Merge pull request #2785 from devitocodes/JDBetteridge/more_nvidia_docker_images

Add separate images for CUDA 12

Author: mloubout

.github/workflows/docker-bases.yml

@@ -190,6 +190,7 @@ jobs:
     strategy:
       matrix:
         arch: [amd64, arm64]
+        cuda: ['12', '13']
         include:
           - arch: amd64
             platform: linux/amd64
@@ -199,6 +200,12 @@ jobs:
           - arch: arm64
             platform: linux/arm64
             runner: ubuntu-24.04-arm
+          - cuda: '12'
+            nvhpc_version: 'nvhpc-25-7'
+            extra_tag: '12'
+          - cuda: '13'
+            nvhpc_version: 'nvhpc'
+            extra_tag: ''
 
     steps:
       - name: Checkout devito
@@ -229,10 +236,10 @@ jobs:
           platforms: ${{ matrix.platform }}
           build-args: |
             arch=nvc
-            ver=nvhpc-25-7
+            ver=${{ matrix.nvhpc_version }}
           # Label (not tag) with runner name for traceability without changing image tags
           labels: builder-runner=${{ runner.name }}
-          tags: "devitocodes/bases:nvidia-nvc-${{ matrix.arch }}"
+          tags: "devitocodes/bases:nvidia-nvc${{ matrix.extra_tag }}-${{ matrix.arch }}"
 
       - name: NVCC image
         uses: docker/build-push-action@v6
@@ -244,9 +251,9 @@ jobs:
           platforms: ${{ matrix.platform }}
           build-args: |
             arch=nvcc
-            ver=nvhpc-25-7
+            ver=${{ matrix.nvhpc_version }}
           labels: builder-runner=${{ runner.name }}
-          tags: "devitocodes/bases:nvidia-nvcc-${{ matrix.arch }}"
+          tags: "devitocodes/bases:nvidia-nvcc${{ matrix.extra_tag }}-${{ matrix.arch }}"
 
       - name: NVC host image
         uses: docker/build-push-action@v6
@@ -258,9 +265,9 @@ jobs:
           platforms: ${{ matrix.platform }}
           build-args: |
             arch=nvc-host
-            ver=nvhpc-25-7
+            ver=${{ matrix.nvhpc_version }}
           labels: builder-runner=${{ runner.name }}
-          tags: "devitocodes/bases:cpu-nvc-${{ matrix.arch }}"
+          tags: "devitocodes/bases:cpu-nvc${{ matrix.extra_tag }}-${{ matrix.arch }}"
 
   deploy-nvidia-bases-manifest:
     if: inputs.nvidia
@@ -272,7 +279,7 @@ jobs:
 
     strategy:
       matrix:
-        final_tag: [nvidia-nvc, nvidia-nvcc, cpu-nvc]
+        final_tag: [nvidia-nvc, nvidia-nvcc, cpu-nvc, nvidia-nvc12, nvidia-nvcc12, cpu-nvc12]
 
     steps:
       - name: Set up Docker Buildx

.github/workflows/docker-devito.yml

@@ -44,6 +44,28 @@ jobs:
             test: ''
             runner: ubuntu-24.04-arm
 
+          - base: 'bases:nvidia-nvc12'
+            tag: 'nvidia-nvc12'
+            arch: amd64
+            platform: linux/amd64
+            run_tests: false
+            tag_suffix: '-amd64'
+            # Respect CUDA_VISIBLE_DEVICES set by the runner and hard-limit docker to that device.
+            # (--gpus maps only the selected device from CUDA_VISIBLE_DEVICES)
+            flag: --init --gpus "device=${CUDA_VISIBLE_DEVICES:-all}"
+            test: 'tests/test_gpu_openacc.py tests/test_gpu_common.py'
+            runner: ["self-hosted", "nvidiagpu"]
+
+          - base: 'bases:nvidia-nvc12'
+            tag: 'nvidia-nvc12'
+            arch: arm64
+            platform: linux/arm64
+            run_tests: false
+            tag_suffix: '-arm64'
+            flag: --init --gpus "device=${CUDA_VISIBLE_DEVICES:-all}"
+            test: ''
+            runner: ubuntu-24.04-arm
+
           # Runtime gpu flags from https://hub.docker.com/r/rocm/tensorflow/
           - base: 'bases:amd'
             tag: 'amd'
@@ -127,9 +149,9 @@ jobs:
             type=raw,value=latest,enable=${{ matrix.base == 'bases:cpu-gcc' }},suffix=${{ matrix.tag_suffix }}
             type=semver,pattern={{raw}},prefix=${{ matrix.tag }}-,enable=${{ github.event_name == 'release' }},suffix=${{ matrix.tag_suffix }}
             # Legacy "gpu" tag
-            type=raw,value=gpu-dev,enable=${{ matrix.base == 'bases:nvidia-nvc' }},suffix=${{ matrix.tag_suffix }}
-            type=semver,pattern={{raw}},prefix=gpu-,enable=${{ github.event_name == 'release' &&  matrix.base == 'bases:nvidia-nvc' }},suffix=${{ matrix.tag_suffix }}
-            type=semver,pattern={{raw}},value=gpu-latest,enable=${{ github.event_name == 'release' &&  matrix.base == 'bases:nvidia-nvc' }},suffix=${{ matrix.tag_suffix }}
+            type=raw,value=gpu-dev,enable=${{ contains(matrix.base, 'bases:nvidia-nvc') }},suffix=${{ matrix.tag_suffix }}
+            type=semver,pattern={{raw}},prefix=gpu-,enable=${{ github.event_name == 'release' &&  contains(matrix.base, 'bases:nvidia-nvc') }},suffix=${{ matrix.tag_suffix }}
+            type=semver,pattern={{raw}},value=gpu-latest,enable=${{ github.event_name == 'release' &&  contains(matrix.base, 'bases:nvidia-nvc') }},suffix=${{ matrix.tag_suffix }}
             # Legacy "cpu" tag
             type=raw,value=cpu-dev,enable=${{ matrix.base == 'bases:cpu-gcc' }},suffix=${{ matrix.tag_suffix }}
             type=semver,pattern={{raw}},prefix=cpu-,enable=${{ github.event_name == 'release' && matrix.base == 'bases:cpu-gcc' }},suffix=${{ matrix.tag_suffix }}
@@ -170,6 +192,9 @@ jobs:
           - tag: 'nvidia-nvc'
             base: 'bases:nvidia-nvc'
             arches: "amd64 arm64"
+          - tag: 'nvidia-nvc12'
+            base: 'bases:nvidia-nvc12'
+            arches: "amd64 arm64"
           - tag: 'gcc'
             base: 'bases:cpu-gcc'
             arches: "amd64 arm64"
@@ -198,9 +223,9 @@ jobs:
             type=raw,value=latest,enable=${{ matrix.base == 'bases:cpu-gcc' }}
             type=semver,pattern={{raw}},prefix=${{ matrix.tag }}-,enable=${{ github.event_name == 'release' }}
             # Legacy "gpu" tag
-            type=raw,value=gpu-dev,enable=${{ matrix.base == 'bases:nvidia-nvc' }}
-            type=semver,pattern={{raw}},prefix=gpu-,enable=${{ github.event_name == 'release' &&  matrix.base == 'bases:nvidia-nvc' }}
-            type=semver,pattern={{raw}},value=gpu-latest,enable=${{ github.event_name == 'release' &&  matrix.base == 'bases:nvidia-nvc' }}
+            type=raw,value=gpu-dev,enable=${{ contains(matrix.base, 'bases:nvidia-nvc') }}
+            type=semver,pattern={{raw}},prefix=gpu-,enable=${{ github.event_name == 'release' &&  contains(matrix.base, 'bases:nvidia-nvc') }}
+            type=semver,pattern={{raw}},value=gpu-latest,enable=${{ github.event_name == 'release' &&  contains(matrix.base, 'bases:nvidia-nvc') }}
             # Legacy "cpu" tag
             type=raw,value=cpu-dev,enable=${{ matrix.base == 'bases:cpu-gcc' }}
             type=semver,pattern={{raw}},prefix=cpu-,enable=${{ github.event_name == 'release' && matrix.base == 'bases:cpu-gcc' }}
@@ -239,6 +264,10 @@ jobs:
             flag: --init --gpus "device=${CUDA_VISIBLE_DEVICES:-all}"
             test: 'tests/test_gpu_openacc.py tests/test_gpu_common.py'
             runner: ["self-hosted", "nvidiagpu"]
+          - tag: 'nvidia-nvc12'
+            flag: --init --gpus "device=${CUDA_VISIBLE_DEVICES:-all}"
+            test: 'tests/test_gpu_openacc.py tests/test_gpu_common.py'
+            runner: ["self-hosted", "nvidiagpu"]
           - tag: 'gcc'
             flag: '--init -t'
             test: 'tests/test_operator.py'

.github/workflows/pytest-gpu.yml

@@ -47,7 +47,7 @@ jobs:
           # -------------------- NVIDIA job --------------------
           - name: pytest-gpu-acc-nvidia
             test_files: "tests/test_adjoint.py tests/test_gpu_common.py tests/test_gpu_openacc.py tests/test_operator.py::TestEstimateMemory"
-            base: "devitocodes/bases:nvidia-nvc"
+            base: "devitocodes/bases:nvidia-nvc12"
             runner_label: nvidiagpu
             test_drive_cmd: "nvidia-smi"
             # Respect CUDA_VISIBLE_DEVICES and also hard-limit Docker to that device.