Support CUDA 13

.github/workflows/ci.yml

@@ -105,7 +105,7 @@ jobs:
     strategy:
       matrix:
         platform: [amd64, arm64]
-        cuda_version: ["12.6"]
+        cuda_version: ["12.6", "13.0"]
       fail-fast: false
     uses: ./.github/workflows/dev_environment.yml
     secrets:
@@ -132,7 +132,7 @@ jobs:
     strategy:
       matrix:
         platform: [amd64, arm64]
-        cuda_version: ["12.6"]
+        cuda_version: ["12.6", "13.0"]
       fail-fast: false
     uses: ./.github/workflows/dev_environment.yml
     secrets:
@@ -235,7 +235,7 @@ jobs:
       matrix:
         platform: [amd64, arm64]
         python_version: ['3.11', '3.13']
-        cuda_version: ["12.6"]
+        cuda_version: ["12.6", "13.0"]
       fail-fast: false
     uses: ./.github/workflows/python_wheels.yml
     secrets:
@@ -256,7 +256,7 @@ jobs:
     with:
       cudaq_version: ${{ needs.python_wheels.outputs.cudaq_version }}
       python_versions: "['3.11', '3.13']"
-      cuda_versions: "['', '12.6']"
+      cuda_versions: "['', '12.6', '13.0']"
       wheel_artifacts: 'pycudaq-*'
 
   binaries:
@@ -265,7 +265,7 @@ jobs:
     strategy:
       matrix:
         platform: [amd64, arm64]
-        cuda_version: ["12.6"]
+        cuda_version: ["12.6", "13.0"]
       fail-fast: false
     uses: ./.github/workflows/prebuilt_binaries.yml
     secrets:

.github/workflows/config/gitlab_commits.txt

@@ -1,2 +1,2 @@
 nvidia-mgpu-repo: cuda-quantum/cuquantum-mgpu.git
-nvidia-mgpu-commit: 8346e5d327318e49d1d6a117fccdd2eea78a803d
+nvidia-mgpu-commit: bb46a64f7a17f4aafac019adbdde78edd792b16e

.github/workflows/config/validation_config.json

@@ -75,6 +75,41 @@
                 "cuda_distribution": "fedora42"
             }
         },
+        {
+            "platform": "linux/amd64",
+            "cuda_major": 13,
+            "operating_systems": 
+            [
+                "ubuntu:22.04",
+                "debian:12",
+                "opensuse/leap:15.5",
+                "fedora:42"
+            ],
+            "ubuntu:22.04": 
+            {
+                "libcdev_package": "libc6-dev",
+                "cudart_version": "13.0",
+                "cuda_distribution": "ubuntu2204"
+            },
+            "debian:12": 
+            {
+                "libcdev_package": "libc6-dev",
+                "cudart_version": "13.0",
+                "cuda_distribution": "debian12"
+            },
+            "opensuse/leap:15.5": 
+            {
+                "libcdev_package": "glibc-devel",
+                "cudart_version": "13.0",
+                "cuda_distribution": "opensuse15"
+            },
+            "fedora:42":
+            {
+                "libcdev_package": "glibc-devel",
+                "cudart_version": "13.0",
+                "cuda_distribution": "fedora42"
+            }
+        },
         {
             "platform": "linux/arm64",
             "cuda_major": 12,
@@ -95,6 +130,20 @@
                 "cudart_version": "12.6",
                 "cuda_distribution": "rhel9"
             }
+        },
+        {
+            "platform": "linux/arm64",
+            "cuda_major": 13,
+            "operating_systems": 
+            [
+                "ubuntu:22.04"
+            ],
+            "ubuntu:22.04": 
+            {
+                "libcdev_package": "libc6-dev",
+                "cudart_version": "13.0",
+                "cuda_distribution": "ubuntu2204"
+            }
         }
     ]
 }

.github/workflows/create_cache_command.yml

@@ -112,7 +112,7 @@ jobs:
     strategy:
       matrix:
         platform: [amd64, arm64]
-        cuda_version: ["12.6"]
+        cuda_version: ["12.6", "13.0"]
       fail-fast: false
     uses: ./.github/workflows/dev_environment.yml
     secrets:
@@ -141,7 +141,7 @@ jobs:
     strategy:
       matrix:
         platform: [amd64, arm64]
-        cuda_version: ["12.6"]
+        cuda_version: ["12.6", "13.0"]
       fail-fast: false
     uses: ./.github/workflows/dev_environment.yml
     secrets:

.github/workflows/deployments.yml

@@ -205,7 +205,7 @@ jobs:
         # There are currently no multi-platform manylinux images available.
         # See https://github.com/pypa/manylinux/issues/1306.
         platform: ${{ fromJson(needs.metadata.outputs.platforms).ids }}
-        cuda_version: ["12.6"]
+        cuda_version: ["12.6", "13.0"]
       fail-fast: false
     uses: ./.github/workflows/dev_environment.yml
     secrets:
@@ -237,7 +237,7 @@ jobs:
     strategy:
       matrix:
         platform: ${{ fromJson(needs.metadata.outputs.platforms).ids }}
-        cuda_version: ["12.6"]
+        cuda_version: ["12.6", "13.0"]
       fail-fast: false
     uses: ./.github/workflows/dev_environment.yml
     secrets:
@@ -268,7 +268,7 @@ jobs:
     strategy:
       matrix:
         platform: ${{ fromJson(needs.metadata.outputs.platforms).ids }}
-        cuda_version: ["12.6"]
+        cuda_version: ["12.6", "13.0"]
       fail-fast: false
     uses: ./.github/workflows/dev_environment.yml
     secrets:
@@ -354,7 +354,7 @@ jobs:
     strategy:
       matrix:
         platform: ${{ fromJson(needs.metadata.outputs.platforms).ids }}
-        cuda_version: ["12.6"]
+        cuda_version: ["12.6", "13.0"]
       fail-fast: false
     uses: ./.github/workflows/dev_environment.yml
     secrets:
@@ -402,7 +402,7 @@ jobs:
     strategy:
       matrix:
         platform: ${{ fromJson(needs.metadata.outputs.platforms).ids }}
-        cuda_version: ["12.6"]
+        cuda_version: ["12.6", "13.0"]
       fail-fast: false
     uses: ./.github/workflows/docker_images.yml
     secrets:
@@ -443,7 +443,7 @@ jobs:
     runs-on: ubuntu-latest
     strategy:
       matrix:
-        cuda: ["12.6"]
+        cuda: ["12.6", "13.0"]
 
     environment: ghcr-deployment
 
@@ -666,7 +666,7 @@ jobs:
           STITCHED_JSON='${{ needs.aggregate_stitched_images.outputs.json }}'
           echo "$STITCHED_JSON" | jq .
 
-          for cuda in 12.6; do
+          for cuda in 12.6 13.0; do
             cuda_major="${cuda%%.*}"
             artifact_name="image_cu${cuda_major}_publishing"            
             info_file="$artifact_name.txt"
@@ -704,14 +704,22 @@ jobs:
           retention-days: 30
           if-no-files-found: error
 
+      - name: Upload image_cu13_publishing
+        uses: actions/upload-artifact@v4
+        with:
+          name: image_cu13_publishing
+          path: image_cu13_publishing.txt
+          retention-days: 30
+          if-no-files-found: error
+
   python_wheels:
     name: Create Python wheels
     needs: [metadata, config]
     strategy:
       matrix:
         platform: ${{ fromJson(needs.metadata.outputs.platforms).ids }}
         python_version: ['3.11', '3.12', '3.13']
-        cuda_version: ["12.6"]
+        cuda_version: ["12.6", "13.0"]
       fail-fast: false
     uses: ./.github/workflows/python_wheels.yml
     secrets:
@@ -730,7 +738,7 @@ jobs:
     strategy:
       matrix:
         platform: ${{ fromJson(needs.metadata.outputs.platforms).ids }}
-        cuda_version: ["12.6"]
+        cuda_version: ["12.6", "13.0"]
       fail-fast: false
     uses: ./.github/workflows/prebuilt_binaries.yml
     secrets:

.github/workflows/publishing.yml

@@ -783,7 +783,7 @@ jobs:
     with:
       cudaq_version: ${{ needs.assets.outputs.cudaq_version }}
       python_versions: "['3.11', '3.12', '3.13']"
-      cuda_versions: "['', '12.6']"
+      cuda_versions: "['', '12.6', '13.0']"
       wheel_artifacts: '*-wheels'
       github_commit: ${{ inputs.github_commit || needs.assets.outputs.github_commit }}
 
@@ -925,7 +925,7 @@ jobs:
     strategy:
       matrix:
         os_image: ['redhat/ubi8:8.10', 'ubuntu:24.04']
-        cuda_version: ['12.6']
+        cuda_version: ['12.6', '13.0']
       fail-fast: false
 
     container:
@@ -1143,7 +1143,7 @@ jobs:
 
     strategy:
       matrix:
-        cuda_version: ['11.8', '12.4']
+        cuda_version: ['12.4', '13.0']
       fail-fast: false
 
     # Must have environment to access environment secreats

.github/workflows/python_metapackages.yml

@@ -58,7 +58,7 @@ jobs:
 
             echo "Creating README.md for cudaq package"
             package_name=cudaq
-            cuda_version_requirement="11.x (where x \>= 8) or 12.x"
+            cuda_version_requirement="12.x or 13.x"
             cuda_version_conda=12.4.0 # only used as example in the install script
             deprecation_notice="**Note**: Support for CUDA 11 will be removed in future releases. Please update to CUDA 12."
             cat python/README.md.in > python/metapackages/README.md
@@ -201,9 +201,9 @@ jobs:
               set -e && check_package=cuda-quantum
             ;; 
             *) 
-              # Skip conflict test entirely on Python 3.13 (cuda-quantum==0.8.0 not available)
-              if [ "${{ matrix.python_version }}" = "3.13" ]; then
-                echo "::warning::Skipping conflict test on Python 3.13 (cuda-quantum==0.8.0 unavailable)."
+              # Skip conflict test entirely on Python 3.{13,12} (cuda-quantum==0.8.0 not available)
+              if [ "${{ matrix.python_version }}" = "3.12" ] || [ "${{ matrix.python_version }}" = "3.13" ]; then
+                echo "::warning::Skipping conflict test on Python 3.{12,13} (cuda-quantum==0.8.0 unavailable)."
                 test_conflicting="<skipped>"
                 check_package=cudaq
               else

CMakeLists.txt

@@ -537,8 +537,17 @@ endfunction()
 
 if(CMAKE_CUDA_COMPILER)
   if (NOT CUDA_TARGET_ARCHS)
-    # Volta, Ampere, Hopper
-    set(CUDA_TARGET_ARCHS  "70;80;90")
+    find_package(CUDAToolkit REQUIRED)
+
+    message(STATUS "Found CUDA Toolkit version: ${CUDAToolkit_VERSION}")
+
+    if (CUDAToolkit_VERSION VERSION_LESS 13.0)
+      # Turing, Ampere, Hopper
+      set(CUDA_TARGET_ARCHS  "75;80;90")
+    else()
+      # Turing, Ampere, Hopper, Blackwell
+      set(CUDA_TARGET_ARCHS  "75;80;90;100")
+    endif()
   endif()
   CUDA_get_gencode_args(CUDA_gencode_flags ${CUDA_TARGET_ARCHS})
   set(CMAKE_CUDA_FLAGS "${CMAKE_CUDA_FLAGS} -shared -std=c++17 ${CUDA_gencode_flags} --compiler-options -fPIC")

docker/build/assets.Dockerfile

@@ -187,15 +187,20 @@ RUN dnf install -y --nobest --setopt=install_weak_deps=False ${PYTHON}-devel &&
     ${PYTHON} -m pip install numpy build auditwheel patchelf
 
 RUN cd /cuda-quantum && source scripts/configure_build.sh && \
-    if [ "${CUDA_VERSION#11.}" != "${CUDA_VERSION}" ]; then \
-        cublas_version=11.11 && \
-        cusparse_version=11.7 && \
-        sed -i "s/-cu12/-cu11/g" pyproject.toml && \
-        sed -i "s/-cuda12/-cuda11/g" pyproject.toml && \
+    if [ "${CUDA_VERSION#12.}" != "${CUDA_VERSION}" ]; then \
+        cublas_version=12.0 && \
+        cusolver_version=11.4 && \
+        cuda_runtime_version=12.0 && \
+        cuda_nvrtc_version=12.0 && \
+        cupy_version=13.4.1 && \
+        sed -i "s/-cu13/-cu12/g" pyproject.toml && \
+        sed -i "s/-cuda13/-cuda12/g" pyproject.toml && \
+        sed -i -E "s/cupy-cuda[0-9]+x/cupy-cuda12x/g" pyproject.toml && \
+        sed -i -E "s/(cupy-cuda[0-9]+x? ~= )[0-9\.]*/\1${cupy_version}/g" pyproject.toml && \
         sed -i -E "s/(nvidia-cublas-cu[0-9]* ~= )[0-9\.]*/\1${cublas_version}/g" pyproject.toml && \
-        sed -i -E "s/(nvidia-cusparse-cu[0-9]* ~= )[0-9\.]*/\1${cusparse_version}/g" pyproject.toml && \
-        sed -i -E "s/(nvidia-cuda-nvrtc-cu[0-9]* ~= )[0-9\.]*/\1${CUDA_VERSION}/g" pyproject.toml && \
-        sed -i -E "s/(nvidia-cuda-runtime-cu[0-9]* ~= )[0-9\.]*/\1${CUDA_VERSION}/g" pyproject.toml; \
+        sed -i -E "s/(nvidia-cusolver-cu[0-9]* ~= )[0-9\.]*/\1${cusolver_version}/g" pyproject.toml && \
+        sed -i -E "s/(nvidia-cuda-nvrtc-cu[0-9]* ~= )[0-9\.]*/\1${cuda_nvrtc_version}/g" pyproject.toml && \
+        sed -i -E "s/(nvidia-cuda-runtime-cu[0-9]* ~= )[0-9\.]*/\1${cuda_runtime_version}/g" pyproject.toml; \
     fi && \
     # Needed to retrigger the LLVM build, since the MLIR Python bindings
     # are not built in the prereqs stage.

docker/build/devdeps.ext.Dockerfile

@@ -175,7 +175,12 @@ ENV PIP_BREAK_SYSTEM_PACKAGES=1
 RUN apt-get update && apt-get install -y --no-install-recommends \
         python3 python3-pip && \
     apt-get autoremove -y && apt-get clean && rm -rf /var/lib/apt/lists/* && \
-    python3 -m pip install --break-system-packages cupy-cuda$(echo $CUDA_VERSION | cut -d . -f1)x==13.4.1 cuquantum-cu$(echo $CUDA_VERSION | cut -d . -f1)==25.09 && \
+    if [ "$(echo $CUDA_VERSION | cut -d . -f1)" = "13" ]; then \
+        cupy_version=13.6.0; \
+    else \
+        cupy_version=13.4.1; \
+    fi && \
+    python3 -m pip install --break-system-packages cupy-cuda$(echo $CUDA_VERSION | cut -d . -f1)x==${cupy_version} cuquantum-cu$(echo $CUDA_VERSION | cut -d . -f1)==25.09 && \
     if [ "$(python3 --version | grep -o [0-9\.]* | cut -d . -f -2)" != "3.12" ]; then \
         echo "expecting Python version 3.12"; \
     fi

docker/build/devdeps.manylinux.Dockerfile

@@ -95,7 +95,7 @@ RUN LLVM_PROJECTS='clang;mlir' LLVM_SOURCE=/llvm-project \
 
 # Install CUDA
 
-ARG cuda_version=11.8
+ARG cuda_version=12.6
 ENV CUDA_VERSION=${cuda_version}
 
 RUN arch_folder=$([ "$(uname -m)" == "aarch64" ] && echo sbsa || echo x86_64) \

docker/build/devdeps.ompi.Dockerfile

@@ -81,12 +81,12 @@ RUN apt-get update -y && apt-get install -y --no-install-recommends \
 ENV CPATH="$GDRCOPY_INSTALL_PREFIX/include:$CPATH"
 ENV LIBRARY_PATH="$GDRCOPY_INSTALL_PREFIX/lib64:$LIBRARY_PATH"
 
-# 5 - Install UCX version v1.16.0
+# 5 - Install UCX version v1.19.0
 
 ENV UCX_INSTALL_PREFIX=/usr/local/ucx
 RUN mkdir -p /var/tmp && cd /var/tmp \
     && git clone https://github.com/openucx/ucx.git ucx && cd /var/tmp/ucx \
-    && git checkout v1.16.0 \
+    && git checkout v1.19.0 \
     && ./autogen.sh \
     && export common_flags=$([ "$TARGETARCH" == "arm64" ] && echo "$COMMON_COMPILER_FLAGS_ARM" || echo "$COMMON_COMPILER_FLAGS") \
     &&  CC=gcc CFLAGS="$common_flags" \