simd_half: TD-SIMD-8 — F16C-vectorized F16↔f32 batch conversion #403

Workflow file for this run

	on:
	pull_request:
	paths-ignore:
	- '.github/workflows/latest-deps.yaml'
	merge_group:
	push:
	branches:
	- master
	- main

	name: Continuous integration

	env:
	CARGO_TERM_COLOR: always
	HOST: x86_64-unknown-linux-gnu
	FEATURES: "approx,serde,rayon"
	# `-C target-cpu=x86-64-v3` was removed from the global env. It conflicts
	# with the cross_test matrix (`i686-unknown-linux-gnu` is 32-bit, `s390x`
	# isn't even x86) and contradicts the design intent recorded in
	# `.cargo/config.toml`: per-function `#[target_feature]` + runtime
	# `LazyLock<Tier>` detection means one binary, all ISAs. Jobs that
	# specifically need a higher target-cpu can opt in via per-job env.
	RUSTFLAGS: "-D warnings"
	MSRV: 1.95.0
	BLAS_MSRV: 1.95.0

	jobs:
	pass-msrv:
	runs-on: ubuntu-latest
	name: Pass MSRV values to other jobs
	outputs:
	MSRV: ${{ env.MSRV }}
	BLAS_MSRV: ${{ env.BLAS_MSRV }}
	steps:
	- name: Pass MSRV
	run: \|
	echo "MSRV=${{ env.MSRV }}" >> $GITHUB_OUTPUT
	echo "BLAS_MSRV=${{ env.BLAS_MSRV }}" >> $GITHUB_OUTPUT

	clippy:
	runs-on: ubuntu-latest
	strategy:
	matrix:
	# Pinned to 1.95.0 to match `rust-toolchain.toml`. The workspace
	# bumped from 1.94.1 → 1.95.0 in commit
	# https://github.com/AdaWorldAPI/ndarray (this PR) to align with
	# bevy (edition 2024 → 1.95 MSRV) and lance-graph. 1.95 added the
	# `clippy::manual_checked_ops` lint which fires on
	# `impl_owned_array.rs::into_scalar` — fixed in this same PR.
	rust:
	- "1.95.0"
	name: clippy/${{ matrix.rust }}
	steps:
	- uses: actions/checkout@v4
	- uses: dtolnay/rust-toolchain@1.95.0
	with:
	components: clippy
	- uses: Swatinem/rust-cache@v2
	- run: cargo clippy --features approx,serde,rayon -- -D warnings
	- run: cargo clippy --features native -- -D warnings

	format:
	runs-on: ubuntu-latest
	name: format/stable
	# Blocking — no continue-on-error. `rustfmt.toml` was stripped of all
	# nightly-only directives in PR #133 (`7ca1cf3d chore(fmt): strip
	# unstable rustfmt options, keep stable-only config` + `ed59b3b1
	# style: normalize codebase to stable rustfmt 1.94.1`) and the
	# codebase normalized — so stable rustfmt enforces the project's
	# style losslessly. The previous `continue-on-error: true` +
	# `dtolnay/rust-toolchain@nightly` setup was a stale band-aid from
	# when `brace_style = AlwaysNextLine`, `imports_granularity =
	# Preserve`, `unstable_features = true` still lived in rustfmt.toml.
	# That state no longer exists; the band-aid is removed in this PR.
	steps:
	- uses: actions/checkout@v4
	# Stable rustfmt 1.95.0 — pinned in `rust-toolchain.toml`. No
	# nightly dependency since rustfmt.toml is stable-clean post-PR
	# #133.
	- uses: dtolnay/rust-toolchain@1.95.0
	with:
	components: rustfmt
	- run: cargo fmt --all --check

	nostd:
	runs-on: ubuntu-latest
	continue-on-error: ${{ matrix.experimental }}
	strategy:
	matrix:
	include:
	- rust: stable
	experimental: false
	target: thumbv6m-none-eabi

	name: nostd/${{ matrix.target }}/${{ matrix.rust }}
	steps:
	- uses: actions/checkout@v4
	- uses: dtolnay/rust-toolchain@stable
	with:
	toolchain: ${{ matrix.rust }}
	targets: ${{ matrix.target }}
	# rust-toolchain.toml pins 1.95.0 — install the cross-compile target
	# for that toolchain too, since dtolnay/rust-toolchain only installs
	# for the matrix value which may differ from the pinned version.
	- run: rustup target add ${{ matrix.target }}
	- name: Tests
	run: \|
	# Scope to `-p ndarray` so workspace dev-deps (criterion → serde_core
	# → getrandom) don't get evaluated against the no_std target. The
	# library itself builds cleanly under no_std + portable-atomic-
	# critical-section; only its dev-dependency tree pulls std-requiring
	# crates that have no business existing in the nostd build.
	cargo rustc -p ndarray "--target=${{ matrix.target }}" --no-default-features --features portable-atomic-critical-section

	tests:
	runs-on: ubuntu-latest
	needs: pass-msrv
	strategy:
	matrix:
	rust:
	- stable
	- beta
	- ${{ needs.pass-msrv.outputs.MSRV }}

	name: tests/${{ matrix.rust }}
	steps:
	- uses: actions/checkout@v4
	- uses: dtolnay/rust-toolchain@master
	with:
	toolchain: ${{ matrix.rust }}
	- uses: rui314/setup-mold@v1
	- uses: Swatinem/rust-cache@v2
	- uses: taiki-e/install-action@nextest
	- name: Install openblas
	run: sudo apt-get install libopenblas-dev gfortran
	- run: ./scripts/all-tests.sh "$FEATURES" ${{ matrix.rust }}

	native-backend:
	runs-on: ubuntu-latest
	name: native-backend/stable
	steps:
	- uses: actions/checkout@v4
	- uses: dtolnay/rust-toolchain@stable
	- uses: Swatinem/rust-cache@v2
	- uses: taiki-e/install-action@nextest
	- name: Build native backend
	run: cargo build --features native
	- name: Test native backend
	run: cargo nextest run -p ndarray --features native
	- name: Test native + approx
	run: cargo nextest run -p ndarray --features native,approx

	hpc-stream-parallel:
	# D-CSV-17 (sprint-13 W-I4): rayon par_* variants for hpc::stream.
	# This job co-ships with the par_* implementation so the rayon feature
	# gate is always exercised alongside the code it guards — prevents
	# silent-dead-code drift (spec §0 + worker-template-v2 §5 CI ownership).
	runs-on: ubuntu-latest
	name: hpc-stream-parallel/rayon
	steps:
	- uses: actions/checkout@v4
	- uses: dtolnay/rust-toolchain@1.95.0
	- uses: Swatinem/rust-cache@v2
	- uses: taiki-e/install-action@nextest
	- name: cargo check (no rayon — scalar path unchanged)
	run: cargo check -p ndarray
	- name: cargo check --features rayon
	run: cargo check -p ndarray --features rayon
	- name: par_* stream tests (--features rayon, hpc::stream filter)
	run: cargo nextest run -p ndarray --features rayon -E 'test(hpc::stream)'
	- name: clippy --features rayon
	run: cargo clippy -p ndarray --features rayon --lib -- -D warnings

	tier4-avx512-check:
	# Phase 6 of the SIMD integration plan (.claude/knowledge/
	# simd-dispatch-architecture.md). Exercises the AVX-512 dispatch
	# arm (`#[cfg(target_feature = "avx512f")]` in `src/simd.rs`) so it
	# doesn't bit-rot under the v3-default cargo config.
	#
	# `cargo check` rather than `cargo test`: GH-hosted `ubuntu-latest`
	# runners may not have AVX-512 silicon (intermittent across SKUs),
	# so a v4-baked binary would SIGILL at run time. `check` compiles
	# the AVX-512 code path through the type checker + borrow checker
	# + monomorphization without producing a runnable artifact —
	# catches type mismatches and dispatch-arm holes that the v3
	# default never touches.
	#
	# # Why `CARGO_TARGET_<triple>_RUSTFLAGS` instead of plain `RUSTFLAGS`:
	#
	# The first iteration used `env: RUSTFLAGS: "-Ctarget-cpu=x86-64-v4"`
	# and failed in ~23 s — RUSTFLAGS env applies to BOTH the target
	# compilation AND host build scripts (`build.rs` artifacts that
	# cargo runs natively). On a GH runner without AVX-512 silicon,
	# those v4-baked build scripts SIGILL during the dep build.
	#
	# `CARGO_TARGET_X86_64_UNKNOWN_LINUX_GNU_RUSTFLAGS` is documented to
	# apply only when cargo is producing artifacts for that triple, NOT
	# to host build scripts. Combined with explicit `--target` (so cargo
	# distinguishes host from target even when they're the same triple),
	# this gives us "v4 for our crate, baseline for build scripts."
	runs-on: ubuntu-latest
	name: tier4-avx512-check
	env:
	CARGO_TARGET_X86_64_UNKNOWN_LINUX_GNU_RUSTFLAGS: "-D warnings -Ctarget-cpu=x86-64-v4"
	steps:
	- uses: actions/checkout@v4
	- uses: dtolnay/rust-toolchain@stable
	- uses: Swatinem/rust-cache@v2
	- name: cargo check (v4 / AVX-512 dispatch arm)
	run: cargo check --target=x86_64-unknown-linux-gnu -p ndarray --features approx,serde,rayon
	- name: cargo check (v4 / AVX-512 + hpc-extras)
	run: cargo check --target=x86_64-unknown-linux-gnu -p ndarray --features approx,serde,rayon,hpc-extras

	nightly-simd-polyfill:
	# TD-SIMD-9 from .claude/knowledge/simd-dispatch-architecture.md.
	# Exercises the `feature = "nightly-simd"` dispatch arm in
	# `src/simd.rs` (added in Phase 2 / PR #173) so the portable
	# `core::simd::*` backend doesn't bit-rot between PRs.
	#
	# Why nightly: `src/simd_nightly/*` uses `#![feature(portable_simd)]`
	# which only compiles on nightly rustc. The stable `cargo check`
	# paths in tests/clippy never reach this arm.
	#
	# `cargo check` rather than full build/test — pure type/borrow/cfg
	# verification. Catches arm shadowing bugs and the kind of API-break
	# codex flagged on PR #173 (lowercase aliases dropped).
	#
	# Runs only on merge_group + push (not every PR) to keep nightly
	# toolchain pulls out of the per-PR critical path. Mirrors the
	# `miri` job's gating strategy.
	if: github.event_name == 'merge_group' \|\| github.event_name == 'push'
	runs-on: ubuntu-latest
	name: nightly-simd-polyfill
	steps:
	- uses: actions/checkout@v4
	- uses: dtolnay/rust-toolchain@nightly
	- uses: Swatinem/rust-cache@v2
	- name: cargo check (nightly + nightly-simd feature)
	run: cargo check -p ndarray --features approx,serde,rayon,nightly-simd
	- name: cargo check (nightly + nightly-simd + hpc-extras)
	run: cargo check -p ndarray --features approx,serde,rayon,nightly-simd,hpc-extras

	blas-msrv:
	runs-on: ubuntu-latest
	name: blas-msrv
	needs: pass-msrv
	steps:
	- uses: actions/checkout@v4
	- uses: dtolnay/rust-toolchain@master
	with:
	toolchain: ${{ needs.pass-msrv.outputs.BLAS_MSRV }}
	- uses: rui314/setup-mold@v1
	- uses: Swatinem/rust-cache@v2
	- uses: taiki-e/install-action@nextest
	- name: Install openblas
	run: sudo apt-get install libopenblas-dev gfortran
	- run: ./scripts/blas-integ-tests.sh $BLAS_MSRV

	miri:
	# Nightly-only. Cannot run SIMD intrinsics or FFI.
	# Only runs on merge queue / push to main — never blocks PR checks.
	if: github.event_name == 'merge_group' \|\| github.event_name == 'push'
	runs-on: ubuntu-latest
	name: miri
	steps:
	- uses: actions/checkout@v4
	- uses: dtolnay/rust-toolchain@nightly
	with:
	components: miri
	- uses: Swatinem/rust-cache@v2
	- uses: taiki-e/install-action@nextest
	- run: ./scripts/miri-tests.sh

	cross_test:
	# Gated on merge_group only — cross-compile via docker (cross-rs) for
	# s390x / i686 is slow, flaky on the s390x docker image's toolchain
	# resolution (rust-toolchain.toml's 1.94.1 pin doesn't resolve cleanly
	# inside the s390x cross container), and reliably caught by the
	# `tests/{stable,beta,1.94.0}` jobs on every PR push. Reserve cross
	# validation for the merge queue where it can fail loudly without
	# gating individual PRs on infra flakiness. The commented `if:` was
	# the original intent (per the pre-existing comment) — uncommenting
	# per the PR #143 codex thread that surfaced this consistently.
	if: ${{ github.event_name == 'merge_group' }}
	runs-on: ubuntu-latest
	strategy:
	matrix:
	include:
	- rust: stable
	target: s390x-unknown-linux-gnu
	- rust: stable
	target: i686-unknown-linux-gnu

	name: cross_test/${{ matrix.target }}/${{ matrix.rust }}
	steps:
	- uses: actions/checkout@v4
	- uses: dtolnay/rust-toolchain@master
	with:
	toolchain: ${{ matrix.rust }}
	targets: ${{ matrix.target }}
	- uses: rui314/setup-mold@v1
	- uses: Swatinem/rust-cache@v2
	- name: Install cross
	run: cargo install cross
	- run: ./scripts/cross-tests.sh "approx,serde,rayon" ${{ matrix.rust }} ${{ matrix.target }}

	cargo-careful:
	# Nightly-only. Only runs on merge queue / push — never blocks PR checks.
	if: github.event_name == 'merge_group' \|\| github.event_name == 'push'
	runs-on: ubuntu-latest
	name: cargo-careful
	steps:
	- uses: actions/checkout@v4
	- uses: dtolnay/rust-toolchain@nightly
	- uses: Swatinem/rust-cache@v2
	- uses: taiki-e/install-action@nextest
	- name: Install cargo-careful
	run: cargo install cargo-careful
	- run: cargo careful nextest run -Zcareful-sanitizer --features="$FEATURES"

	docs:
	# Nightly needed for --cfg docsrs. Only on merge queue / push.
	if: github.event_name == 'merge_group' \|\| github.event_name == 'push'
	runs-on: ubuntu-latest
	name: docs/nightly
	env:
	RUSTDOCFLAGS: "-Dwarnings --cfg docsrs"
	steps:
	- uses: actions/checkout@v4
	- uses: dtolnay/rust-toolchain@nightly
	- run: cargo doc --no-deps --all-features

	conclusion:
	needs:
	- clippy
	- format
	- nostd
	- tests
	- native-backend
	- hpc-stream-parallel
	- tier4-avx512-check
	- nightly-simd-polyfill
	- miri
	- cross_test
	- cargo-careful
	- docs
	runs-on: ubuntu-latest
	steps:
	- name: Result
	run: \|
	jq -C <<< "${needs}"
	# Check if all needs were successful or skipped.
	"$(jq -r 'all(.result as $result \| (["success", "skipped"] \| contains([$result])))' <<< "${needs}")"
	env:
	needs: ${{ toJson(needs) }}

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

simd_half: TD-SIMD-8 — F16C-vectorized F16↔f32 batch conversion #403

Workflow file

simd_half: TD-SIMD-8 — F16C-vectorized F16↔f32 batch conversion #403

Uh oh!

Workflow file for this run