Merge remote-tracking branch 'rapidsai/main' into ltoir-reqs

.github/workflows/pr.yaml

@@ -0,0 +1,120 @@
+# Copyright (c) 2023, NVIDIA CORPORATION.
+name: pr
+
+on:
+ push:
+ branches:
+ - "pull-request/[0-9]+"
+
+concurrency:
+ group: ${{ github.workflow }}-${{ github.ref }}
+ cancel-in-progress: true
+
+jobs:
+ pr-builder:
+ needs:
+ - checks
+ - wheel-build-amd64
+ - wheel-build-arm64
+ - wheel-test-amd64
+ - wheel-test-arm64
+ secrets: inherit
+ uses: rapidsai/shared-action-workflows/.github/workflows/pr-builder.yaml@branch-23.12
+ checks:
+ secrets: inherit
+ uses: rapidsai/shared-action-workflows/.github/workflows/checks.yaml@branch-23.12
+ with:
+ enable_check_generated_files: false
+ wheel-build-amd64:
+ runs-on: linux-amd64-cpu4
+ strategy:
+ matrix:
+ python-version: ["3.9", "3.10"]
+ container:
+ image: "rapidsai/ci-wheel:cuda12.0.1-centos7-py${{ matrix.python-version }}"
+ steps:
+ - name: Checkout code
+ uses: actions/checkout@v3
+ with:
+ fetch-depth: 0
+ - name: Get PR Info
+ id: get-pr-info
+ uses: rapidsai/shared-actions/get-pr-info@main
+ - name: Run build_wheel.sh
+ run: ci/build_wheel.sh
+ - name: Upload Python Wheel
+ uses: actions/upload-artifact@v2
+ with:
+ path: ./wheel-build
+ wheel-build-arm64:
+ runs-on: linux-arm64-cpu4
+ strategy:
+ matrix:
+ python-version: ["3.9", "3.10"]
+ container:
+ image: "rapidsai/ci-wheel:cuda12.0.1-rockylinux8-py${{ matrix.python-version }}"
+ steps:
+ - name: Checkout code
+ uses: actions/checkout@v3
+ with:
+ fetch-depth: 0
+ - name: Get PR Info
+ id: get-pr-info
+ uses: rapidsai/shared-actions/get-pr-info@main
+ - name: Run build_wheel.sh
+ run: ci/build_wheel.sh
+ - name: Upload Python Wheel
+ uses: actions/upload-artifact@v2
+ with:
+ path: ./wheel-build
+ wheel-test-amd64:
+ needs:
+ - wheel-build-amd64
+ runs-on: linux-amd64-gpu-v100-latest-1
+ strategy:
+ matrix:
+ python-version: ["3.9", "3.10"]
+ linux-version: ["ubuntu20.04", "ubuntu18.04"]
+ container:
+ image: "rapidsai/citestwheel:cuda12.0.1-${{ matrix.linux-version }}-py${{ matrix.python-version }}"
+ env:
+ NVIDIA_VISIBLE_DEVICES: ${{ env.NVIDIA_VISIBLE_DEVICES }}
+ steps:
+ - name: Checkout code
+ uses: actions/checkout@v3
+ with:
+ fetch-depth: 0
+ - name: Get PR Info
+ id: get-pr-info
+ uses: rapidsai/shared-actions/get-pr-info@main
+ - name: Download Python Wheel
+ uses: actions/download-artifact@v2
+ with:
+ path: ./
+ - name: Run test_wheel.sh
+ run: ci/test_wheel.sh 
+ wheel-test-arm64:
+ needs:
+ - wheel-build-arm64
+ runs-on: linux-arm64-gpu-a100-latest-1
+ strategy:
+ matrix:
+ python-version: ["3.9", "3.10"]
+ container:
+ image: "rapidsai/citestwheel:cuda12.0.1-ubuntu20.04-py${{ matrix.python-version }}"
+ env:
+ NVIDIA_VISIBLE_DEVICES: ${{ env.NVIDIA_VISIBLE_DEVICES }}
+ steps:
+ - name: Checkout code
+ uses: actions/checkout@v3
+ with:
+ fetch-depth: 0
+ - name: Get PR Info
+ id: get-pr-info
+ uses: rapidsai/shared-actions/get-pr-info@main
+ - name: Download Python Wheel
+ uses: actions/download-artifact@v2
+ with:
+ path: ./
+ - name: run test_wheel.sh
+ run: ci/test_wheel.sh 

.gitignore

@@ -3,3 +3,7 @@ build
 *.so
 *.egg-info
 .*.swp
+
+# compiled test files
+pynvjitlink/tests/test_device_functions.*
+pynvjitlink/tests/undefined_extern.cubin

CMakeLists.txt

@@ -3,7 +3,7 @@ cmake_minimum_required(VERSION 3.26.4 FATAL_ERROR)
 project(
  pynvjitlink
  VERSION 0.1.0
- LANGUAGES CXX
+ LANGUAGES CXX CUDA
 )
 
 find_package(Python COMPONENTS Interpreter Development REQUIRED)
@@ -17,4 +17,6 @@ target_link_libraries(_nvjitlinklib PRIVATE CUDA::nvJitLink_static CUDA::nvptxco
 
 target_compile_options(_nvjitlinklib PRIVATE -Werror -Wall)
 
+target_compile_features(_nvjitlinklib PRIVATE cxx_std_11)
+
 install(TARGETS _nvjitlinklib LIBRARY DESTINATION pynvjitlink)

ci/build_wheel.sh

@@ -0,0 +1,17 @@
+#!/bin/bash
+# Copyright (c) 2023, NVIDIA CORPORATION
+
+
+rapids-logger "Install CUDA Toolkit"
+yum update
+yum install -y epel-release
+yum install -y nvidia-driver-latest-dkms
+yum install -y cuda-toolkit-12-3
+
+rapids-logger "Check conda environment"
+conda list
+
+rapids-logger "Build wheel"
+export SCCACHE_S3_NO_CREDENTIALS=1
+mkdir -p ./wheel-build
+pip wheel . --wheel-dir=./wheel-build -vvv

ci/check_style.sh

@@ -0,0 +1,9 @@
+#!/bin/bash
+# Copyright (c) 2023, NVIDIA CORPORATION
+
+set -euo pipefail
+
+pip install pre-commit
+
+# Run pre-commit checks
+pre-commit run --all-files --show-diff-on-failure

ci/test_wheel.sh

@@ -0,0 +1,18 @@
+#!/bin/bash
+# Copyright (c) 2023, NVIDIA CORPORATION
+
+set -e
+
+rapids-logger "Search artifact directory"
+ls ./artifact
+
+rapids-logger "Install wheel"
+pip install --find-links ./artifact pynvjitlink
+
+rapids-logger "Build Tests"
+cd test_binary_generation && make
+
+rapids-logger "Run Tests"
+cd ..
+pip install pytest
+pytest pynvjitlink/tests

pynvjitlink/VERSION

@@ -0,0 +1 @@
+0.1.0

pynvjitlink/tests/test_pynvjitlink.py

@@ -48,28 +48,27 @@ def device_functions_ptx():
 def undefined_extern_cubin():
  return read_test_file('undefined_extern.cubin')
 
-@pytest.mark.skip
+
 def test_create_no_arch_error():
  # nvjitlink expects at least the architecture to be specified.
  with pytest.raises(RuntimeError,
  match='NVJITLINK_ERROR_MISSING_ARCH error'):
  _nvjitlinklib.create()
 
 
-@pytest.mark.skip('Causes fatal error and exit(1)')
 def test_invalid_arch_error():
  # sm_XX is not a valid architecture
  with pytest.raises(RuntimeError,
  match='NVJITLINK_ERROR_UNRECOGNIZED_OPTION error'):
  _nvjitlinklib.create('-arch=sm_XX')
 
-@pytest.mark.skip
+
 def test_unrecognized_option_error():
  with pytest.raises(RuntimeError,
  match='NVJITLINK_ERROR_UNRECOGNIZED_OPTION error'):
  _nvjitlinklib.create('-fictitious_option')
 
-@pytest.mark.skip
+
 def test_invalid_option_type_error():
  with pytest.raises(TypeError,
  match='Expecting only strings'):
@@ -97,9 +96,7 @@ def test_complete_empty():
  marks=pytest.mark.xfail),
  ('device_functions_ptx', InputType.PTX),
  ('device_functions_object', InputType.OBJECT),
- # XXX: Archive type needs debugging - results in a segfault.
- pytest.param('device_functions_archive', InputType.LIBRARY,
- marks=pytest.mark.skip),
+ ('device_functions_archive', InputType.LIBRARY),
 ])
 def test_add_file(input_file, input_type, request):
  filename, data = request.getfixturevalue(input_file)
@@ -108,7 +105,7 @@ def test_add_file(input_file, input_type, request):
  _nvjitlinklib.add_data(handle, input_type.value, data, filename)
  _nvjitlinklib.destroy(handle)
 
-@pytest.mark.skip
+
 def test_get_error_log(undefined_extern_cubin):
  handle = _nvjitlinklib.create('-arch=sm_75')
  filename, data = undefined_extern_cubin
@@ -118,11 +115,8 @@ def test_get_error_log(undefined_extern_cubin):
  _nvjitlinklib.complete(handle)
  error_log = _nvjitlinklib.get_error_log(handle)
  _nvjitlinklib.destroy(handle)
- # XXX: The error message in the log is strange. The actual expected error
- # message appears on the terminal:
- # error : Undefined reference to '_Z5undefff' in
- # 'undefined_extern.cubin'
- assert "ERROR 9: finish" in error_log
+ assert "Undefined reference to '_Z5undefff' " \
+ "in 'undefined_extern.cubin'" in error_log
 
 
 def test_get_info_log(device_functions_cubin):
@@ -188,7 +182,7 @@ def test_get_linked_ptx_from_lto(device_functions_ltoir):
  handle = _nvjitlinklib.create('-arch=sm_75', '-lto', '-ptx')
  _nvjitlinklib.add_data(handle, input_type, data, filename)
  _nvjitlinklib.complete(handle)
- ptx = _nvjitlinklib.get_linked_ptx(handle)
+ _nvjitlinklib.get_linked_ptx(handle)
  _nvjitlinklib.destroy(handle)