llama-cpp: fix build due to openblas update; 1483 -> 1538

[ghost]

pkg-confg file for openblas moved from openblas.pc -> openblas64.pc

add patch so build checks for openblas64 module. submitted patch upstream ggml-org/llama.cpp#4134

#255443 updated from openblas 0.3.21 -> 0.3.24. openblas 0.3.22 moved openblas.pc -> openblas64.pc

OpenMathLib/OpenBLAS#3790

updated the package as well.

https://hydra.nixos.org/build/241259310
ZHF: #265948

Result of nixpkgs-review pr 268449 run on x86_64-linux 1

2 packages built:

• llama-cpp
• ollama

Description of changes

Things done

• Built on platform(s)
  • x86_64-linux
  • aarch64-linux
  • x86_64-darwin
  • aarch64-darwin
• For non-Linux: Is sandboxing enabled in nix.conf? (See Nix manual)
  • sandbox = relaxed
  • sandbox = true
• Tested, as applicable:
  • NixOS test(s) (look inside nixos/tests)
  • and/or package tests
  • or, for functions and "core" functionality, tests in lib/tests or pkgs/test
  • made sure NixOS tests are linked to the relevant packages
• Tested compilation of all packages that depend on this change using nix-shell -p nixpkgs-review --run "nixpkgs-review rev HEAD". Note: all changes have to be committed, also see nixpkgs-review usage
• Tested basic functionality of all binary files (usually in ./result/bin/)
• 23.11 Release Notes (or backporting 23.05 Release notes)
  • (Package updates) Added a release notes entry if the change is major or breaking
  • (Module updates) Added a release notes entry if the change is significant
  • (Module addition) Added a release notes entry if adding a new NixOS module
• Fits CONTRIBUTING.md.

[elohmeier]

LGTM

[ghost]

thanks!

[ghost]

@elohmeier I had mistakenly assumed that the 64 was revering to pointer size when i made this change -- it doesn't. the 64 refers to the integer size, so openblas64.pc is paired with the ILP64 openblas, and openblas.pc is paired with the LP64.

is this package trying to use the ILP64 openblas or the LP64 openblas?

the nix flake also uses the ILP64 openblas -- there was a comment
in my upstream PR which made me think that the LP64 was expected but I think i misunderstood.

[elohmeier]

TBH I'm not using the OpenBLAS implementation personally and did not test it in depth. I don't really have an opinion on which one to use here, but would suggest to copy the upstream choice from their nix flake.