Workflows: get Cuda arch from nvidia-smi in eamxx-sa-XYZ.yml

E3SM-Project · Nov 15, 2024 · 74dfb81 · 74dfb81
1 parent 2354936
commit 74dfb81
Show file tree

Hide file tree

Showing 3 changed files with 99 additions and 3 deletions.
diff --git a/.github/workflows/eamxx-sa-coverage.yml b/.github/workflows/eamxx-sa-coverage.yml
@@ -48,11 +48,43 @@ jobs:
           submodules: recursive
       - name: Show action trigger
         uses: ./.github/actions/show-workflow-trigger
+      - name: Get CUDA Arch
+        run: |
+          # Ensure nvidia-smi is available
+          if ! command -v nvidia-smi &> /dev/null; then
+              echo "nvidia-smi could not be found. Please ensure you have Nvidia drivers installed."
+              exit 1
+          fi
+
+          # Get the GPU model from nvidia-smi
+          gpu_model=$(nvidia-smi --query-gpu=name --format=csv,noheader | head -n 1)
+          case "$gpu_model" in
+              *"H100"*)
+                  Hopper=ON
+                  ARCH=89
+                  ;;
+              *"A100"*)
+                  Amper=ON
+                  ARCH=80
+                  ;;
+              *"V100"*)
+                  Volta=ON
+                  ARCH=70
+                  ;;
+              *)
+                  echo "Unsupported GPU model: $gpu_model"
+                  exit 1
+                  ;;
+          esac
+
+          # Set the output variables for the next step
+          echo "KOKKOS_ARCH=${KOKKOS_ARCH}" >> $GITHUB_ENV
+          echo "CMAKE_CUDA_ARCHITECTURES=${CMAKE_CUDA_ARCHITECTURES}" >> $GITHUB_ENV
       - name: Run tests
         uses: ./.github/actions/test-all-scream
         with:
           build_type: cov
           machine: ghci-snl-cuda
           generate: false
           submit: ${{ env.submit }}
-          cmake-configs: Kokkos_ARCH_VOLTA70=ON;CMAKE_CUDA_ARCHITECTURES=70
+          cmake-configs: Kokkos_ARCH_HOPPER89=${{ env.Hopper }};Kokkos_ARCH_AMPERE80=${{ env.Ampere }};Kokkos_ARCH_VOLTA70=${{ env.Volta }};CMAKE_CUDA_ARCHITECTURES=${{ env.ARCH }}
diff --git a/.github/workflows/eamxx-sa-sanitizer.yml b/.github/workflows/eamxx-sa-sanitizer.yml
@@ -52,11 +52,43 @@ jobs:
           submodules: recursive
       - name: Show action trigger
         uses: ./.github/actions/show-workflow-trigger
+      - name: Get CUDA Arch
+        run: |
+          # Ensure nvidia-smi is available
+          if ! command -v nvidia-smi &> /dev/null; then
+              echo "nvidia-smi could not be found. Please ensure you have Nvidia drivers installed."
+              exit 1
+          fi
+
+          # Get the GPU model from nvidia-smi
+          gpu_model=$(nvidia-smi --query-gpu=name --format=csv,noheader | head -n 1)
+          case "$gpu_model" in
+              *"H100"*)
+                  Hopper=ON
+                  ARCH=89
+                  ;;
+              *"A100"*)
+                  Amper=ON
+                  ARCH=80
+                  ;;
+              *"V100"*)
+                  Volta=ON
+                  ARCH=70
+                  ;;
+              *)
+                  echo "Unsupported GPU model: $gpu_model"
+                  exit 1
+                  ;;
+          esac
+
+          # Set the output variables for the next step
+          echo "KOKKOS_ARCH=${KOKKOS_ARCH}" >> $GITHUB_ENV
+          echo "CMAKE_CUDA_ARCHITECTURES=${CMAKE_CUDA_ARCHITECTURES}" >> $GITHUB_ENV
       - name: Run tests
         uses: ./.github/actions/test-all-scream
         with:
           build_type: ${{ matrix.build_type }}
           machine: ghci-snl-cuda
           generate: false
           submit: ${{ env.submit }}
-          cmake-configs: Kokkos_ARCH_VOLTA70=ON;CMAKE_CUDA_ARCHITECTURES=70
+          cmake-configs: Kokkos_ARCH_HOPPER89=${{ env.Hopper }};Kokkos_ARCH_AMPERE80=${{ env.Ampere }};Kokkos_ARCH_VOLTA70=${{ env.Volta }};CMAKE_CUDA_ARCHITECTURES=${{ env.ARCH }}
diff --git a/.github/workflows/eamxx-sa-testing.yml b/.github/workflows/eamxx-sa-testing.yml
@@ -172,11 +172,43 @@ jobs:
               echo "generate=true" >> $GITHUB_ENV
             fi
           fi
+      - name: Get CUDA Arch
+        run: |
+          # Ensure nvidia-smi is available
+          if ! command -v nvidia-smi &> /dev/null; then
+              echo "nvidia-smi could not be found. Please ensure you have Nvidia drivers installed."
+              exit 1
+          fi
+
+          # Get the GPU model from nvidia-smi
+          gpu_model=$(nvidia-smi --query-gpu=name --format=csv,noheader | head -n 1)
+          case "$gpu_model" in
+              *"H100"*)
+                  Hopper=ON
+                  ARCH=89
+                  ;;
+              *"A100"*)
+                  Amper=ON
+                  ARCH=80
+                  ;;
+              *"V100"*)
+                  Volta=ON
+                  ARCH=70
+                  ;;
+              *)
+                  echo "Unsupported GPU model: $gpu_model"
+                  exit 1
+                  ;;
+          esac
+
+          # Set the output variables for the next step
+          echo "KOKKOS_ARCH=${KOKKOS_ARCH}" >> $GITHUB_ENV
+          echo "CMAKE_CUDA_ARCHITECTURES=${CMAKE_CUDA_ARCHITECTURES}" >> $GITHUB_ENV
       - name: Run tests
         uses: ./.github/actions/test-all-scream
         with:
           build_type: ${{ matrix.build_type }}
           machine: ghci-snl-cuda
           generate: ${{ env.generate }}
           submit: ${{ env.submit }}
-          cmake-configs: Kokkos_ARCH_VOLTA70=ON;CMAKE_CUDA_ARCHITECTURES=70
+          cmake-configs: Kokkos_ARCH_HOPPER89=${{ env.Hopper }};Kokkos_ARCH_AMPERE80=${{ env.Ampere }};Kokkos_ARCH_VOLTA70=${{ env.Volta }};CMAKE_CUDA_ARCHITECTURES=${{ env.ARCH }}