Merge branch 'main' into nvvm-discovery

Author: rwgk

.github/workflows/ci.yml

@@ -221,6 +221,7 @@ jobs:
     uses: ./.github/workflows/build-docs.yml
     with:
       build-ctk-ver: ${{ needs.ci-vars.outputs.CUDA_BUILD_VER }}
+      is-release: ${{ github.ref_type == 'tag' }}
 
   checks:
     name: Check job status

.github/workflows/release-cuda-pathfinder.yml

@@ -4,8 +4,8 @@
 
 # One-click release workflow for cuda-pathfinder.
 #
-# Provide a version number and commit SHA.  The workflow automatically finds
-# the CI run and creates the git tag, creates a draft GitHub release with the standard
+# Provide a release tag. The workflow finds
+# the CI run, creates a draft GitHub release with the standard
 # body, builds versioned docs, uploads source archive + wheels to the
 # release, publishes to TestPyPI, verifies the install, publishes to PyPI,
 # verifies again, and finally marks the release as published.
@@ -15,12 +15,8 @@ name: "Release: cuda-pathfinder"
 on:
   workflow_dispatch:
     inputs:
-      version:
-        description: "Version to release (e.g. 1.3.5)"
-        required: true
-        type: string
-      commit:
-        description: "Commit SHA to release (must be on default branch)"
+      tag:
+        description: "Release tag to publish (e.g. cuda-pathfinder-v1.3.5)"
         required: true
         type: string
 
@@ -34,7 +30,7 @@ defaults:
 
 jobs:
   # --------------------------------------------------------------------------
-  # Validate inputs, find the CI run, create the tag + draft release.
+  # Collect release metadata, find the CI run, create a draft release.
   # --------------------------------------------------------------------------
   prepare:
     runs-on: ubuntu-latest
@@ -53,52 +49,22 @@ jobs:
             exit 1
           fi
 
-      - name: Validate version
+      - name: Set release variables
         id: vars
         env:
-          VERSION_INPUT: ${{ inputs.version }}
+          TAG_INPUT: ${{ inputs.tag }}
         run: |
-          # Strip leading "v" if present (common typo)
-          version="${VERSION_INPUT#v}"
-          if [[ ! "${version}" =~ ^[0-9]+\.[0-9]+\.[0-9]+$ ]]; then
-            echo "::error::Version must be MAJOR.MINOR.PATCH, got: ${version}"
-            exit 1
-          fi
-          tag="cuda-pathfinder-v${version}"
+          version="${TAG_INPUT#cuda-pathfinder-v}"
           {
-            echo "tag=${tag}"
+            echo "tag=${TAG_INPUT}"
             echo "version=${version}"
           } >> "$GITHUB_OUTPUT"
 
-      - name: Validate commit input format
-        env:
-          COMMIT_INPUT: ${{ inputs.commit }}
-        run: |
-          # Require a full SHA to avoid ambiguity and accidental releases.
-          if [[ ! "${COMMIT_INPUT}" =~ ^[0-9a-fA-F]{40}$ ]]; then
-            echo "::error::Commit must be a full 40-character SHA, got: ${COMMIT_INPUT}"
-            exit 1
-          fi
-
       - uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd  # v6.0.2
         with:
           fetch-depth: 0
-          ref: ${{ inputs.commit }}
-
-      - name: Validate and resolve commit
-        id: commit
-        env:
-          DEFAULT_BRANCH: ${{ github.event.repository.default_branch }}
-        run: |
-          git fetch --no-tags origin "${DEFAULT_BRANCH}"
-          commit=$(git rev-parse --verify "HEAD^{commit}")
-
-          if ! git merge-base --is-ancestor "${commit}" "origin/${DEFAULT_BRANCH}"; then
-            echo "::error::Commit ${commit} is not reachable from origin/${DEFAULT_BRANCH}"
-            exit 1
-          fi
-
-          echo "commit=${commit}" >> "$GITHUB_OUTPUT"
+          # Resolve only the exact tag ref; checkout fails if the tag does not exist.
+          ref: refs/tags/${{ steps.vars.outputs.tag }}
 
       - name: Check release notes exist
         env:
@@ -117,22 +83,6 @@ jobs:
           ctk_ver=$(yq '.cuda.build.version' ci/versions.yml)
           echo "ctk-ver=${ctk_ver}" >> "$GITHUB_OUTPUT"
 
-      - name: Create tag
-        env:
-          TAG: ${{ steps.vars.outputs.tag }}
-          TARGET_COMMIT: ${{ steps.commit.outputs.commit }}
-        run: |
-          if git rev-parse "${TAG}" >/dev/null 2>&1; then
-            existing_commit=$(git rev-parse "${TAG}^{commit}")
-            if [[ "${existing_commit}" != "${TARGET_COMMIT}" ]]; then
-              echo "::error::Tag ${TAG} already exists at ${existing_commit}, expected ${TARGET_COMMIT}"
-              exit 1
-            fi
-          else
-            git tag "${TAG}" "${TARGET_COMMIT}"
-            git push origin "${TAG}"
-          fi
-
       - name: Detect CI run ID
         id: detect-run
         env:
@@ -179,6 +129,7 @@ jobs:
             --repo "${{ github.repository }}" \
             --draft \
             --latest=false \
+            --verify-tag \
             --title "cuda-pathfinder v${VERSION}" \
             --notes-file /tmp/release-body.md
 

cuda_bindings/cuda/bindings/driver.pyx.in

@@ -1,7 +1,7 @@
 # SPDX-FileCopyrightText: Copyright (c) 2021-2026 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
 # SPDX-License-Identifier: LicenseRef-NVIDIA-SOFTWARE-LICENSE
 
-# This code was automatically generated with version 13.1.0, generator version fd3f910. Do not modify it directly.
+# This code was automatically generated with version 13.1.0, generator version c185cc3. Do not modify it directly.
 from typing import Any, Optional
 import cython
 import ctypes
@@ -26111,7 +26111,7 @@ def cuDeviceGetHostAtomicCapabilities(operations : Optional[tuple[CUatomicOperat
         cycapabilities = <unsigned int*>calloc(count, sizeof(unsigned int))
         if cycapabilities is NULL:
             raise MemoryError('Failed to allocate length x size memory: ' + str(count) + 'x' + str(sizeof(unsigned int)))
-    cdef vector[cydriver.CUatomicOperation] cyoperations = [int(pyoperations) for pyoperations in (operations)]
+    cdef vector[cydriver.CUatomicOperation] cyoperations = operations
     if count > len(operations): raise RuntimeError("List is too small: " + str(len(operations)) + " < " + str(count))
     with nogil:
         err = cydriver.cuDeviceGetHostAtomicCapabilities(cycapabilities, cyoperations.data(), count, cydev)
@@ -28374,7 +28374,7 @@ def cuModuleLoadDataEx(image, unsigned int numOptions, options : Optional[tuple[
     cdef void* cyimage = _helper_input_void_ptr(image, &cyimageHelper)
     if numOptions > len(options): raise RuntimeError("List is too small: " + str(len(options)) + " < " + str(numOptions))
     if numOptions > len(optionValues): raise RuntimeError("List is too small: " + str(len(optionValues)) + " < " + str(numOptions))
-    cdef vector[cydriver.CUjit_option] cyoptions = [int(pyoptions) for pyoptions in (options)]
+    cdef vector[cydriver.CUjit_option] cyoptions = options
     pylist = [_HelperCUjit_option(pyoptions, pyoptionValues) for pyoptions, pyoptionValues in zip(options, optionValues)]
     cdef _InputVoidPtrPtrHelper voidStarHelperoptionValues = _InputVoidPtrPtrHelper(pylist)
     cdef void** cyoptionValues_ptr = <void**><void_ptr>voidStarHelperoptionValues.cptr
@@ -28744,7 +28744,7 @@ def cuLinkCreate(unsigned int numOptions, options : Optional[tuple[CUjit_option]
         raise TypeError("Argument 'options' is not instance of type (expected tuple[cydriver.CUjit_option] or list[cydriver.CUjit_option]")
     if numOptions > len(options): raise RuntimeError("List is too small: " + str(len(options)) + " < " + str(numOptions))
     if numOptions > len(optionValues): raise RuntimeError("List is too small: " + str(len(optionValues)) + " < " + str(numOptions))
-    cdef vector[cydriver.CUjit_option] cyoptions = [int(pyoptions) for pyoptions in (options)]
+    cdef vector[cydriver.CUjit_option] cyoptions = options
     pylist = [_HelperCUjit_option(pyoptions, pyoptionValues) for pyoptions, pyoptionValues in zip(options, optionValues)]
     cdef _InputVoidPtrPtrHelper voidStarHelperoptionValues = _InputVoidPtrPtrHelper(pylist)
     cdef void** cyoptionValues_ptr = <void**><void_ptr>voidStarHelperoptionValues.cptr
@@ -28824,7 +28824,7 @@ def cuLinkAddData(state, typename not None : CUjitInputType, data, size_t size,
     cdef void* cydata = _helper_input_void_ptr(data, &cydataHelper)
     if numOptions > len(options): raise RuntimeError("List is too small: " + str(len(options)) + " < " + str(numOptions))
     if numOptions > len(optionValues): raise RuntimeError("List is too small: " + str(len(optionValues)) + " < " + str(numOptions))
-    cdef vector[cydriver.CUjit_option] cyoptions = [int(pyoptions) for pyoptions in (options)]
+    cdef vector[cydriver.CUjit_option] cyoptions = options
     pylist = [_HelperCUjit_option(pyoptions, pyoptionValues) for pyoptions, pyoptionValues in zip(options, optionValues)]
     cdef _InputVoidPtrPtrHelper voidStarHelperoptionValues = _InputVoidPtrPtrHelper(pylist)
     cdef void** cyoptionValues_ptr = <void**><void_ptr>voidStarHelperoptionValues.cptr
@@ -28895,7 +28895,7 @@ def cuLinkAddFile(state, typename not None : CUjitInputType, char* path, unsigne
     cdef cydriver.CUjitInputType cytypename = int(typename)
     if numOptions > len(options): raise RuntimeError("List is too small: " + str(len(options)) + " < " + str(numOptions))
     if numOptions > len(optionValues): raise RuntimeError("List is too small: " + str(len(optionValues)) + " < " + str(numOptions))
-    cdef vector[cydriver.CUjit_option] cyoptions = [int(pyoptions) for pyoptions in (options)]
+    cdef vector[cydriver.CUjit_option] cyoptions = options
     pylist = [_HelperCUjit_option(pyoptions, pyoptionValues) for pyoptions, pyoptionValues in zip(options, optionValues)]
     cdef _InputVoidPtrPtrHelper voidStarHelperoptionValues = _InputVoidPtrPtrHelper(pylist)
     cdef void** cyoptionValues_ptr = <void**><void_ptr>voidStarHelperoptionValues.cptr
@@ -29160,13 +29160,13 @@ def cuLibraryLoadData(code, jitOptions : Optional[tuple[CUjit_option] | list[CUj
     cdef CUlibrary library = CUlibrary()
     cdef _HelperInputVoidPtrStruct cycodeHelper
     cdef void* cycode = _helper_input_void_ptr(code, &cycodeHelper)
-    cdef vector[cydriver.CUjit_option] cyjitOptions = [int(pyjitOptions) for pyjitOptions in (jitOptions)]
+    cdef vector[cydriver.CUjit_option] cyjitOptions = jitOptions
     pylist = [_HelperCUjit_option(pyoptions, pyoptionValues) for pyoptions, pyoptionValues in zip(jitOptions, jitOptionsValues)]
     cdef _InputVoidPtrPtrHelper voidStarHelperjitOptionsValues = _InputVoidPtrPtrHelper(pylist)
     cdef void** cyjitOptionsValues_ptr = <void**><void_ptr>voidStarHelperjitOptionsValues.cptr
     if numJitOptions > len(jitOptions): raise RuntimeError("List is too small: " + str(len(jitOptions)) + " < " + str(numJitOptions))
     if numJitOptions > len(jitOptionsValues): raise RuntimeError("List is too small: " + str(len(jitOptionsValues)) + " < " + str(numJitOptions))
-    cdef vector[cydriver.CUlibraryOption] cylibraryOptions = [int(pylibraryOptions) for pylibraryOptions in (libraryOptions)]
+    cdef vector[cydriver.CUlibraryOption] cylibraryOptions = libraryOptions
     pylist = [_HelperCUlibraryOption(pyoptions, pyoptionValues) for pyoptions, pyoptionValues in zip(libraryOptions, libraryOptionValues)]
     cdef _InputVoidPtrPtrHelper voidStarHelperlibraryOptionValues = _InputVoidPtrPtrHelper(pylist)
     cdef void** cylibraryOptionValues_ptr = <void**><void_ptr>voidStarHelperlibraryOptionValues.cptr
@@ -29258,13 +29258,13 @@ def cuLibraryLoadFromFile(char* fileName, jitOptions : Optional[tuple[CUjit_opti
     if not all(isinstance(_x, (CUjit_option)) for _x in jitOptions):
         raise TypeError("Argument 'jitOptions' is not instance of type (expected tuple[cydriver.CUjit_option] or list[cydriver.CUjit_option]")
     cdef CUlibrary library = CUlibrary()
-    cdef vector[cydriver.CUjit_option] cyjitOptions = [int(pyjitOptions) for pyjitOptions in (jitOptions)]
+    cdef vector[cydriver.CUjit_option] cyjitOptions = jitOptions
     pylist = [_HelperCUjit_option(pyoptions, pyoptionValues) for pyoptions, pyoptionValues in zip(jitOptions, jitOptionsValues)]
     cdef _InputVoidPtrPtrHelper voidStarHelperjitOptionsValues = _InputVoidPtrPtrHelper(pylist)
     cdef void** cyjitOptionsValues_ptr = <void**><void_ptr>voidStarHelperjitOptionsValues.cptr
     if numJitOptions > len(jitOptions): raise RuntimeError("List is too small: " + str(len(jitOptions)) + " < " + str(numJitOptions))
     if numJitOptions > len(jitOptionsValues): raise RuntimeError("List is too small: " + str(len(jitOptionsValues)) + " < " + str(numJitOptions))
-    cdef vector[cydriver.CUlibraryOption] cylibraryOptions = [int(pylibraryOptions) for pylibraryOptions in (libraryOptions)]
+    cdef vector[cydriver.CUlibraryOption] cylibraryOptions = libraryOptions
     pylist = [_HelperCUlibraryOption(pyoptions, pyoptionValues) for pyoptions, pyoptionValues in zip(libraryOptions, libraryOptionValues)]
     cdef _InputVoidPtrPtrHelper voidStarHelperlibraryOptionValues = _InputVoidPtrPtrHelper(pylist)
     cdef void** cylibraryOptionValues_ptr = <void**><void_ptr>voidStarHelperlibraryOptionValues.cptr
@@ -38566,7 +38566,7 @@ def cuMemRangeGetAttributes(dataSizes : tuple[int] | list[int], attributes : Opt
     cdef _InputVoidPtrPtrHelper voidStarHelperdata = _InputVoidPtrPtrHelper(pylist)
     cdef void** cyvoidStarHelper_ptr = <void**><void_ptr>voidStarHelperdata.cptr
     cdef vector[size_t] cydataSizes = dataSizes
-    cdef vector[cydriver.CUmem_range_attribute] cyattributes = [int(pyattributes) for pyattributes in (attributes)]
+    cdef vector[cydriver.CUmem_range_attribute] cyattributes = attributes
     if numAttributes > <size_t>len(dataSizes): raise RuntimeError("List is too small: " + str(len(dataSizes)) + " < " + str(numAttributes))
     if numAttributes > <size_t>len(attributes): raise RuntimeError("List is too small: " + str(len(attributes)) + " < " + str(numAttributes))
     with nogil:
@@ -38715,7 +38715,7 @@ def cuPointerGetAttributes(unsigned int numAttributes, attributes : Optional[tup
     if not all(isinstance(_x, (CUpointer_attribute)) for _x in attributes):
         raise TypeError("Argument 'attributes' is not instance of type (expected tuple[cydriver.CUpointer_attribute] or list[cydriver.CUpointer_attribute]")
     if numAttributes > len(attributes): raise RuntimeError("List is too small: " + str(len(attributes)) + " < " + str(numAttributes))
-    cdef vector[cydriver.CUpointer_attribute] cyattributes = [int(pyattributes) for pyattributes in (attributes)]
+    cdef vector[cydriver.CUpointer_attribute] cyattributes = attributes
     pylist = [_HelperCUpointer_attribute(pyattributes, 0, is_getter=True) for pyattributes in attributes]
     cdef _InputVoidPtrPtrHelper voidStarHelperdata = _InputVoidPtrPtrHelper(pylist)
     cdef void** cyvoidStarHelper_ptr = <void**><void_ptr>voidStarHelperdata.cptr
@@ -52586,7 +52586,7 @@ def cuDeviceGetP2PAtomicCapabilities(operations : Optional[tuple[CUatomicOperati
         cycapabilities = <unsigned int*>calloc(count, sizeof(unsigned int))
         if cycapabilities is NULL:
             raise MemoryError('Failed to allocate length x size memory: ' + str(count) + 'x' + str(sizeof(unsigned int)))
-    cdef vector[cydriver.CUatomicOperation] cyoperations = [int(pyoperations) for pyoperations in (operations)]
+    cdef vector[cydriver.CUatomicOperation] cyoperations = operations
     if count > len(operations): raise RuntimeError("List is too small: " + str(len(operations)) + " < " + str(count))
     with nogil:
         err = cydriver.cuDeviceGetP2PAtomicCapabilities(cycapabilities, cyoperations.data(), count, cysrcDevice, cydstDevice)