From 0141cbfee4494e479c6b3b768a9c73727429fed7 Mon Sep 17 00:00:00 2001
From: Leo Fang <leof@nvidia.com>
Date: Thu, 5 Dec 2024 05:03:48 +0000
Subject: [PATCH 01/65] switch to cibuildwheel + some cleanups

---
 .github/actions/build/action.yml        | 48 +++++++------------------
 .github/workflows/gh-build-and-test.yml | 15 ++++----
 .github/workflows/gh-build.yml          |  7 +---
 3 files changed, 20 insertions(+), 50 deletions(-)

diff --git a/.github/actions/build/action.yml b/.github/actions/build/action.yml
index 952fb9cd1..e6d85bb7f 100644
--- a/.github/actions/build/action.yml
+++ b/.github/actions/build/action.yml
@@ -13,12 +13,6 @@ inputs:
   host-platform:
     required: true
     type: string
-  use-container:
-    required: true
-    type: boolean
-  docker-image:
-    type: string
-    required: true
   upload-enabled:
     required: true
     type: boolean
@@ -29,36 +23,18 @@ inputs:
 runs:
   using: composite
   steps:
-
-    - if: ${{ inputs.use-container }}
-      name: Build (in container)
-      shell: bash --noprofile --norc -xeuo pipefail {0}
-      run: |
-
-        docker run \
-          -e AWS_REGION \
-          -e AWS_SESSION_TOKEN \
-          -e AWS_ACCESS_KEY_ID \
-          -e AWS_SECRET_ACCESS_KEY \
-          -e GITHUB_TOKEN \
-          -e BINDINGS_ARTIFACTS_DIR="$BINDINGS_ARTIFACTS_DIR" \
-          -e CORE_ARTIFACTS_DIR="$CORE_ARTIFACTS_DIR" \
-          -e UPLOAD_ENABLED="$UPLOAD_ENABLED" \
-          -e USE_CUDA="$USE_CUDA" \
-          -e REPO_DIR="$REPO_DIR" \
-          -e LEGATE_CORE_BUILD_MODE="$LEGATE_CORE_BUILD_MODE" \
-          -e PYTHON_VERSION="$PYTHON_VERSION" \
-          -v "${{ env.REPO_DIR }}:${{ env.REPO_DIR }}" \
-          -v "${{ env.BINDINGS_ARTIFACTS_DIR }}:${{ env.BINDINGS_ARTIFACTS_DIR }}" \
-          -v "${{ env.CORE_ARTIFACTS_DIR }}:${{ env.CORE_ARTIFACTS_DIR }}" \
-          --rm "${{ inputs.docker-image }}" \
-          /bin/bash -c "${{ env.REPO_DIR }}/continuous_integration/scripts/entrypoint ${{ env.REPO_DIR }}/continuous_integration/scripts/build ${{ inputs.build-type}} ${{ inputs.target-device }}"
-
-    - if: ${{ !inputs.use-container }}
-      name: Build (without container)
-      shell: bash --noprofile --norc -xeuo pipefail {0}
-      run: |
-        "${{ env.REPO_DIR }}/continuous_integration/scripts/entrypoint" "${{ env.REPO_DIR }}/continuous_integration/scripts/build" "${{ inputs.build-type}}" "${{ inputs.target-device }}"
+    - name: Check out the repo
+      uses: actions/checkout@v4
+
+    - name: Build cuda.core wheel
+      uses: pypa/cibuildwheel@v2.22.0
+      env:
+        CIBW_BUILD: cp312-*  # FIXME: Only build on CPython 3.12
+        CIBW_ARCHS_LINUX: "auto"
+        CIBW_REPAIR_WHEEL_COMMAND: ""  # no repair
+      with:
+        package-dir: ./cuda_core/
+        output-dir: ${{ env.BINDINGS_ARTIFACTS_DIR }}
 
     - name: Display structure of the bindings artifacts folder (post build)
       shell: bash --noprofile --norc -xeuo pipefail {0}
diff --git a/.github/workflows/gh-build-and-test.yml b/.github/workflows/gh-build-and-test.yml
index ffc6f9594..eae1c25a1 100644
--- a/.github/workflows/gh-build-and-test.yml
+++ b/.github/workflows/gh-build-and-test.yml
@@ -1,16 +1,16 @@
 on:
   workflow_call:
     inputs:
-      host-platform:
+      target-device:
         type: string
         required: true
-      target-device:
+      build-type:
         type: string
         required: true
-      build-mode:
+      host-platform:
         type: string
         required: true
-      build-type:
+      build-mode:
         type: string
         required: true
       upload-enabled:
@@ -28,13 +28,12 @@ jobs:
       client-repo: ${{ github.event.repository.name }}
       target-device: ${{ inputs.target-device }}
       runs-on: ${{ (inputs.host-platform == 'linux-x64' && 'linux-amd64-cpu8') ||
-                   (inputs.host-platform == 'linux-aarch64' && 'linux-arm64-cpu8') }}
+                   (inputs.host-platform == 'linux-aarch64' && 'linux-arm64-cpu8') ||
+                   (inputs.host-platform == 'win-x64' && 'linux-amd64-cpu8') }}
       build-type: ${{ inputs.build-type }}
-      use-container: ${{ inputs.host-platform == 'linux-x64' ||
-                         inputs.host-platform == 'linux-aarch64'}}
       host-platform: ${{ inputs.host-platform }}
-      dependencies-file: ""
       build-mode: ${{ inputs.build-mode }}
       upload-enabled: ${{ inputs.upload-enabled }}
       python-version: ${{ inputs.python-version }}
+      dependencies-file: ""
     secrets: inherit
diff --git a/.github/workflows/gh-build.yml b/.github/workflows/gh-build.yml
index c60e0c2ac..3a0d2dea3 100644
--- a/.github/workflows/gh-build.yml
+++ b/.github/workflows/gh-build.yml
@@ -16,9 +16,6 @@ on:
         required: true
         type: string
         description: One of ci / release
-      use-container:
-        required: true
-        type: boolean
       host-platform:
         required: true
         type: string
@@ -38,7 +35,7 @@ on:
 
 jobs:
   build:
-    name: Build (${{ inputs.host-platform }}, ${{ inputs.target-device }}, ${{ inputs.build-type }}, CMake build-mode=${{ inputs.build-mode }}, Python "${{ inputs.python-version }}", Use container=${{ inputs.use-container }} )
+    name: Build (${{ inputs.host-platform }}, ${{ inputs.target-device }}, ${{ inputs.build-type }}, build-mode=${{ inputs.build-mode }}, Python "${{ inputs.python-version }}")
 
     permissions:
       id-token: write # This is required for configure-aws-credentials
@@ -69,6 +66,4 @@ jobs:
           build-type: ${{ inputs.build-type }}
           target-device: "${{ inputs.target-device }}"
           host-platform: ${{ inputs.host-platform }}
-          use-container: ${{ inputs.use-container }}
-          docker-image: "condaforge/miniforge3:latest"
           upload-enabled: ${{ inputs.upload-enabled }}

From 79fd67afcbdc392532efeba67668363b42cd306e Mon Sep 17 00:00:00 2001
From: Leo Fang <leof@nvidia.com>
Date: Thu, 5 Dec 2024 05:23:24 +0000
Subject: [PATCH 02/65] try setting up Python manually

---
 .github/actions/build/action.yml | 7 +++++++
 1 file changed, 7 insertions(+)

diff --git a/.github/actions/build/action.yml b/.github/actions/build/action.yml
index e6d85bb7f..b2c870268 100644
--- a/.github/actions/build/action.yml
+++ b/.github/actions/build/action.yml
@@ -26,10 +26,17 @@ runs:
     - name: Check out the repo
       uses: actions/checkout@v4
 
+    - name: Set up Python
+      id: setup-python
+      uses: actions/setup-python@v5
+      with:
+        python-version: ${{ inputs.python-version }}
+
     - name: Build cuda.core wheel
       uses: pypa/cibuildwheel@v2.22.0
       env:
         CIBW_BUILD: cp312-*  # FIXME: Only build on CPython 3.12
+        CIBW_PROJECT_REQUIRES_PYTHON: ">=3.6"
         CIBW_ARCHS_LINUX: "auto"
         CIBW_REPAIR_WHEEL_COMMAND: ""  # no repair
       with:

From 92283dc47791b9b622a7fa2f480f26e1a6133588 Mon Sep 17 00:00:00 2001
From: Leo Fang <leof@nvidia.com>
Date: Thu, 5 Dec 2024 05:29:06 +0000
Subject: [PATCH 03/65] comment out un-needed code + propagate python-version

---
 .github/workflows/gh-build.yml | 29 +++++++++++++++--------------
 1 file changed, 15 insertions(+), 14 deletions(-)

diff --git a/.github/workflows/gh-build.yml b/.github/workflows/gh-build.yml
index 3a0d2dea3..90d7a4d62 100644
--- a/.github/workflows/gh-build.yml
+++ b/.github/workflows/gh-build.yml
@@ -44,21 +44,21 @@ jobs:
     runs-on: ${{ inputs.runs-on }}
 
     steps:
-      - name: Checkout ${{ inputs.client-repo }}
-        uses: actions/checkout@v4
-        with:
-          fetch-depth: 0
+      #- name: Checkout ${{ inputs.client-repo }}
+      #  uses: actions/checkout@v4
+      #  with:
+      #    fetch-depth: 0
 
-      - name: Setup
-        uses: ./.github/actions/setup
-        with:
-          client-repo: ${{ inputs.client-repo }}
-          build-type: ${{ inputs.build-type }}
-          target-device: "${{ inputs.target-device }}"
-          host-platform: ${{ inputs.host-platform }}
-          build-mode: ${{ inputs.build-mode }}
-          upload-enabled: ${{ inputs.upload-enabled }}
-          python-version: ${{ inputs.python-version }}
+      #- name: Setup
+      #  uses: ./.github/actions/setup
+      #  with:
+      #    client-repo: ${{ inputs.client-repo }}
+      #    build-type: ${{ inputs.build-type }}
+      #    target-device: "${{ inputs.target-device }}"
+      #    host-platform: ${{ inputs.host-platform }}
+      #    build-mode: ${{ inputs.build-mode }}
+      #    upload-enabled: ${{ inputs.upload-enabled }}
+      #    python-version: ${{ inputs.python-version }}
 
       - name: Call build action
         uses: ./.github/actions/build
@@ -67,3 +67,4 @@ jobs:
           target-device: "${{ inputs.target-device }}"
           host-platform: ${{ inputs.host-platform }}
           upload-enabled: ${{ inputs.upload-enabled }}
+          python-version: ${{ inputs.python-version }}

From 6666e905714f16d6d1d7973071cadc51b2abb2aa Mon Sep 17 00:00:00 2001
From: Leo Fang <leof@nvidia.com>
Date: Thu, 5 Dec 2024 05:31:26 +0000
Subject: [PATCH 04/65] fix: need to check out first

---
 .github/actions/build/action.yml | 5 +----
 .github/workflows/gh-build.yml   | 8 ++++----
 2 files changed, 5 insertions(+), 8 deletions(-)

diff --git a/.github/actions/build/action.yml b/.github/actions/build/action.yml
index b2c870268..2f5d5f6c5 100644
--- a/.github/actions/build/action.yml
+++ b/.github/actions/build/action.yml
@@ -23,10 +23,7 @@ inputs:
 runs:
   using: composite
   steps:
-    - name: Check out the repo
-      uses: actions/checkout@v4
-
-    - name: Set up Python
+    - name: Set up Python ${{ inputs.python-version }}
       id: setup-python
       uses: actions/setup-python@v5
       with:
diff --git a/.github/workflows/gh-build.yml b/.github/workflows/gh-build.yml
index 90d7a4d62..5a26691ef 100644
--- a/.github/workflows/gh-build.yml
+++ b/.github/workflows/gh-build.yml
@@ -44,10 +44,10 @@ jobs:
     runs-on: ${{ inputs.runs-on }}
 
     steps:
-      #- name: Checkout ${{ inputs.client-repo }}
-      #  uses: actions/checkout@v4
-      #  with:
-      #    fetch-depth: 0
+      - name: Checkout ${{ inputs.client-repo }}
+        uses: actions/checkout@v4
+        with:
+          fetch-depth: 0
 
       #- name: Setup
       #  uses: ./.github/actions/setup

From c96fef3b901c1ba64db23f2292ff4cb19c10baf8 Mon Sep 17 00:00:00 2001
From: Leo Fang <leof@nvidia.com>
Date: Thu, 5 Dec 2024 05:42:08 +0000
Subject: [PATCH 05/65] only build natively; add -v; ensure targeting manylinux

---
 .github/actions/build/action.yml | 8 ++++++--
 1 file changed, 6 insertions(+), 2 deletions(-)

diff --git a/.github/actions/build/action.yml b/.github/actions/build/action.yml
index 2f5d5f6c5..52ce69a03 100644
--- a/.github/actions/build/action.yml
+++ b/.github/actions/build/action.yml
@@ -32,10 +32,14 @@ runs:
     - name: Build cuda.core wheel
       uses: pypa/cibuildwheel@v2.22.0
       env:
-        CIBW_BUILD: cp312-*  # FIXME: Only build on CPython 3.12
+        CIBW_BUILD: ${{ (inputs.python-version == '3.12' && "cp312-manylinux*") ||
+                        (inputs.python-version == '3.11' && "cp311-manylinux*") ||
+                        (inputs.python-version == '3.10' && "cp310-manylinux*") ||
+                        (inputs.python-version == '3.9' && "cp39-manylinux*") }}
         CIBW_PROJECT_REQUIRES_PYTHON: ">=3.6"
-        CIBW_ARCHS_LINUX: "auto"
+        CIBW_ARCHS_LINUX: "native"
         CIBW_REPAIR_WHEEL_COMMAND: ""  # no repair
+        CIBW_BUILD_VERBOSITY: 1
       with:
         package-dir: ./cuda_core/
         output-dir: ${{ env.BINDINGS_ARTIFACTS_DIR }}

From ea775722d50cc1a47d57a01a4f6c535f0ea92663 Mon Sep 17 00:00:00 2001
From: Leo Fang <leof@nvidia.com>
Date: Thu, 5 Dec 2024 05:44:34 +0000
Subject: [PATCH 06/65] single quotes

---
 .github/actions/build/action.yml | 8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)

diff --git a/.github/actions/build/action.yml b/.github/actions/build/action.yml
index 52ce69a03..7a7c6cc01 100644
--- a/.github/actions/build/action.yml
+++ b/.github/actions/build/action.yml
@@ -32,10 +32,10 @@ runs:
     - name: Build cuda.core wheel
       uses: pypa/cibuildwheel@v2.22.0
       env:
-        CIBW_BUILD: ${{ (inputs.python-version == '3.12' && "cp312-manylinux*") ||
-                        (inputs.python-version == '3.11' && "cp311-manylinux*") ||
-                        (inputs.python-version == '3.10' && "cp310-manylinux*") ||
-                        (inputs.python-version == '3.9' && "cp39-manylinux*") }}
+        CIBW_BUILD: ${{ (inputs.python-version == '3.12' && 'cp312-manylinux*') ||
+                        (inputs.python-version == '3.11' && 'cp311-manylinux*') ||
+                        (inputs.python-version == '3.10' && 'cp310-manylinux*') ||
+                        (inputs.python-version == '3.9' && 'cp39-manylinux*') }}
         CIBW_PROJECT_REQUIRES_PYTHON: ">=3.6"
         CIBW_ARCHS_LINUX: "native"
         CIBW_REPAIR_WHEEL_COMMAND: ""  # no repair

From 508a12606a4a13b04511fda6387159ae3fb77840 Mon Sep 17 00:00:00 2001
From: Leo Fang <leof@nvidia.com>
Date: Thu, 5 Dec 2024 05:52:59 +0000
Subject: [PATCH 07/65] restore env setup to get artifact dir

---
 .github/actions/setup/action.yml |  5 -----
 .github/workflows/gh-build.yml   | 20 ++++++++++----------
 2 files changed, 10 insertions(+), 15 deletions(-)

diff --git a/.github/actions/setup/action.yml b/.github/actions/setup/action.yml
index c2a8407c3..34111d99d 100644
--- a/.github/actions/setup/action.yml
+++ b/.github/actions/setup/action.yml
@@ -57,15 +57,10 @@ runs:
           PKG_DIR="${BUILD_MODE}/${TARGET_PLATFORM}"
         fi
 
-        PYTHON_VERSION_FORMATTED=$(echo '${{ inputs.python-version }}' | tr -d '.')
-
-        echo "BINDINGS_ARTIFACT_NAME=${{ inputs.host-platform }}-${{ inputs.build-type }}-cuda_bindings-python${PYTHON_VERSION_FORMATTED}-${{ inputs.target-device }}${BUILD_MODE_STR}${WITH_TESTS_STR}-${{ github.sha }}" >> $GITHUB_ENV
         echo "BINDINGS_ARTIFACTS_DIR=$(realpath "$(pwd)/cuda_bindings/dist")" >> $GITHUB_ENV
-        echo "CORE_ARTIFACT_NAME=${{ inputs.host-platform }}-${{ inputs.build-type }}-cuda_core-python${PYTHON_VERSION_FORMATTED}-${{ inputs.target-device }}${BUILD_MODE_STR}${WITH_TESTS_STR}-${{ github.sha }}" >> $GITHUB_ENV
         echo "CORE_ARTIFACTS_DIR=$(realpath "$(pwd)/cuda_core/dist")" >> $GITHUB_ENV
         echo "USE_CUDA=${{ (inputs.target-device == 'cpu' && 'OFF') || 'ON' }}" >> $GITHUB_ENV
         echo "UPLOAD_ENABLED=${{ (inputs.upload-enabled  == 'true' && 'ON') || 'OFF' }}" >> $GITHUB_ENV
-        echo "LEGATE_CORE_BUILD_MODE=${BUILD_MODE}" >> $GITHUB_ENV
         echo "BUILD_DATE=$(date +%Y%m%d)" >> $GITHUB_ENV
         echo "TARGET_PLATFORM=${TARGET_PLATFORM}" >> $GITHUB_ENV
         echo "PKG_DIR=${PKG_DIR}" >> $GITHUB_ENV
diff --git a/.github/workflows/gh-build.yml b/.github/workflows/gh-build.yml
index 5a26691ef..70f7772d6 100644
--- a/.github/workflows/gh-build.yml
+++ b/.github/workflows/gh-build.yml
@@ -49,16 +49,16 @@ jobs:
         with:
           fetch-depth: 0
 
-      #- name: Setup
-      #  uses: ./.github/actions/setup
-      #  with:
-      #    client-repo: ${{ inputs.client-repo }}
-      #    build-type: ${{ inputs.build-type }}
-      #    target-device: "${{ inputs.target-device }}"
-      #    host-platform: ${{ inputs.host-platform }}
-      #    build-mode: ${{ inputs.build-mode }}
-      #    upload-enabled: ${{ inputs.upload-enabled }}
-      #    python-version: ${{ inputs.python-version }}
+      - name: Setup environment variables
+        uses: ./.github/actions/setup
+        with:
+          client-repo: ${{ inputs.client-repo }}
+          build-type: ${{ inputs.build-type }}
+          target-device: "${{ inputs.target-device }}"
+          host-platform: ${{ inputs.host-platform }}
+          build-mode: ${{ inputs.build-mode }}
+          upload-enabled: ${{ inputs.upload-enabled }}
+          python-version: ${{ inputs.python-version }}
 
       - name: Call build action
         uses: ./.github/actions/build

From beff50d94d44abbb2e0a724384fe788b6a909e2f Mon Sep 17 00:00:00 2001
From: Leo Fang <leof@nvidia.com>
Date: Thu, 5 Dec 2024 06:00:52 +0000
Subject: [PATCH 08/65] fix artifact dir

---
 .github/actions/build/action.yml | 34 ++++++++++++++++----------------
 .github/actions/setup/action.yml |  4 ++--
 2 files changed, 19 insertions(+), 19 deletions(-)

diff --git a/.github/actions/build/action.yml b/.github/actions/build/action.yml
index 7a7c6cc01..8081548a5 100644
--- a/.github/actions/build/action.yml
+++ b/.github/actions/build/action.yml
@@ -42,28 +42,28 @@ runs:
         CIBW_BUILD_VERBOSITY: 1
       with:
         package-dir: ./cuda_core/
-        output-dir: ${{ env.BINDINGS_ARTIFACTS_DIR }}
+        output-dir: ${{ env.CUDA_CORE_ARTIFACTS_DIR }}
 
-    - name: Display structure of the bindings artifacts folder (post build)
-      shell: bash --noprofile --norc -xeuo pipefail {0}
-      run: |
-        sudo chown -R $(whoami) ${{ env.BINDINGS_ARTIFACTS_DIR }}
-        ls -lahR ${{ env.BINDINGS_ARTIFACTS_DIR }}
+    #- name: Display structure of the bindings artifacts folder (post build)
+    #  shell: bash --noprofile --norc -xeuo pipefail {0}
+    #  run: |
+    #    sudo chown -R $(whoami) ${{ env.CUDA_BINDINGS_ARTIFACTS_DIR }}
+    #    ls -lahR ${{ env.CUDA_BINDINGS_ARTIFACTS_DIR }}
 
-    - name: Upload bindings build artifacts
-      uses: actions/upload-artifact@v4
-      with:
-        name: ${{ env.BINDINGS_ARTIFACT_NAME }}
-        path: ${{ env.BINDINGS_ARTIFACTS_DIR }}
+    #- name: Upload bindings build artifacts
+    #  uses: actions/upload-artifact@v4
+    #  with:
+    #    name: ${{ env.CUDA_BINDINGS_ARTIFACT_NAME }}
+    #    path: ${{ env.CUDA_BINDINGS_ARTIFACTS_DIR }}
 
-    - name: Display structure of the core artifacts folder (post build)
+    - name: List the cuda.core artifacts directory
       shell: bash --noprofile --norc -xeuo pipefail {0}
       run: |
-        sudo chown -R $(whoami) ${{ env.CORE_ARTIFACTS_DIR }}
-        ls -lahR ${{ env.CORE_ARTIFACTS_DIR }}
+        sudo chown -R $(whoami) ${{ env.CUDA_CORE_ARTIFACTS_DIR }}
+        ls -lahR ${{ env.CUDA_CORE_ARTIFACTS_DIR }}
 
-    - name: Upload core build artifacts
+    - name: Upload cuda.core build artifacts
       uses: actions/upload-artifact@v4
       with:
-        name: ${{ env.CORE_ARTIFACT_NAME }}
-        path: ${{ env.CORE_ARTIFACTS_DIR }}
+        name: ${{ env.CUDA_CORE_ARTIFACT_NAME }}
+        path: ${{ env.CUDA_CORE_ARTIFACTS_DIR }}
diff --git a/.github/actions/setup/action.yml b/.github/actions/setup/action.yml
index 34111d99d..c4ab05b33 100644
--- a/.github/actions/setup/action.yml
+++ b/.github/actions/setup/action.yml
@@ -57,8 +57,8 @@ runs:
           PKG_DIR="${BUILD_MODE}/${TARGET_PLATFORM}"
         fi
 
-        echo "BINDINGS_ARTIFACTS_DIR=$(realpath "$(pwd)/cuda_bindings/dist")" >> $GITHUB_ENV
-        echo "CORE_ARTIFACTS_DIR=$(realpath "$(pwd)/cuda_core/dist")" >> $GITHUB_ENV
+        echo "CUDA_BINDINGS_ARTIFACTS_DIR=$(realpath "$(pwd)/cuda_bindings/dist")" >> $GITHUB_ENV
+        echo "CUDA_CORE_ARTIFACTS_DIR=$(realpath "$(pwd)/cuda_core/dist")" >> $GITHUB_ENV
         echo "USE_CUDA=${{ (inputs.target-device == 'cpu' && 'OFF') || 'ON' }}" >> $GITHUB_ENV
         echo "UPLOAD_ENABLED=${{ (inputs.upload-enabled  == 'true' && 'ON') || 'OFF' }}" >> $GITHUB_ENV
         echo "BUILD_DATE=$(date +%Y%m%d)" >> $GITHUB_ENV

From a935da2ea266608107630a56ff046b52fb5cc88d Mon Sep 17 00:00:00 2001
From: Leo Fang <leof@nvidia.com>
Date: Thu, 5 Dec 2024 06:04:37 +0000
Subject: [PATCH 09/65] fix artifact name

---
 .github/actions/build/action.yml | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/.github/actions/build/action.yml b/.github/actions/build/action.yml
index 8081548a5..960b84193 100644
--- a/.github/actions/build/action.yml
+++ b/.github/actions/build/action.yml
@@ -65,5 +65,5 @@ runs:
     - name: Upload cuda.core build artifacts
       uses: actions/upload-artifact@v4
       with:
-        name: ${{ env.CUDA_CORE_ARTIFACT_NAME }}
+        name: cuda_core*.whl
         path: ${{ env.CUDA_CORE_ARTIFACTS_DIR }}

From df982e5e5bc821d2355e3906b7db395c8bad76a4 Mon Sep 17 00:00:00 2001
From: Leo Fang <leof@nvidia.com>
Date: Thu, 5 Dec 2024 06:24:03 +0000
Subject: [PATCH 10/65] restore & fix artifact name again

---
 .github/actions/build/action.yml |  6 ++++--
 .github/actions/setup/action.yml | 10 +---------
 2 files changed, 5 insertions(+), 11 deletions(-)

diff --git a/.github/actions/build/action.yml b/.github/actions/build/action.yml
index 960b84193..bf121c1b6 100644
--- a/.github/actions/build/action.yml
+++ b/.github/actions/build/action.yml
@@ -65,5 +65,7 @@ runs:
     - name: Upload cuda.core build artifacts
       uses: actions/upload-artifact@v4
       with:
-        name: cuda_core*.whl
-        path: ${{ env.CUDA_CORE_ARTIFACTS_DIR }}
+        name: ${{ env.CUDA_CORE_ARTIFACT_NAME }}
+        path: ${{ env.CUDA_CORE_ARTIFACTS_DIR }}/*.whl
+        if-no-files-found: error
+        overwrite: 'true'
diff --git a/.github/actions/setup/action.yml b/.github/actions/setup/action.yml
index c4ab05b33..fa24536d8 100644
--- a/.github/actions/setup/action.yml
+++ b/.github/actions/setup/action.yml
@@ -35,21 +35,12 @@ runs:
     - name: Set environment variables
       shell: bash --noprofile --norc -xeuo pipefail {0}
       run: |
-
-        WITH_TESTS_STR=''
-        if [[ ("${{ inputs.upload-enabled }}" == "false") && ("${{ inputs.build-type }}" != "ci") ]]; then
-          WITH_TESTS_STR='-with_tests'
-        fi
-
         TARGET_PLATFORM='linux-64'
         if [[ "${{ inputs.host-platform }}" == "linux-aarch64" ]]; then
           TARGET_PLATFORM='linux-aarch64'
         fi
 
         BUILD_MODE="${{ inputs.build-mode }}"
-        BUILD_MODE_STR=""
-        [ -n "${BUILD_MODE}" ] && BUILD_MODE_STR="-${BUILD_MODE}"
-
         if [[ ("${BUILD_MODE}" == "") || ("${BUILD_MODE}" == "release") ]]; then
           # We upload release versions in the default folder.
           PKG_DIR="${TARGET_PLATFORM}"
@@ -58,6 +49,7 @@ runs:
         fi
 
         echo "CUDA_BINDINGS_ARTIFACTS_DIR=$(realpath "$(pwd)/cuda_bindings/dist")" >> $GITHUB_ENV
+        echo "CUDA_CORE_ARTIFACT_NAME=cuda-core-python${{ inputs.python-version }}-${{ inputs.host-platform }}-${{ inputs.build-type }}-${{ github.sha }}" >> $GITHUB_ENV
         echo "CUDA_CORE_ARTIFACTS_DIR=$(realpath "$(pwd)/cuda_core/dist")" >> $GITHUB_ENV
         echo "USE_CUDA=${{ (inputs.target-device == 'cpu' && 'OFF') || 'ON' }}" >> $GITHUB_ENV
         echo "UPLOAD_ENABLED=${{ (inputs.upload-enabled  == 'true' && 'ON') || 'OFF' }}" >> $GITHUB_ENV

From 426cd583ae489a307de497a964521aa83cb6eb1a Mon Sep 17 00:00:00 2001
From: Leo Fang <leof@nvidia.com>
Date: Thu, 5 Dec 2024 06:34:15 +0000
Subject: [PATCH 11/65] build on all platforms!

---
 .github/actions/setup/action.yml | 10 ++++++++--
 .github/workflows/ci-gh.yml      |  2 ++
 2 files changed, 10 insertions(+), 2 deletions(-)

diff --git a/.github/actions/setup/action.yml b/.github/actions/setup/action.yml
index fa24536d8..a48d2f51e 100644
--- a/.github/actions/setup/action.yml
+++ b/.github/actions/setup/action.yml
@@ -35,9 +35,15 @@ runs:
     - name: Set environment variables
       shell: bash --noprofile --norc -xeuo pipefail {0}
       run: |
-        TARGET_PLATFORM='linux-64'
-        if [[ "${{ inputs.host-platform }}" == "linux-aarch64" ]]; then
+        # TODO: just align host-platform names with TARGET_PLATFORM...
+        if [[ "${{ inputs.host-platform }}" == "linux-x64" ]]; then
+          TARGET_PLATFORM='linux-64'
+        elif [[ "${{ inputs.host-platform }}" == "linux-aarch64" ]]; then
           TARGET_PLATFORM='linux-aarch64'
+        elif [[ "${{ inputs.host-platform }}" == "win-x64" ]]; then
+          TARGET_PLATFORM='win-64'
+        else
+          exit 1
         fi
 
         BUILD_MODE="${{ inputs.build-mode }}"
diff --git a/.github/workflows/ci-gh.yml b/.github/workflows/ci-gh.yml
index d38cb8e30..cf1683257 100644
--- a/.github/workflows/ci-gh.yml
+++ b/.github/workflows/ci-gh.yml
@@ -18,6 +18,8 @@ jobs:
       matrix:
         host-platform:
           - linux-x64
+          - linux-aarch64
+          - win-x64
         target-device:
           - gpu
         build-mode:

From d167a5536815d393c6d2dcaa1838965daa133bfa Mon Sep 17 00:00:00 2001
From: Leo Fang <leof@nvidia.com>
Date: Thu, 5 Dec 2024 06:43:01 +0000
Subject: [PATCH 12/65] fix CIBW_BUILD for windows

---
 .github/actions/build/action.yml | 12 ++++++++----
 1 file changed, 8 insertions(+), 4 deletions(-)

diff --git a/.github/actions/build/action.yml b/.github/actions/build/action.yml
index bf121c1b6..bf095323b 100644
--- a/.github/actions/build/action.yml
+++ b/.github/actions/build/action.yml
@@ -32,10 +32,14 @@ runs:
     - name: Build cuda.core wheel
       uses: pypa/cibuildwheel@v2.22.0
       env:
-        CIBW_BUILD: ${{ (inputs.python-version == '3.12' && 'cp312-manylinux*') ||
-                        (inputs.python-version == '3.11' && 'cp311-manylinux*') ||
-                        (inputs.python-version == '3.10' && 'cp310-manylinux*') ||
-                        (inputs.python-version == '3.9' && 'cp39-manylinux*') }}
+        CIBW_BUILD: ${{ (inputs.python-version == '3.12' && inputs.host-platform == 'linux*' && 'cp312-manylinux*') ||
+                        (inputs.python-version == '3.11' && inputs.host-platform == 'linux*' && 'cp311-manylinux*') ||
+                        (inputs.python-version == '3.10' && inputs.host-platform == 'linux*' && 'cp310-manylinux*') ||
+                        (inputs.python-version == '3.9'  && inputs.hott-platform == 'linux*' &&  'cp39-manylinux*') ||
+                        (inputs.python-version == '3.12' && inputs.host-platform == 'win*' && 'cp312-win*') ||
+                        (inputs.python-version == '3.11' && inputs.host-platform == 'win*' && 'cp311-win*') ||
+                        (inputs.python-version == '3.10' && inputs.host-platform == 'win*' && 'cp310-win*') ||
+                        (inputs.python-version == '3.9'  && inputs.hott-platform == 'win*' &&  'cp39-win*') }}
         CIBW_PROJECT_REQUIRES_PYTHON: ">=3.6"
         CIBW_ARCHS_LINUX: "native"
         CIBW_REPAIR_WHEEL_COMMAND: ""  # no repair

From 9cd08038c960e0dd649cf2af59f9c6f3e968de60 Mon Sep 17 00:00:00 2001
From: Leo Fang <leof@nvidia.com>
Date: Thu, 5 Dec 2024 06:45:45 +0000
Subject: [PATCH 13/65] fix typo

---
 .github/actions/build/action.yml | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/.github/actions/build/action.yml b/.github/actions/build/action.yml
index bf095323b..1ab7ce0ed 100644
--- a/.github/actions/build/action.yml
+++ b/.github/actions/build/action.yml
@@ -35,11 +35,11 @@ runs:
         CIBW_BUILD: ${{ (inputs.python-version == '3.12' && inputs.host-platform == 'linux*' && 'cp312-manylinux*') ||
                         (inputs.python-version == '3.11' && inputs.host-platform == 'linux*' && 'cp311-manylinux*') ||
                         (inputs.python-version == '3.10' && inputs.host-platform == 'linux*' && 'cp310-manylinux*') ||
-                        (inputs.python-version == '3.9'  && inputs.hott-platform == 'linux*' &&  'cp39-manylinux*') ||
+                        (inputs.python-version == '3.9'  && inputs.host-platform == 'linux*' &&  'cp39-manylinux*') ||
                         (inputs.python-version == '3.12' && inputs.host-platform == 'win*' && 'cp312-win*') ||
                         (inputs.python-version == '3.11' && inputs.host-platform == 'win*' && 'cp311-win*') ||
                         (inputs.python-version == '3.10' && inputs.host-platform == 'win*' && 'cp310-win*') ||
-                        (inputs.python-version == '3.9'  && inputs.hott-platform == 'win*' &&  'cp39-win*') }}
+                        (inputs.python-version == '3.9'  && inputs.host-platform == 'win*' &&  'cp39-win*') }}
         CIBW_PROJECT_REQUIRES_PYTHON: ">=3.6"
         CIBW_ARCHS_LINUX: "native"
         CIBW_REPAIR_WHEEL_COMMAND: ""  # no repair

From 261b1e5e197ae7b33f731ddf21cbb7c8391bf3cd Mon Sep 17 00:00:00 2001
From: Leo Fang <leof@nvidia.com>
Date: Thu, 5 Dec 2024 06:53:26 +0000
Subject: [PATCH 14/65] no quotes for wildcard matching

---
 .github/actions/build/action.yml | 16 ++++++++--------
 1 file changed, 8 insertions(+), 8 deletions(-)

diff --git a/.github/actions/build/action.yml b/.github/actions/build/action.yml
index 1ab7ce0ed..4b891977b 100644
--- a/.github/actions/build/action.yml
+++ b/.github/actions/build/action.yml
@@ -32,14 +32,14 @@ runs:
     - name: Build cuda.core wheel
       uses: pypa/cibuildwheel@v2.22.0
       env:
-        CIBW_BUILD: ${{ (inputs.python-version == '3.12' && inputs.host-platform == 'linux*' && 'cp312-manylinux*') ||
-                        (inputs.python-version == '3.11' && inputs.host-platform == 'linux*' && 'cp311-manylinux*') ||
-                        (inputs.python-version == '3.10' && inputs.host-platform == 'linux*' && 'cp310-manylinux*') ||
-                        (inputs.python-version == '3.9'  && inputs.host-platform == 'linux*' &&  'cp39-manylinux*') ||
-                        (inputs.python-version == '3.12' && inputs.host-platform == 'win*' && 'cp312-win*') ||
-                        (inputs.python-version == '3.11' && inputs.host-platform == 'win*' && 'cp311-win*') ||
-                        (inputs.python-version == '3.10' && inputs.host-platform == 'win*' && 'cp310-win*') ||
-                        (inputs.python-version == '3.9'  && inputs.host-platform == 'win*' &&  'cp39-win*') }}
+        CIBW_BUILD: ${{ (inputs.python-version == '3.12' && inputs.host-platform == linux* && 'cp312-manylinux*') ||
+                        (inputs.python-version == '3.11' && inputs.host-platform == linux* && 'cp311-manylinux*') ||
+                        (inputs.python-version == '3.10' && inputs.host-platform == linux* && 'cp310-manylinux*') ||
+                        (inputs.python-version == '3.9'  && inputs.host-platform == linux* &&  'cp39-manylinux*') ||
+                        (inputs.python-version == '3.12' && inputs.host-platform == win* && 'cp312-win*') ||
+                        (inputs.python-version == '3.11' && inputs.host-platform == win* && 'cp311-win*') ||
+                        (inputs.python-version == '3.10' && inputs.host-platform == win* && 'cp310-win*') ||
+                        (inputs.python-version == '3.9'  && inputs.host-platform == win* &&  'cp39-win*') }}
         CIBW_PROJECT_REQUIRES_PYTHON: ">=3.6"
         CIBW_ARCHS_LINUX: "native"
         CIBW_REPAIR_WHEEL_COMMAND: ""  # no repair

From 9134f898953a02b32f14d32d04d912e6ae67d7d7 Mon Sep 17 00:00:00 2001
From: Leo Fang <leof@nvidia.com>
Date: Thu, 5 Dec 2024 06:58:42 +0000
Subject: [PATCH 15/65] move CIBW_BUILD logic to script

---
 .github/actions/build/action.yml |  9 +--------
 .github/actions/setup/action.yml | 21 +++++++++++++++++++++
 2 files changed, 22 insertions(+), 8 deletions(-)

diff --git a/.github/actions/build/action.yml b/.github/actions/build/action.yml
index 4b891977b..42d47f541 100644
--- a/.github/actions/build/action.yml
+++ b/.github/actions/build/action.yml
@@ -32,14 +32,7 @@ runs:
     - name: Build cuda.core wheel
       uses: pypa/cibuildwheel@v2.22.0
       env:
-        CIBW_BUILD: ${{ (inputs.python-version == '3.12' && inputs.host-platform == linux* && 'cp312-manylinux*') ||
-                        (inputs.python-version == '3.11' && inputs.host-platform == linux* && 'cp311-manylinux*') ||
-                        (inputs.python-version == '3.10' && inputs.host-platform == linux* && 'cp310-manylinux*') ||
-                        (inputs.python-version == '3.9'  && inputs.host-platform == linux* &&  'cp39-manylinux*') ||
-                        (inputs.python-version == '3.12' && inputs.host-platform == win* && 'cp312-win*') ||
-                        (inputs.python-version == '3.11' && inputs.host-platform == win* && 'cp311-win*') ||
-                        (inputs.python-version == '3.10' && inputs.host-platform == win* && 'cp310-win*') ||
-                        (inputs.python-version == '3.9'  && inputs.host-platform == win* &&  'cp39-win*') }}
+        CIBW_BUILD: ${{ env.CIBW_BUILD }}
         CIBW_PROJECT_REQUIRES_PYTHON: ">=3.6"
         CIBW_ARCHS_LINUX: "native"
         CIBW_REPAIR_WHEEL_COMMAND: ""  # no repair
diff --git a/.github/actions/setup/action.yml b/.github/actions/setup/action.yml
index a48d2f51e..29fbe3581 100644
--- a/.github/actions/setup/action.yml
+++ b/.github/actions/setup/action.yml
@@ -46,6 +46,26 @@ runs:
           exit 1
         fi
 
+        if [[ "${{ inputs.python-version }}" == "3.12" && "${{ inputs.host-platform }}" == linux* ]]; then
+          CIBW_BUILD='cp312-manylinux*'
+        elif [[ "${{ inputs.python-version }}" == "3.11" && "${{ inputs.host-platform }}" == linux* ]]; then
+          CIBW_BUILD='cp311-manylinux*'
+        elif [[ "${{ inputs.python-version }}" == "3.10" && "${{ inputs.host-platform }}" == linux* ]]; then
+          CIBW_BUILD='cp310-manylinux*'
+        elif [[ "${{ inputs.python-version }}" == "3.9" && "${{ inputs.host-platform }}" == linux* ]]; then
+          CIBW_BUILD='cp39-manylinux*'
+        elif [[ "${{ inputs.python-version }}" == "3.12" && "${{ inputs.host-platform }}" == win* ]]; then
+          CIBW_BUILD='cp312-win*'
+        elif [[ "${{ inputs.python-version }}" == "3.11" && "${{ inputs.host-platform }}" == win* ]]; then
+          CIBW_BUILD='cp311-win*'
+        elif [[ "${{ inputs.python-version }}" == "3.10" && "${{ inputs.host-platform }}" == win* ]]; then
+          CIBW_BUILD='cp310-win*'
+        elif [[ "${{ inputs.python-version }}" == "3.9" && "${{ inputs.host-platform }}" == win* ]]; then
+          CIBW_BUILD='cp39-win*'
+        else
+          exit 1
+        fi
+
         BUILD_MODE="${{ inputs.build-mode }}"
         if [[ ("${BUILD_MODE}" == "") || ("${BUILD_MODE}" == "release") ]]; then
           # We upload release versions in the default folder.
@@ -63,3 +83,4 @@ runs:
         echo "TARGET_PLATFORM=${TARGET_PLATFORM}" >> $GITHUB_ENV
         echo "PKG_DIR=${PKG_DIR}" >> $GITHUB_ENV
         echo "PYTHON_VERSION=${{ inputs.python-version }}" >> $GITHUB_ENV
+        echo "CIBW_BUILD=${CIBW_BUILD}" >> $GITHUB_ENV

From 762ad764666b6f14d6fc4e184787ae9ce8b2bc38 Mon Sep 17 00:00:00 2001
From: Leo Fang <leof@nvidia.com>
Date: Thu, 5 Dec 2024 07:01:49 +0000
Subject: [PATCH 16/65] fix win runner name

---
 .github/workflows/gh-build-and-test.yml | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/.github/workflows/gh-build-and-test.yml b/.github/workflows/gh-build-and-test.yml
index eae1c25a1..867cffde0 100644
--- a/.github/workflows/gh-build-and-test.yml
+++ b/.github/workflows/gh-build-and-test.yml
@@ -29,7 +29,7 @@ jobs:
       target-device: ${{ inputs.target-device }}
       runs-on: ${{ (inputs.host-platform == 'linux-x64' && 'linux-amd64-cpu8') ||
                    (inputs.host-platform == 'linux-aarch64' && 'linux-arm64-cpu8') ||
-                   (inputs.host-platform == 'win-x64' && 'linux-amd64-cpu8') }}
+                   (inputs.host-platform == 'win-x64' && 'windows-amd64-cpu8') }}
       build-type: ${{ inputs.build-type }}
       host-platform: ${{ inputs.host-platform }}
       build-mode: ${{ inputs.build-mode }}

From 0d3c32bd3b4ec8fdfecadebf6b570d6e63c0804c Mon Sep 17 00:00:00 2001
From: Leo Fang <leof@nvidia.com>
Date: Thu, 5 Dec 2024 07:40:53 +0000
Subject: [PATCH 17/65] try to find where pwsh is

---
 .github/actions/setup/action.yml | 6 ++++++
 1 file changed, 6 insertions(+)

diff --git a/.github/actions/setup/action.yml b/.github/actions/setup/action.yml
index 29fbe3581..dde7862b8 100644
--- a/.github/actions/setup/action.yml
+++ b/.github/actions/setup/action.yml
@@ -74,6 +74,12 @@ runs:
           PKG_DIR="${BUILD_MODE}/${TARGET_PLATFORM}"
         fi
 
+        if [[ "${{ inputs.host-platform }}" == win* ]]; then
+          ls $PROGRAMFILES
+          ls $PROGRAMFILES\PowerShell
+          ls C:\Windows\System32\
+        fi
+
         echo "CUDA_BINDINGS_ARTIFACTS_DIR=$(realpath "$(pwd)/cuda_bindings/dist")" >> $GITHUB_ENV
         echo "CUDA_CORE_ARTIFACT_NAME=cuda-core-python${{ inputs.python-version }}-${{ inputs.host-platform }}-${{ inputs.build-type }}-${{ github.sha }}" >> $GITHUB_ENV
         echo "CUDA_CORE_ARTIFACTS_DIR=$(realpath "$(pwd)/cuda_core/dist")" >> $GITHUB_ENV

From 37006e80c295ed7b055f482bc74aafed98c06add Mon Sep 17 00:00:00 2001
From: Leo Fang <leof@nvidia.com>
Date: Thu, 5 Dec 2024 07:46:06 +0000
Subject: [PATCH 18/65] try to escape

---
 .github/actions/setup/action.yml | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/.github/actions/setup/action.yml b/.github/actions/setup/action.yml
index dde7862b8..77670f6ee 100644
--- a/.github/actions/setup/action.yml
+++ b/.github/actions/setup/action.yml
@@ -75,9 +75,9 @@ runs:
         fi
 
         if [[ "${{ inputs.host-platform }}" == win* ]]; then
-          ls $PROGRAMFILES
-          ls $PROGRAMFILES\PowerShell
-          ls C:\Windows\System32\
+          ls "${PROGRAMFILES}"
+          ls "${PROGRAMFILES}\\PowerShell"
+          ls "C:\\Windows\\System32\\"
         fi
 
         echo "CUDA_BINDINGS_ARTIFACTS_DIR=$(realpath "$(pwd)/cuda_bindings/dist")" >> $GITHUB_ENV

From 0ab50675bb16e7e168af76b786cf438a726671ec Mon Sep 17 00:00:00 2001
From: Leo Fang <leof@nvidia.com>
Date: Thu, 5 Dec 2024 07:51:22 +0000
Subject: [PATCH 19/65] continue hunting..

---
 .github/actions/setup/action.yml | 3 +--
 1 file changed, 1 insertion(+), 2 deletions(-)

diff --git a/.github/actions/setup/action.yml b/.github/actions/setup/action.yml
index 77670f6ee..44095f588 100644
--- a/.github/actions/setup/action.yml
+++ b/.github/actions/setup/action.yml
@@ -75,8 +75,7 @@ runs:
         fi
 
         if [[ "${{ inputs.host-platform }}" == win* ]]; then
-          ls "${PROGRAMFILES}"
-          ls "${PROGRAMFILES}\\PowerShell"
+          ls "${PROGRAMFILES}\\WindowsPowerShell"
           ls "C:\\Windows\\System32\\"
         fi
 

From d42f8ddd4bf93ea1180c77a8e87cc680fb1d1591 Mon Sep 17 00:00:00 2001
From: Leo Fang <leof@nvidia.com>
Date: Thu, 5 Dec 2024 08:02:48 +0000
Subject: [PATCH 20/65] try to overwrite shell

---
 .github/actions/build/action.yml | 1 +
 .github/actions/setup/action.yml | 6 ++++--
 2 files changed, 5 insertions(+), 2 deletions(-)

diff --git a/.github/actions/build/action.yml b/.github/actions/build/action.yml
index 42d47f541..6ba09c46f 100644
--- a/.github/actions/build/action.yml
+++ b/.github/actions/build/action.yml
@@ -40,6 +40,7 @@ runs:
       with:
         package-dir: ./cuda_core/
         output-dir: ${{ env.CUDA_CORE_ARTIFACTS_DIR }}
+      shell: ${{ WAR_SHELL }}
 
     #- name: Display structure of the bindings artifacts folder (post build)
     #  shell: bash --noprofile --norc -xeuo pipefail {0}
diff --git a/.github/actions/setup/action.yml b/.github/actions/setup/action.yml
index 44095f588..3f9c146fc 100644
--- a/.github/actions/setup/action.yml
+++ b/.github/actions/setup/action.yml
@@ -75,8 +75,10 @@ runs:
         fi
 
         if [[ "${{ inputs.host-platform }}" == win* ]]; then
-          ls "${PROGRAMFILES}\\WindowsPowerShell"
-          ls "C:\\Windows\\System32\\"
+          ls "C:\\Windows\\System32\\WindowsPowerShell\\v1.0\\"
+          echo "WAR_SHELL=powershell" >> $GITHUB_ENV
+        else
+          echo "WAR_SHELL=bash" >> $GITHUB_ENV
         fi
 
         echo "CUDA_BINDINGS_ARTIFACTS_DIR=$(realpath "$(pwd)/cuda_bindings/dist")" >> $GITHUB_ENV

From 55c03705f0276775fe31032efd44f6d02e509582 Mon Sep 17 00:00:00 2001
From: Leo Fang <leof@nvidia.com>
Date: Thu, 5 Dec 2024 15:37:59 +0000
Subject: [PATCH 21/65] try to install ps

---
 .github/actions/build/action.yml |  1 -
 .github/actions/setup/action.yml | 17 ++++++++++-------
 2 files changed, 10 insertions(+), 8 deletions(-)

diff --git a/.github/actions/build/action.yml b/.github/actions/build/action.yml
index 6ba09c46f..42d47f541 100644
--- a/.github/actions/build/action.yml
+++ b/.github/actions/build/action.yml
@@ -40,7 +40,6 @@ runs:
       with:
         package-dir: ./cuda_core/
         output-dir: ${{ env.CUDA_CORE_ARTIFACTS_DIR }}
-      shell: ${{ WAR_SHELL }}
 
     #- name: Display structure of the bindings artifacts folder (post build)
     #  shell: bash --noprofile --norc -xeuo pipefail {0}
diff --git a/.github/actions/setup/action.yml b/.github/actions/setup/action.yml
index 3f9c146fc..38aab81f7 100644
--- a/.github/actions/setup/action.yml
+++ b/.github/actions/setup/action.yml
@@ -26,6 +26,16 @@ inputs:
 runs:
   using: composite
   steps:
+    - name: Install PowerShell
+      if: ${{ startsWith(inputs.host-platform, 'win') }}
+      env:
+        # doesn't seem there's an easy way to avoid hard-coding it?
+        PS_MSI_URL: https://github.com/PowerShell/PowerShell/releases/download/v7.4.6/PowerShell-7.4.6-win-x64.msi
+      run: |
+        Invoke-WebRequest -Uri "$env:PS_MSI_URL" -OutFile "ps_installer.msi"
+        Start-Process msiexec.exe -Wait -Verbose -ArgumentList '/i "ps_installer.msi" /qn REGISTER_MANIFEST=1'
+        echo "$env:ProgramFiles\\PowerShell\\7" >> $env:GITHUB_PATH
+
     - name: Set REPO_DIR and Dump environment
       shell: bash --noprofile --norc -xeuo pipefail {0}
       run: |
@@ -74,13 +84,6 @@ runs:
           PKG_DIR="${BUILD_MODE}/${TARGET_PLATFORM}"
         fi
 
-        if [[ "${{ inputs.host-platform }}" == win* ]]; then
-          ls "C:\\Windows\\System32\\WindowsPowerShell\\v1.0\\"
-          echo "WAR_SHELL=powershell" >> $GITHUB_ENV
-        else
-          echo "WAR_SHELL=bash" >> $GITHUB_ENV
-        fi
-
         echo "CUDA_BINDINGS_ARTIFACTS_DIR=$(realpath "$(pwd)/cuda_bindings/dist")" >> $GITHUB_ENV
         echo "CUDA_CORE_ARTIFACT_NAME=cuda-core-python${{ inputs.python-version }}-${{ inputs.host-platform }}-${{ inputs.build-type }}-${{ github.sha }}" >> $GITHUB_ENV
         echo "CUDA_CORE_ARTIFACTS_DIR=$(realpath "$(pwd)/cuda_core/dist")" >> $GITHUB_ENV

From f3356f44cf9788944e7267e16b36ec847b5ab232 Mon Sep 17 00:00:00 2001
From: Leo Fang <leof@nvidia.com>
Date: Thu, 5 Dec 2024 15:42:49 +0000
Subject: [PATCH 22/65] be explicit about shell (why?)

---
 .github/actions/setup/action.yml | 1 +
 1 file changed, 1 insertion(+)

diff --git a/.github/actions/setup/action.yml b/.github/actions/setup/action.yml
index 38aab81f7..ecc5eb56d 100644
--- a/.github/actions/setup/action.yml
+++ b/.github/actions/setup/action.yml
@@ -28,6 +28,7 @@ runs:
   steps:
     - name: Install PowerShell
       if: ${{ startsWith(inputs.host-platform, 'win') }}
+      shell: powershell
       env:
         # doesn't seem there's an easy way to avoid hard-coding it?
         PS_MSI_URL: https://github.com/PowerShell/PowerShell/releases/download/v7.4.6/PowerShell-7.4.6-win-x64.msi

From 5480f10c8de0d818ff933c9c50390c1da4e318f5 Mon Sep 17 00:00:00 2001
From: Leo Fang <leof@nvidia.com>
Date: Thu, 5 Dec 2024 15:56:52 +0000
Subject: [PATCH 23/65] only build for win 64 bits

---
 .github/actions/setup/action.yml | 8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)

diff --git a/.github/actions/setup/action.yml b/.github/actions/setup/action.yml
index ecc5eb56d..739977713 100644
--- a/.github/actions/setup/action.yml
+++ b/.github/actions/setup/action.yml
@@ -66,13 +66,13 @@ runs:
         elif [[ "${{ inputs.python-version }}" == "3.9" && "${{ inputs.host-platform }}" == linux* ]]; then
           CIBW_BUILD='cp39-manylinux*'
         elif [[ "${{ inputs.python-version }}" == "3.12" && "${{ inputs.host-platform }}" == win* ]]; then
-          CIBW_BUILD='cp312-win*'
+          CIBW_BUILD='cp312-win_amd64'
         elif [[ "${{ inputs.python-version }}" == "3.11" && "${{ inputs.host-platform }}" == win* ]]; then
-          CIBW_BUILD='cp311-win*'
+          CIBW_BUILD='cp311-win_amd64'
         elif [[ "${{ inputs.python-version }}" == "3.10" && "${{ inputs.host-platform }}" == win* ]]; then
-          CIBW_BUILD='cp310-win*'
+          CIBW_BUILD='cp310-win_amd64'
         elif [[ "${{ inputs.python-version }}" == "3.9" && "${{ inputs.host-platform }}" == win* ]]; then
-          CIBW_BUILD='cp39-win*'
+          CIBW_BUILD='cp39-win_amd64'
         else
           exit 1
         fi

From 4b1a43188860f9bb26bcd2ad13a9c0e9de5b6fa4 Mon Sep 17 00:00:00 2001
From: Leo Fang <leof@nvidia.com>
Date: Thu, 5 Dec 2024 16:26:04 +0000
Subject: [PATCH 24/65] try to install msvc

---
 .github/actions/build/action.yml |  9 ---------
 .github/actions/setup/action.yml | 14 +++++++++++++-
 .github/workflows/gh-build.yml   |  1 -
 3 files changed, 13 insertions(+), 11 deletions(-)

diff --git a/.github/actions/build/action.yml b/.github/actions/build/action.yml
index 42d47f541..a61b4222f 100644
--- a/.github/actions/build/action.yml
+++ b/.github/actions/build/action.yml
@@ -16,19 +16,10 @@ inputs:
   upload-enabled:
     required: true
     type: boolean
-  python-version:
-    required: true
-    type: string
 
 runs:
   using: composite
   steps:
-    - name: Set up Python ${{ inputs.python-version }}
-      id: setup-python
-      uses: actions/setup-python@v5
-      with:
-        python-version: ${{ inputs.python-version }}
-
     - name: Build cuda.core wheel
       uses: pypa/cibuildwheel@v2.22.0
       env:
diff --git a/.github/actions/setup/action.yml b/.github/actions/setup/action.yml
index 739977713..94c6c0a5e 100644
--- a/.github/actions/setup/action.yml
+++ b/.github/actions/setup/action.yml
@@ -26,7 +26,19 @@ inputs:
 runs:
   using: composite
   steps:
-    - name: Install PowerShell
+    - name: Set up Python ${{ inputs.python-version }}
+      id: setup-python
+      uses: actions/setup-python@v5
+      with:
+        python-version: ${{ inputs.python-version }}
+
+    - name: Set up MSVC
+      if: ${{ startsWith(inputs.host-platform, 'win') }}
+      uses: ilammy/msvc-dev-cmd@v1
+      with:
+        vsversion: 2017
+
+    - name: Set up new PowerShell (pwsh)
       if: ${{ startsWith(inputs.host-platform, 'win') }}
       shell: powershell
       env:
diff --git a/.github/workflows/gh-build.yml b/.github/workflows/gh-build.yml
index 70f7772d6..f0f419c6c 100644
--- a/.github/workflows/gh-build.yml
+++ b/.github/workflows/gh-build.yml
@@ -67,4 +67,3 @@ jobs:
           target-device: "${{ inputs.target-device }}"
           host-platform: ${{ inputs.host-platform }}
           upload-enabled: ${{ inputs.upload-enabled }}
-          python-version: ${{ inputs.python-version }}

From 7044c8b7b265dab24bcc76dd04fe75de04ac7222 Mon Sep 17 00:00:00 2001
From: Leo Fang <leof@nvidia.com>
Date: Thu, 5 Dec 2024 17:33:51 +0000
Subject: [PATCH 25/65] install msvc ourselves

---
 .github/actions/setup/action.yml              | 42 ++++++++++++---
 continuous_integration/scripts/install-vs.ps1 | 53 +++++++++++++++++++
 2 files changed, 89 insertions(+), 6 deletions(-)
 create mode 100644 continuous_integration/scripts/install-vs.ps1

diff --git a/.github/actions/setup/action.yml b/.github/actions/setup/action.yml
index 94c6c0a5e..b7cdddca1 100644
--- a/.github/actions/setup/action.yml
+++ b/.github/actions/setup/action.yml
@@ -32,12 +32,6 @@ runs:
       with:
         python-version: ${{ inputs.python-version }}
 
-    - name: Set up MSVC
-      if: ${{ startsWith(inputs.host-platform, 'win') }}
-      uses: ilammy/msvc-dev-cmd@v1
-      with:
-        vsversion: 2017
-
     - name: Set up new PowerShell (pwsh)
       if: ${{ startsWith(inputs.host-platform, 'win') }}
       shell: powershell
@@ -49,6 +43,42 @@ runs:
         Start-Process msiexec.exe -Wait -Verbose -ArgumentList '/i "ps_installer.msi" /qn REGISTER_MANIFEST=1'
         echo "$env:ProgramFiles\\PowerShell\\7" >> $env:GITHUB_PATH
 
+    - name: Install MSVC
+      if: ${{ startsWith(inputs.host-platform, 'win') }}
+      shell: pwsh
+      env:
+      run: |
+        # Download & install
+        .\\continuous_integration\\scripts\\install-vs.ps1 -msvcVersion 15 -clVersion 14.16
+
+        # Verify
+        # Note: this assumes vs2017, e.g. see _find_vc2017():
+        # https://github.com/pypa/setuptools/blob/9692cde009af4651819d18a1e839d3b6e3fcd77d/setuptools/_distutils/_msvccompiler.py#L67
+        $vsPath = & "${env:ProgramFiles(x86)}\\Microsoft Visual Studio\\Installer\\vswhere.exe" `
+                  -latest `
+                  -products * `
+                  -requires Microsoft.VisualStudio.Component.VC.Tools.x86.x64 `
+                  -property installationPath
+        $clPath = Join-Path $vsPath "VC\\Tools\\MSVC\\*\\bin\\Hostx64\\x64\\cl.exe"
+        $clPath = (Get-ChildItem $clPath).FullName
+        echo "found cl.exe: $clPath"
+        # For some reason below just doesn't work in the CI...
+        Start-Process -NoNewWindow -RedirectStandardError cl.out -FilePath "$clPath"
+        $CL_VERSION_STRING = & type cl.out
+        if (($CL_VERSION_STRING -join " ") -match "Version (\\d+\\.\\d+)\\.\\d+") {
+            $CL_VERSION = $matches[1]
+            echo "Detected cl.exe version: $CL_VERSION"
+        }
+
+        # Delete temporary items created by installers and such
+        Remove-Item -Recurse $ENV:TEMP\*
+
+    - name: Set up MSVC
+      if: ${{ startsWith(inputs.host-platform, 'win') }}
+      uses: ilammy/msvc-dev-cmd@v1
+      with:
+        vsversion: 2017
+
     - name: Set REPO_DIR and Dump environment
       shell: bash --noprofile --norc -xeuo pipefail {0}
       run: |
diff --git a/continuous_integration/scripts/install-vs.ps1 b/continuous_integration/scripts/install-vs.ps1
new file mode 100644
index 000000000..3b70b3034
--- /dev/null
+++ b/continuous_integration/scripts/install-vs.ps1
@@ -0,0 +1,53 @@
+# from https://github.com/rapidsai/devcontainers/blob/branch-25.02/windows/image/installers/install-vs.ps1
+
+Param(
+    [Parameter(Mandatory=$true)]
+    [ValidateSet('15', '16', '17')]
+    [string]
+    $msvcVersion,
+    [Parameter(Mandatory=$false)]
+    [string]
+    $clVersion
+)
+
+$msvcPath = "C:\msbuild\$msvcVersion"
+
+$vsComponentsMap = @{
+    "15"     = "Microsoft.VisualStudio.Component.Windows10SDK.17763"
+    "16"     = "Microsoft.VisualStudio.Component.Windows10SDK.20348"
+    "17"     = "Microsoft.VisualStudio.Component.Windows10SDK.20348"
+    "14.14"  = "Microsoft.VisualStudio.Component.VC.Tools.14.14"
+    "14.15"  = "Microsoft.VisualStudio.Component.VC.Tools.14.15"
+    "14.16"  = "Microsoft.VisualStudio.Component.VC.Tools.14.16"
+    "14.27"  = "Microsoft.VisualStudio.Component.VC.14.27.x86.x64"
+    "14.28"  = "Microsoft.VisualStudio.Component.VC.14.28.x86.x64"
+    "14.29"  = "Microsoft.VisualStudio.Component.VC.14.29.x86.x64"
+    "14.34"  = "Microsoft.VisualStudio.Component.VC.14.34.17.4.x86.x64"
+    "14.35"  = "Microsoft.VisualStudio.Component.VC.14.35.17.5.x86.x64"
+    "14.36"  = "Microsoft.VisualStudio.Component.VC.14.36.17.6.x86.x64"
+    "14.37"  = "Microsoft.VisualStudio.Component.VC.14.37.17.7.x86.x64"
+    "14.38"  = "Microsoft.VisualStudio.Component.VC.14.38.17.8.x86.x64"
+    "14.39"  = "Microsoft.VisualStudio.Component.VC.14.39.17.9.x86.x64"
+    "latest" = "Microsoft.VisualStudio.Component.VC.Tools.x86.x64"
+}
+
+# Always install/update core VC tools
+$vsComponent = $vsComponentsMap[$msvcVersion]
+$vsComponentString = "--add Microsoft.VisualStudio.Component.VC.Tools.x86.x64 --add $vsComponent"
+
+if ($clVersion) {
+    $clComponent = $vsComponentsMap[$clVersion]
+    $vsComponentString = "$vsComponentString --add $clComponent"
+}
+
+# 14.16 alone does not install build tools
+if ($clVersion -eq "14.16") {
+    $clComponent = $vsComponentsMap["14.15"]
+    $vsComponentString = "$vsComponentString --add $clComponent"
+}
+
+Invoke-WebRequest -Uri "https://aka.ms/vs/$msvcVersion/release/vs_buildtools.exe" -UseBasicParsing -OutFile .\vs_buildtools.exe
+Write-Output "Installing components: $vsComponentString"
+Start-Process -NoNewWindow -PassThru -Wait -FilePath .\vs_buildtools.exe -ArgumentList "install --installWhileDownloading --installPath $msvcPath --wait --norestart --nocache --quiet $vsComponentString"
+
+Remove-Item .\vs_buildtools.exe

From 9fbdff24c9ae826bc763dc149eeb1b723a7f8c9f Mon Sep 17 00:00:00 2001
From: Leo Fang <leof@nvidia.com>
Date: Thu, 5 Dec 2024 17:35:28 +0000
Subject: [PATCH 26/65] fix typo

---
 .github/actions/setup/action.yml | 1 -
 1 file changed, 1 deletion(-)

diff --git a/.github/actions/setup/action.yml b/.github/actions/setup/action.yml
index b7cdddca1..aa77bff4c 100644
--- a/.github/actions/setup/action.yml
+++ b/.github/actions/setup/action.yml
@@ -46,7 +46,6 @@ runs:
     - name: Install MSVC
       if: ${{ startsWith(inputs.host-platform, 'win') }}
       shell: pwsh
-      env:
       run: |
         # Download & install
         .\\continuous_integration\\scripts\\install-vs.ps1 -msvcVersion 15 -clVersion 14.16

From 8f77f1752164eb3f5b8ee4c96ac74d286c169d3e Mon Sep 17 00:00:00 2001
From: Leo Fang <leof@nvidia.com>
Date: Thu, 5 Dec 2024 17:57:35 +0000
Subject: [PATCH 27/65] skip custom cl ver check

---
 .github/actions/setup/action.yml | 17 +++++++++--------
 .github/workflows/gh-build.yml   |  2 +-
 2 files changed, 10 insertions(+), 9 deletions(-)

diff --git a/.github/actions/setup/action.yml b/.github/actions/setup/action.yml
index aa77bff4c..715ed65fb 100644
--- a/.github/actions/setup/action.yml
+++ b/.github/actions/setup/action.yml
@@ -61,21 +61,22 @@ runs:
         $clPath = Join-Path $vsPath "VC\\Tools\\MSVC\\*\\bin\\Hostx64\\x64\\cl.exe"
         $clPath = (Get-ChildItem $clPath).FullName
         echo "found cl.exe: $clPath"
-        # For some reason below just doesn't work in the CI...
-        Start-Process -NoNewWindow -RedirectStandardError cl.out -FilePath "$clPath"
-        $CL_VERSION_STRING = & type cl.out
-        if (($CL_VERSION_STRING -join " ") -match "Version (\\d+\\.\\d+)\\.\\d+") {
-            $CL_VERSION = $matches[1]
-            echo "Detected cl.exe version: $CL_VERSION"
-        }
+        # # For some reason below just doesn't work in the CI...
+        # Start-Process -NoNewWindow -RedirectStandardError cl.out -FilePath "$clPath"
+        # $CL_VERSION_STRING = & type cl.out
+        # if (($CL_VERSION_STRING -join " ") -match "Version (\\d+\\.\\d+)\\.\\d+") {
+        #     $CL_VERSION = $matches[1]
+        #     echo "Detected cl.exe version: $CL_VERSION"
+        # }
 
         # Delete temporary items created by installers and such
-        Remove-Item -Recurse $ENV:TEMP\*
+        Remove-Item -Recurse "${env:TEMP}\\*"
 
     - name: Set up MSVC
       if: ${{ startsWith(inputs.host-platform, 'win') }}
       uses: ilammy/msvc-dev-cmd@v1
       with:
+        toolset: 14.16
         vsversion: 2017
 
     - name: Set REPO_DIR and Dump environment
diff --git a/.github/workflows/gh-build.yml b/.github/workflows/gh-build.yml
index f0f419c6c..494517692 100644
--- a/.github/workflows/gh-build.yml
+++ b/.github/workflows/gh-build.yml
@@ -49,7 +49,7 @@ jobs:
         with:
           fetch-depth: 0
 
-      - name: Setup environment variables
+      - name: Set up build environment
         uses: ./.github/actions/setup
         with:
           client-repo: ${{ inputs.client-repo }}

From 29e9e2ed27f4f57aa6fd5c6dedb40ebc0629e313 Mon Sep 17 00:00:00 2001
From: Leo Fang <leof@nvidia.com>
Date: Thu, 5 Dec 2024 18:29:02 +0000
Subject: [PATCH 28/65] install to standard location

---
 continuous_integration/scripts/install-vs.ps1 | 6 ++----
 1 file changed, 2 insertions(+), 4 deletions(-)

diff --git a/continuous_integration/scripts/install-vs.ps1 b/continuous_integration/scripts/install-vs.ps1
index 3b70b3034..3b10a6472 100644
--- a/continuous_integration/scripts/install-vs.ps1
+++ b/continuous_integration/scripts/install-vs.ps1
@@ -1,4 +1,4 @@
-# from https://github.com/rapidsai/devcontainers/blob/branch-25.02/windows/image/installers/install-vs.ps1
+# Modified from https://github.com/rapidsai/devcontainers/blob/branch-25.02/windows/image/installers/install-vs.ps1
 
 Param(
     [Parameter(Mandatory=$true)]
@@ -10,8 +10,6 @@ Param(
     $clVersion
 )
 
-$msvcPath = "C:\msbuild\$msvcVersion"
-
 $vsComponentsMap = @{
     "15"     = "Microsoft.VisualStudio.Component.Windows10SDK.17763"
     "16"     = "Microsoft.VisualStudio.Component.Windows10SDK.20348"
@@ -48,6 +46,6 @@ if ($clVersion -eq "14.16") {
 
 Invoke-WebRequest -Uri "https://aka.ms/vs/$msvcVersion/release/vs_buildtools.exe" -UseBasicParsing -OutFile .\vs_buildtools.exe
 Write-Output "Installing components: $vsComponentString"
-Start-Process -NoNewWindow -PassThru -Wait -FilePath .\vs_buildtools.exe -ArgumentList "install --installWhileDownloading --installPath $msvcPath --wait --norestart --nocache --quiet $vsComponentString"
+Start-Process -NoNewWindow -PassThru -Wait -FilePath .\vs_buildtools.exe -ArgumentList "install --installWhileDownloading --wait --norestart --nocache --quiet $vsComponentString"
 
 Remove-Item .\vs_buildtools.exe

From e2cffcea8d260934a1caed3b3b15657f4e740b59 Mon Sep 17 00:00:00 2001
From: Leo Fang <leof@nvidia.com>
Date: Thu, 5 Dec 2024 19:24:16 +0000
Subject: [PATCH 29/65] try to locate Python include path

---
 .github/actions/build/action.yml | 3 +++
 1 file changed, 3 insertions(+)

diff --git a/.github/actions/build/action.yml b/.github/actions/build/action.yml
index a61b4222f..cfe8f66c8 100644
--- a/.github/actions/build/action.yml
+++ b/.github/actions/build/action.yml
@@ -28,6 +28,9 @@ runs:
         CIBW_ARCHS_LINUX: "native"
         CIBW_REPAIR_WHEEL_COMMAND: ""  # no repair
         CIBW_BUILD_VERBOSITY: 1
+        CIBW_BEFORE_BUILD_WINDOWS: >
+          # ensure Python.h & co can be found
+          python -c "import sysconfig; print(sysconfig.get_path('include'))" >> $env:INCLUDE
       with:
         package-dir: ./cuda_core/
         output-dir: ${{ env.CUDA_CORE_ARTIFACTS_DIR }}

From a7157b5608583f5996cecf48617507c24b7a1619 Mon Sep 17 00:00:00 2001
From: Leo Fang <leof@nvidia.com>
Date: Thu, 5 Dec 2024 20:02:26 +0000
Subject: [PATCH 30/65] switch to public windows runner for now

---
 .github/actions/build/action.yml        |  6 +-
 .github/actions/setup/action.yml        | 78 ++++++++++++-------------
 .github/workflows/gh-build-and-test.yml |  3 +-
 3 files changed, 44 insertions(+), 43 deletions(-)

diff --git a/.github/actions/build/action.yml b/.github/actions/build/action.yml
index cfe8f66c8..e93d11140 100644
--- a/.github/actions/build/action.yml
+++ b/.github/actions/build/action.yml
@@ -28,9 +28,9 @@ runs:
         CIBW_ARCHS_LINUX: "native"
         CIBW_REPAIR_WHEEL_COMMAND: ""  # no repair
         CIBW_BUILD_VERBOSITY: 1
-        CIBW_BEFORE_BUILD_WINDOWS: >
-          # ensure Python.h & co can be found
-          python -c "import sysconfig; print(sysconfig.get_path('include'))" >> $env:INCLUDE
+        # # ensure Python.h & co can be found
+        # CIBW_BEFORE_BUILD_WINDOWS: >
+        #   python -c "import sysconfig; print(sysconfig.get_path('include'))" >> $env:INCLUDE
       with:
         package-dir: ./cuda_core/
         output-dir: ${{ env.CUDA_CORE_ARTIFACTS_DIR }}
diff --git a/.github/actions/setup/action.yml b/.github/actions/setup/action.yml
index 715ed65fb..6f9ed1dc8 100644
--- a/.github/actions/setup/action.yml
+++ b/.github/actions/setup/action.yml
@@ -32,52 +32,52 @@ runs:
       with:
         python-version: ${{ inputs.python-version }}
 
-    - name: Set up new PowerShell (pwsh)
-      if: ${{ startsWith(inputs.host-platform, 'win') }}
-      shell: powershell
-      env:
-        # doesn't seem there's an easy way to avoid hard-coding it?
-        PS_MSI_URL: https://github.com/PowerShell/PowerShell/releases/download/v7.4.6/PowerShell-7.4.6-win-x64.msi
-      run: |
-        Invoke-WebRequest -Uri "$env:PS_MSI_URL" -OutFile "ps_installer.msi"
-        Start-Process msiexec.exe -Wait -Verbose -ArgumentList '/i "ps_installer.msi" /qn REGISTER_MANIFEST=1'
-        echo "$env:ProgramFiles\\PowerShell\\7" >> $env:GITHUB_PATH
+    # - name: Set up new PowerShell (pwsh)
+    #   if: ${{ startsWith(inputs.host-platform, 'win') }}
+    #   shell: powershell
+    #   env:
+    #     # doesn't seem there's an easy way to avoid hard-coding it?
+    #     PS_MSI_URL: https://github.com/PowerShell/PowerShell/releases/download/v7.4.6/PowerShell-7.4.6-win-x64.msi
+    #   run: |
+    #     Invoke-WebRequest -Uri "$env:PS_MSI_URL" -OutFile "ps_installer.msi"
+    #     Start-Process msiexec.exe -Wait -Verbose -ArgumentList '/i "ps_installer.msi" /qn REGISTER_MANIFEST=1'
+    #     echo "$env:ProgramFiles\\PowerShell\\7" >> $env:GITHUB_PATH
 
-    - name: Install MSVC
-      if: ${{ startsWith(inputs.host-platform, 'win') }}
-      shell: pwsh
-      run: |
-        # Download & install
-        .\\continuous_integration\\scripts\\install-vs.ps1 -msvcVersion 15 -clVersion 14.16
+    # - name: Install MSVC
+    #   if: ${{ startsWith(inputs.host-platform, 'win') }}
+    #   shell: pwsh
+    #   run: |
+    #     # Download & install
+    #     .\\continuous_integration\\scripts\\install-vs.ps1 -msvcVersion 15 -clVersion 14.16
 
-        # Verify
-        # Note: this assumes vs2017, e.g. see _find_vc2017():
-        # https://github.com/pypa/setuptools/blob/9692cde009af4651819d18a1e839d3b6e3fcd77d/setuptools/_distutils/_msvccompiler.py#L67
-        $vsPath = & "${env:ProgramFiles(x86)}\\Microsoft Visual Studio\\Installer\\vswhere.exe" `
-                  -latest `
-                  -products * `
-                  -requires Microsoft.VisualStudio.Component.VC.Tools.x86.x64 `
-                  -property installationPath
-        $clPath = Join-Path $vsPath "VC\\Tools\\MSVC\\*\\bin\\Hostx64\\x64\\cl.exe"
-        $clPath = (Get-ChildItem $clPath).FullName
-        echo "found cl.exe: $clPath"
-        # # For some reason below just doesn't work in the CI...
-        # Start-Process -NoNewWindow -RedirectStandardError cl.out -FilePath "$clPath"
-        # $CL_VERSION_STRING = & type cl.out
-        # if (($CL_VERSION_STRING -join " ") -match "Version (\\d+\\.\\d+)\\.\\d+") {
-        #     $CL_VERSION = $matches[1]
-        #     echo "Detected cl.exe version: $CL_VERSION"
-        # }
+    #     # Verify
+    #     # Note: this assumes vs2017, e.g. see _find_vc2017():
+    #     # https://github.com/pypa/setuptools/blob/9692cde009af4651819d18a1e839d3b6e3fcd77d/setuptools/_distutils/_msvccompiler.py#L67
+    #     $vsPath = & "${env:ProgramFiles(x86)}\\Microsoft Visual Studio\\Installer\\vswhere.exe" `
+    #               -latest `
+    #               -products * `
+    #               -requires Microsoft.VisualStudio.Component.VC.Tools.x86.x64 `
+    #               -property installationPath
+    #     $clPath = Join-Path $vsPath "VC\\Tools\\MSVC\\*\\bin\\Hostx64\\x64\\cl.exe"
+    #     $clPath = (Get-ChildItem $clPath).FullName
+    #     echo "found cl.exe: $clPath"
+    #     # # For some reason below just doesn't work in the CI...
+    #     # Start-Process -NoNewWindow -RedirectStandardError cl.out -FilePath "$clPath"
+    #     # $CL_VERSION_STRING = & type cl.out
+    #     # if (($CL_VERSION_STRING -join " ") -match "Version (\\d+\\.\\d+)\\.\\d+") {
+    #     #     $CL_VERSION = $matches[1]
+    #     #     echo "Detected cl.exe version: $CL_VERSION"
+    #     # }
 
-        # Delete temporary items created by installers and such
-        Remove-Item -Recurse "${env:TEMP}\\*"
+    #     # Delete temporary items created by installers and such
+    #     Remove-Item -Recurse "${env:TEMP}\\*"
 
     - name: Set up MSVC
       if: ${{ startsWith(inputs.host-platform, 'win') }}
       uses: ilammy/msvc-dev-cmd@v1
-      with:
-        toolset: 14.16
-        vsversion: 2017
+    #  with:
+    #    toolset: 14.16
+    #    vsversion: 2017
 
     - name: Set REPO_DIR and Dump environment
       shell: bash --noprofile --norc -xeuo pipefail {0}
diff --git a/.github/workflows/gh-build-and-test.yml b/.github/workflows/gh-build-and-test.yml
index 867cffde0..8d1c1b97d 100644
--- a/.github/workflows/gh-build-and-test.yml
+++ b/.github/workflows/gh-build-and-test.yml
@@ -29,7 +29,8 @@ jobs:
       target-device: ${{ inputs.target-device }}
       runs-on: ${{ (inputs.host-platform == 'linux-x64' && 'linux-amd64-cpu8') ||
                    (inputs.host-platform == 'linux-aarch64' && 'linux-arm64-cpu8') ||
-                   (inputs.host-platform == 'win-x64' && 'windows-amd64-cpu8') }}
+                   (inputs.host-platform == 'win-x64' && 'windows-2019') }}
+                 #  (inputs.host-platform == 'win-x64' && 'windows-amd64-cpu8') }}
       build-type: ${{ inputs.build-type }}
       host-platform: ${{ inputs.host-platform }}
       build-mode: ${{ inputs.build-mode }}

From 965b23f04e8eb8b88a26f251742ccbc0cd2bb852 Mon Sep 17 00:00:00 2001
From: Leo Fang <leof@nvidia.com>
Date: Thu, 5 Dec 2024 20:18:46 +0000
Subject: [PATCH 31/65] windows image does not have sudo

---
 .github/actions/build/action.yml | 7 ++++++-
 1 file changed, 6 insertions(+), 1 deletion(-)

diff --git a/.github/actions/build/action.yml b/.github/actions/build/action.yml
index e93d11140..0ed4ab192 100644
--- a/.github/actions/build/action.yml
+++ b/.github/actions/build/action.yml
@@ -50,7 +50,12 @@ runs:
     - name: List the cuda.core artifacts directory
       shell: bash --noprofile --norc -xeuo pipefail {0}
       run: |
-        sudo chown -R $(whoami) ${{ env.CUDA_CORE_ARTIFACTS_DIR }}
+        if [[ "${{ inputs.host-platform }}" == win* ]]; then
+          export CHOWN=chown
+        else
+          export CHOWN="sudo chown"
+        fi
+        $CHOWN -R $(whoami) ${{ env.CUDA_CORE_ARTIFACTS_DIR }}
         ls -lahR ${{ env.CUDA_CORE_ARTIFACTS_DIR }}
 
     - name: Upload cuda.core build artifacts

From 8ae7bd2e633ede890ba536063321539dcbfcf921 Mon Sep 17 00:00:00 2001
From: Leo Fang <leof@nvidia.com>
Date: Thu, 5 Dec 2024 20:51:04 +0000
Subject: [PATCH 32/65] pwd on Windows Bash does not use Windows path format

---
 .github/actions/setup/action.yml | 11 ++++++++---
 1 file changed, 8 insertions(+), 3 deletions(-)

diff --git a/.github/actions/setup/action.yml b/.github/actions/setup/action.yml
index 6f9ed1dc8..3fd9e06d7 100644
--- a/.github/actions/setup/action.yml
+++ b/.github/actions/setup/action.yml
@@ -82,7 +82,12 @@ runs:
     - name: Set REPO_DIR and Dump environment
       shell: bash --noprofile --norc -xeuo pipefail {0}
       run: |
-        echo "REPO_DIR=$(pwd)" >> $GITHUB_ENV
+        if [[ "${{ inputs.host-platform }}" == linux* ]]; then
+          echo "REPO_DIR=$(pwd)" >> $GITHUB_ENV
+        else
+          PWD=$(pwd)
+          echo "REPO_DIR=$(cygpath -w $PWD)" >> $GITHUB_ENV
+        fi
         env
 
     - name: Set environment variables
@@ -127,9 +132,9 @@ runs:
           PKG_DIR="${BUILD_MODE}/${TARGET_PLATFORM}"
         fi
 
-        echo "CUDA_BINDINGS_ARTIFACTS_DIR=$(realpath "$(pwd)/cuda_bindings/dist")" >> $GITHUB_ENV
+        echo "CUDA_BINDINGS_ARTIFACTS_DIR=$(realpath "$REPO_DIR/cuda_bindings/dist")" >> $GITHUB_ENV
         echo "CUDA_CORE_ARTIFACT_NAME=cuda-core-python${{ inputs.python-version }}-${{ inputs.host-platform }}-${{ inputs.build-type }}-${{ github.sha }}" >> $GITHUB_ENV
-        echo "CUDA_CORE_ARTIFACTS_DIR=$(realpath "$(pwd)/cuda_core/dist")" >> $GITHUB_ENV
+        echo "CUDA_CORE_ARTIFACTS_DIR=$(realpath "$REPO_DIR/cuda_core/dist")" >> $GITHUB_ENV
         echo "USE_CUDA=${{ (inputs.target-device == 'cpu' && 'OFF') || 'ON' }}" >> $GITHUB_ENV
         echo "UPLOAD_ENABLED=${{ (inputs.upload-enabled  == 'true' && 'ON') || 'OFF' }}" >> $GITHUB_ENV
         echo "BUILD_DATE=$(date +%Y%m%d)" >> $GITHUB_ENV

From 2175ccbcc2db39621497b471dd69d3a8900bd6dd Mon Sep 17 00:00:00 2001
From: Leo Fang <leof@nvidia.com>
Date: Thu, 5 Dec 2024 20:59:34 +0000
Subject: [PATCH 33/65] cover all Python versions!

---
 .github/workflows/ci-gh.yml | 5 ++++-
 1 file changed, 4 insertions(+), 1 deletion(-)

diff --git a/.github/workflows/ci-gh.yml b/.github/workflows/ci-gh.yml
index cf1683257..43cb16bbe 100644
--- a/.github/workflows/ci-gh.yml
+++ b/.github/workflows/ci-gh.yml
@@ -27,8 +27,11 @@ jobs:
         upload-enabled:
           - false
         python-version:
-          #TODO cover the whole python and cuda matrix
           - 3.12
+          - 3.11
+          - 3.10
+          - 3.9
+        # TODO cover the whole cuda matrix
     uses:
       ./.github/workflows/gh-build-and-test.yml
     with:

From 50eb0e732d94f43ce5e424d94ad434357bad89bd Mon Sep 17 00:00:00 2001
From: Leo Fang <leof@nvidia.com>
Date: Thu, 5 Dec 2024 21:06:35 +0000
Subject: [PATCH 34/65] add quotes

---
 .github/workflows/ci-gh.yml | 8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)

diff --git a/.github/workflows/ci-gh.yml b/.github/workflows/ci-gh.yml
index 43cb16bbe..62bb036cd 100644
--- a/.github/workflows/ci-gh.yml
+++ b/.github/workflows/ci-gh.yml
@@ -27,10 +27,10 @@ jobs:
         upload-enabled:
           - false
         python-version:
-          - 3.12
-          - 3.11
-          - 3.10
-          - 3.9
+          - "3.12"
+          - "3.11"
+          - "3.10"
+          - "3.9"
         # TODO cover the whole cuda matrix
     uses:
       ./.github/workflows/gh-build-and-test.yml

From b0046ac5cc6d9b42ff381b07c8dbdab3640e3d4d Mon Sep 17 00:00:00 2001
From: Leo Fang <leof@nvidia.com>
Date: Thu, 5 Dec 2024 21:13:17 +0000
Subject: [PATCH 35/65] align the python version installed via GHA vs used at
 build time

---
 .github/actions/build/action.yml | 2 +-
 .github/actions/setup/action.yml | 1 +
 2 files changed, 2 insertions(+), 1 deletion(-)

diff --git a/.github/actions/build/action.yml b/.github/actions/build/action.yml
index 0ed4ab192..e42be1c9c 100644
--- a/.github/actions/build/action.yml
+++ b/.github/actions/build/action.yml
@@ -24,7 +24,7 @@ runs:
       uses: pypa/cibuildwheel@v2.22.0
       env:
         CIBW_BUILD: ${{ env.CIBW_BUILD }}
-        CIBW_PROJECT_REQUIRES_PYTHON: ">=3.6"
+        CIBW_PROJECT_REQUIRES_PYTHON: ${{ env.CIBW_PROJECT_REQUIRES_PYTHON }}
         CIBW_ARCHS_LINUX: "native"
         CIBW_REPAIR_WHEEL_COMMAND: ""  # no repair
         CIBW_BUILD_VERBOSITY: 1
diff --git a/.github/actions/setup/action.yml b/.github/actions/setup/action.yml
index 3fd9e06d7..95953799e 100644
--- a/.github/actions/setup/action.yml
+++ b/.github/actions/setup/action.yml
@@ -142,3 +142,4 @@ runs:
         echo "PKG_DIR=${PKG_DIR}" >> $GITHUB_ENV
         echo "PYTHON_VERSION=${{ inputs.python-version }}" >> $GITHUB_ENV
         echo "CIBW_BUILD=${CIBW_BUILD}" >> $GITHUB_ENV
+        echo "CIBW_PROJECT_REQUIRES_PYTHON=${{ inputs.python-version }}" >> $GITHUB_ENV

From 493266739f19ff99a99855526f14a02fdedc921b Mon Sep 17 00:00:00 2001
From: Leo Fang <leof@nvidia.com>
Date: Thu, 5 Dec 2024 21:16:49 +0000
Subject: [PATCH 36/65] fix constraint syntax

---
 .github/actions/setup/action.yml | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/.github/actions/setup/action.yml b/.github/actions/setup/action.yml
index 95953799e..f3624dd0c 100644
--- a/.github/actions/setup/action.yml
+++ b/.github/actions/setup/action.yml
@@ -142,4 +142,4 @@ runs:
         echo "PKG_DIR=${PKG_DIR}" >> $GITHUB_ENV
         echo "PYTHON_VERSION=${{ inputs.python-version }}" >> $GITHUB_ENV
         echo "CIBW_BUILD=${CIBW_BUILD}" >> $GITHUB_ENV
-        echo "CIBW_PROJECT_REQUIRES_PYTHON=${{ inputs.python-version }}" >> $GITHUB_ENV
+        echo "CIBW_PROJECT_REQUIRES_PYTHON=\"==${{ inputs.python-version }}\"" >> $GITHUB_ENV

From 7cb84c3a147dfbd5a60923bb4db4045020ca0395 Mon Sep 17 00:00:00 2001
From: Leo Fang <leof@nvidia.com>
Date: Thu, 5 Dec 2024 21:30:04 +0000
Subject: [PATCH 37/65] check if setup-python is causing interference

---
 .github/actions/build/action.yml |  2 +-
 .github/actions/setup/action.yml | 28 +++++++++++++++-------------
 2 files changed, 16 insertions(+), 14 deletions(-)

diff --git a/.github/actions/build/action.yml b/.github/actions/build/action.yml
index e42be1c9c..1b58ba6c2 100644
--- a/.github/actions/build/action.yml
+++ b/.github/actions/build/action.yml
@@ -24,7 +24,7 @@ runs:
       uses: pypa/cibuildwheel@v2.22.0
       env:
         CIBW_BUILD: ${{ env.CIBW_BUILD }}
-        CIBW_PROJECT_REQUIRES_PYTHON: ${{ env.CIBW_PROJECT_REQUIRES_PYTHON }}
+        #CIBW_PROJECT_REQUIRES_PYTHON: ${{ env.CIBW_PROJECT_REQUIRES_PYTHON }}
         CIBW_ARCHS_LINUX: "native"
         CIBW_REPAIR_WHEEL_COMMAND: ""  # no repair
         CIBW_BUILD_VERBOSITY: 1
diff --git a/.github/actions/setup/action.yml b/.github/actions/setup/action.yml
index f3624dd0c..7b90e8ee7 100644
--- a/.github/actions/setup/action.yml
+++ b/.github/actions/setup/action.yml
@@ -26,11 +26,11 @@ inputs:
 runs:
   using: composite
   steps:
-    - name: Set up Python ${{ inputs.python-version }}
-      id: setup-python
-      uses: actions/setup-python@v5
-      with:
-        python-version: ${{ inputs.python-version }}
+    # - name: Set up Python ${{ inputs.python-version }}
+    #   id: setup-python
+    #   uses: actions/setup-python@v5
+    #   with:
+    #     python-version: ${{ inputs.python-version }}
 
     # - name: Set up new PowerShell (pwsh)
     #   if: ${{ startsWith(inputs.host-platform, 'win') }}
@@ -79,15 +79,9 @@ runs:
     #    toolset: 14.16
     #    vsversion: 2017
 
-    - name: Set REPO_DIR and Dump environment
+    - name: Dump environment
       shell: bash --noprofile --norc -xeuo pipefail {0}
       run: |
-        if [[ "${{ inputs.host-platform }}" == linux* ]]; then
-          echo "REPO_DIR=$(pwd)" >> $GITHUB_ENV
-        else
-          PWD=$(pwd)
-          echo "REPO_DIR=$(cygpath -w $PWD)" >> $GITHUB_ENV
-        fi
         env
 
     - name: Set environment variables
@@ -132,6 +126,14 @@ runs:
           PKG_DIR="${BUILD_MODE}/${TARGET_PLATFORM}"
         fi
 
+        if [[ "${{ inputs.host-platform }}" == linux* ]]; then
+          REPO_DIR=$(pwd)"
+        else
+          PWD=$(pwd)
+          REPO_DIR=$(cygpath -w $PWD)"
+        fi
+
+        echo "REPO_DIR=$REPO_DIR" >> $GITHUB_ENV
         echo "CUDA_BINDINGS_ARTIFACTS_DIR=$(realpath "$REPO_DIR/cuda_bindings/dist")" >> $GITHUB_ENV
         echo "CUDA_CORE_ARTIFACT_NAME=cuda-core-python${{ inputs.python-version }}-${{ inputs.host-platform }}-${{ inputs.build-type }}-${{ github.sha }}" >> $GITHUB_ENV
         echo "CUDA_CORE_ARTIFACTS_DIR=$(realpath "$REPO_DIR/cuda_core/dist")" >> $GITHUB_ENV
@@ -142,4 +144,4 @@ runs:
         echo "PKG_DIR=${PKG_DIR}" >> $GITHUB_ENV
         echo "PYTHON_VERSION=${{ inputs.python-version }}" >> $GITHUB_ENV
         echo "CIBW_BUILD=${CIBW_BUILD}" >> $GITHUB_ENV
-        echo "CIBW_PROJECT_REQUIRES_PYTHON=\"==${{ inputs.python-version }}\"" >> $GITHUB_ENV
+        #echo "CIBW_PROJECT_REQUIRES_PYTHON=\"==${{ inputs.python-version }}\"" >> $GITHUB_ENV

From d97fbddb7a583e70b95182065f18292b962c2c7a Mon Sep 17 00:00:00 2001
From: Leo Fang <leof@nvidia.com>
Date: Thu, 5 Dec 2024 21:35:20 +0000
Subject: [PATCH 38/65] fix typo

---
 .github/actions/setup/action.yml | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/.github/actions/setup/action.yml b/.github/actions/setup/action.yml
index 7b90e8ee7..bf7022f3d 100644
--- a/.github/actions/setup/action.yml
+++ b/.github/actions/setup/action.yml
@@ -127,10 +127,10 @@ runs:
         fi
 
         if [[ "${{ inputs.host-platform }}" == linux* ]]; then
-          REPO_DIR=$(pwd)"
+          REPO_DIR=$(pwd)
         else
           PWD=$(pwd)
-          REPO_DIR=$(cygpath -w $PWD)"
+          REPO_DIR=$(cygpath -w $PWD)
         fi
 
         echo "REPO_DIR=$REPO_DIR" >> $GITHUB_ENV

From 59e4e0a464498346c5f4ee311969e2c20cd5711a Mon Sep 17 00:00:00 2001
From: Leo Fang <leof@nvidia.com>
Date: Thu, 5 Dec 2024 22:32:54 +0000
Subject: [PATCH 39/65] apply a WAR on Linux

---
 .github/actions/setup/action.yml | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)

diff --git a/.github/actions/setup/action.yml b/.github/actions/setup/action.yml
index bf7022f3d..d8e29bccf 100644
--- a/.github/actions/setup/action.yml
+++ b/.github/actions/setup/action.yml
@@ -128,6 +128,9 @@ runs:
 
         if [[ "${{ inputs.host-platform }}" == linux* ]]; then
           REPO_DIR=$(pwd)
+          # WAR: setup-python is not relocatable...
+          # see https://github.com/actions/setup-python/issues/871
+          echo "LD_LIBRARY_PATH=/opt/hostedtoolcache/Python/lib:$LD_LIBRARY_PATH" >> $GITHUB_ENV
         else
           PWD=$(pwd)
           REPO_DIR=$(cygpath -w $PWD)
@@ -144,4 +147,3 @@ runs:
         echo "PKG_DIR=${PKG_DIR}" >> $GITHUB_ENV
         echo "PYTHON_VERSION=${{ inputs.python-version }}" >> $GITHUB_ENV
         echo "CIBW_BUILD=${CIBW_BUILD}" >> $GITHUB_ENV
-        #echo "CIBW_PROJECT_REQUIRES_PYTHON=\"==${{ inputs.python-version }}\"" >> $GITHUB_ENV

From 2ab0877b0ceb0d670ac0ebc5732b33da01a9c5fd Mon Sep 17 00:00:00 2001
From: Leo Fang <leof@nvidia.com>
Date: Thu, 5 Dec 2024 22:35:51 +0000
Subject: [PATCH 40/65] fix unbound var

---
 .github/actions/setup/action.yml | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/.github/actions/setup/action.yml b/.github/actions/setup/action.yml
index d8e29bccf..876803897 100644
--- a/.github/actions/setup/action.yml
+++ b/.github/actions/setup/action.yml
@@ -130,7 +130,7 @@ runs:
           REPO_DIR=$(pwd)
           # WAR: setup-python is not relocatable...
           # see https://github.com/actions/setup-python/issues/871
-          echo "LD_LIBRARY_PATH=/opt/hostedtoolcache/Python/lib:$LD_LIBRARY_PATH" >> $GITHUB_ENV
+          echo "LD_LIBRARY_PATH=/opt/hostedtoolcache/Python/lib" >> $GITHUB_ENV
         else
           PWD=$(pwd)
           REPO_DIR=$(cygpath -w $PWD)

From cac9de8b6c485d25898e9702271e69501a05bd31 Mon Sep 17 00:00:00 2001
From: Leo Fang <leof@nvidia.com>
Date: Thu, 5 Dec 2024 22:48:03 +0000
Subject: [PATCH 41/65] detect Python path after it's installed (by CIBW)

---
 .github/actions/build/action.yml | 4 ++++
 .github/actions/setup/action.yml | 3 ---
 2 files changed, 4 insertions(+), 3 deletions(-)

diff --git a/.github/actions/build/action.yml b/.github/actions/build/action.yml
index 1b58ba6c2..84b36683e 100644
--- a/.github/actions/build/action.yml
+++ b/.github/actions/build/action.yml
@@ -28,6 +28,10 @@ runs:
         CIBW_ARCHS_LINUX: "native"
         CIBW_REPAIR_WHEEL_COMMAND: ""  # no repair
         CIBW_BUILD_VERBOSITY: 1
+        # WAR: setup-python is not relocatable...
+        # see https://github.com/actions/setup-python/issues/871
+        CIBW_BEFORE_BUILD_LINUX: >
+          export LD_LIBRARY_PATH="$(dirname $(which python))/../lib:${LD_LIBRARY_PATH:-}"
         # # ensure Python.h & co can be found
         # CIBW_BEFORE_BUILD_WINDOWS: >
         #   python -c "import sysconfig; print(sysconfig.get_path('include'))" >> $env:INCLUDE
diff --git a/.github/actions/setup/action.yml b/.github/actions/setup/action.yml
index 876803897..09ab9f139 100644
--- a/.github/actions/setup/action.yml
+++ b/.github/actions/setup/action.yml
@@ -128,9 +128,6 @@ runs:
 
         if [[ "${{ inputs.host-platform }}" == linux* ]]; then
           REPO_DIR=$(pwd)
-          # WAR: setup-python is not relocatable...
-          # see https://github.com/actions/setup-python/issues/871
-          echo "LD_LIBRARY_PATH=/opt/hostedtoolcache/Python/lib" >> $GITHUB_ENV
         else
           PWD=$(pwd)
           REPO_DIR=$(cygpath -w $PWD)

From a3ef630d7bc627aa53cf4f56875dfd12752d7f36 Mon Sep 17 00:00:00 2001
From: Leo Fang <leof@nvidia.com>
Date: Thu, 5 Dec 2024 22:52:39 +0000
Subject: [PATCH 42/65] try CIBW_BEFORE_ALL_LINUX

---
 .github/actions/build/action.yml | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/.github/actions/build/action.yml b/.github/actions/build/action.yml
index 84b36683e..925dd2caa 100644
--- a/.github/actions/build/action.yml
+++ b/.github/actions/build/action.yml
@@ -30,7 +30,7 @@ runs:
         CIBW_BUILD_VERBOSITY: 1
         # WAR: setup-python is not relocatable...
         # see https://github.com/actions/setup-python/issues/871
-        CIBW_BEFORE_BUILD_LINUX: >
+        CIBW_BEFORE_ALL_LINUX: >
           export LD_LIBRARY_PATH="$(dirname $(which python))/../lib:${LD_LIBRARY_PATH:-}"
         # # ensure Python.h & co can be found
         # CIBW_BEFORE_BUILD_WINDOWS: >

From 26af84d139605a86aa1276e156abaa3262e56573 Mon Sep 17 00:00:00 2001
From: Leo Fang <leof@nvidia.com>
Date: Thu, 5 Dec 2024 23:36:55 +0000
Subject: [PATCH 43/65] try to restore the pre-py-span setup...

---
 .github/actions/build/action.yml |  4 ----
 .github/actions/setup/action.yml | 13 ++++++++-----
 2 files changed, 8 insertions(+), 9 deletions(-)

diff --git a/.github/actions/build/action.yml b/.github/actions/build/action.yml
index 925dd2caa..1b58ba6c2 100644
--- a/.github/actions/build/action.yml
+++ b/.github/actions/build/action.yml
@@ -28,10 +28,6 @@ runs:
         CIBW_ARCHS_LINUX: "native"
         CIBW_REPAIR_WHEEL_COMMAND: ""  # no repair
         CIBW_BUILD_VERBOSITY: 1
-        # WAR: setup-python is not relocatable...
-        # see https://github.com/actions/setup-python/issues/871
-        CIBW_BEFORE_ALL_LINUX: >
-          export LD_LIBRARY_PATH="$(dirname $(which python))/../lib:${LD_LIBRARY_PATH:-}"
         # # ensure Python.h & co can be found
         # CIBW_BEFORE_BUILD_WINDOWS: >
         #   python -c "import sysconfig; print(sysconfig.get_path('include'))" >> $env:INCLUDE
diff --git a/.github/actions/setup/action.yml b/.github/actions/setup/action.yml
index 09ab9f139..d478718c3 100644
--- a/.github/actions/setup/action.yml
+++ b/.github/actions/setup/action.yml
@@ -26,11 +26,14 @@ inputs:
 runs:
   using: composite
   steps:
-    # - name: Set up Python ${{ inputs.python-version }}
-    #   id: setup-python
-    #   uses: actions/setup-python@v5
-    #   with:
-    #     python-version: ${{ inputs.python-version }}
+    # WAR: setup-python is not relocatable...
+    # see https://github.com/actions/setup-python/issues/871
+    - name: Set up Python ${{ inputs.python-version }}
+      if: ${{ startsWith(inputs.host-platform, 'linux') }}
+      id: setup-python
+      uses: actions/setup-python@v5
+      with:
+        python-version: "3.12"
 
     # - name: Set up new PowerShell (pwsh)
     #   if: ${{ startsWith(inputs.host-platform, 'win') }}

From cdbdfb5598f4e747bf93b27426abc99aff9f4a3d Mon Sep 17 00:00:00 2001
From: Leo Fang <leof@nvidia.com>
Date: Fri, 6 Dec 2024 00:08:59 +0000
Subject: [PATCH 44/65] reduce build matrix to experiment with cuda.bindings
 builds

---
 .github/actions/build/action.yml        | 46 ++++++++++++++++++-------
 .github/actions/setup/action.yml        | 30 +++++++++++++++-
 .github/workflows/ci-gh.yml             | 16 +++++----
 .github/workflows/gh-build-and-test.yml |  5 +++
 .github/workflows/gh-build.yml          |  6 +++-
 5 files changed, 82 insertions(+), 21 deletions(-)

diff --git a/.github/actions/build/action.yml b/.github/actions/build/action.yml
index 1b58ba6c2..befef9b2c 100644
--- a/.github/actions/build/action.yml
+++ b/.github/actions/build/action.yml
@@ -24,7 +24,6 @@ runs:
       uses: pypa/cibuildwheel@v2.22.0
       env:
         CIBW_BUILD: ${{ env.CIBW_BUILD }}
-        #CIBW_PROJECT_REQUIRES_PYTHON: ${{ env.CIBW_PROJECT_REQUIRES_PYTHON }}
         CIBW_ARCHS_LINUX: "native"
         CIBW_REPAIR_WHEEL_COMMAND: ""  # no repair
         CIBW_BUILD_VERBOSITY: 1
@@ -35,18 +34,6 @@ runs:
         package-dir: ./cuda_core/
         output-dir: ${{ env.CUDA_CORE_ARTIFACTS_DIR }}
 
-    #- name: Display structure of the bindings artifacts folder (post build)
-    #  shell: bash --noprofile --norc -xeuo pipefail {0}
-    #  run: |
-    #    sudo chown -R $(whoami) ${{ env.CUDA_BINDINGS_ARTIFACTS_DIR }}
-    #    ls -lahR ${{ env.CUDA_BINDINGS_ARTIFACTS_DIR }}
-
-    #- name: Upload bindings build artifacts
-    #  uses: actions/upload-artifact@v4
-    #  with:
-    #    name: ${{ env.CUDA_BINDINGS_ARTIFACT_NAME }}
-    #    path: ${{ env.CUDA_BINDINGS_ARTIFACTS_DIR }}
-
     - name: List the cuda.core artifacts directory
       shell: bash --noprofile --norc -xeuo pipefail {0}
       run: |
@@ -65,3 +52,36 @@ runs:
         path: ${{ env.CUDA_CORE_ARTIFACTS_DIR }}/*.whl
         if-no-files-found: error
         overwrite: 'true'
+
+    - name: Build cuda.bindings wheel
+      uses: pypa/cibuildwheel@v2.22.0
+      env:
+        CIBW_BUILD: ${{ env.CIBW_BUILD }}
+        CIBW_ARCHS_LINUX: "native"
+        CIBW_REPAIR_WHEEL_COMMAND: ""  # no repair
+        CIBW_BUILD_VERBOSITY: 1
+        # # ensure Python.h & co can be found
+        # CIBW_BEFORE_BUILD_WINDOWS: >
+        #   python -c "import sysconfig; print(sysconfig.get_path('include'))" >> $env:INCLUDE
+      with:
+        package-dir: ./cuda_bindings/
+        output-dir: ${{ env.CUDA_BINDINGS_ARTIFACTS_DIR }}
+
+    - name: List the cuda.bindings artifacts directory
+      shell: bash --noprofile --norc -xeuo pipefail {0}
+      run: |
+        if [[ "${{ inputs.host-platform }}" == win* ]]; then
+          export CHOWN=chown
+        else
+          export CHOWN="sudo chown"
+        fi
+        $CHOWN -R $(whoami) ${{ env.CUDA_BINDINGS_ARTIFACTS_DIR }}
+        ls -lahR ${{ env.CUDA_BINDINGS_ARTIFACTS_DIR }}
+
+    - name: Upload cuda.bindings build artifacts
+      uses: actions/upload-artifact@v4
+      with:
+        name: ${{ env.CUDA_BINDINGS_ARTIFACT_NAME }}
+        path: ${{ env.CUDA_BINDINGS_ARTIFACTS_DIR }}/*.whl
+        if-no-files-found: error
+        overwrite: 'true'
diff --git a/.github/actions/setup/action.yml b/.github/actions/setup/action.yml
index d478718c3..49fec033a 100644
--- a/.github/actions/setup/action.yml
+++ b/.github/actions/setup/action.yml
@@ -22,6 +22,9 @@ inputs:
   python-version:
     required: true
     type: string
+  cuda-version:
+    required: true
+    type: string
 
 runs:
   using: composite
@@ -87,6 +90,29 @@ runs:
       run: |
         env
 
+    - name: Get CUDA headers
+      shell: bash --noprofile --norc -xeuo pipefail {0}
+      run: |
+        CUDA_PATH="cuda-${{ inputs.cuda-version }}"
+        mkdir -p $CUDA_PATH/include
+
+        git clone -b "cuda-${{ inputs.cuda-version }}" https://gitlab.com/nvidia/headers/cuda-individual/cudart.git
+        rm -rf cudart/.git
+        mv cudart/* $CUDA_PATH/include
+        rmdir cudart
+
+        git clone -b "cuda-${{ inputs.cuda-version }}" https://gitlab.com/nvidia/headers/cuda-individual/nvcc.git
+        rm -rf nvcc/.git
+        mv nvcc/* $CUDA_PATH/include
+        rmdir nvcc
+
+        git clone -b "cuda-${{ inputs.cuda-version }}" https://gitlab.com/nvidia/headers/cuda-individual/nvrtc.git
+        rm -rf nvrtc/.git
+        mv nvrtc/* $CUDA_PATH/include
+        rmdir nvrtc
+
+        echo "CUDA_PATH=$CUDA_PATH" >> $GITHUB_ENV
+
     - name: Set environment variables
       shell: bash --noprofile --norc -xeuo pipefail {0}
       run: |
@@ -136,10 +162,12 @@ runs:
           REPO_DIR=$(cygpath -w $PWD)
         fi
 
+        echo "PARALLEL_LEVEL=${nproc}" >> $GITHUB_ENV
         echo "REPO_DIR=$REPO_DIR" >> $GITHUB_ENV
-        echo "CUDA_BINDINGS_ARTIFACTS_DIR=$(realpath "$REPO_DIR/cuda_bindings/dist")" >> $GITHUB_ENV
         echo "CUDA_CORE_ARTIFACT_NAME=cuda-core-python${{ inputs.python-version }}-${{ inputs.host-platform }}-${{ inputs.build-type }}-${{ github.sha }}" >> $GITHUB_ENV
         echo "CUDA_CORE_ARTIFACTS_DIR=$(realpath "$REPO_DIR/cuda_core/dist")" >> $GITHUB_ENV
+        echo "CUDA_BINDINGS_ARTIFACT_NAME=cuda-bindings-python${{ inputs.python-version }}-cuda${{ inputs.cuda-version }}-${{ inputs.host-platform }}-${{ inputs.build-type }}-${{ github.sha }}" >> $GITHUB_ENV
+        echo "CUDA_BINDINGS_ARTIFACTS_DIR=$(realpath "$REPO_DIR/cuda_bindings/dist")" >> $GITHUB_ENV
         echo "USE_CUDA=${{ (inputs.target-device == 'cpu' && 'OFF') || 'ON' }}" >> $GITHUB_ENV
         echo "UPLOAD_ENABLED=${{ (inputs.upload-enabled  == 'true' && 'ON') || 'OFF' }}" >> $GITHUB_ENV
         echo "BUILD_DATE=$(date +%Y%m%d)" >> $GITHUB_ENV
diff --git a/.github/workflows/ci-gh.yml b/.github/workflows/ci-gh.yml
index 62bb036cd..fd702f5a3 100644
--- a/.github/workflows/ci-gh.yml
+++ b/.github/workflows/ci-gh.yml
@@ -18,8 +18,8 @@ jobs:
       matrix:
         host-platform:
           - linux-x64
-          - linux-aarch64
-          - win-x64
+          #- linux-aarch64
+          #- win-x64
         target-device:
           - gpu
         build-mode:
@@ -28,10 +28,13 @@ jobs:
           - false
         python-version:
           - "3.12"
-          - "3.11"
-          - "3.10"
-          - "3.9"
-        # TODO cover the whole cuda matrix
+          #- "3.11"
+          #- "3.10"
+          #- "3.9"
+        cuda-version:
+          # Note: this is for build-time only; the test-time matrix needs to be
+          # defined separately.
+          - "12.6.2"
     uses:
       ./.github/workflows/gh-build-and-test.yml
     with:
@@ -41,4 +44,5 @@ jobs:
       build-type: ci
       upload-enabled: ${{ matrix.upload-enabled }}
       python-version: ${{ matrix.python-version }}
+      cuda-version: ${{ matrix.cuda-version }}
     secrets: inherit
diff --git a/.github/workflows/gh-build-and-test.yml b/.github/workflows/gh-build-and-test.yml
index 8d1c1b97d..a9a711d4f 100644
--- a/.github/workflows/gh-build-and-test.yml
+++ b/.github/workflows/gh-build-and-test.yml
@@ -19,6 +19,10 @@ on:
       python-version:
         type: string
         required: true
+      cuda-version:
+        type: string
+        required: true
+
 jobs:
   build:
     if: ${{ github.repository_owner == 'nvidia' }}
@@ -36,5 +40,6 @@ jobs:
       build-mode: ${{ inputs.build-mode }}
       upload-enabled: ${{ inputs.upload-enabled }}
       python-version: ${{ inputs.python-version }}
+      cuda-version: ${{ inputs.cuda-version }}
       dependencies-file: ""
     secrets: inherit
diff --git a/.github/workflows/gh-build.yml b/.github/workflows/gh-build.yml
index 494517692..7a9f03ce3 100644
--- a/.github/workflows/gh-build.yml
+++ b/.github/workflows/gh-build.yml
@@ -32,10 +32,13 @@ on:
       python-version:
         required: true
         type: string
+      cuda-version:
+        required: true
+        type: string
 
 jobs:
   build:
-    name: Build (${{ inputs.host-platform }}, ${{ inputs.target-device }}, ${{ inputs.build-type }}, build-mode=${{ inputs.build-mode }}, Python "${{ inputs.python-version }}")
+    name: Build (${{ inputs.host-platform }}, ${{ inputs.build-type }}, ${{ inputs.build-mode }}, Python "${{ inputs.python-version }}")
 
     permissions:
       id-token: write # This is required for configure-aws-credentials
@@ -59,6 +62,7 @@ jobs:
           build-mode: ${{ inputs.build-mode }}
           upload-enabled: ${{ inputs.upload-enabled }}
           python-version: ${{ inputs.python-version }}
+          cuda-version: ${{ inputs.cuda-version }}
 
       - name: Call build action
         uses: ./.github/actions/build

From 8a61031631904847f6633cfb1e38756ffe0a0f86 Mon Sep 17 00:00:00 2001
From: Leo Fang <leof@nvidia.com>
Date: Fri, 6 Dec 2024 00:11:06 +0000
Subject: [PATCH 45/65] fix parenthesis

---
 .github/actions/setup/action.yml | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/.github/actions/setup/action.yml b/.github/actions/setup/action.yml
index 49fec033a..75d746953 100644
--- a/.github/actions/setup/action.yml
+++ b/.github/actions/setup/action.yml
@@ -162,7 +162,7 @@ runs:
           REPO_DIR=$(cygpath -w $PWD)
         fi
 
-        echo "PARALLEL_LEVEL=${nproc}" >> $GITHUB_ENV
+        echo "PARALLEL_LEVEL=$(nproc)" >> $GITHUB_ENV
         echo "REPO_DIR=$REPO_DIR" >> $GITHUB_ENV
         echo "CUDA_CORE_ARTIFACT_NAME=cuda-core-python${{ inputs.python-version }}-${{ inputs.host-platform }}-${{ inputs.build-type }}-${{ github.sha }}" >> $GITHUB_ENV
         echo "CUDA_CORE_ARTIFACTS_DIR=$(realpath "$REPO_DIR/cuda_core/dist")" >> $GITHUB_ENV

From 35b88265853449ec52cf860711df68ab5094d111 Mon Sep 17 00:00:00 2001
From: Leo Fang <leof@nvidia.com>
Date: Fri, 6 Dec 2024 00:16:13 +0000
Subject: [PATCH 46/65] use abs path

---
 .github/actions/setup/action.yml | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/.github/actions/setup/action.yml b/.github/actions/setup/action.yml
index 75d746953..7f4f5e9b5 100644
--- a/.github/actions/setup/action.yml
+++ b/.github/actions/setup/action.yml
@@ -111,7 +111,7 @@ runs:
         mv nvrtc/* $CUDA_PATH/include
         rmdir nvrtc
 
-        echo "CUDA_PATH=$CUDA_PATH" >> $GITHUB_ENV
+        echo "CUDA_PATH=$(realpath $CUDA_PATH)" >> $GITHUB_ENV
 
     - name: Set environment variables
       shell: bash --noprofile --norc -xeuo pipefail {0}

From 8e42dc41e5caacfa4d17359edc15d2402dc28695 Mon Sep 17 00:00:00 2001
From: Leo Fang <leof@nvidia.com>
Date: Fri, 6 Dec 2024 00:24:30 +0000
Subject: [PATCH 47/65] defer CUDA_PATH setting

---
 .github/actions/build/action.yml | 2 ++
 .github/actions/setup/action.yml | 3 ++-
 2 files changed, 4 insertions(+), 1 deletion(-)

diff --git a/.github/actions/build/action.yml b/.github/actions/build/action.yml
index befef9b2c..7a0ae5e25 100644
--- a/.github/actions/build/action.yml
+++ b/.github/actions/build/action.yml
@@ -60,6 +60,8 @@ runs:
         CIBW_ARCHS_LINUX: "native"
         CIBW_REPAIR_WHEEL_COMMAND: ""  # no repair
         CIBW_BUILD_VERBOSITY: 1
+        CIBW_BEFORE_BUILD_LINUX: >
+          export CUDA_PATH="$(realpath ./cuda-*)"  # TODO: use cuda-version here?
         # # ensure Python.h & co can be found
         # CIBW_BEFORE_BUILD_WINDOWS: >
         #   python -c "import sysconfig; print(sysconfig.get_path('include'))" >> $env:INCLUDE
diff --git a/.github/actions/setup/action.yml b/.github/actions/setup/action.yml
index 7f4f5e9b5..484af8204 100644
--- a/.github/actions/setup/action.yml
+++ b/.github/actions/setup/action.yml
@@ -111,7 +111,8 @@ runs:
         mv nvrtc/* $CUDA_PATH/include
         rmdir nvrtc
 
-        echo "CUDA_PATH=$(realpath $CUDA_PATH)" >> $GITHUB_ENV
+        # Note: the headers will be copied into the cibuildwheel manylinux container,
+        # so setting the CUDA_PATH env var here is meaningless.
 
     - name: Set environment variables
       shell: bash --noprofile --norc -xeuo pipefail {0}

From 1588e575e4f004285c3fa5e038e8fb40f8576bdc Mon Sep 17 00:00:00 2001
From: Leo Fang <leof@nvidia.com>
Date: Fri, 6 Dec 2024 01:07:32 +0000
Subject: [PATCH 48/65] use CIBW_ENVIRONMENT to pass env var

---
 .github/actions/build/action.yml | 4 ++--
 .github/actions/setup/action.yml | 2 +-
 2 files changed, 3 insertions(+), 3 deletions(-)

diff --git a/.github/actions/build/action.yml b/.github/actions/build/action.yml
index 7a0ae5e25..fb5236a4a 100644
--- a/.github/actions/build/action.yml
+++ b/.github/actions/build/action.yml
@@ -60,8 +60,8 @@ runs:
         CIBW_ARCHS_LINUX: "native"
         CIBW_REPAIR_WHEEL_COMMAND: ""  # no repair
         CIBW_BUILD_VERBOSITY: 1
-        CIBW_BEFORE_BUILD_LINUX: >
-          export CUDA_PATH="$(realpath ./cuda-*)"  # TODO: use cuda-version here?
+        CIBW_ENVIRONMENT: >
+          CUDA_PATH="$(realpath ./cuda_toolkit)"
         # # ensure Python.h & co can be found
         # CIBW_BEFORE_BUILD_WINDOWS: >
         #   python -c "import sysconfig; print(sysconfig.get_path('include'))" >> $env:INCLUDE
diff --git a/.github/actions/setup/action.yml b/.github/actions/setup/action.yml
index 484af8204..b33bb4d62 100644
--- a/.github/actions/setup/action.yml
+++ b/.github/actions/setup/action.yml
@@ -93,7 +93,7 @@ runs:
     - name: Get CUDA headers
       shell: bash --noprofile --norc -xeuo pipefail {0}
       run: |
-        CUDA_PATH="cuda-${{ inputs.cuda-version }}"
+        CUDA_PATH="./cuda_toolkit"
         mkdir -p $CUDA_PATH/include
 
         git clone -b "cuda-${{ inputs.cuda-version }}" https://gitlab.com/nvidia/headers/cuda-individual/cudart.git

From e13854df68de7c133bc16d9835e89f1d77e72409 Mon Sep 17 00:00:00 2001
From: Leo Fang <leof@nvidia.com>
Date: Fri, 6 Dec 2024 01:43:43 +0000
Subject: [PATCH 49/65] fetch cuda-profiler-api headers

---
 .github/actions/setup/action.yml | 27 +++++++++++++++++++++++++++
 1 file changed, 27 insertions(+)

diff --git a/.github/actions/setup/action.yml b/.github/actions/setup/action.yml
index b33bb4d62..a3b6bfbab 100644
--- a/.github/actions/setup/action.yml
+++ b/.github/actions/setup/action.yml
@@ -111,6 +111,33 @@ runs:
         mv nvrtc/* $CUDA_PATH/include
         rmdir nvrtc
 
+        # cuda-profiler-api aren't updated on Gitlab...
+        # TODO: consider applying the same approach for all components? The binary archives (redist) are
+        # guaranteed to be updated as part of the release posting.
+        CTK_BASE_URL="https://developer.download.nvidia.com/compute/cuda/redist/"
+        CTK_JSON_URL="$CTK_BASE_URL/redistrib_${{ inputs.cuda-version }}.json"
+        if [[ "${{ inputs.host-platform }}" == "linux-x64" ]]; then
+          CTK_SUBDIR="linux-x86_64"
+        elif [[ "${{ inputs.host-platform }}" == "linux-aarch64" ]]; then
+          CTK_SUBDIR="linux-sbsa"
+        elif [[ "${{ inputs.host-platform }}" == "win-x64" ]]; then
+          CTK_SUBDIR="windows-x86_64"
+        else
+          exit 1
+        fi
+        CTK_COMPONENT="cuda_profiler_api"
+        CTK_COMPONENT_REL_PATH="$(curl -s $CTK_JSON_URL |
+            python -c "import sys, json; print(json.load(sys.stdin)['${CTK_COMPONENT}']['${CTK_SUBDIR}']['relative_path'])")"
+        CTK_COMPONENT_URL="${CTK_BASE_URL}/${CTK_COMPONENT_REL_PATH}"
+        wget $CTK_COMPONENT_URL
+        CTK_COMPONENT_COMPONENT_FILENAME="$(basename $CTK_COMPONENT_REL_PATH)"
+        if [[ "${{ inputs.host-platform }}" == linux* ]]; then
+          tar -xvf $CTK_COMPONENT_COMPONENT_FILENAME -C $CUDA_PATH --strip-components=1
+        else
+          exit 1  # FIXME
+        fi
+        rm $CTK_COMPONENT_COMPONENT_FILENAME
+
         # Note: the headers will be copied into the cibuildwheel manylinux container,
         # so setting the CUDA_PATH env var here is meaningless.
 

From 4f735090b4864794e62cd4788534ad7fdd609a1e Mon Sep 17 00:00:00 2001
From: Leo Fang <leof@nvidia.com>
Date: Fri, 6 Dec 2024 02:00:31 +0000
Subject: [PATCH 50/65] only rely on redist

---
 .github/actions/setup/action.yml | 54 +++++++++++++-------------------
 1 file changed, 21 insertions(+), 33 deletions(-)

diff --git a/.github/actions/setup/action.yml b/.github/actions/setup/action.yml
index a3b6bfbab..d067dabb2 100644
--- a/.github/actions/setup/action.yml
+++ b/.github/actions/setup/action.yml
@@ -90,30 +90,13 @@ runs:
       run: |
         env
 
-    - name: Get CUDA headers
+    - name: Get CUDA components
       shell: bash --noprofile --norc -xeuo pipefail {0}
       run: |
         CUDA_PATH="./cuda_toolkit"
-        mkdir -p $CUDA_PATH/include
-
-        git clone -b "cuda-${{ inputs.cuda-version }}" https://gitlab.com/nvidia/headers/cuda-individual/cudart.git
-        rm -rf cudart/.git
-        mv cudart/* $CUDA_PATH/include
-        rmdir cudart
-
-        git clone -b "cuda-${{ inputs.cuda-version }}" https://gitlab.com/nvidia/headers/cuda-individual/nvcc.git
-        rm -rf nvcc/.git
-        mv nvcc/* $CUDA_PATH/include
-        rmdir nvcc
-
-        git clone -b "cuda-${{ inputs.cuda-version }}" https://gitlab.com/nvidia/headers/cuda-individual/nvrtc.git
-        rm -rf nvrtc/.git
-        mv nvrtc/* $CUDA_PATH/include
-        rmdir nvrtc
-
-        # cuda-profiler-api aren't updated on Gitlab...
-        # TODO: consider applying the same approach for all components? The binary archives (redist) are
-        # guaranteed to be updated as part of the release posting.
+        mkdir $CUDA_PATH
+
+        # The binary archives (redist) are guaranteed to be updated as part of the release posting.
         CTK_BASE_URL="https://developer.download.nvidia.com/compute/cuda/redist/"
         CTK_JSON_URL="$CTK_BASE_URL/redistrib_${{ inputs.cuda-version }}.json"
         if [[ "${{ inputs.host-platform }}" == "linux-x64" ]]; then
@@ -125,18 +108,23 @@ runs:
         else
           exit 1
         fi
-        CTK_COMPONENT="cuda_profiler_api"
-        CTK_COMPONENT_REL_PATH="$(curl -s $CTK_JSON_URL |
-            python -c "import sys, json; print(json.load(sys.stdin)['${CTK_COMPONENT}']['${CTK_SUBDIR}']['relative_path'])")"
-        CTK_COMPONENT_URL="${CTK_BASE_URL}/${CTK_COMPONENT_REL_PATH}"
-        wget $CTK_COMPONENT_URL
-        CTK_COMPONENT_COMPONENT_FILENAME="$(basename $CTK_COMPONENT_REL_PATH)"
-        if [[ "${{ inputs.host-platform }}" == linux* ]]; then
-          tar -xvf $CTK_COMPONENT_COMPONENT_FILENAME -C $CUDA_PATH --strip-components=1
-        else
-          exit 1  # FIXME
-        fi
-        rm $CTK_COMPONENT_COMPONENT_FILENAME
+        function populate_cuda_path() {
+            # take the component name as a argument
+            CTK_COMPONENT=$1
+            CTK_COMPONENT_REL_PATH="$(curl -s $CTK_JSON_URL |
+                python -c "import sys, json; print(json.load(sys.stdin)['${CTK_COMPONENT}']['${CTK_SUBDIR}']['relative_path'])")"
+            CTK_COMPONENT_URL="${CTK_BASE_URL}/${CTK_COMPONENT_REL_PATH}"
+            wget $CTK_COMPONENT_URL
+            CTK_COMPONENT_COMPONENT_FILENAME="$(basename $CTK_COMPONENT_REL_PATH)"
+            tar -xvf $CTK_COMPONENT_COMPONENT_FILENAME -C $CUDA_PATH --strip-components=1
+            rm $CTK_COMPONENT_COMPONENT_FILENAME
+        }
+
+        # Get headers and shared libraries in place
+        populate_cuda_path cuda_nvcc
+        populate_cuda_path cuda_cudart
+        populate_cuda_path cuda_nvrtc
+        populate_cuda_path cuda_profiler_api
 
         # Note: the headers will be copied into the cibuildwheel manylinux container,
         # so setting the CUDA_PATH env var here is meaningless.

From 128ab1241caddf80765d2358111c221bcbcf3b17 Mon Sep 17 00:00:00 2001
From: Leo Fang <leof@nvidia.com>
Date: Fri, 6 Dec 2024 02:08:35 +0000
Subject: [PATCH 51/65] allow wheel repair to fix the triplet tags

---
 .github/actions/build/action.yml | 2 --
 1 file changed, 2 deletions(-)

diff --git a/.github/actions/build/action.yml b/.github/actions/build/action.yml
index fb5236a4a..c53970700 100644
--- a/.github/actions/build/action.yml
+++ b/.github/actions/build/action.yml
@@ -25,7 +25,6 @@ runs:
       env:
         CIBW_BUILD: ${{ env.CIBW_BUILD }}
         CIBW_ARCHS_LINUX: "native"
-        CIBW_REPAIR_WHEEL_COMMAND: ""  # no repair
         CIBW_BUILD_VERBOSITY: 1
         # # ensure Python.h & co can be found
         # CIBW_BEFORE_BUILD_WINDOWS: >
@@ -58,7 +57,6 @@ runs:
       env:
         CIBW_BUILD: ${{ env.CIBW_BUILD }}
         CIBW_ARCHS_LINUX: "native"
-        CIBW_REPAIR_WHEEL_COMMAND: ""  # no repair
         CIBW_BUILD_VERBOSITY: 1
         CIBW_ENVIRONMENT: >
           CUDA_PATH="$(realpath ./cuda_toolkit)"

From 6087a54def96154f39c7c3413e69d149bfa98537 Mon Sep 17 00:00:00 2001
From: Leo Fang <leof@nvidia.com>
Date: Fri, 6 Dec 2024 02:23:16 +0000
Subject: [PATCH 52/65] restore full build matrix!

---
 .github/workflows/ci-gh.yml | 10 +++++-----
 1 file changed, 5 insertions(+), 5 deletions(-)

diff --git a/.github/workflows/ci-gh.yml b/.github/workflows/ci-gh.yml
index fd702f5a3..1975c3b52 100644
--- a/.github/workflows/ci-gh.yml
+++ b/.github/workflows/ci-gh.yml
@@ -18,8 +18,8 @@ jobs:
       matrix:
         host-platform:
           - linux-x64
-          #- linux-aarch64
-          #- win-x64
+          - linux-aarch64
+          - win-x64
         target-device:
           - gpu
         build-mode:
@@ -28,9 +28,9 @@ jobs:
           - false
         python-version:
           - "3.12"
-          #- "3.11"
-          #- "3.10"
-          #- "3.9"
+          - "3.11"
+          - "3.10"
+          - "3.9"
         cuda-version:
           # Note: this is for build-time only; the test-time matrix needs to be
           # defined separately.

From b05e95dad7386e18eccef1f2e6378636ba6459e9 Mon Sep 17 00:00:00 2001
From: Leo Fang <leof@nvidia.com>
Date: Fri, 6 Dec 2024 02:45:19 +0000
Subject: [PATCH 53/65] fix wget on Windows; pass PARALLEL_LEVEL to CIBW

---
 .github/actions/build/action.yml |  1 +
 .github/actions/setup/action.yml | 12 ++++++------
 2 files changed, 7 insertions(+), 6 deletions(-)

diff --git a/.github/actions/build/action.yml b/.github/actions/build/action.yml
index c53970700..abad93c26 100644
--- a/.github/actions/build/action.yml
+++ b/.github/actions/build/action.yml
@@ -60,6 +60,7 @@ runs:
         CIBW_BUILD_VERBOSITY: 1
         CIBW_ENVIRONMENT: >
           CUDA_PATH="$(realpath ./cuda_toolkit)"
+          PARALLEL_LEVEL=${{ env.PARALLEL_LEVEL }}
         # # ensure Python.h & co can be found
         # CIBW_BEFORE_BUILD_WINDOWS: >
         #   python -c "import sysconfig; print(sysconfig.get_path('include'))" >> $env:INCLUDE
diff --git a/.github/actions/setup/action.yml b/.github/actions/setup/action.yml
index d067dabb2..656858e71 100644
--- a/.github/actions/setup/action.yml
+++ b/.github/actions/setup/action.yml
@@ -105,9 +105,11 @@ runs:
           CTK_SUBDIR="linux-sbsa"
         elif [[ "${{ inputs.host-platform }}" == "win-x64" ]]; then
           CTK_SUBDIR="windows-x86_64"
-        else
-          exit 1
+          apt update; apt install wget
         fi
+        function extractor() {
+            tar -xvf $1 -C $CUDA_PATH --strip-components=1
+        }
         function populate_cuda_path() {
             # take the component name as a argument
             CTK_COMPONENT=$1
@@ -116,7 +118,7 @@ runs:
             CTK_COMPONENT_URL="${CTK_BASE_URL}/${CTK_COMPONENT_REL_PATH}"
             wget $CTK_COMPONENT_URL
             CTK_COMPONENT_COMPONENT_FILENAME="$(basename $CTK_COMPONENT_REL_PATH)"
-            tar -xvf $CTK_COMPONENT_COMPONENT_FILENAME -C $CUDA_PATH --strip-components=1
+            extractor $CTK_COMPONENT_COMPONENT_FILENAME
             rm $CTK_COMPONENT_COMPONENT_FILENAME
         }
 
@@ -180,14 +182,12 @@ runs:
 
         echo "PARALLEL_LEVEL=$(nproc)" >> $GITHUB_ENV
         echo "REPO_DIR=$REPO_DIR" >> $GITHUB_ENV
+        echo "PKG_DIR=${PKG_DIR}" >> $GITHUB_ENV
         echo "CUDA_CORE_ARTIFACT_NAME=cuda-core-python${{ inputs.python-version }}-${{ inputs.host-platform }}-${{ inputs.build-type }}-${{ github.sha }}" >> $GITHUB_ENV
         echo "CUDA_CORE_ARTIFACTS_DIR=$(realpath "$REPO_DIR/cuda_core/dist")" >> $GITHUB_ENV
         echo "CUDA_BINDINGS_ARTIFACT_NAME=cuda-bindings-python${{ inputs.python-version }}-cuda${{ inputs.cuda-version }}-${{ inputs.host-platform }}-${{ inputs.build-type }}-${{ github.sha }}" >> $GITHUB_ENV
         echo "CUDA_BINDINGS_ARTIFACTS_DIR=$(realpath "$REPO_DIR/cuda_bindings/dist")" >> $GITHUB_ENV
-        echo "USE_CUDA=${{ (inputs.target-device == 'cpu' && 'OFF') || 'ON' }}" >> $GITHUB_ENV
         echo "UPLOAD_ENABLED=${{ (inputs.upload-enabled  == 'true' && 'ON') || 'OFF' }}" >> $GITHUB_ENV
         echo "BUILD_DATE=$(date +%Y%m%d)" >> $GITHUB_ENV
         echo "TARGET_PLATFORM=${TARGET_PLATFORM}" >> $GITHUB_ENV
-        echo "PKG_DIR=${PKG_DIR}" >> $GITHUB_ENV
-        echo "PYTHON_VERSION=${{ inputs.python-version }}" >> $GITHUB_ENV
         echo "CIBW_BUILD=${CIBW_BUILD}" >> $GITHUB_ENV

From 0c91ed8c9106701685f24917f53e510961f375bb Mon Sep 17 00:00:00 2001
From: Leo Fang <leof@nvidia.com>
Date: Fri, 6 Dec 2024 02:57:50 +0000
Subject: [PATCH 54/65] switch from wget to curl

---
 .github/actions/setup/action.yml | 28 +++++++++++++++-------------
 1 file changed, 15 insertions(+), 13 deletions(-)

diff --git a/.github/actions/setup/action.yml b/.github/actions/setup/action.yml
index 656858e71..264cd1108 100644
--- a/.github/actions/setup/action.yml
+++ b/.github/actions/setup/action.yml
@@ -105,21 +105,23 @@ runs:
           CTK_SUBDIR="linux-sbsa"
         elif [[ "${{ inputs.host-platform }}" == "win-x64" ]]; then
           CTK_SUBDIR="windows-x86_64"
-          apt update; apt install wget
         fi
-        function extractor() {
-            tar -xvf $1 -C $CUDA_PATH --strip-components=1
-        }
         function populate_cuda_path() {
-            # take the component name as a argument
-            CTK_COMPONENT=$1
-            CTK_COMPONENT_REL_PATH="$(curl -s $CTK_JSON_URL |
-                python -c "import sys, json; print(json.load(sys.stdin)['${CTK_COMPONENT}']['${CTK_SUBDIR}']['relative_path'])")"
-            CTK_COMPONENT_URL="${CTK_BASE_URL}/${CTK_COMPONENT_REL_PATH}"
-            wget $CTK_COMPONENT_URL
-            CTK_COMPONENT_COMPONENT_FILENAME="$(basename $CTK_COMPONENT_REL_PATH)"
-            extractor $CTK_COMPONENT_COMPONENT_FILENAME
-            rm $CTK_COMPONENT_COMPONENT_FILENAME
+          # take the component name as a argument
+          function download() {
+            curl -kLSs $1 -o $2
+          }
+          function extract() {
+            tar -xvf $1 -C $CUDA_PATH --strip-components=1
+          }
+          CTK_COMPONENT=$1
+          CTK_COMPONENT_REL_PATH="$(curl -s $CTK_JSON_URL |
+              python -c "import sys, json; print(json.load(sys.stdin)['${CTK_COMPONENT}']['${CTK_SUBDIR}']['relative_path'])")"
+          CTK_COMPONENT_URL="${CTK_BASE_URL}/${CTK_COMPONENT_REL_PATH}"
+          CTK_COMPONENT_COMPONENT_FILENAME="$(basename $CTK_COMPONENT_REL_PATH)"
+          download $CTK_COMPONENT_URL $CTK_COMPONENT_COMPONENT_FILENAME
+          extract $CTK_COMPONENT_COMPONENT_FILENAME
+          rm $CTK_COMPONENT_COMPONENT_FILENAME
         }
 
         # Get headers and shared libraries in place

From 33f35c336cb1720eb4cbeb27fe4676cb50456fdd Mon Sep 17 00:00:00 2001
From: Leo Fang <leof@nvidia.com>
Date: Fri, 6 Dec 2024 03:15:59 +0000
Subject: [PATCH 55/65] windows needs unzip not tar

---
 .github/actions/setup/action.yml | 22 +++++++++++++++-------
 1 file changed, 15 insertions(+), 7 deletions(-)

diff --git a/.github/actions/setup/action.yml b/.github/actions/setup/action.yml
index 264cd1108..a5a0ab296 100644
--- a/.github/actions/setup/action.yml
+++ b/.github/actions/setup/action.yml
@@ -99,21 +99,29 @@ runs:
         # The binary archives (redist) are guaranteed to be updated as part of the release posting.
         CTK_BASE_URL="https://developer.download.nvidia.com/compute/cuda/redist/"
         CTK_JSON_URL="$CTK_BASE_URL/redistrib_${{ inputs.cuda-version }}.json"
-        if [[ "${{ inputs.host-platform }}" == "linux-x64" ]]; then
-          CTK_SUBDIR="linux-x86_64"
-        elif [[ "${{ inputs.host-platform }}" == "linux-aarch64" ]]; then
-          CTK_SUBDIR="linux-sbsa"
+        if [[ "${{ inputs.host-platform }}" == linux* ]]; then
+          if [[ "${{ inputs.host-platform }}" == "linux-x64" ]]; then
+            CTK_SUBDIR="linux-x86_64"
+          elif [[ "${{ inputs.host-platform }}" == "linux-aarch64" ]]; then
+            CTK_SUBDIR="linux-sbsa"
+          fi
+          function extract() {
+            tar -xvf $1 -C $CUDA_PATH --strip-components=1
+          }
         elif [[ "${{ inputs.host-platform }}" == "win-x64" ]]; then
           CTK_SUBDIR="windows-x86_64"
+          function extract() {
+            _TEMP_DIR_=$(mktemp -d)
+            unzip $1 -d $_TEMP_DIR_
+            mv $_TEMP_DIR_/*/* $CUDA_PATH
+            rmdir $_TEMP_DIR_/* $_TEMP_DIR_
+          }
         fi
         function populate_cuda_path() {
           # take the component name as a argument
           function download() {
             curl -kLSs $1 -o $2
           }
-          function extract() {
-            tar -xvf $1 -C $CUDA_PATH --strip-components=1
-          }
           CTK_COMPONENT=$1
           CTK_COMPONENT_REL_PATH="$(curl -s $CTK_JSON_URL |
               python -c "import sys, json; print(json.load(sys.stdin)['${CTK_COMPONENT}']['${CTK_SUBDIR}']['relative_path'])")"

From 4b3341a80bcf9b1145e69bdcd20084b4abf05cdb Mon Sep 17 00:00:00 2001
From: Leo Fang <leof@nvidia.com>
Date: Fri, 6 Dec 2024 03:23:09 +0000
Subject: [PATCH 56/65] mv -> rsync

---
 .github/actions/setup/action.yml | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/.github/actions/setup/action.yml b/.github/actions/setup/action.yml
index a5a0ab296..0ad9cadf6 100644
--- a/.github/actions/setup/action.yml
+++ b/.github/actions/setup/action.yml
@@ -113,7 +113,7 @@ runs:
           function extract() {
             _TEMP_DIR_=$(mktemp -d)
             unzip $1 -d $_TEMP_DIR_
-            mv $_TEMP_DIR_/*/* $CUDA_PATH
+            rsync -avh --remove-source-files $_TEMP_DIR_/*/* $CUDA_PATH
             rmdir $_TEMP_DIR_/* $_TEMP_DIR_
           }
         fi

From 3503dae21a8b524fe754acd5d252b9025f975bc2 Mon Sep 17 00:00:00 2001
From: Leo Fang <leof@nvidia.com>
Date: Fri, 6 Dec 2024 03:32:35 +0000
Subject: [PATCH 57/65] git bash has no wget or rsync...

---
 .github/actions/setup/action.yml | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/.github/actions/setup/action.yml b/.github/actions/setup/action.yml
index 0ad9cadf6..c61c87da5 100644
--- a/.github/actions/setup/action.yml
+++ b/.github/actions/setup/action.yml
@@ -113,8 +113,8 @@ runs:
           function extract() {
             _TEMP_DIR_=$(mktemp -d)
             unzip $1 -d $_TEMP_DIR_
-            rsync -avh --remove-source-files $_TEMP_DIR_/*/* $CUDA_PATH
-            rmdir $_TEMP_DIR_/* $_TEMP_DIR_
+            cp -r $_TEMP_DIR_/* $CUDA_PATH
+            rm -r $_TEMP_DIR_
           }
         fi
         function populate_cuda_path() {

From dc12840f8243c2621d9910508444a3e8230aca39 Mon Sep 17 00:00:00 2001
From: Leo Fang <leof@nvidia.com>
Date: Fri, 6 Dec 2024 03:39:39 +0000
Subject: [PATCH 58/65] ensure win-style path on win

---
 .github/actions/build/action.yml | 5 ++++-
 .github/actions/setup/action.yml | 2 +-
 2 files changed, 5 insertions(+), 2 deletions(-)

diff --git a/.github/actions/build/action.yml b/.github/actions/build/action.yml
index abad93c26..91c47ca0e 100644
--- a/.github/actions/build/action.yml
+++ b/.github/actions/build/action.yml
@@ -58,9 +58,12 @@ runs:
         CIBW_BUILD: ${{ env.CIBW_BUILD }}
         CIBW_ARCHS_LINUX: "native"
         CIBW_BUILD_VERBOSITY: 1
-        CIBW_ENVIRONMENT: >
+        CIBW_ENVIRONMENT_LINUX: >
           CUDA_PATH="$(realpath ./cuda_toolkit)"
           PARALLEL_LEVEL=${{ env.PARALLEL_LEVEL }}
+        CIBW_ENVIRONMENT_WINDOWS: >
+          CUDA_HOME="$(cygpath -w ./cuda_toolkit)"
+          PARALLEL_LEVEL=${{ env.PARALLEL_LEVEL }}
         # # ensure Python.h & co can be found
         # CIBW_BEFORE_BUILD_WINDOWS: >
         #   python -c "import sysconfig; print(sysconfig.get_path('include'))" >> $env:INCLUDE
diff --git a/.github/actions/setup/action.yml b/.github/actions/setup/action.yml
index c61c87da5..6c21e221f 100644
--- a/.github/actions/setup/action.yml
+++ b/.github/actions/setup/action.yml
@@ -114,7 +114,7 @@ runs:
             _TEMP_DIR_=$(mktemp -d)
             unzip $1 -d $_TEMP_DIR_
             cp -r $_TEMP_DIR_/* $CUDA_PATH
-            rm -r $_TEMP_DIR_
+            rm -rf $_TEMP_DIR_
           }
         fi
         function populate_cuda_path() {

From c55bdc55ad610fd16c4387d70318bca3c2a7e65f Mon Sep 17 00:00:00 2001
From: Leo Fang <leof@nvidia.com>
Date: Fri, 6 Dec 2024 03:43:56 +0000
Subject: [PATCH 59/65] convert abs path

---
 .github/actions/build/action.yml | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/.github/actions/build/action.yml b/.github/actions/build/action.yml
index 91c47ca0e..fc9ff6014 100644
--- a/.github/actions/build/action.yml
+++ b/.github/actions/build/action.yml
@@ -62,7 +62,7 @@ runs:
           CUDA_PATH="$(realpath ./cuda_toolkit)"
           PARALLEL_LEVEL=${{ env.PARALLEL_LEVEL }}
         CIBW_ENVIRONMENT_WINDOWS: >
-          CUDA_HOME="$(cygpath -w ./cuda_toolkit)"
+          CUDA_HOME="$(cygpath -w $(realpath ./cuda_toolkit))"
           PARALLEL_LEVEL=${{ env.PARALLEL_LEVEL }}
         # # ensure Python.h & co can be found
         # CIBW_BEFORE_BUILD_WINDOWS: >

From bdb35999c1e7b2fd09e7bf9337ba1a9c2fab9aed Mon Sep 17 00:00:00 2001
From: Leo Fang <leof@nvidia.com>
Date: Fri, 6 Dec 2024 03:54:22 +0000
Subject: [PATCH 60/65] debug

---
 .github/actions/setup/action.yml | 52 +++++++++++++++-----------------
 1 file changed, 25 insertions(+), 27 deletions(-)

diff --git a/.github/actions/setup/action.yml b/.github/actions/setup/action.yml
index 6c21e221f..67b51a2d2 100644
--- a/.github/actions/setup/action.yml
+++ b/.github/actions/setup/action.yml
@@ -138,6 +138,8 @@ runs:
         populate_cuda_path cuda_nvrtc
         populate_cuda_path cuda_profiler_api
 
+        ls -l $CUDA_PATH
+        tree $CUDA_PATH
         # Note: the headers will be copied into the cibuildwheel manylinux container,
         # so setting the CUDA_PATH env var here is meaningless.
 
@@ -151,28 +153,31 @@ runs:
           TARGET_PLATFORM='linux-aarch64'
         elif [[ "${{ inputs.host-platform }}" == "win-x64" ]]; then
           TARGET_PLATFORM='win-64'
-        else
-          exit 1
         fi
 
-        if [[ "${{ inputs.python-version }}" == "3.12" && "${{ inputs.host-platform }}" == linux* ]]; then
-          CIBW_BUILD='cp312-manylinux*'
-        elif [[ "${{ inputs.python-version }}" == "3.11" && "${{ inputs.host-platform }}" == linux* ]]; then
-          CIBW_BUILD='cp311-manylinux*'
-        elif [[ "${{ inputs.python-version }}" == "3.10" && "${{ inputs.host-platform }}" == linux* ]]; then
-          CIBW_BUILD='cp310-manylinux*'
-        elif [[ "${{ inputs.python-version }}" == "3.9" && "${{ inputs.host-platform }}" == linux* ]]; then
-          CIBW_BUILD='cp39-manylinux*'
-        elif [[ "${{ inputs.python-version }}" == "3.12" && "${{ inputs.host-platform }}" == win* ]]; then
-          CIBW_BUILD='cp312-win_amd64'
-        elif [[ "${{ inputs.python-version }}" == "3.11" && "${{ inputs.host-platform }}" == win* ]]; then
-          CIBW_BUILD='cp311-win_amd64'
-        elif [[ "${{ inputs.python-version }}" == "3.10" && "${{ inputs.host-platform }}" == win* ]]; then
-          CIBW_BUILD='cp310-win_amd64'
-        elif [[ "${{ inputs.python-version }}" == "3.9" && "${{ inputs.host-platform }}" == win* ]]; then
-          CIBW_BUILD='cp39-win_amd64'
-        else
-          exit 1
+        if [[ "${{ inputs.host-platform }}" == linux* ]]; then
+          if [[ "${{ inputs.python-version }}" == "3.12" ]]; then
+            CIBW_BUILD='cp312-manylinux*'
+          elif [[ "${{ inputs.python-version }}" == "3.11" ]]; then
+            CIBW_BUILD='cp311-manylinux*'
+          elif [[ "${{ inputs.python-version }}" == "3.10" ]]; then
+            CIBW_BUILD='cp310-manylinux*'
+          elif [[ "${{ inputs.python-version }}" == "3.9" ]]; then
+            CIBW_BUILD='cp39-manylinux*'
+          fi
+          REPO_DIR=$(pwd)
+        elif [[ "${{ inputs.host-platform }}" == win* ]]; then
+          if [[ "${{ inputs.python-version }}" == "3.12" ]]; then
+            CIBW_BUILD='cp312-win_amd64'
+          elif [[ "${{ inputs.python-version }}" == "3.11" ]]; then
+            CIBW_BUILD='cp311-win_amd64'
+          elif [[ "${{ inputs.python-version }}" == "3.10" ]]; then
+            CIBW_BUILD='cp310-win_amd64'
+          elif [[ "${{ inputs.python-version }}" == "3.9" ]]; then
+           CIBW_BUILD='cp39-win_amd64'
+          fi
+          PWD=$(pwd)
+          REPO_DIR=$(cygpath -w $PWD)
         fi
 
         BUILD_MODE="${{ inputs.build-mode }}"
@@ -183,13 +188,6 @@ runs:
           PKG_DIR="${BUILD_MODE}/${TARGET_PLATFORM}"
         fi
 
-        if [[ "${{ inputs.host-platform }}" == linux* ]]; then
-          REPO_DIR=$(pwd)
-        else
-          PWD=$(pwd)
-          REPO_DIR=$(cygpath -w $PWD)
-        fi
-
         echo "PARALLEL_LEVEL=$(nproc)" >> $GITHUB_ENV
         echo "REPO_DIR=$REPO_DIR" >> $GITHUB_ENV
         echo "PKG_DIR=${PKG_DIR}" >> $GITHUB_ENV

From c8fd323954f94bd8783ee82c375b3e34da1c61b5 Mon Sep 17 00:00:00 2001
From: Leo Fang <leof@nvidia.com>
Date: Fri, 6 Dec 2024 03:57:29 +0000
Subject: [PATCH 61/65] another level down

---
 .github/actions/setup/action.yml | 5 ++---
 1 file changed, 2 insertions(+), 3 deletions(-)

diff --git a/.github/actions/setup/action.yml b/.github/actions/setup/action.yml
index 67b51a2d2..db525c75f 100644
--- a/.github/actions/setup/action.yml
+++ b/.github/actions/setup/action.yml
@@ -113,7 +113,7 @@ runs:
           function extract() {
             _TEMP_DIR_=$(mktemp -d)
             unzip $1 -d $_TEMP_DIR_
-            cp -r $_TEMP_DIR_/* $CUDA_PATH
+            cp -r $_TEMP_DIR_/*/* $CUDA_PATH
             rm -rf $_TEMP_DIR_
           }
         fi
@@ -137,9 +137,8 @@ runs:
         populate_cuda_path cuda_cudart
         populate_cuda_path cuda_nvrtc
         populate_cuda_path cuda_profiler_api
-
         ls -l $CUDA_PATH
-        tree $CUDA_PATH
+
         # Note: the headers will be copied into the cibuildwheel manylinux container,
         # so setting the CUDA_PATH env var here is meaningless.
 

From 0d554f0945b1bd1b7797170ece2a13765ab6dbe8 Mon Sep 17 00:00:00 2001
From: Leo Fang <leof@nvidia.com>
Date: Fri, 6 Dec 2024 04:23:29 +0000
Subject: [PATCH 62/65] check if it is a race condition on win

---
 .github/actions/build/action.yml | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/.github/actions/build/action.yml b/.github/actions/build/action.yml
index fc9ff6014..e1552ae83 100644
--- a/.github/actions/build/action.yml
+++ b/.github/actions/build/action.yml
@@ -63,7 +63,7 @@ runs:
           PARALLEL_LEVEL=${{ env.PARALLEL_LEVEL }}
         CIBW_ENVIRONMENT_WINDOWS: >
           CUDA_HOME="$(cygpath -w $(realpath ./cuda_toolkit))"
-          PARALLEL_LEVEL=${{ env.PARALLEL_LEVEL }}
+        #  PARALLEL_LEVEL=${{ env.PARALLEL_LEVEL }}
         # # ensure Python.h & co can be found
         # CIBW_BEFORE_BUILD_WINDOWS: >
         #   python -c "import sysconfig; print(sysconfig.get_path('include'))" >> $env:INCLUDE

From 19445d58143f0dc58193a39080c5808a4e52d220 Mon Sep 17 00:00:00 2001
From: Leo Fang <leof@nvidia.com>
Date: Fri, 6 Dec 2024 15:41:32 -0500
Subject: [PATCH 63/65] clean up unused (but still relevant) code

---
 .github/actions/setup/action.yml | 43 --------------------------------
 1 file changed, 43 deletions(-)

diff --git a/.github/actions/setup/action.yml b/.github/actions/setup/action.yml
index db525c75f..63c120877 100644
--- a/.github/actions/setup/action.yml
+++ b/.github/actions/setup/action.yml
@@ -38,52 +38,9 @@ runs:
       with:
         python-version: "3.12"
 
-    # - name: Set up new PowerShell (pwsh)
-    #   if: ${{ startsWith(inputs.host-platform, 'win') }}
-    #   shell: powershell
-    #   env:
-    #     # doesn't seem there's an easy way to avoid hard-coding it?
-    #     PS_MSI_URL: https://github.com/PowerShell/PowerShell/releases/download/v7.4.6/PowerShell-7.4.6-win-x64.msi
-    #   run: |
-    #     Invoke-WebRequest -Uri "$env:PS_MSI_URL" -OutFile "ps_installer.msi"
-    #     Start-Process msiexec.exe -Wait -Verbose -ArgumentList '/i "ps_installer.msi" /qn REGISTER_MANIFEST=1'
-    #     echo "$env:ProgramFiles\\PowerShell\\7" >> $env:GITHUB_PATH
-
-    # - name: Install MSVC
-    #   if: ${{ startsWith(inputs.host-platform, 'win') }}
-    #   shell: pwsh
-    #   run: |
-    #     # Download & install
-    #     .\\continuous_integration\\scripts\\install-vs.ps1 -msvcVersion 15 -clVersion 14.16
-
-    #     # Verify
-    #     # Note: this assumes vs2017, e.g. see _find_vc2017():
-    #     # https://github.com/pypa/setuptools/blob/9692cde009af4651819d18a1e839d3b6e3fcd77d/setuptools/_distutils/_msvccompiler.py#L67
-    #     $vsPath = & "${env:ProgramFiles(x86)}\\Microsoft Visual Studio\\Installer\\vswhere.exe" `
-    #               -latest `
-    #               -products * `
-    #               -requires Microsoft.VisualStudio.Component.VC.Tools.x86.x64 `
-    #               -property installationPath
-    #     $clPath = Join-Path $vsPath "VC\\Tools\\MSVC\\*\\bin\\Hostx64\\x64\\cl.exe"
-    #     $clPath = (Get-ChildItem $clPath).FullName
-    #     echo "found cl.exe: $clPath"
-    #     # # For some reason below just doesn't work in the CI...
-    #     # Start-Process -NoNewWindow -RedirectStandardError cl.out -FilePath "$clPath"
-    #     # $CL_VERSION_STRING = & type cl.out
-    #     # if (($CL_VERSION_STRING -join " ") -match "Version (\\d+\\.\\d+)\\.\\d+") {
-    #     #     $CL_VERSION = $matches[1]
-    #     #     echo "Detected cl.exe version: $CL_VERSION"
-    #     # }
-
-    #     # Delete temporary items created by installers and such
-    #     Remove-Item -Recurse "${env:TEMP}\\*"
-
     - name: Set up MSVC
       if: ${{ startsWith(inputs.host-platform, 'win') }}
       uses: ilammy/msvc-dev-cmd@v1
-    #  with:
-    #    toolset: 14.16
-    #    vsversion: 2017
 
     - name: Dump environment
       shell: bash --noprofile --norc -xeuo pipefail {0}

From 15626a8e781adf8417754556fc6fba6fd7511fa9 Mon Sep 17 00:00:00 2001
From: Leo Fang <leof@nvidia.com>
Date: Fri, 6 Dec 2024 15:41:51 -0500
Subject: [PATCH 64/65] clean up unused (but still relevant) code - cont'd

---
 continuous_integration/scripts/install-vs.ps1 | 51 -------------------
 1 file changed, 51 deletions(-)
 delete mode 100644 continuous_integration/scripts/install-vs.ps1

diff --git a/continuous_integration/scripts/install-vs.ps1 b/continuous_integration/scripts/install-vs.ps1
deleted file mode 100644
index 3b10a6472..000000000
--- a/continuous_integration/scripts/install-vs.ps1
+++ /dev/null
@@ -1,51 +0,0 @@
-# Modified from https://github.com/rapidsai/devcontainers/blob/branch-25.02/windows/image/installers/install-vs.ps1
-
-Param(
-    [Parameter(Mandatory=$true)]
-    [ValidateSet('15', '16', '17')]
-    [string]
-    $msvcVersion,
-    [Parameter(Mandatory=$false)]
-    [string]
-    $clVersion
-)
-
-$vsComponentsMap = @{
-    "15"     = "Microsoft.VisualStudio.Component.Windows10SDK.17763"
-    "16"     = "Microsoft.VisualStudio.Component.Windows10SDK.20348"
-    "17"     = "Microsoft.VisualStudio.Component.Windows10SDK.20348"
-    "14.14"  = "Microsoft.VisualStudio.Component.VC.Tools.14.14"
-    "14.15"  = "Microsoft.VisualStudio.Component.VC.Tools.14.15"
-    "14.16"  = "Microsoft.VisualStudio.Component.VC.Tools.14.16"
-    "14.27"  = "Microsoft.VisualStudio.Component.VC.14.27.x86.x64"
-    "14.28"  = "Microsoft.VisualStudio.Component.VC.14.28.x86.x64"
-    "14.29"  = "Microsoft.VisualStudio.Component.VC.14.29.x86.x64"
-    "14.34"  = "Microsoft.VisualStudio.Component.VC.14.34.17.4.x86.x64"
-    "14.35"  = "Microsoft.VisualStudio.Component.VC.14.35.17.5.x86.x64"
-    "14.36"  = "Microsoft.VisualStudio.Component.VC.14.36.17.6.x86.x64"
-    "14.37"  = "Microsoft.VisualStudio.Component.VC.14.37.17.7.x86.x64"
-    "14.38"  = "Microsoft.VisualStudio.Component.VC.14.38.17.8.x86.x64"
-    "14.39"  = "Microsoft.VisualStudio.Component.VC.14.39.17.9.x86.x64"
-    "latest" = "Microsoft.VisualStudio.Component.VC.Tools.x86.x64"
-}
-
-# Always install/update core VC tools
-$vsComponent = $vsComponentsMap[$msvcVersion]
-$vsComponentString = "--add Microsoft.VisualStudio.Component.VC.Tools.x86.x64 --add $vsComponent"
-
-if ($clVersion) {
-    $clComponent = $vsComponentsMap[$clVersion]
-    $vsComponentString = "$vsComponentString --add $clComponent"
-}
-
-# 14.16 alone does not install build tools
-if ($clVersion -eq "14.16") {
-    $clComponent = $vsComponentsMap["14.15"]
-    $vsComponentString = "$vsComponentString --add $clComponent"
-}
-
-Invoke-WebRequest -Uri "https://aka.ms/vs/$msvcVersion/release/vs_buildtools.exe" -UseBasicParsing -OutFile .\vs_buildtools.exe
-Write-Output "Installing components: $vsComponentString"
-Start-Process -NoNewWindow -PassThru -Wait -FilePath .\vs_buildtools.exe -ArgumentList "install --installWhileDownloading --wait --norestart --nocache --quiet $vsComponentString"
-
-Remove-Item .\vs_buildtools.exe

From 07b8d1d97291dbff3cf1cc6f2cde065d48ab788f Mon Sep 17 00:00:00 2001
From: Leo Fang <leof@nvidia.com>
Date: Fri, 6 Dec 2024 21:24:22 +0000
Subject: [PATCH 65/65] consolidate with PYTHON_VERSION_FORMATTED

---
 .github/actions/setup/action.yml | 25 +++++--------------------
 1 file changed, 5 insertions(+), 20 deletions(-)

diff --git a/.github/actions/setup/action.yml b/.github/actions/setup/action.yml
index 63c120877..e00cf27ff 100644
--- a/.github/actions/setup/action.yml
+++ b/.github/actions/setup/action.yml
@@ -111,27 +111,12 @@ runs:
           TARGET_PLATFORM='win-64'
         fi
 
+        PYTHON_VERSION_FORMATTED=$(echo '${{ inputs.python-version }}' | tr -d '.')
         if [[ "${{ inputs.host-platform }}" == linux* ]]; then
-          if [[ "${{ inputs.python-version }}" == "3.12" ]]; then
-            CIBW_BUILD='cp312-manylinux*'
-          elif [[ "${{ inputs.python-version }}" == "3.11" ]]; then
-            CIBW_BUILD='cp311-manylinux*'
-          elif [[ "${{ inputs.python-version }}" == "3.10" ]]; then
-            CIBW_BUILD='cp310-manylinux*'
-          elif [[ "${{ inputs.python-version }}" == "3.9" ]]; then
-            CIBW_BUILD='cp39-manylinux*'
-          fi
+          CIBW_BUILD="cp${PYTHON_VERSION_FORMATTED}-manylinux*"
           REPO_DIR=$(pwd)
         elif [[ "${{ inputs.host-platform }}" == win* ]]; then
-          if [[ "${{ inputs.python-version }}" == "3.12" ]]; then
-            CIBW_BUILD='cp312-win_amd64'
-          elif [[ "${{ inputs.python-version }}" == "3.11" ]]; then
-            CIBW_BUILD='cp311-win_amd64'
-          elif [[ "${{ inputs.python-version }}" == "3.10" ]]; then
-            CIBW_BUILD='cp310-win_amd64'
-          elif [[ "${{ inputs.python-version }}" == "3.9" ]]; then
-           CIBW_BUILD='cp39-win_amd64'
-          fi
+          CIBW_BUILD="cp${PYTHON_VERSION_FORMATTED}-win_amd64"
           PWD=$(pwd)
           REPO_DIR=$(cygpath -w $PWD)
         fi
@@ -147,9 +132,9 @@ runs:
         echo "PARALLEL_LEVEL=$(nproc)" >> $GITHUB_ENV
         echo "REPO_DIR=$REPO_DIR" >> $GITHUB_ENV
         echo "PKG_DIR=${PKG_DIR}" >> $GITHUB_ENV
-        echo "CUDA_CORE_ARTIFACT_NAME=cuda-core-python${{ inputs.python-version }}-${{ inputs.host-platform }}-${{ inputs.build-type }}-${{ github.sha }}" >> $GITHUB_ENV
+        echo "CUDA_CORE_ARTIFACT_NAME=cuda-core-python${PYTHON_VERSION_FORMATTED}-${{ inputs.host-platform }}-${{ inputs.build-type }}-${{ github.sha }}" >> $GITHUB_ENV
         echo "CUDA_CORE_ARTIFACTS_DIR=$(realpath "$REPO_DIR/cuda_core/dist")" >> $GITHUB_ENV
-        echo "CUDA_BINDINGS_ARTIFACT_NAME=cuda-bindings-python${{ inputs.python-version }}-cuda${{ inputs.cuda-version }}-${{ inputs.host-platform }}-${{ inputs.build-type }}-${{ github.sha }}" >> $GITHUB_ENV
+        echo "CUDA_BINDINGS_ARTIFACT_NAME=cuda-bindings-python${PYTHON_VERSION_FORMATTED}-cuda${{ inputs.cuda-version }}-${{ inputs.host-platform }}-${{ inputs.build-type }}-${{ github.sha }}" >> $GITHUB_ENV
         echo "CUDA_BINDINGS_ARTIFACTS_DIR=$(realpath "$REPO_DIR/cuda_bindings/dist")" >> $GITHUB_ENV
         echo "UPLOAD_ENABLED=${{ (inputs.upload-enabled  == 'true' && 'ON') || 'OFF' }}" >> $GITHUB_ENV
         echo "BUILD_DATE=$(date +%Y%m%d)" >> $GITHUB_ENV