Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

[OpenBLAS] Ugrade to v0.3.26 #52762

Merged
merged 1 commit into from
Jan 6, 2024
Merged

Conversation

giordano
Copy link
Contributor

@giordano giordano commented Jan 5, 2024

Memo to self:

  • update version number in stdlib/OpenBLAS_jll/Project.toml
  • update version number and sha in deps/openblas.version
  • refresh checksums with make -f contrib/refresh_checksums.mk -j openblas

In the release notes of v0.3.26 one of the more interesting point is perhaps

included support for Apple M1 and newer targets in DYNAMIC_ARCH builds

Quoting from JuliaPackaging/Yggdrasil#7911 (comment)

Some quick benchmarks on M1: with OpenBLAS 0.3.25:

julia> using LinearAlgebra, BenchmarkTools

julia> peakflops()
1.7049262964078418e11

julia> x = randn(1_000); y = randn(size(x));

julia> @btime dot($x, $y);
  402.705 ns (0 allocations: 0 bytes)

with OpenBLAS v0.3.26:

julia> using LinearAlgebra, BenchmarkTools

julia> peakflops()
1.8042546290642157e11

julia> x = randn(1_000); y = randn(size(x));

julia> @btime dot($x, $y);
  143.777 ns (0 allocations: 0 bytes)

Up to OpenBLAS v0.3.25 you can get similar performance on Apple Silicon by exporting the environment variable

OPENBLAS_CORETYPE=NEOVERSEN1

@giordano giordano added domain:building Build system, or building Julia or its dependencies external dependencies Involves LLVM, OpenBLAS, or other linked libraries stdlib:JLLs labels Jan 5, 2024
@giordano giordano merged commit b682592 into JuliaLang:master Jan 6, 2024
10 checks passed
@giordano giordano deleted the mg/openblas branch January 6, 2024 08:41
@PallHaraldsson PallHaraldsson mentioned this pull request Jan 6, 2024
33 tasks
@KristofferC KristofferC added the backport 1.10 Change should be backported to the 1.10 release label Jan 17, 2024
KristofferC pushed a commit that referenced this pull request Jan 24, 2024
Memo to self:

* update version number in `stdlib/OpenBLAS_jll/Project.toml`
* update version number and sha in `deps/openblas.version`
* refresh checksums with `make -f contrib/refresh_checksums.mk -j
openblas`

In the [release notes of
v0.3.26](https://github.com/OpenMathLib/OpenBLAS/releases/tag/v0.3.26)
one of the more interesting point is perhaps

> included support for Apple M1 and newer targets in DYNAMIC_ARCH builds

Quoting from
JuliaPackaging/Yggdrasil#7911 (comment)

> Some quick benchmarks on M1: with OpenBLAS 0.3.25:
> ```julia
> julia> using LinearAlgebra, BenchmarkTools
>
> julia> peakflops()
> 1.7049262964078418e11
>
> julia> x = randn(1_000); y = randn(size(x));
>
> julia> @Btime dot($x, $y);
>   402.705 ns (0 allocations: 0 bytes)
> ```
> with OpenBLAS v0.3.26:
> ```julia
> julia> using LinearAlgebra, BenchmarkTools
>
> julia> peakflops()
> 1.8042546290642157e11
>
> julia> x = randn(1_000); y = randn(size(x));
>
> julia> @Btime dot($x, $y);
>   143.777 ns (0 allocations: 0 bytes)
> ```
> Up to OpenBLAS v0.3.25 you can get similar performance on Apple
Silicon by exporting the environment variable
> ```sh
> OPENBLAS_CORETYPE=NEOVERSEN1
> ```

(cherry picked from commit b682592)
@KristofferC KristofferC removed the backport 1.10 Change should be backported to the 1.10 release label Jan 24, 2024
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
domain:building Build system, or building Julia or its dependencies external dependencies Involves LLVM, OpenBLAS, or other linked libraries stdlib:JLLs
Projects
None yet
Development

Successfully merging this pull request may close these issues.

2 participants