Visible to Intel only — GUID: GUID-527DC177-2228-45D4-8269-48E8699A3DBA
Visible to Intel only — GUID: GUID-527DC177-2228-45D4-8269-48E8699A3DBA
CMake Config for oneMKL
If you want to integrate oneMKL into your CMake projects, starting with the Intel® oneAPI Math Kernel Library (oneMKL) 2021.3 release, MKLConfig.cmake is provided as part of the package and installation. MKLConfig.cmake supports all oneMKL configurations, compilers, and runtimes, as the oneMKL product itself. Help/usage is provided in the top section of MKLConfig.cmake.
Example
my_test/ |_ build/ <-- Out-of-source build directory |_ CMakeLists.txt <-- User side project's CMakeLists.txt |_ app.c <-- Source file that uses oneMKL API
app.c
#include <stdio.h> #include "mkl.h" int main() { MKLVersion mkl_version; mkl_get_version(&mkl_version); printf("You are using oneMKL %d.%d\n", mkl_version.MajorVersion, mkl_version.UpdateVersion); return 0; }
CMakeLists.txt
cmake_minimum_required(VERSION 3.13) project(oneMKL_Example LANGUAGES C) find_package(MKL CONFIG REQUIRED PATHS $ENV{MKLROOT}) add_executable(myapp app.c) target_link_libraries(myapp PUBLIC MKL::MKL)
Command line
# Call the oneAPI setvars.bat/oneapi-vars.bat script beforehand build>cmake .. -G Ninja -DCMAKE_C_COMPILER=icx build>cmake --build .
If your application requires a customized setting, the following CMakeLists.txt demonstrates how to achieve a more fine-grained control.
CMakeLists.txt
cmake_minimum_required(VERSION 3.13) project(oneMKL_Example LANGUAGES C) find_package(MKL CONFIG REQUIRED PATHS $ENV{MKLROOT}) message(STATUS "Imported oneMKL targets: ${MKL_IMPORTED_TARGETS}") add_executable(myapp app.c) target_compile_options(myapp PUBLIC $<TARGET_PROPERTY:MKL::MKL,INTERFACE_COMPILE_OPTIONS>) target_include_directories(myapp PUBLIC $<TARGET_PROPERTY:MKL::MKL,INTERFACE_INCLUDE_DIRECTORIES>) target_link_libraries(myapp PUBLIC $<LINK_ONLY:MKL::MKL>)
Ninja and NMake Makefiles are supported on Windows. Other CMake Generators may work but are not guaranteed.
When the Ninja build system is in use, Ninja 1.10.2 or later is required for Fortran support.