chore: add CUDA code to lint + primitive test CIs (#1997)

Author: stephenh-axiom-xyz

.github/workflows/lints.yml

@@ -86,3 +86,39 @@ jobs:
         run: |
           cargo binstall --no-confirm --force cargo-audit
           cargo audit
+
+  lint-cuda:
+    name: Lint CUDA
+    runs-on:
+      - runs-on=${{ github.run_id }}
+      - family=g6.xlarge
+      - image=ubuntu24-gpu-x64
+      - extras=s3-cache
+
+    steps:
+      - uses: runs-on/action@v1
+      - uses: actions/checkout@v4
+
+      - uses: dtolnay/rust-toolchain@stable
+      - uses: Swatinem/rust-cache@v2
+        with:
+          cache-on-failure: true
+
+      - name: Verify CUDA setup
+        run: |
+          nvidia-smi
+          nvidia-smi -L
+          nvcc --version
+          
+      - name: Run fmt
+        working-directory: crates
+        run: |
+          rustup install nightly
+          rustup component add --toolchain nightly rustfmt
+          cargo +nightly fmt --all -- --check
+
+      - name: Run clippy
+        run: |
+          # includes GPU specific features cuda and touchemall
+          cargo clippy --all-targets --all --tests --features "aggregation bls12_381 bn254 build-elfs cuda default entrypoint evm-prove evm-verify export-intrinsics export-libm function-span getrandom-unsupported halo2-compiler halo2curves heap-embedded-alloc jemalloc jemalloc-prof metrics nightly-features panic-handler parallel perf-metrics rust-runtime static-verifier std test-utils touchemall" -- -D warnings
+          cargo clippy --all-targets --all --tests --no-default-features --features "jemalloc jemalloc-prof" -- -D warnings

.github/workflows/primitives.yml

@@ -23,10 +23,16 @@ env:
 
 jobs:
   build:
+    strategy:
+      matrix:
+        config:
+          - { family: "c7g.8xlarge", image: "ubuntu24-full-arm64" }
+          - { family: "g6.xlarge", image: "ubuntu24-gpu-x64" }
+
     runs-on:
       - runs-on=${{ github.run_id }}
-      - runner=32cpu-linux-arm64
-      - image=ubuntu24-full-arm64
+      - family=${{ matrix.config.family }}
+      - image=${{ matrix.config.image }}
       - extras=s3-cache
 
     steps:
@@ -38,22 +44,40 @@ jobs:
           cache-on-failure: true
       - uses: taiki-e/install-action@nextest
 
+      - name: Set CPU tests environment variables
+        if: matrix.config.family != 'g6.xlarge'
+        run: |
+          echo "NEXTEST_ENV=" >> $GITHUB_ENV
+          echo "NEXTEST_ARGS=--cargo-profile fast --features parallel" >> $GITHUB_ENV
+
+      - name: Check CUDA status and set environment variables
+        if: matrix.config.family == 'g6.xlarge'
+        run: |
+          nvidia-smi
+          nvidia-smi -L
+          nvcc --version
+
+          echo "NEXTEST_ENV=CUDA_OPT_LEVEL=1" >> $GITHUB_ENV
+          echo "NEXTEST_ARGS=cuda --features parallel,cuda,touchemall" >> $GITHUB_ENV
+
       - name: Run tests for primitives
         working-directory: crates/circuits/primitives
         run: |
-          cargo nextest run --cargo-profile fast --features parallel
+          ${{ env.NEXTEST_ENV }} cargo nextest run ${{ env.NEXTEST_ARGS }} 
 
       - name: Run tests for poseidon2-air
         working-directory: crates/circuits/poseidon2-air
         run: |
-          cargo nextest run --cargo-profile fast --features parallel
+          ${{ env.NEXTEST_ENV }} cargo nextest run ${{ env.NEXTEST_ARGS }} 
 
       - name: Run tests for sha256-air
+        if: matrix.config.family != 'g6.xlarge'
         working-directory: crates/circuits/sha256-air
         run: |
-          cargo nextest run --cargo-profile fast --features parallel
+          ${{ env.NEXTEST_ENV }} cargo nextest run ${{ env.NEXTEST_ARGS }}
 
       - name: Run tests for mod-builder
+        if: matrix.config.family != 'g6.xlarge'
         working-directory: crates/circuits/mod-builder
         run: |
-          cargo nextest run --cargo-profile fast --features parallel
+          ${{ env.NEXTEST_ENV }} cargo nextest run ${{ env.NEXTEST_ARGS }}

Cargo.lock

@@ -34,3 +34,4 @@ openvm-cuda-builder.workspace = true
 default = ["parallel"]
 parallel = ["openvm-stark-backend/parallel"]
 cuda = ["dep:openvm-cuda-common", "dep:openvm-cuda-backend"]
+touchemall = ["openvm-cuda-backend/touchemall"]

crates/circuits/poseidon2-air/Cargo.toml

@@ -32,3 +32,4 @@ openvm-cuda-builder.workspace = true
 default = ["parallel"]
 parallel = ["openvm-stark-backend/parallel"]
 cuda = ["dep:openvm-cuda-common", "dep:openvm-cuda-backend"]
+touchemall = ["openvm-cuda-backend/touchemall"]