:05 |
| pie_ | tpw_rules: if I ever manage to get any sort of graphics going, do I need nixGL? I don't know how the graphics stack works so idk if I can just throw xvfb-run at it and get sw rendering | 05:06:40 |
 tpw_rules | no, nixGL is only necessary to link up with the hardware graphics drivers | 05:07:26 |
| pie_ | Hm. I didnt hink through how I'm going to deal with UnityPlayer.so requiring glibc 2.14 or 2.15 | 05:27:18 |
| pie_ | * Hm. I didnt think through how I'm going to deal with UnityPlayer.so requiring glibc 2.14 or 2.15 | 05:27:26 |
| pie_ | though technically I think thats old enough to run? | 05:27:37 |
| pie_ | I might have done a lot more work than needed for this binary at least | 05:28:14 |
| pie_ | it might not be trivial to build against an old glibc though | 05:50:48 |
| pie_ | https://discourse.nixos.org/t/compiling-with-old-glibc/12054/4 | 05:51:02 |
| 18 Dec 2022 |
 Madoura | So apparently mainline clang can compile and link HIP executables just fine without hipcc.
It still makes sense to follow the official ROCm software stack for nixpkgs, but dang. This would have been a lot easier to just use regular clang. | 15:48:46 |
| pie_ | heh | 15:49:57 |
| pie_ | I still want to work on it but I haven't gotten my little project working yet. Also exams. | 15:50:23 |
| Madoura | Ah I get it.
I've had a lot of free time lately so I've been spending it (without sleeping...) working on porting ROCm to nixpkgs lol.
Theoretically since we have rocmClangStdenv now switching away from hipcc if/when it's needed should be easy for all the ROCm packages. | 15:51:35 |
| 20 Dec 2022 |
|  Pascal joined the room. | 00:24:20 |
| 24 Dec 2022 |
| Madoura | We have torch now. https://github.com/Madouura/nixpkgs/commit/df71e711026a37178f9a258f236db0e1a66e2f0b | 14:36:29 |
| 29 Dec 2022 |
|  luizirber joined the room. | 19:48:22 |
| 21 Feb 2023 |
|  John ✒️ joined the room. | 10:48:27 |
| 28 Feb 2023 |
|  Ido Samuelson joined the room. | 05:43:18 |
| 14 Mar 2023 |
|  pbsds joined the room. | 13:15:12 |
| 24 Mar 2023 |
|  htran joined the room. | 02:20:37 |
| 25 Mar 2023 |
|  @valconius:matrix.org joined the room. | 23:44:55 |
|  rbutani joined the room. | 23:55:19 |
| 27 Mar 2023 |
|  Ryan Swart joined the room. | 09:41:42 |
| 29 Mar 2023 |
|  SomeoneSerge (utc+3) joined the room. | 19:39:45 |
| SomeoneSerge (utc+3) | Hi! Just wanted to mention that torchWithRocm is going to require allowUnfree = true when https://github.com/NixOS/nixpkgs/pull/222273 gets merged. I think this needs to be addressed upstream in openai/triton (they unconditionally download and vendor a copy of ptxas from cuda), but I haven't the capacity to handle this now | 19:54:53 |
| 31 Mar 2023 |
| Ryan Swart | Kind of a meta question - I am trying to run static nix on an HPC cluster and am running into some roadblocks (with a custom path and disabling the sandbox as much as possible I am still running into permission denied errors - can't chroot or run bubblewrap); given that I have the ability to run arbitrary binaries - is there a way to disable all "special" path modification etc in such a way that common HPC access limitations can allow nix to still build derivations? | 02:28:46 |
