Skip to content

Build the no-PerfWorks configuration in CI so link errors surface on … #152

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
CodingInAVan merged 1 commit into from
Aug 8, 2026
Merged

Build the no-PerfWorks configuration in CI so link errors surface on … #152

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
12 changes: 10 additions & 2 deletions .github/workflows/build.yml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -28,20 +28,28 @@ on:

jobs:
build:
name: Build on ${{ matrix.os }}
name: Build on ${{ matrix.os }} py${{ matrix.python-version }} perfworks=${{ matrix.perfworks }}
runs-on: ${{ matrix.os }}
strategy:
matrix:
# Pin windows-2022 (VS 2022): windows-latest moved to VS 2026 (Jun
# 2026), whose MSVC toolset CUDA 13.x nvcc doesn't support.
os: [ubuntu-22.04, windows-2022]
python-version: ["3.12", "3.13"]
perfworks: ["ON"] # quoted: unquoted ON/OFF are YAML 1.1 booleans
include:
# Release wheels build without PerfWorks (manylinux ships no nvperf)
# and nothing else here does. One job is enough for a link error.
- os: ubuntu-22.04
python-version: "3.12"
perfworks: "OFF"

env:
CMAKE_ARGS: >-
-DGPUFL_ENABLE_NVIDIA=ON
-DGPUFL_ENABLE_AMD=OFF
-DBUILD_TESTING=OFF
-DGPUFL_ENABLE_PERFWORKS=${{ matrix.perfworks }}
-DBUILD_TESTING=OFF

steps:
- uses: actions/checkout@v4
Expand Down
46 changes: 28 additions & 18 deletions CMakeLists.txt
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -29,6 +29,9 @@ set(CMAKE_CXX_EXTENSIONS OFF)
option(GPUFL_ENABLE_NVIDIA "Enable NVIDIA backends (CUDA + NVML when available)" ON)
option(GPUFL_ENABLE_AMD "Enable AMD backends (ROCm when available)" OFF)
option(GPUFL_ENABLE_AMD_ROCPROFILER "Enable AMD rocprofiler-sdk tracing backend when available" ON)
# OFF builds the configuration release wheels ship (manylinux has no nvperf),
# so CI can catch #if GPUFL_HAS_PERFWORKS mistakes.
option(GPUFL_ENABLE_PERFWORKS "Link PerfWorks (NVPERF) for hardware-counter scopes" ON)

option(BUILD_GPUFL_EXAMPLE "Build gpufl example application" ON)
option(BUILD_GPUFL_MONITOR "Build gpufl-monitor standalone daemon" ON)
Expand Down Expand Up @@ -455,26 +458,33 @@ if(GPUFL_ENABLE_NVIDIA)
# --------------------------------------------------------
# NVPERF Support (for GFL_PERF_SCOPE hardware counters)
# --------------------------------------------------------
find_library(NVPERF_HOST_LIBRARY NAMES nvperf_host nvperf_host_static
HINTS "${CUDAToolkit_ROOT}/extras/CUPTI/lib64"
"${CUDAToolkit_ROOT}/targets/x86_64-linux/lib"
"$ENV{CUDA_PATH}/extras/CUPTI/lib64"
"$ENV{CUDA_PATH}/targets/x86_64-linux/lib"
)
find_library(NVPERF_TARGET_LIBRARY NAMES nvperf_target
HINTS "${CUDAToolkit_ROOT}/extras/CUPTI/lib64"
"${CUDAToolkit_ROOT}/targets/x86_64-linux/lib"
"$ENV{CUDA_PATH}/extras/CUPTI/lib64"
"$ENV{CUDA_PATH}/targets/x86_64-linux/lib"
)
if(NVPERF_HOST_LIBRARY AND NVPERF_TARGET_LIBRARY)
set(GPUFL_HAS_PERFWORKS 1)
target_link_libraries(gpufl PRIVATE
${NVPERF_HOST_LIBRARY} ${NVPERF_TARGET_LIBRARY})
message(STATUS "Found NVPERF: ${NVPERF_HOST_LIBRARY}")
# Gated on the option, not just detection: find_library caches, so
# there is no command-line way to force a miss.
if(GPUFL_ENABLE_PERFWORKS)
find_library(NVPERF_HOST_LIBRARY NAMES nvperf_host nvperf_host_static
HINTS "${CUDAToolkit_ROOT}/extras/CUPTI/lib64"
"${CUDAToolkit_ROOT}/targets/x86_64-linux/lib"
"$ENV{CUDA_PATH}/extras/CUPTI/lib64"
"$ENV{CUDA_PATH}/targets/x86_64-linux/lib"
)
find_library(NVPERF_TARGET_LIBRARY NAMES nvperf_target
HINTS "${CUDAToolkit_ROOT}/extras/CUPTI/lib64"
"${CUDAToolkit_ROOT}/targets/x86_64-linux/lib"
"$ENV{CUDA_PATH}/extras/CUPTI/lib64"
"$ENV{CUDA_PATH}/targets/x86_64-linux/lib"
)
if(NVPERF_HOST_LIBRARY AND NVPERF_TARGET_LIBRARY)
set(GPUFL_HAS_PERFWORKS 1)
target_link_libraries(gpufl PRIVATE
${NVPERF_HOST_LIBRARY} ${NVPERF_TARGET_LIBRARY})
message(STATUS "Found NVPERF: ${NVPERF_HOST_LIBRARY}")
else()
set(GPUFL_HAS_PERFWORKS 0)
message(WARNING "NVPERF not found - GFL_PERF_SCOPE disabled at runtime")
endif()
else()
set(GPUFL_HAS_PERFWORKS 0)
message(WARNING "NVPERF not found - GFL_PERF_SCOPE disabled at runtime")
message(STATUS "PerfWorks disabled by GPUFL_ENABLE_PERFWORKS=OFF")
endif()
endif()
endif()
Expand Down
Loading