dlib SIMD tests

fhog_simd_ex.cpp

#include <dlib/image_io.h>
#include <dlib/image_transforms.h>


using namespace std;
using namespace dlib;


int main(int argc, char** argv)
{
    try
    {
        int numRuns = 100;
        double sum = 0.0;
        for (int i = 0; i < numRuns; ++i)
        {
            matrix<unsigned char, 1536, 2048> img;
            dlib::array<array2d<double>> hog;
            chrono::system_clock::time_point start = chrono::system_clock::now();
            impl_fhog::impl_extract_fhog_features(img, hog, 8, 1, 1);
            chrono::system_clock::time_point end = chrono::system_clock::now();
            double msec = std::chrono::duration_cast<chrono::milliseconds>(end - start).count();
            sum += msec;
        }

        cout << "~" << (sum / numRuns) << " ms"  << endl;
    }
    catch (exception& e)
    {
        cout << "exception thrown: " << e.what() << endl;
    }
}

Place this in your your dlib/examples folder and compile and run with something like ...

cd dlib/examples
// copy the file in to this directory

Note that this will run 100 times and output the average time in milliseconds, so on a really slow processor it could take a while.

macOS

with default

g++ -DDLIB_NO_GUI_SUPPORT -DNO_MAKEFILE -std=c++11 -O3 -I.. -lpthread fhog_simd_ex.cpp && ./a.out 

with avx

g++ -DDLIB_NO_GUI_SUPPORT -DNO_MAKEFILE -std=c++11 -O3 -mavx -I.. -lpthread fhog_simd_ex.cpp && ./a.out

Ubuntu

with default

g++ -DDLIB_NO_GUI_SUPPORT -DNO_MAKEFILE -std=c++11 -O3 -I.. -pthread fhog_simd_ex.cpp && ./a.out 

with avx

g++ -DDLIB_NO_GUI_SUPPORT -DNO_MAKEFILE -std=c++11 -O3 -mavx -I.. -pthread fhog_simd_ex.cpp && ./a.out

Raspberry Pi 3

with default

g++ -DDLIB_NO_GUI_SUPPORT -DNO_MAKEFILE -std=c++11 -O3 -I.. -lpthread fhog_simd_ex.cpp && ./a.out 

with neon

g++ -DDLIB_NO_GUI_SUPPORT -DNO_MAKEFILE -std=c++11 -O3 -mfpu=neon -I.. -lpthread fhog_simd_ex.cpp && ./a.out

Orange Pi Zero plus 2 (armbian)

neon not worked on aarch64

g++ -DDLIB_NO_GUI_SUPPORT -DNO_MAKEFILE -std=c++11 -O3 -I.. -lpthread fhog_simd_ex.cpp && ./a.out 

result single core

~450.77 ms