Intel® Math Kernel Library 2019 Developer Reference - C
The following sample program shows how to employ internal OpenMP* threading in Intel MKL for FFT computation.
To specify the number of threads inside Intel MKL, use the following settings:
set MKL_NUM_THREADS = 1 for one-threaded mode;
set MKL_NUM_THREADS = 4 for multi-threaded mode.
#include "mkl_dfti.h"
int main ()
{
float x[200][100];
DFTI_DESCRIPTOR_HANDLE fft;
MKL_LONG len[2] = {200, 100};
// initialize x
DftiCreateDescriptor ( &fft, DFTI_SINGLE, DFTI_REAL, 2, len );
DftiCommitDescriptor ( fft );
DftiComputeForward ( fft, x );
DftiFreeDescriptor ( &fft );
return 0;
}
The following Example “Using Parallel Mode with Multiple Descriptors Initialized in a Parallel Region” and Example “Using Parallel Mode with Multiple Descriptors Initialized in One Thread” illustrate a parallel customer program with each descriptor instance used only in a single thread.
Specify the number of OpenMP threads for Example “Using Parallel Mode with Multiple Descriptors Initialized in a Parallel Region” like this:
set MKL_NUM_THREADS = 1 for Intel MKL to work in the single-threaded mode (recommended);
set OMP_NUM_THREADS = 4 for the customer program to work in the multi-threaded mode.
Note that in this example, the program can be transformed to become single-threaded at the customer level but using parallel mode within Intel MKL. To achieve this, you need to set the parameter DFTI_NUMBER_OF_TRANSFORMS = 4 and to set the corresponding parameter DFTI_INPUT_DISTANCE = 5000.
#include "mkl_dfti.h"
#include <omp.h>
#define ARRAY_LEN(a) sizeof(a)/sizeof(a[0])
int main ()
{
// 4 OMP threads, each does 2D FFT 50x100 points
MKL_Complex8 x[4][50][100];
int nth = ARRAY_LEN(x);
MKL_LONG len[2] = {ARRAY_LEN(x[0]), ARRAY_LEN(x[0][0])};
int th;
// assume x is initialized and do 2D FFTs
#pragma omp parallel for shared(len, x)
for (th = 0; th < nth; th++)
{
DFTI_DESCRIPTOR_HANDLE myFFT;
DftiCreateDescriptor (&myFFT, DFTI_SINGLE, DFTI_COMPLEX, 2, len);
DftiCommitDescriptor (myFFT);
DftiComputeForward (myFFT, x[th]);
DftiFreeDescriptor (&myFFT);
}
return 0;
}
Specify the number of OpenMP threads for Example “Using Parallel Mode with Multiple Descriptors Initialized in One Thread” like this:
set MKL_NUM_THREADS = 1 for Intel MKL to work in the single-threaded mode (obligatory);
set OMP_NUM_THREADS = 4 for the customer program to work in the multi-threaded mode.
#include "mkl_dfti.h"
#include <omp.h>
#define ARRAY_LEN(a) sizeof(a)/sizeof(a[0])
int main ()
{
// 4 OMP threads, each does 2D FFT 50x100 points
MKL_Complex8 x[4][50][100];
int nth = ARRAY_LEN(x);
MKL_LONG len[2] = {ARRAY_LEN(x[0]), ARRAY_LEN(x[0][0])};
DFTI_DESCRIPTOR_HANDLE FFT[ARRAY_LEN(x)];
int th;
for (th = 0; th < nth; th++)
DftiCreateDescriptor (&FFT[th], DFTI_SINGLE, DFTI_COMPLEX, 2, len);
for (th = 0; th < nth; th++)
DftiCommitDescriptor (FFT[th]);
// assume x is initialized and do 2D FFTs
#pragma omp parallel for shared(FFT, x)
for (th = 0; th < nth; th++)
DftiComputeForward (FFT[th], x[th]);
for (th = 0; th < nth; th++)
DftiFreeDescriptor (&FFT[th]);
return 0;
}
The following Example “Using Parallel Mode with a Common Descriptor” illustrates a parallel customer program with a common descriptor used in several threads.
#include "mkl_dfti.h"
#include <omp.h>
#define ARRAY_LEN(a) sizeof(a)/sizeof(a[0])
int main ()
{
// 4 OMP threads, each does 2D FFT 50x100 points
MKL_Complex8 x[4][50][100];
int nth = ARRAY_LEN(x);
MKL_LONG len[2] = {ARRAY_LEN(x[0]), ARRAY_LEN(x[0][0])};
DFTI_DESCRIPTOR_HANDLE FFT;
int th;
DftiCreateDescriptor (&FFT, DFTI_SINGLE, DFTI_COMPLEX, 2, len);
DftiCommitDescriptor (FFT);
// assume x is initialized and do 2D FFTs
#pragma omp parallel for shared(FFT, x)
for (th = 0; th < nth; th++)
DftiComputeForward (FFT, x[th]);
DftiFreeDescriptor (&FFT);
return 0;
}