Update LLVM Version from 16->22.1

.github/required-checks.yml

@@ -8,10 +8,10 @@
 # An empty list (or missing key) for an event type disables enforcement
 # for that event — useful for bootstrapping.
 merge_group: &full_matrix
-- Build and test (arm64, gcc11, openmpi) / Dev environment (Debug)
-- Build and test (arm64, gcc11, openmpi) / Dev environment (Python)
-- Build and test (arm64, gcc12, openmpi) / Dev environment (Debug)
-- Build and test (arm64, gcc12, openmpi) / Dev environment (Python)
+- Build and test (amd64, llvm, openmpi) / Dev environment (Debug)
+- Build and test (amd64, llvm, openmpi) / Dev environment (Python)
+- Build and test (arm64, llvm, openmpi) / Dev environment (Debug)
+- Build and test (arm64, llvm, openmpi) / Dev environment (Python)
 - Create CUDA Quantum installer (amd64, 12.6) / Build CUDA Quantum assets
 - Create CUDA Quantum installer (amd64, 12.6) / Minimal OpenMPI installation
 - Create CUDA Quantum installer (amd64, 12.6) / Validate installer (debian:12)
@@ -44,13 +44,40 @@ merge_group: &full_matrix
 - Create Python wheels (arm64, 3.11, 12.6) / Validate wheel (redhat/ubi8:8.10, --user)
 - Create Python wheels (arm64, 3.11, 12.6) / Validate wheel (redhat/ubi8:8.10)
 push:
-- Build and test (amd64, clang16, openmpi) / Dev environment (Debug)
-- Build and test (amd64, clang16, openmpi) / Dev environment (Python)
-- Build and test (amd64, gcc11, openmpi) / Dev environment (Debug)
-- Build and test (amd64, gcc11, openmpi) / Dev environment (Python)
-- Build and test (amd64, gcc12, openmpi) / Dev environment (Debug)
-- Build and test (amd64, gcc12, openmpi) / Dev environment (Python)
-- Build and test (arm64, clang16, openmpi) / Dev environment (Debug)
-- Build and test (arm64, clang16, openmpi) / Dev environment (Python)
+- Build and test (amd64, llvm, openmpi) / Dev environment (Debug)
+- Build and test (amd64, llvm, openmpi) / Dev environment (Python)
+- Build and test (arm64, llvm, openmpi) / Dev environment (Debug)
+- Build and test (arm64, llvm, openmpi) / Dev environment (Python)
+- Create CUDA Quantum installer (amd64, 12.6) / Build CUDA Quantum assets
+- Create CUDA Quantum installer (amd64, 12.6) / Minimal OpenMPI installation
+- Create CUDA Quantum installer (amd64, 12.6) / Validate installer (debian:12)
+- Create CUDA Quantum installer (amd64, 12.6) / Validate installer (fedora:42)
+- Create CUDA Quantum installer (amd64, 12.6) / Validate installer (opensuse/leap:15.5)
+- Create CUDA Quantum installer (amd64, 12.6) / Validate installer (redhat/ubi9:9.6)
+- Create CUDA Quantum installer (amd64, 12.6) / Validate installer (ubuntu:22.04)
+- Create CUDA Quantum installer (arm64, 12.6) / Build CUDA Quantum assets
+- Create CUDA Quantum installer (arm64, 12.6) / Minimal OpenMPI installation
+- Create CUDA Quantum installer (arm64, 12.6) / Validate installer (redhat/ubi9:9.6)
+- Create CUDA Quantum installer (arm64, 12.6) / Validate installer (ubuntu:22.04)
+- Create Docker images (amd64) / Documentation
+- Create Docker images (amd64) / Validation
+- Create Docker images (arm64) / Validation
+- Create Python metapackages / Build Python metapackages
+- Create Python metapackages / Test Python metapackages (3.11)
+- Create Python metapackages / Test Python metapackages (3.13)
+- Create Python metapackages / Test Python metapackages (12.6, 3.11)
+- Create Python metapackages / Test Python metapackages (12.6, 3.13)
+- Create Python wheels (amd64, 3.11, 12.6) / Validate wheel (debian:12, --user)
+- Create Python wheels (amd64, 3.11, 12.6) / Validate wheel (debian:12)
+- Create Python wheels (amd64, 3.11, 12.6) / Validate wheel (fedora:42, --user)
+- Create Python wheels (amd64, 3.11, 12.6) / Validate wheel (fedora:42)
+- Create Python wheels (amd64, 3.11, 12.6) / Validate wheel (redhat/ubi8:8.10, --user)
+- Create Python wheels (amd64, 3.11, 12.6) / Validate wheel (redhat/ubi8:8.10)
+- Create Python wheels (arm64, 3.11, 12.6) / Validate wheel (debian:12, --user)
+- Create Python wheels (arm64, 3.11, 12.6) / Validate wheel (debian:12)
+- Create Python wheels (arm64, 3.11, 12.6) / Validate wheel (fedora:42, --user)
+- Create Python wheels (arm64, 3.11, 12.6) / Validate wheel (fedora:42)
+- Create Python wheels (arm64, 3.11, 12.6) / Validate wheel (redhat/ubi8:8.10, --user)
+- Create Python wheels (arm64, 3.11, 12.6) / Validate wheel (redhat/ubi8:8.10)
 workflow_dispatch: *full_matrix
 schedule: *full_matrix

.github/workflows/build_package_sources.yml

@@ -67,7 +67,7 @@ jobs:
           else
             # cudaqx: devcontainer base, cudaqx target
             cu_tag=$(echo "${{ matrix.cuda }}" | tr -d .)
-            echo "base_image=ghcr.io/nvidia/cuda-quantum-devcontainer:amd64-cu${{ matrix.cuda }}-gcc11-main" | tee -a $GITHUB_OUTPUT
+            echo "base_image=ghcr.io/nvidia/cuda-quantum-devcontainer:amd64-cu${{ matrix.cuda }}-llvm-main" | tee -a $GITHUB_OUTPUT
             echo "target_image=ghcr.io/nvidia/cudaqx:cu${cuda_major}-latest" | tee -a $GITHUB_OUTPUT
           fi
 

.github/workflows/ci.yml

@@ -42,6 +42,7 @@ jobs:
       cache_base: ${{ steps.pr_info.outputs.pr_base }}
       llvm_commit: ${{ steps.repo_info.outputs.llvm_commit }}
       pybind11_commit: ${{ steps.repo_info.outputs.pybind11_commit }}
+      nanobind_commit: ${{ steps.repo_info.outputs.nanobind_commit }}
       platform_config: ${{ steps.config.outputs.platforms }}
       build_test_matrix: ${{ steps.config.outputs.build_test_matrix }}
 
@@ -65,19 +66,13 @@ jobs:
           # (expensive, redundant with amd64/gcc). merge_group/dispatch run all.
           if [ "${{ github.event_name }}" = "push" ]; then
             build_test_matrix='{"include":[
-              {"platform":"amd64","toolchain":"clang16","mpi":"openmpi"},
-              {"platform":"amd64","toolchain":"gcc11","mpi":"openmpi"},
-              {"platform":"amd64","toolchain":"gcc12","mpi":"openmpi"},
-              {"platform":"arm64","toolchain":"clang16","mpi":"openmpi"}
+              {"platform":"amd64","toolchain":"llvm","mpi":"openmpi"},
+              {"platform":"arm64","toolchain":"llvm","mpi":"openmpi"}
             ]}'
           else
             build_test_matrix='{"include":[
-              {"platform":"amd64","toolchain":"clang16","mpi":"openmpi"},
-              {"platform":"amd64","toolchain":"gcc11","mpi":"openmpi"},
-              {"platform":"amd64","toolchain":"gcc12","mpi":"openmpi"},
-              {"platform":"arm64","toolchain":"clang16","mpi":"openmpi"},
-              {"platform":"arm64","toolchain":"gcc11","mpi":"openmpi"},
-              {"platform":"arm64","toolchain":"gcc12","mpi":"openmpi"}
+              {"platform":"amd64","toolchain":"llvm","mpi":"openmpi"},
+              {"platform":"arm64","toolchain":"llvm","mpi":"openmpi"},
             ]}'
           fi
           echo "build_test_matrix=$(echo "$build_test_matrix" | jq -c .)" >> $GITHUB_OUTPUT
@@ -114,14 +109,14 @@ jobs:
         run: |
           echo "llvm_commit=$(git rev-parse @:./tpls/llvm)" >> $GITHUB_OUTPUT
           echo "pybind11_commit=$(git rev-parse @:./tpls/pybind11)" >> $GITHUB_OUTPUT
+          echo "nanobind_commit=$(git rev-parse @:./tpls/nanobind)" >> $GITHUB_OUTPUT
 
   devdeps:
     name: Load dependencies
     needs: metadata
     strategy:
       matrix:
         platform: [amd64, arm64]
-        toolchain: [clang16, gcc11, gcc12]
       fail-fast: false
     uses: ./.github/workflows/dev_environment.yml
     secrets:
@@ -130,14 +125,12 @@ jobs:
     with:
       platforms: linux/${{ matrix.platform }}
       dockerfile: build/devdeps.Dockerfile
-      build_config_id: ${{ matrix.toolchain }}
-      build_args: |
-        toolchain=${{ matrix.toolchain }}
+      build_config_id: llvm
       registry_cache_from: ${{ inputs.cache_base || needs.metadata.outputs.cache_base }}
       checkout_submodules: true
       environment: ghcr-ci
       # needed only for the cloudposse GitHub action
-      matrix_key: ${{ matrix.platform }}-${{ matrix.toolchain }}
+      matrix_key: ${{ matrix.platform }}-llvm
       matrix_step_name: dev_environment_devdeps
 
   wheeldeps:
@@ -155,14 +148,15 @@ jobs:
     with:
       platforms: linux/${{ matrix.platform }}
       dockerfile: build/devdeps.manylinux.Dockerfile
-      build_config_id: cu${{ matrix.cuda_version }}-gcc11
+      build_config_id: cu${{ matrix.cuda_version }}-gcc12
       build_args: |
         base_image=ghcr.io/nvidia/pypa/manylinux_2_28${{ (matrix.platform == 'arm64' && '_aarch64') || (matrix.platform == 'amd64' && '_x86_64') || '' }}:latest
         cuda_version=${{ matrix.cuda_version }}
-        toolchain=gcc11
+        toolchain=gcc12
         distro=rhel8
         llvm_commit=${{ needs.metadata.outputs.llvm_commit }}
         pybind11_commit=${{ needs.metadata.outputs.pybind11_commit }}
+        nanobind_commit=${{ needs.metadata.outputs.nanobind_commit }}
       registry_cache_from: ${{ inputs.cache_base || needs.metadata.outputs.cache_base }}
       environment: ghcr-ci
       # needed only for the cloudposse GitHub action
@@ -259,16 +253,15 @@ jobs:
     strategy:
       matrix:
         platform: [amd64]
-        toolchain: [clang16]
       fail-fast: false
     uses: ./.github/workflows/generate_cc.yml
     secrets:
       CODECOV_TOKEN: ${{ secrets.CODECOV_TOKEN }}
     with:
       platform: linux/${{ matrix.platform }}
-      devdeps_image: ${{ fromJson(needs.config_devdeps.outputs.json).image_hash[format('{0}-{1}', matrix.platform, matrix.toolchain)] }}
-      devdeps_cache: ${{ fromJson(needs.config_devdeps.outputs.json).cache_key[format('{0}-{1}', matrix.platform, matrix.toolchain)] }}
-      devdeps_archive: ${{ fromJson(needs.config_devdeps.outputs.json).tar_archive[format('{0}-{1}', matrix.platform, matrix.toolchain)] }}
+      devdeps_image: ${{ fromJson(needs.config_devdeps.outputs.json).image_hash[format('{0}-llvm', matrix.platform)] }}
+      devdeps_cache: ${{ fromJson(needs.config_devdeps.outputs.json).cache_key[format('{0}-llvm', matrix.platform)] }}
+      devdeps_archive: ${{ fromJson(needs.config_devdeps.outputs.json).tar_archive[format('{0}-llvm', matrix.platform)] }}
       export_environment: ${{ github.event_name == 'workflow_dispatch' && inputs.export_environment }}
 
   # Docker images are packaging, not correctness — only built on merge_group/dispatch.
@@ -286,9 +279,9 @@ jobs:
       DOCKERHUB_READONLY_TOKEN: ${{ secrets.DOCKERHUB_READONLY_TOKEN }}
     with:
       platforms: linux/${{ matrix.platform }}
-      devdeps_image: ${{ fromJson(needs.config_devdeps.outputs.json).image_hash[format('{0}-gcc11', matrix.platform)] }}
-      devdeps_cache: ${{ fromJson(needs.config_devdeps.outputs.json).cache_key[format('{0}-gcc11', matrix.platform)] }}
-      devdeps_archive: ${{ fromJson(needs.config_devdeps.outputs.json).tar_archive[format('{0}-gcc11', matrix.platform)] }}
+      devdeps_image: ${{ fromJson(needs.config_devdeps.outputs.json).image_hash[format('{0}-llvm', matrix.platform)] }}
+      devdeps_cache: ${{ fromJson(needs.config_devdeps.outputs.json).cache_key[format('{0}-llvm', matrix.platform)] }}
+      devdeps_archive: ${{ fromJson(needs.config_devdeps.outputs.json).tar_archive[format('{0}-llvm', matrix.platform)] }}
       environment: ghcr-ci
 
   python_wheels:

.github/workflows/clean_caches.yml

@@ -125,14 +125,14 @@ jobs:
       - name: Delete build caches for MPI asset (ARM64)
         uses: actions/delete-package-versions@v5
         with: 
-          package-name: buildcache-cuda-quantum-assets-openmpi-gcc11-arm64
+          package-name: buildcache-cuda-quantum-assets-openmpi-llvm-arm64
           package-type: 'container'
           min-versions-to-keep: 1 # the used action does not support 0 here
 
       - name: Delete build caches for MPI asset (AMD64)
         uses: actions/delete-package-versions@v5
         with: 
-          package-name: buildcache-cuda-quantum-assets-openmpi-gcc11-amd64
+          package-name: buildcache-cuda-quantum-assets-openmpi-llvm-amd64
           package-type: 'container'
           min-versions-to-keep: 1 # the used action does not support 0 here
 

.github/workflows/create_cache_command.yml

@@ -56,6 +56,7 @@ jobs:
     outputs:
       llvm_commit: ${{ steps.repo_info.outputs.llvm_commit }}
       pybind11_commit: ${{ steps.repo_info.outputs.pybind11_commit }}
+      nanobind_commit: ${{ steps.repo_info.outputs.nanobind_commit }}
       platform_config: ${{ steps.config.outputs.platforms }}
 
     steps:
@@ -80,14 +81,14 @@ jobs:
         run: |
           echo "llvm_commit=$(git rev-parse @:./tpls/llvm)" >> $GITHUB_OUTPUT
           echo "pybind11_commit=$(git rev-parse @:./tpls/pybind11)" >> $GITHUB_OUTPUT
+          echo "nanobind_commit=$(git rev-parse @:./tpls/nanobind)" >> $GITHUB_OUTPUT
 
   devdeps_caches:
     name: Cache dev dependencies
     needs: [pr_info, metadata]
     strategy:
       matrix:
         platform: [amd64, arm64]
-        toolchain: [clang16, gcc11, gcc12]
       fail-fast: false
     uses: ./.github/workflows/dev_environment.yml
     secrets:
@@ -96,15 +97,13 @@ jobs:
     with:
       platforms: linux/${{ matrix.platform }}
       dockerfile: build/devdeps.Dockerfile
-      build_config_id: ${{ matrix.toolchain }}
-      build_args: |
-        toolchain=${{ matrix.toolchain }}
+      build_config_id: llvm
       create_local_cache: true
       registry_cache_from: ${{ needs.pr_info.outputs.target_branch }}
       pull_request_number: ${{ needs.pr_info.outputs.pull_request_number }}
       checkout_submodules: true
       # needed only for the cloudposse GitHub action
-      matrix_key: ${{ matrix.platform }}-${{ matrix.toolchain }}
+      matrix_key: ${{ matrix.platform }}-llvm
 
   wheeldeps_caches:
     name: Cache wheel dependencies
@@ -121,14 +120,15 @@ jobs:
     with:
       platforms: linux/${{ matrix.platform }}
       dockerfile: build/devdeps.manylinux.Dockerfile
-      build_config_id: cu${{ matrix.cuda_version }}-gcc11
+      build_config_id: cu${{ matrix.cuda_version }}-gcc12
       build_args: |
         base_image=ghcr.io/nvidia/pypa/manylinux_2_28${{ (matrix.platform == 'arm64' && '_aarch64') || (matrix.platform == 'amd64' && '_x86_64') || '' }}:latest
         cuda_version=${{ matrix.cuda_version }}
-        toolchain=gcc11
+        toolchain=gcc12
         distro=rhel8
         llvm_commit=${{ needs.metadata.outputs.llvm_commit }}
         pybind11_commit=${{ needs.metadata.outputs.pybind11_commit }}
+        nanobind_commit=${{ needs.metadata.outputs.nanobind_commit }}
       create_local_cache: true
       registry_cache_from: ${{ needs.pr_info.outputs.target_branch }}
       pull_request_number: ${{ needs.pr_info.outputs.pull_request_number }}

.github/workflows/deployments.yml

@@ -77,6 +77,7 @@ jobs:
       pull_request_commit: ${{ steps.pr_info.outputs.merge_commit }}
       llvm_commit: ${{ steps.build_config.outputs.llvm_commit }}
       pybind11_commit: ${{ steps.build_config.outputs.pybind11_commit }}
+      nanobind_commit: ${{ steps.build_config.outputs.nanobind_commit }}
       cache_base: ${{ steps.build_info.outputs.cache_base }}
       cache_target: ${{ steps.build_info.outputs.cache_target }}
       multi_platform: ${{ steps.build_info.outputs.multi_platform }}
@@ -188,6 +189,7 @@ jobs:
         run: |
           echo "llvm_commit=$(git rev-parse @:./tpls/llvm)" >> $GITHUB_OUTPUT
           echo "pybind11_commit=$(git rev-parse @:./tpls/pybind11)" >> $GITHUB_OUTPUT
+          echo "nanobind_commit=$(git rev-parse @:./tpls/nanobind)" >> $GITHUB_OUTPUT
 
           if ${{ github.event_name != 'workflow_run' || steps.pr_info.outputs.pr_number != '' }}; then
             echo "build_dependencies=true" >> $GITHUB_OUTPUT
@@ -203,7 +205,7 @@ jobs:
     strategy:
       matrix:
         platform: ${{ fromJson(needs.metadata.outputs.platforms).ids }}
-        toolchain: [clang16, gcc11, gcc12]
+        toolchain: [llvm]
       fail-fast: false
     uses: ./.github/workflows/dev_environment.yml
     secrets:
@@ -242,14 +244,15 @@ jobs:
     with:
       platforms: ${{ fromJson(needs.metadata.outputs.platforms)[format('{0}', matrix.platform)].docker_flag }}
       dockerfile: build/devdeps.manylinux.Dockerfile
-      build_config_id: cu${{ matrix.cuda_version }}-gcc11
+      build_config_id: cu${{ matrix.cuda_version }}-gcc12
       build_args: |
         base_image=ghcr.io/nvidia/pypa/manylinux_2_28${{ (matrix.platform == 'arm64' && '_aarch64') || (matrix.platform == 'amd64' && '_x86_64') || '' }}:latest
         cuda_version=${{ matrix.cuda_version }}
-        toolchain=gcc11
+        toolchain=gcc12
         distro=rhel8
         llvm_commit=${{ needs.metadata.outputs.llvm_commit }}
         pybind11_commit=${{ needs.metadata.outputs.pybind11_commit }}
+        nanobind_commit=${{ needs.metadata.outputs.nanobind_commit }}
       registry_cache_from: ${{ needs.metadata.outputs.cache_base }}
       update_registry_cache: ${{ needs.metadata.outputs.cache_target }}
       pull_request_number: ${{ needs.metadata.outputs.pull_request_number }}
@@ -368,7 +371,7 @@ jobs:
     strategy:
       matrix:
         platform: [amd64]
-        toolchain: [clang16]
+        toolchain: [llvm]
       fail-fast: false
     uses: ./.github/workflows/generate_cc.yml
     secrets:
@@ -393,10 +396,10 @@ jobs:
     with:
       platforms: ${{ fromJson(needs.metadata.outputs.platforms)[format('{0}', matrix.platform)].docker_flag }}
       dockerfile: build/devcontainer.Dockerfile
-      build_config_id: cu${{ matrix.cuda_version }}-gcc11
+      build_config_id: cu${{ matrix.cuda_version }}-llvm
       build_args: |
         cuda_version=${{ matrix.cuda_version }}
-        base_image=${{ fromJson(needs.config.outputs.json).image_hash[format('{0}-gcc11', matrix.platform)] }}
+        base_image=${{ fromJson(needs.config.outputs.json).image_hash[format('{0}-llvm', matrix.platform)] }}
         ompidev_image=${{ fromJson(needs.config.outputs.json).image_hash[format('{0}-cu{1}-ompi', matrix.platform, matrix.cuda_version)] }}
       registry_cache_from: ${{ needs.metadata.outputs.cache_base }}
       update_registry_cache: ${{ needs.metadata.outputs.cache_target }}

.github/workflows/dev_environment_macos.yml

@@ -95,6 +95,7 @@ jobs:
             scripts/build_llvm.sh \
             scripts/set_env_defaults.sh \
             .github/workflows/dev_environment_macos.yml \
+            tpls/customizations/llvm/*.diff \
             | sha256sum | cut -c1-8)
           echo "scripts_hash=$scripts_hash" >> $GITHUB_OUTPUT
 
@@ -181,6 +182,9 @@ jobs:
           # cache already has MLIR_ENABLE_BINDINGS_PYTHON=ON.  Downstream
           # wheel jobs only change Python3_EXECUTABLE, which keeps ninja's
           # incremental rebuild scoped to the binding targets.
+          # Initialize nanobind submodule which are needed for MLIR Python bindings
+          git submodule update --init --recursive tpls/nanobind
+
           source scripts/set_env_defaults.sh
           export LLVM_PROJECTS='clang;lld;mlir;openmp;python-bindings'
 

.github/workflows/test_in_devenv.yml

@@ -297,6 +297,7 @@ jobs:
               echo "::error file=test_in_devenv.yml:: Pip install of CUDA Quantum failed with status $pyinstall_status."
               exit 1
             fi
+
             python -m pip install pytest pytest-xdist
             python -m pytest -v --durations=0 -n auto python/tests/ \
               --ignore python/tests/backends \

.gitmodules

@@ -51,3 +51,4 @@
 [submodule "tpls/nanobind"]
 	path = tpls/nanobind
 	url = https://github.com/wjakob/nanobind.git
+	ignore = dirty

.pre-commit-config.yaml

@@ -53,7 +53,6 @@ repos:
       - id: markdownlint
         name: Markdown linting
         files: '\.md$'
-        exclude: '^tpls/'
         args: ['--config', '.github/pre-commit/md_lint_config.yml']
 
   # Standard quality checks

Building.md

@@ -69,7 +69,10 @@ CUDA-Q can be built on macOS for development purposes. Note that:
 
 - **ARM64 only**: Only Apple silicon Macs are supported; Intel Macs are not supported
 - **CPU-only**: No CUDA/GPU support is available on macOS
-- **Apple Clang**: Uses the system compiler (no need to install GCC or LLVM separately)
+- **LLVM 22.1 toolchain**: CUDA-Q is compiled with the `Clang/LLD/libomp` built
+  by `scripts/build_llvm.sh` — the same toolchain used on Linux. Xcode Command
+  Line Tools are still required for the macOS `SDK/sysroot` (headers, frameworks,
+  `xcrun`), but Apple Clang is no longer used to compile CUDA-Q itself.
 - **Prerequisites required**: You must use `-p` to install LLVM and other dependencies
 
 Before building, complete the macOS setup steps in