Merge branch 'main' into faster_set_current

Author: leofang

.github/BACKPORT_BRANCH

@@ -1 +1 @@
-11.8.x
+12.9.x

.github/actions/fetch_ctk/action.yml

@@ -17,7 +17,7 @@ inputs:
     description: "A list of the CTK components to install as a comma-separated list. e.g. 'cuda_nvcc,cuda_nvrtc,cuda_cudart'"
     required: false
     type: string
-    default: "cuda_nvcc,cuda_cudart,cuda_nvrtc,cuda_profiler_api,cuda_cccl,libnvjitlink,libcufile"
+    default: "cuda_nvcc,cuda_cudart,cuda_crt,libnvvm,cuda_nvrtc,cuda_profiler_api,cuda_cccl,libnvjitlink,libcufile"
 
 runs:
   using: composite
@@ -32,15 +32,16 @@ runs:
         if [[ "$CUDA_MAJOR_VER" -lt 12 ]]; then
           CTK_CACHE_COMPONENTS="${CTK_CACHE_COMPONENTS//libnvjitlink/}"
         fi
+        # Conditionally strip out cuda_crt and libnvvm for CUDA versions < 13
+        CUDA_MAJOR_VER="$(cut -d '.' -f 1 <<< ${{ inputs.cuda-version }})"
+        if [[ "$CUDA_MAJOR_VER" -lt 13 ]]; then
+          CTK_CACHE_COMPONENTS="${CTK_CACHE_COMPONENTS//cuda_crt/}"
+          CTK_CACHE_COMPONENTS="${CTK_CACHE_COMPONENTS//libnvvm/}"
+        fi
         # Conditionally strip out libcufile since it does not support Windows
         if [[ "${{ inputs.host-platform }}" == win-* ]]; then
           CTK_CACHE_COMPONENTS="${CTK_CACHE_COMPONENTS//libcufile/}"
         fi
-        # Conditionally strip out libcufile for CUDA versions < 12.2.0 + aarch64 (redist not available)
-        CUDA_MINOR_VER="$(cut -d '.' -f 2 <<< ${{ inputs.cuda-version }})"
-        if [[ ("$CUDA_MAJOR_VER" -lt 12 || "$CUDA_MINOR_VER" -lt 2) && "${{ inputs.host-platform }}" == "linux-aarch64" ]]; then
-          CTK_CACHE_COMPONENTS="${CTK_CACHE_COMPONENTS//libcufile/}"
-        fi
         # Cleanup stray commas after removing components
         CTK_CACHE_COMPONENTS="${CTK_CACHE_COMPONENTS//,,/,}"
 

.github/workflows/build-wheel.yml

@@ -107,7 +107,7 @@ jobs:
           if-no-files-found: error
 
       - name: Build cuda.core wheel
-        uses: pypa/cibuildwheel@95d2f3a92fbf80abe066b09418bbf128a8923df2  # v3.0.1
+        uses: pypa/cibuildwheel@e6de07ed3921b51089aae6981989889cf1eddd0c  # v3.1.1
         env:
           CIBW_BUILD: ${{ env.CIBW_BUILD }}
           CIBW_ARCHS_LINUX: "native"
@@ -149,7 +149,7 @@ jobs:
           cuda-version: ${{ inputs.cuda-version }}
 
       - name: Build cuda.bindings wheel
-        uses: pypa/cibuildwheel@95d2f3a92fbf80abe066b09418bbf128a8923df2  # v3.0.1
+        uses: pypa/cibuildwheel@e6de07ed3921b51089aae6981989889cf1eddd0c  # v3.1.1
         env:
           CIBW_BUILD: ${{ env.CIBW_BUILD }}
           CIBW_ARCHS_LINUX: "native"

.github/workflows/codeql.yml

@@ -31,13 +31,13 @@ jobs:
       uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683  # v4.2.2
 
     - name: Initialize CodeQL
-      uses: github/codeql-action/init@d6bbdef45e766d081b84a2def353b0055f728d3e  # v3.29.3
+      uses: github/codeql-action/init@4e828ff8d448a8a6e532957b1811f387a63867e8  # v3.29.4
       with:
         languages: ${{ matrix.language }}
         build-mode: ${{ matrix.build-mode }}
         queries: security-extended
 
     - name: Perform CodeQL Analysis
-      uses: github/codeql-action/analyze@d6bbdef45e766d081b84a2def353b0055f728d3e  # v3.29.3
+      uses: github/codeql-action/analyze@4e828ff8d448a8a6e532957b1811f387a63867e8  # v3.29.4
       with:
         category: "/language:${{matrix.language}}"

.github/workflows/guess_latest.sh

@@ -6,18 +6,27 @@
 # URL to search
 URL="https://developer.download.nvidia.com/compute/cuda/redist/"
 
+# Ensure exactly one argument is provided
+if [ "$#" -ne 1 ]; then
+    echo "Usage: $0 <CUDA_major_version>"
+    exit 1
+fi
+
+# Accept major version as the first argument
+MAJOR_VERSION="$1"
+
 # Fetch the directory listing and extract the latest version number
 get_latest_version() {
     # Get the HTML content of the page
     local html_content=$(wget -q -O - "$URL")
 
     # Extract links matching the pattern redistrib_?.?.?.json
-    local files=$(echo "$html_content" | grep -oP 'redistrib_[0-9]+\.[0-9]+\.[0-9]+\.json' | cut -d'"' -f2)
+    local files=$(echo "$html_content" | grep -oP "redistrib_${MAJOR_VERSION}\.[0-9]+\.[0-9]+\.json" | cut -d'"' -f2)
 
     # If files were found, extract the version numbers and find the latest
     if [ -n "$files" ]; then
         # Extract just the version numbers using regex
-        local versions=$(echo "$files" | grep -oP 'redistrib_\K[0-9]+\.[0-9]+\.[0-9]+(?=\.json)')
+        local versions=$(echo "$files" | grep -oP "redistrib_\K${MAJOR_VERSION}\.[0-9]+\.[0-9]+(?=\.json)")
 
         # Sort the versions and get the latest
         local latest_version=$(echo "$versions" | sort -V | tail -n 1)

.github/workflows/install_gpu_driver.ps1

@@ -6,9 +6,9 @@
 function Install-Driver {
 
     # Set the correct URL, filename, and arguments to the installer
-    # This driver is picked to support Windows 11 & CUDA 12.8
-    $url = 'https://us.download.nvidia.com/tesla/572.13/572.13-data-center-tesla-desktop-win10-win11-64bit-dch-international.exe';
-    $file_dir = 'C:\NVIDIA-Driver\572.13-data-center-tesla-desktop-win10-win11-64bit-dch-international.exe';
+    # This driver is picked to support Windows 11 & CUDA 13.0
+    $url = 'https://us.download.nvidia.com/tesla/580.88/580.88-data-center-tesla-desktop-win10-win11-64bit-dch-international.exe';
+    $file_dir = 'C:\NVIDIA-Driver\580.88-data-center-tesla-desktop-win10-win11-64bit-dch-international.exe';
     $install_args = '/s /noeula /noreboot';
 
     # Create the folder for the driver download

.github/workflows/test-wheel-linux.yml

@@ -51,7 +51,7 @@ jobs:
           # Add a special entry for the H100 runner on amd64.
           special_runner=""
           if [[ "${ARCH}" == "amd64" ]]; then
-            special_runner="- { ARCH: ${ARCH}, PY_VER: '3.13', CUDA_VER: '12.9.0', LOCAL_CTK: '1', GPU: 'H100', DRIVER: 'latest' }"
+            special_runner="- { ARCH: ${ARCH}, PY_VER: '3.13', CUDA_VER: '13.0.0', LOCAL_CTK: '1', GPU: 'H100', DRIVER: 'latest' }"
           fi
 
           # Please keep the matrices sorted in ascending order by the following:
@@ -62,18 +62,16 @@ jobs:
           #
           export MATRICES="
             pull-request:
-              - { ARCH: ${ARCH}, PY_VER: '3.9',  CUDA_VER: '11.8.0', LOCAL_CTK: '0', GPU: ${gpu}, DRIVER: 'earliest' }
-              - { ARCH: ${ARCH}, PY_VER: '3.9',  CUDA_VER: '12.0.1', LOCAL_CTK: '1', GPU: ${gpu}, DRIVER: 'latest' }
               - { ARCH: ${ARCH}, PY_VER: '3.9',  CUDA_VER: '12.9.0', LOCAL_CTK: '0', GPU: ${gpu}, DRIVER: 'latest' }
-              - { ARCH: ${ARCH}, PY_VER: '3.10', CUDA_VER: '11.8.0', LOCAL_CTK: '1', GPU: ${gpu}, DRIVER: 'earliest' }
+              - { ARCH: ${ARCH}, PY_VER: '3.9',  CUDA_VER: '13.0.0', LOCAL_CTK: '1', GPU: ${gpu}, DRIVER: 'latest' }
               - { ARCH: ${ARCH}, PY_VER: '3.10', CUDA_VER: '12.9.0', LOCAL_CTK: '1', GPU: ${gpu}, DRIVER: 'latest' }
-              - { ARCH: ${ARCH}, PY_VER: '3.11', CUDA_VER: '11.8.0', LOCAL_CTK: '1', GPU: ${gpu}, DRIVER: 'latest' }
+              - { ARCH: ${ARCH}, PY_VER: '3.10', CUDA_VER: '13.0.0', LOCAL_CTK: '0', GPU: ${gpu}, DRIVER: 'latest' }
               - { ARCH: ${ARCH}, PY_VER: '3.11', CUDA_VER: '12.9.0', LOCAL_CTK: '0', GPU: ${gpu}, DRIVER: 'latest' }
-              - { ARCH: ${ARCH}, PY_VER: '3.12', CUDA_VER: '12.0.1', LOCAL_CTK: '1', GPU: ${gpu}, DRIVER: 'earliest' }
+              - { ARCH: ${ARCH}, PY_VER: '3.11', CUDA_VER: '13.0.0', LOCAL_CTK: '1', GPU: ${gpu}, DRIVER: 'latest' }
               - { ARCH: ${ARCH}, PY_VER: '3.12', CUDA_VER: '12.9.0', LOCAL_CTK: '1', GPU: ${gpu}, DRIVER: 'latest' }
-              - { ARCH: ${ARCH}, PY_VER: '3.13', CUDA_VER: '11.8.0', LOCAL_CTK: '0', GPU: ${gpu}, DRIVER: 'latest' }
-              - { ARCH: ${ARCH}, PY_VER: '3.13', CUDA_VER: '12.0.1', LOCAL_CTK: '1', GPU: ${gpu}, DRIVER: 'latest' }
+              - { ARCH: ${ARCH}, PY_VER: '3.12', CUDA_VER: '13.0.0', LOCAL_CTK: '0', GPU: ${gpu}, DRIVER: 'latest' }
               - { ARCH: ${ARCH}, PY_VER: '3.13', CUDA_VER: '12.9.0', LOCAL_CTK: '0', GPU: ${gpu}, DRIVER: 'latest' }
+              - { ARCH: ${ARCH}, PY_VER: '3.13', CUDA_VER: '13.0.0', LOCAL_CTK: '1', GPU: ${gpu}, DRIVER: 'latest' }
               ${special_runner}
             nightly:
               - { ARCH: ${ARCH}, PY_VER: '3.9',  CUDA_VER: '11.8.0', LOCAL_CTK: '0', GPU: ${gpu}, DRIVER: 'earliest' }
@@ -200,13 +198,14 @@ jobs:
 
           OLD_BRANCH=$(cat .github/BACKPORT_BRANCH)
           OLD_BASENAME="cuda-bindings-python${PYTHON_VERSION_FORMATTED}-cuda*-${{ inputs.host-platform }}*"
-          LATEST_PRIOR_RUN_ID=$(gh run list -b ${OLD_BRANCH} -L 1 -w "build-and-test.yml" -s completed -R NVIDIA/cuda-python --json databaseId | jq '.[]| .databaseId')
+          LATEST_PRIOR_RUN_ID=$(gh run list -b ${OLD_BRANCH} -L 1 -w "ci.yml" -s completed -R NVIDIA/cuda-python --json databaseId | jq '.[]| .databaseId')
           if [[ "$LATEST_PRIOR_RUN_ID" == "" ]]; then
             echo "LATEST_PRIOR_RUN_ID not found!"
             exit 1
           fi
 
           gh run download $LATEST_PRIOR_RUN_ID -p ${OLD_BASENAME} -R NVIDIA/cuda-python
+          rm -rf ${OLD_BASENAME}-tests  # exclude cython test artifacts
           ls -al $OLD_BASENAME
           mkdir -p "${{ env.CUDA_BINDINGS_ARTIFACTS_DIR }}"
           mv $OLD_BASENAME/*.whl "${{ env.CUDA_BINDINGS_ARTIFACTS_DIR }}"/
@@ -318,16 +317,16 @@ jobs:
             pip install $(ls cuda_python*.whl)[all]
           fi
 
-      - name: Install cuda.pathfinder nvidia_wheels_cu12
-        if: startsWith(matrix.CUDA_VER, '12.')
+      - name: Install cuda.pathfinder nvidia_wheels_cu13
+        if: startsWith(matrix.CUDA_VER, '13.')
         run: |
           pushd cuda_pathfinder
-          pip install -v .[nvidia_wheels_cu12]
+          pip install -v .[nvidia_wheels_cu13]
           pip freeze
           popd
 
       - name: Run cuda.pathfinder tests with all_must_work
-        if: startsWith(matrix.CUDA_VER, '12.')
+        if: startsWith(matrix.CUDA_VER, '13.')
         env:
           CUDA_PATHFINDER_TEST_LOAD_NVIDIA_DYNAMIC_LIB_STRICTNESS: all_must_work
         run: run-tests pathfinder

.github/workflows/test-wheel-windows.yml

@@ -47,10 +47,10 @@ jobs:
           #
           export MATRICES="
             pull-request:
-              - { ARCH: ${ARCH}, PY_VER: '3.12',  CUDA_VER: '11.8.0', LOCAL_CTK: '0' }
-              - { ARCH: ${ARCH}, PY_VER: '3.12',  CUDA_VER: '11.8.0', LOCAL_CTK: '1' }
               - { ARCH: ${ARCH}, PY_VER: '3.12',  CUDA_VER: '12.9.0', LOCAL_CTK: '0' }
               - { ARCH: ${ARCH}, PY_VER: '3.12',  CUDA_VER: '12.9.0', LOCAL_CTK: '1' }
+              - { ARCH: ${ARCH}, PY_VER: '3.13',  CUDA_VER: '13.0.0', LOCAL_CTK: '0' }
+              - { ARCH: ${ARCH}, PY_VER: '3.13',  CUDA_VER: '13.0.0', LOCAL_CTK: '1' }
             nightly:
               - { ARCH: ${ARCH}, PY_VER: '3.12',  CUDA_VER: '11.8.0', LOCAL_CTK: '0' }
               - { ARCH: ${ARCH}, PY_VER: '3.12',  CUDA_VER: '11.8.0', LOCAL_CTK: '1' }
@@ -178,14 +178,15 @@ jobs:
         run: |
           $OLD_BRANCH = Get-Content .github/BACKPORT_BRANCH
           $OLD_BASENAME = "cuda-bindings-python${env:PYTHON_VERSION_FORMATTED}-cuda*-${{ inputs.host-platform }}*"
-          $runData = gh run list -b $OLD_BRANCH -L 1 -w "build-and-test.yml" -s completed -R NVIDIA/cuda-python --json databaseId | ConvertFrom-Json
+          $runData = gh run list -b $OLD_BRANCH -L 1 -w "ci.yml" -s completed -R NVIDIA/cuda-python --json databaseId | ConvertFrom-Json
           if (-not $runData -or $runData.Length -eq 0 -or -not $runData[0].databaseId -or [string]::IsNullOrEmpty($runData[0].databaseId)) {
               Write-Host "LATEST_PRIOR_RUN_ID not found!"
               exit 1
           }
           $LATEST_PRIOR_RUN_ID = $runData[0].databaseId
 
           gh run download $LATEST_PRIOR_RUN_ID -p $OLD_BASENAME -R NVIDIA/cuda-python
+          Remove-Item -Recurse -Force "${OLD_BASENAME}-tests"  # exclude cython test artifacts
           Get-ChildItem -Path $OLD_BASENAME
           New-Item -Path "${{ env.CUDA_BINDINGS_ARTIFACTS_DIR }}" -ItemType Directory -Force
           Move-Item -Path "$OLD_BASENAME/*.whl" -Destination "${{ env.CUDA_BINDINGS_ARTIFACTS_DIR }}"
@@ -285,17 +286,17 @@ jobs:
             pip install "$((Get-ChildItem -Filter cuda_python*.whl).FullName)[all]"
           }
 
-      - name: Install cuda.pathfinder nvidia_wheels_cu12
-        if: startsWith(matrix.CUDA_VER, '12.')
+      - name: Install cuda.pathfinder nvidia_wheels_cu13
+        if: startsWith(matrix.CUDA_VER, '13.')
         shell: bash --noprofile --norc -xeuo pipefail {0}
         run: |
           pushd cuda_pathfinder
-          pip install -v .[nvidia_wheels_cu12]
+          pip install -v .[nvidia_wheels_cu13]
           pip freeze
           popd
 
       - name: Run cuda.pathfinder tests with all_must_work
-        if: startsWith(matrix.CUDA_VER, '12.')
+        if: startsWith(matrix.CUDA_VER, '13.')
         env:
           CUDA_PATHFINDER_TEST_LOAD_NVIDIA_DYNAMIC_LIB_STRICTNESS: all_must_work
         shell: bash --noprofile --norc -xeuo pipefail {0}

README.md

@@ -38,3 +38,4 @@ The list of available interfaces is:
 * NVRTC
 * nvJitLink
 * NVVM
+* cuFile

ci/tools/env-vars

@@ -60,7 +60,7 @@ elif [[ "${1}" == "test" ]]; then
   # We only test compute-sanitizer on python 3.12 arbitrarily; we don't need to use sanitizer on the entire matrix
   # Only local ctk installs have compute-sanitizer; there is no wheel for it
   if [[ "${PY_VER}" == "3.12" && "${CUDA_VER}" != "11.8.0" && "${LOCAL_CTK}" == 1 && "${HOST_PLATFORM}" == linux* ]]; then
-    echo "LATEST_CUDA_VERSION=$(bash .github/workflows/guess_latest.sh)" >> $GITHUB_ENV
+    echo "LATEST_CUDA_VERSION=$(bash .github/workflows/guess_latest.sh $TEST_CUDA_MAJOR)" >> $GITHUB_ENV
     SETUP_SANITIZER=1
   else
     SETUP_SANITIZER=0