| 28 May 2025 |
陈浩南 | So why? | 11:39:16 |
陈浩南 | Where is the code that make difference? | 11:39:32 |
Grimmauld (any/all) | some wrapping magic deep in stdenv, but probably not relevant to your issues anyways | 11:40:14 |
陈浩南 | in top-level.nix I saw the definition of gcc:
default-gcc-version = 14;
gcc = pkgs.${"gcc${toString default-gcc-version}"};
| 11:40:21 |
K900 | Yeah it's not relevant | 11:40:45 |
K900 | Post the full build log please | 11:40:49 |
陈浩南 | Currently Intel fortran compiler ifx works when compiling hdf5 and openmpi, but does not work when compiling VASP (a proprietary software, its source code is only available after purchase, so I could not provide it here).
ifx complains:
ifx: error #10417: Problem setting up the Intel(R) Compiler compilation environment. Requires 'install path' setting gathered from 'gcc'
| 11:46:03 |
K900 | Does it provide any more information than that? | 11:46:35 |
K900 | If not, you might just have to strace it or something | 11:46:41 |
K900 | But either way this is extremely unlikely to be a stdenv issue | 11:46:48 |
陈浩南 | No. I will do it another day. Thank you for your answer. | 11:53:57 |
陈浩南 | * No. I will strace it another day. Thank you for your answer. | 11:54:14 |
Tristan Ross | In reply to @chn:chn.moe
This is my nix repl history:
nix-repl> :lf github:NixOS/nixpkgs/nixos-unstable
Added 17 variables.
nix-repl> outputs.legacyPackages.x86_64-linux.gcc.cc
«derivation /nix/store/sjzbql9ja7p7c08ziagba11cnvj6lsz9-gcc-14.2.1.20250322.drv»
nix-repl> outputs.legacyPackages.x86_64-linux.gcc14.cc
«derivation /nix/store/f3gx5c6q45nsskjyx451amqmzyvasr90-gcc-14.2.1.20250322.drv»
stdenv.cc is built via the last bootstrap stage. gcc is built with stdenv.cc so that's why it's different. | 14:03:21 |
Tristan Ross | In reply to @chn:chn.moe
Currently Intel fortran compiler ifx works when compiling hdf5 and openmpi, but does not work when compiling VASP (a proprietary software, its source code is only available after purchase, so I could not provide it here).
ifx complains:
ifx: error #10417: Problem setting up the Intel(R) Compiler compilation environment. Requires 'install path' setting gathered from 'gcc'
What about using gfortran? | 14:07:41 |
| 29 May 2025 |
陈浩南 | Many fortran programs in my major does not consist with standard fortran, gfortran will refuse to compile them, while restriction in ifx loose a lot. And, not sure why, but program compiled using ifx usually gains a huge performance improvement, comparing to gfortran or compiler from AMD (ROCM), even on AMD CPU. | 03:27:24 |
Tristan Ross | In reply to @chn:chn.moe Many fortran programs in my major does not consist with standard fortran, gfortran will refuse to compile them, while restriction in ifx loose a lot. And, not sure why, but program compiled using ifx usually gains a huge performance improvement, comparing to gfortran or compiler from AMD (ROCM), even on AMD CPU. Oh, that's the only fortran I am aware of being in nixpkgs. We might have flang but I can't recall. What about packaging ifx? | 04:04:28 |
陈浩南 |
What about packaging ifx? That is what I tried to do yesterday.
| 04:06:20 |
陈浩南 | *
What about packaging ifx?
That is what I tried to do yesterday.
| 04:06:29 |
Tristan Ross | Ah | 04:07:52 |
Randy Eckenrode | The Linux stdenv bootstrap doesn’t overlay the versioned gcc attribute. That’s all it is. The Darwin is able to do this for LLVM, so it’s probably possible for Linux, but there’s been no motivating issue to do it for Linux. | 11:53:49 |
Randy Eckenrode | Something like this would probably work for GCC on Linux (but significantly less verbose): https://github.com/NixOS/nixpkgs/blob/adfe0268d8b621c68f0b060fc2e801b5196b5f6a/pkgs/stdenv/darwin/default.nix#L1223-L1246 | 11:55:23 |
| matrixrooms.info mod bot (does NOT read/send messages and/or invites; used for checking reported rooms) joined the room. | 17:28:05 |
| 31 May 2025 |
| @trofi:matrix.org left the room. | 13:47:07 |
| 23 Sep 2025 |
| @rick:matrix.ciphernetics.nl left the room. | 07:35:22 |
| 31 May 2025 |
Grimmauld (any/all) | Wait why can't i invite people here? | 16:33:40 |
hexa | because permission levels are per room | 16:53:35 |
hexa | * because permission levels are a per room setting | 16:53:40 |
Grimmauld (any/all) | Well then. Can someone invite @benjaminsparks:chat.alugha.app ? | 18:33:30 |
| 1 Jun 2025 |
| NixOS Moderation Botchanged room power levels. | 03:03:22 |
uep | Redacted or Malformed Event | 03:04:10 |