Skip to content

simsimd: add build-simsimd.yml for riscv64 wheels - #713

Merged
luhenry merged 1 commit into
mainfrom
simsimd
Aug 28, 2026
Merged

simsimd: add build-simsimd.yml for riscv64 wheels#713
luhenry merged 1 commit into
mainfrom
simsimd

Conversation

@luhenry

@luhenry luhenry commented Aug 28, 2026

Copy link
Copy Markdown
Member

Compiles the SimSIMD C extension, a mixed-precision BLAS-like vector math library. Upstream publishes no riscv64 wheel.

Mirrors upstream's release.yml Linux jobs. SimSIMD has no RISC-V kernels, so the extension builds from the portable serial path and the capability-parametrised tests skip themselves, exactly as for upstream's i686, ppc64le and s390x wheels.

Differs from upstream

  • SIMSIMD_IN_QEMU unset - the riscv64 runners are native hardware.
  • Registry index added so numpy and scipy resolve.

Testing

  • same as upstream

License: wheel bundles libgomp (GPLv3 with the GCC Runtime Library Exception) from the build image, so its sources are published with the release.

Built on cp312; 54 passed, 1168 skipped.

Mirrors the Linux jobs of upstream's release.yml, narrowed to riscv64.
SimSIMD has no RISC-V kernels, so the extension builds from the portable
serial C path; upstream's own i686/ppc64le/s390x wheels are built the same
way.

Drops upstream's SIMSIMD_IN_QEMU marker: the riscv64 runners are native
hardware, so the half-precision and complex-number tests it disables run
here.

The extension links libgomp, which auditwheel vendors into the wheel, so
the workflow publishes the toolchain sources alongside it.
@luhenry
luhenry merged commit 2c2b81f into main Aug 28, 2026
14 of 16 checks passed
@luhenry
luhenry deleted the simsimd branch August 28, 2026 06:40
@luhenry luhenry linked an issue Aug 28, 2026 that may be closed by this pull request
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

simsimd riscv64 support

1 participant