Developer Guide for Intel® oneAPI Math Kernel Library for Windows*
Managing Multi-core Performance
You can obtain best performance on systems with multi-core processors by requiring that threads do not migrate from core to core. To do this, bind threads to the CPU cores by setting an affinity mask to threads. Use one of the following options:
OpenMP facilities (if available), for example, the KMP_AFFINITY environment variable using the Intel OpenMP library
A system function, as explained below
Intel® TBB facilities (if available), for example, the tbb::affinity_partitioner class (for details, see https://www.threadingbuildingblocks.org/documentation )
Consider the following performance issue:
The system has two sockets with two cores each, for a total of four cores (CPUs).
The application sets the number of OpenMP threads to four and calls an Intel® oneAPI Math Kernel Library (oneMKL) LAPACK routine two and calls Intel® oneMKL to perform a Fourier transform . This call takes considerably different amounts of time from run to run.
To resolve this issue, before calling Intel® oneMKL, set an affinity mask for each OpenMP thread using the KMP_AFFINITY environment variable or the SetThreadAffinityMask system function. The following code example shows how to resolve the issue by setting an affinity mask by operating system means using the Intel compiler. The code calls the function SetThreadAffinityMask to bind the threads to appropriate cores, preventing migration of the threads. Then the Intel® oneAPI Math Kernel Library (oneMKL) LAPACK routine is called:
// Set affinity mask
#include <windows.h>
#include <omp.h>
int main(void) {
#pragma omp parallel default(shared)
{
int tid = omp_get_thread_num();
// 2 packages x 2 cores/pkg x 1 threads/core (4 total cores)
DWORD_PTR mask = (1 << (tid == 0 ? 0 : 2 ));
SetThreadAffinityMask( GetCurrentThread(), mask );
}
// Call Intel MKL LAPACK routine
return 0;
}Compile the application with the Intel compiler using the following command:
icx /Qopenmp test_application.cwhere test_application.c is the filename for the application.
Build the application. Run it in four threads, for example, by using the environment variable to set the number of threads:
set OMP_NUM_THREADS=4
test_application.exeSee the Windows API documentation at msdn.microsoft.com/ for the restrictions on the usage of Windows API routines and particulars of the SetThreadAffinityMask function used in the above example.
See also a similar example at en.wikipedia.org/wiki/Affinity_mask.
Product and Performance Information |
|---|
Performance varies by use, configuration and other factors. Learn more at www.Intel.com/PerformanceIndex . Notice revision #20201201 |