-2

I'm doing a project of embedded vision to recognize circles without using color and I have made the program with a total processing time of 600msec but I have observed that only the program block where 4 bidimensional convolutions are executed has a total processing time of 300msec , this means that only that program block demands a time of 50% of the total program time.

This is a simple example of using two-dimensional convolution with the filter2D command of openCV, as we see it arranged sequentially in main function to see how much time it demands when executing four times. the line code: cout << "Execution Time: " << time << endl; in the program display Execution Time: 0.026662sec

#include <stdio.h>
#include <stdlib.h>
#include <opencv2/core/core.hpp>
#include <opencv2/highgui/highgui.hpp>
#include "opencv2/imgproc/imgproc.hpp"
#include <iostream>
#include <ctime>
#include <opencv2/opencv.hpp>
#include <math.h>   

using namespace cv;
using namespace std;


    float buf_in[1280]  = { 1.3456,  -4.75567,  34.4512,  93.65666, -23.564, 0.3445, -20.245,  -23.678,  3.5673, -9.4765, -5.643,  10.37422,  -43.282,  -10.543,   12.567, 23.341, -2.2529,  12.345,   9.5672, -0.2367, 3.3456,  -2.75567,  3.4521,   73.2745,  13.7948, 0.9455, 23.6746,  -36.678,  28.656, -19.4765, 11.3456,  14.75567,  4.45,    13.65666, -2.564,  4.3445, 0.245,    -12.64,  1.5673,  -10.4765,  -34.3456,  -0.75567,  6.45,   -1.4566,  -3.564,  11.3445, -2.245,  -3.678,  31.5673, -1.4765, 71.3456,  -14.75567,  14.45,  3.65666, -0.564,   90.3445,  0.245,  39.678,  -3.5673, -9.4765,   22.3456,  0.75567,  -77.45,  62.65666, -31.564, 1.3445, -0.245,    -4.678,   3.5673,   9.4765, 61.3456,  1.75567,  -8.45,    4.66,    -34.544,  12.3445, -2.245,  -3.678,  54.5673,  9.4765 , 1.3456,  -4.75567,  34.4512,  93.65666, -23.564, 0.3445, -20.245,  -23.678,  3.5673, -9.4765, -5.643,  10.37422,  -43.282,  -10.543,   12.567, 23.341, -2.2529,  12.345,   9.5672, -0.2367, 3.3456,  -2.75567,  3.4521,   73.2745,  13.7948, 0.9455, 23.6746,  -36.678,  28.656, -19.4765, 11.3456,  14.75567,  4.45,    13.65666, -2.564,  4.3445, 0.245,    -12.64,  1.5673,  -10.4765,  -34.3456,  -0.75567,  6.45,   -1.4566,  -3.564,  11.3445, -2.245,  -3.678,  31.5673, -1.4765, 71.3456,  -14.75567,  14.45,  3.65666, -0.564,   90.3445,  0.245,  39.678,  -3.5673, -9.4765,   22.3456,  0.75567,  -77.45,  62.65666, -31.564, 1.3445, -0.245,    -4.678,   3.5673,   9.4765, 61.3456,  1.75567,  -8.45,    4.66,    -34.544,  12.3445, -2.245,  -3.678,  54.5673,  9.4765 , 1.3456,  -4.75567,  34.4512,  93.65666, -23.564, 0.3445, -20.245,  -23.678,  3.5673, -9.4765, -5.643,  10.37422,  -43.282,  -10.543,   12.567, 23.341, -2.2529,  12.345,   9.5672, -0.2367, 3.3456,  -2.75567,  3.4521,   73.2745,  13.7948, 0.9455, 23.6746,  -36.678,  28.656, -19.4765, 11.3456,  14.75567,  4.45,    13.65666, -2.564,  4.3445, 0.245,    -12.64,  1.5673,  -10.4765,  -34.3456,  -0.75567,  6.45,   -1.4566,  -3.564,  11.3445, -2.245,  -3.678,  31.5673, -1.4765, 71.3456,  -14.75567,  14.45,  3.65666, -0.564,   90.3445,  0.245,  39.678,  -3.5673, -9.4765,   22.3456,  0.75567,  -77.45,  62.65666, -31.564, 1.3445, -0.245,    -4.678,   3.5673,   9.4765, 61.3456,  1.75567,  -8.45,    4.66,    -34.544,  12.3445, -2.245,  -3.678,  54.5673,  9.4765 , 1.3456,  -4.75567,  34.4512,  93.65666, -23.564, 0.3445, -20.245,  -23.678,  3.5673, -9.4765, -5.643,  10.37422,  -43.282,  -10.543,   12.567, 23.341, -2.2529,  12.345,   9.5672, -0.2367, 3.3456,  -2.75567,  3.4521,   73.2745,  13.7948, 0.9455, 23.6746,  -36.678,  28.656, -19.4765, 11.3456,  14.75567,  4.45,    13.65666, -2.564,  4.3445, 0.245,    -12.64,  1.5673,  -10.4765,  -34.3456,  -0.75567,  6.45,   -1.4566,  -3.564,  11.3445, -2.245,  -3.678,  31.5673, -1.4765, 71.3456,  -14.75567,  14.45,  3.65666, -0.564,   90.3445,  0.245,  39.678,  -3.5673, -9.4765,   22.3456,  0.75567,  -77.45,  62.65666, -31.564, 1.3445, -0.245,    -4.678,   3.5673,   9.4765, 61.3456,  1.75567,  -8.45,    4.66,    -34.544,  12.3445, -2.245,  -3.678,  54.5673,  9.4765, 1.3456,  -4.75567,  34.4512,  93.65666, -23.564, 0.3445, -20.245,  -23.678,  3.5673, -9.4765, -5.643,  10.37422,  -43.282,  -10.543,   12.567, 23.341, -2.2529,  12.345,   9.5672, -0.2367, 3.3456,  -2.75567,  3.4521,   73.2745,  13.7948, 0.9455, 23.6746,  -36.678,  28.656, -19.4765, 11.3456,  14.75567,  4.45,    13.65666, -2.564,  4.3445, 0.245,    -12.64,  1.5673,  -10.4765,  -34.3456,  -0.75567,  6.45,   -1.4566,  -3.564,  11.3445, -2.245,  -3.678,  31.5673, -1.4765, 71.3456,  -14.75567,  14.45,  3.65666, -0.564,   90.3445,  0.245,  39.678,  -3.5673, -9.4765,   22.3456,  0.75567,  -77.45,  62.65666, -31.564, 1.3445, -0.245,    -4.678,   3.5673,   9.4765, 61.3456,  1.75567,  -8.45,    4.66,    -34.544,  12.3445, -2.245,  -3.678,  54.5673,  9.4765 , 1.3456,  -4.75567,  34.4512,  93.65666, -23.564, 0.3445, -20.245,  -23.678,  3.5673, -9.4765, -5.643,  10.37422,  -43.282,  -10.543,   12.567, 23.341, -2.2529,  12.345,   9.5672, -0.2367, 3.3456,  -2.75567,  3.4521,   73.2745,  13.7948, 0.9455, 23.6746,  -36.678,  28.656, -19.4765, 11.3456,  14.75567,  4.45,    13.65666, -2.564,  4.3445, 0.245,    -12.64,  1.5673,  -10.4765,  -34.3456,  -0.75567,  6.45,   -1.4566,  -3.564,  11.3445, -2.245,  -3.678,  31.5673, -1.4765, 71.3456,  -14.75567,  14.45,  3.65666, -0.564,   90.3445,  0.245,  39.678,  -3.5673, -9.4765,   22.3456,  0.75567,  -77.45,  62.65666, -31.564, 1.3445, -0.245,    -4.678,   3.5673,   9.4765, 61.3456,  1.75567,  -8.45,    4.66,    -34.544,  12.3445, -2.245,  -3.678,  54.5673,  9.4765 , 1.3456,  -4.75567,  34.4512,  93.65666, -23.564, 0.3445, -20.245,  -23.678,  3.5673, -9.4765, -5.643,  10.37422,  -43.282,  -10.543,   12.567, 23.341, -2.2529,  12.345,   9.5672, -0.2367, 3.3456,  -2.75567,  3.4521,   73.2745,  13.7948, 0.9455, 23.6746,  -36.678,  28.656, -19.4765, 11.3456,  14.75567,  4.45,    13.65666, -2.564,  4.3445, 0.245,    -12.64,  1.5673,  -10.4765,  -34.3456,  -0.75567,  6.45,   -1.4566,  -3.564,  11.3445, -2.245,  -3.678,  31.5673, -1.4765, 71.3456,  -14.75567,  14.45,  3.65666, -0.564,   90.3445,  0.245,  39.678,  -3.5673, -9.4765,   22.3456,  0.75567,  -77.45,  62.65666, -31.564, 1.3445, -0.245,    -4.678,   3.5673,   9.4765, 61.3456,  1.75567,  -8.45,    4.66,    -34.544,  12.3445, -2.245,  -3.678,  54.5673,  9.4765 , 1.3456,  -4.75567,  34.4512,  93.65666, -23.564, 0.3445, -20.245,  -23.678,  3.5673, -9.4765, -5.643,  10.37422,  -43.282,  -10.543,   12.567, 23.341, -2.2529,  12.345,   9.5672, -0.2367, 3.3456,  -2.75567,  3.4521,   73.2745,  13.7948, 0.9455, 23.6746,  -36.678,  28.656, -19.4765, 11.3456,  14.75567,  4.45,    13.65666, -2.564,  4.3445, 0.245,    -12.64,  1.5673,  -10.4765,  -34.3456,  -0.75567,  6.45,   -1.4566,  -3.564,  11.3445, -2.245,  -3.678,  31.5673, -1.4765, 71.3456,  -14.75567,  14.45,  3.65666, -0.564,   90.3445,  0.245,  39.678,  -3.5673, -9.4765,   22.3456,  0.75567,  -77.45,  62.65666, -31.564, 1.3445, -0.245,    -4.678,   3.5673,   9.4765, 61.3456,  1.75567,  -8.45,    4.66,    -34.544,  12.3445, -2.245,  -3.678,  54.5673,  9.4765 , 1.3456,  -4.75567,  34.4512,  93.65666, -23.564, 0.3445, -20.245,  -23.678,  3.5673, -9.4765, -5.643,  10.37422,  -43.282,  -10.543,   12.567, 23.341, -2.2529,  12.345,   9.5672, -0.2367, 3.3456,  -2.75567,  3.4521,   73.2745,  13.7948, 0.9455, 23.6746,  -36.678,  28.656, -19.4765, 11.3456,  14.75567,  4.45,    13.65666, -2.564,  4.3445, 0.245,    -12.64,  1.5673,  -10.4765,  -34.3456,  -0.75567,  6.45,   -1.4566,  -3.564,  11.3445, -2.245,  -3.678,  31.5673, -1.4765, 71.3456,  -14.75567,  14.45,  3.65666, -0.564,   90.3445,  0.245,  39.678,  -3.5673, -9.4765,   22.3456,  0.75567,  -77.45,  62.65666, -31.564, 1.3445, -0.245,    -4.678,   3.5673,   9.4765, 61.3456,  1.75567,  -8.45,    4.66,    -34.544,  12.3445, -2.245,  -3.678,  54.5673,  9.4765 , 1.3456,  -4.75567,  34.4512,  93.65666, -23.564, 0.3445, -20.245,  -23.678,  3.5673, -9.4765, -5.643,  10.37422,  -43.282,  -10.543,   12.567, 23.341, -2.2529,  12.345,   9.5672, -0.2367, 3.3456,  -2.75567,  3.4521,   73.2745,  13.7948, 0.9455, 23.6746,  -36.678,  28.656, -19.4765, 11.3456,  14.75567,  4.45,    13.65666, -2.564,  4.3445, 0.245,    -12.64,  1.5673,  -10.4765,  -34.3456,  -0.75567,  6.45,   -1.4566,  -3.564,  11.3445, -2.245,  -3.678,  31.5673, -1.4765, 71.3456,  -14.75567,  14.45,  3.65666, -0.564,   90.3445,  0.245,  39.678,  -3.5673, -9.4765,   22.3456,  0.75567,  -77.45,  62.65666, -31.564, 1.3445, -0.245,    -4.678,   3.5673,   9.4765, 61.3456,  1.75567,  -8.45,    4.66,    -34.544,  12.3445, -2.245,  -3.678,  54.5673,  9.4765 , 1.3456,  -4.75567,  34.4512,  93.65666, -23.564, 0.3445, -20.245,  -23.678,  3.5673, -9.4765, -5.643,  10.37422,  -43.282,  -10.543,   12.567, 23.341, -2.2529,  12.345,   9.5672, -0.2367, 3.3456,  -2.75567,  3.4521,   73.2745,  13.7948, 0.9455, 23.6746,  -36.678,  28.656, -19.4765, 11.3456,  14.75567,  4.45,    13.65666, -2.564,  4.3445, 0.245,    -12.64,  1.5673,  -10.4765,  -34.3456,  -0.75567,  6.45,   -1.4566,  -3.564,  11.3445, -2.245,  -3.678,  31.5673, -1.4765, 71.3456,  -14.75567,  14.45,  3.65666, -0.564,   90.3445,  0.245,  39.678,  -3.5673, -9.4765,   22.3456,  0.75567,  -77.45,  62.65666, -31.564, 1.3445, -0.245,    -4.678,   3.5673,   9.4765, 61.3456,  1.75567,  -8.45,    4.66,    -34.544,  12.3445, -2.245,  -3.678,  54.5673,  9.4765 , 1.3456,  -4.75567,  34.4512,  93.65666, -23.564, 0.3445, -20.245,  -23.678,  3.5673, -9.4765, -5.643,  10.37422,  -43.282,  -10.543,   12.567, 23.341, -2.2529,  12.345,   9.5672, -0.2367, 3.3456,  -2.75567,  3.4521,   73.2745,  13.7948, 0.9455, 23.6746,  -36.678,  28.656, -19.4765, 11.3456,  14.75567,  4.45,    13.65666, -2.564,  4.3445, 0.245,    -12.64,  1.5673,  -10.4765,  -34.3456,  -0.75567,  6.45,   -1.4566,  -3.564,  11.3445, -2.245,  -3.678,  31.5673, -1.4765, 71.3456,  -14.75567,  14.45,  3.65666, -0.564,   90.3445,  0.245,  39.678,  -3.5673, -9.4765,   22.3456,  0.75567,  -77.45,  62.65666, -31.564, 1.3445, -0.245,    -4.678,   3.5673,   9.4765, 61.3456,  1.75567,  -8.45,    4.66,    -34.544,  12.3445, -2.245,  -3.678,  54.5673,  9.4765, 1.3456,  -4.75567,  34.4512,  93.65666, -23.564, 0.3445, -20.245,  -23.678,  3.5673, -9.4765, -5.643,  10.37422,  -43.282,  -10.543,   12.567, 23.341, -2.2529,  12.345,   9.5672, -0.2367, 3.3456,  -2.75567,  3.4521,   73.2745,  13.7948, 0.9455, 23.6746,  -36.678,  28.656, -19.4765, 11.3456,  14.75567,  4.45,    13.65666, -2.564,  4.3445, 0.245,    -12.64,  1.5673,  -10.4765,  -34.3456,  -0.75567,  6.45,   -1.4566,  -3.564,  11.3445, -2.245,  -3.678,  31.5673, -1.4765, 71.3456,  -14.75567,  14.45,  3.65666, -0.564,   90.3445,  0.245,  39.678,  -3.5673, -9.4765,   22.3456,  0.75567,  -77.45,  62.65666, -31.564, 1.3445, -0.245,    -4.678,   3.5673,   9.4765, 61.3456,  1.75567,  -8.45,    4.66,    -34.544,  12.3445, -2.245,  -3.678,  54.5673,  9.4765 , 1.3456,  -4.75567,  34.4512,  93.65666, -23.564, 0.3445, -20.245,  -23.678,  3.5673, -9.4765, -5.643,  10.37422,  -43.282,  -10.543,   12.567, 23.341, -2.2529,  12.345,   9.5672, -0.2367, 3.3456,  -2.75567,  3.4521,   73.2745,  13.7948, 0.9455, 23.6746,  -36.678,  28.656, -19.4765, 11.3456,  14.75567,  4.45,    13.65666, -2.564,  4.3445, 0.245,    -12.64,  1.5673,  -10.4765,  -34.3456,  -0.75567,  6.45,   -1.4566,  -3.564,  11.3445, -2.245,  -3.678,  31.5673, -1.4765, 71.3456,  -14.75567,  14.45,  3.65666, -0.564,   90.3445,  0.245,  39.678,  -3.5673, -9.4765,   22.3456,  0.75567,  -77.45,  62.65666, -31.564, 1.3445, -0.245,    -4.678,   3.5673,   9.4765, 61.3456,  1.75567,  -8.45,    4.66,    -34.544,  12.3445, -2.245,  -3.678,  54.5673,  9.4765 , 1.3456,  -4.75567,  34.4512,  93.65666, -23.564, 0.3445, -20.245,  -23.678,  3.5673, -9.4765, -5.643,  10.37422,  -43.282,  -10.543,   12.567, 23.341, -2.2529,  12.345,   9.5672, -0.2367, 3.3456,  -2.75567,  3.4521,   73.2745,  13.7948, 0.9455, 23.6746,  -36.678,  28.656, -19.4765, 11.3456,  14.75567,  4.45,    13.65666, -2.564,  4.3445, 0.245,    -12.64,  1.5673,  -10.4765,  -34.3456,  -0.75567,  6.45,   -1.4566,  -3.564,  11.3445, -2.245,  -3.678,  31.5673, -1.4765, 71.3456,  -14.75567,  14.45,  3.65666, -0.564,   90.3445,  0.245,  39.678,  -3.5673, -9.4765,   22.3456,  0.75567,  -77.45,  62.65666, -31.564, 1.3445, -0.245,    -4.678,   3.5673,   9.4765, 61.3456,  1.75567,  -8.45,    4.66,    -34.544,  12.3445, -2.245,  -3.678,  54.5673,  9.4765 , 1.3456,  -4.75567,  34.4512,  93.65666, -23.564, 0.3445, -20.245,  -23.678,  3.5673, -9.4765, -5.643,  10.37422,  -43.282,  -10.543,   12.567, 23.341, -2.2529,  12.345,   9.5672, -0.2367, 3.3456,  -2.75567,  3.4521,   73.2745,  13.7948, 0.9455, 23.6746,  -36.678,  28.656, -19.4765, 11.3456,  14.75567,  4.45,    13.65666, -2.564,  4.3445, 0.245,    -12.64,  1.5673,  -10.4765,  -34.3456,  -0.75567,  6.45,   -1.4566,  -3.564,  11.3445, -2.245,  -3.678,  31.5673, -1.4765, 71.3456,  -14.75567,  14.45,  3.65666, -0.564,   90.3445,  0.245,  39.678,  -3.5673, -9.4765,   22.3456,  0.75567,  -77.45,  62.65666, -31.564, 1.3445, -0.245,    -4.678,   3.5673,   9.4765, 61.3456,  1.75567,  -8.45,    4.66,    -34.544,  12.3445, -2.245,  -3.678,  54.5673,  9.4765 };






    float buf_kernel[441] ={ 0.434, 1.4343, 9.434, 5.453, 3.5656, 4.556,   5.776, 7.565, 2.545, 8.65, 1.44,   0.5665, 4.566, 7.666, 4.66, 5.56,    3.455, 6.76,  2.334, 8.65, 6.45,
                           1.44,   0.5665, 4.566, 7.666, 4.66, 4.556,   5.776, 7.565, 2.545, 8.65, 4.556,   5.776, 7.565, 2.545, 8.65, 1.44,   0.5665, 4.566, 7.666, 4.66, 2.53,
                         5.4343, 9.6556, 5,  34.75, 7.56, 5.56,    3.455, 6.76,  2.334, 8.65, 5.56, 3.455, 6.76,  2.334, 8.65, 4.556,   5.776, 7.565, 2.545, 8.65, 1.454,
                         4.556,   5.776, 7.565, 2.545, 8.65, 1.44,   0.5665, 4.566, 7.666, 4.66, 4.556,   5.776, 7.565, 2.545, 8.65, 4.556,   5.776, 7.565, 2.545, 8.65, 4.56,
                         5.56,    3.455, 6.76,  2.334, 8.65, 5.56, 3.455, 6.76,  2.334, 8.65, 4.556,   5.776, 7.565, 2.545, 8.65, 5.4343, 9.6556, 5,  34.75, 7.56, 9.3434,
                         1.44,   0.5665, 4.566, 7.666, 4.66 , 4.556,   5.776, 7.565, 2.545, 8.65, 1.44,   0.5665, 4.566, 7.666, 4.66, 5.56,    3.455, 6.76,  2.334, 8.65, 4.656,
                         1.44,   0.5665, 4.566, 7.666, 4.66, 4.556,   5.776, 7.565, 2.545, 8.65, 4.556,   5.776, 7.565, 2.545, 8.65, 1.44,   0.5665, 4.566, 7.666, 4.66, 3.343,
                         5.4343, 9.6556, 5,  34.75, 7.56, 5.56,    3.455, 6.76,  2.334, 8.65, 5.56, 3.455, 6.76,  2.334, 8.65, 4.556,   5.776, 7.565, 2.545, 8.65, 2.4545,
                         4.556,   5.776, 7.565, 2.545, 8.65, 1.44,   0.5665, 4.566, 7.666, 4.66, 4.556,   5.776, 7.565, 2.545, 8.65, 4.556,   5.776, 7.565, 2.545, 8.65, 1.4334,
                         5.56,    3.455, 6.76,  2.334, 8.65, 5.56, 3.455, 6.76,  2.334, 8.65, 4.556,   5.776, 7.565, 2.545, 8.65, 5.4343, 9.6556, 5,  34.75, 7.56, 7.5665,
                         1.44,   0.5665, 4.566, 7.666, 4.66 , 4.556,   5.776, 7.565, 2.545, 8.65, 1.44,   0.5665, 4.566, 7.666, 4.66, 5.56,    3.455, 6.76,  2.334, 8.65, 9.4554,
                         5.56,    3.455, 6.76,  2.334, 8.65, 5.56, 3.455, 6.76,  2.334, 8.65, 4.556,   5.776, 7.565, 2.545, 8.65, 5.4343, 9.6556, 5,  34.75, 7.56, 3.545,
                         1.44,   0.5665, 4.566, 7.666, 4.66 , 4.556,   5.776, 7.565, 2.545, 8.65, 1.44,   0.5665, 4.566, 7.666, 4.66, 5.56,    3.455, 6.76,  2.334, 8.65, 7.4545,
                         1.44,   0.5665, 4.566, 7.666, 4.66, 4.556,   5.776, 7.565, 2.545, 8.65, 4.556,   5.776, 7.565, 2.545, 8.65, 1.44,   0.5665, 4.566, 7.666, 4.66, 9.445,
                         5.4343, 9.6556, 5,  34.75, 7.56, 5.56,    3.455, 6.76,  2.334, 8.65, 5.56, 3.455, 6.76,  2.334, 8.65, 4.556,   5.776, 7.565, 2.545, 8.65, 3.54545,
                         4.556,   5.776, 7.565, 2.545, 8.65, 1.44,   0.5665, 4.566, 7.666, 4.66, 4.556,   5.776, 7.565, 2.545, 8.65, 4.556,   5.776, 7.565, 2.545, 8.65, 4.546,
                         5.56,    3.455, 6.76,  2.334, 8.65, 5.56, 3.455, 6.76,  2.334, 8.65, 4.556,   5.776, 7.565, 2.545, 8.65, 5.4343, 9.6556, 5,  34.75, 7.56, 6.4545,
                         5.4343, 9.6556, 5,  34.75, 7.56, 5.56,    3.455, 6.76,  2.334, 8.65, 5.56, 3.455, 6.76,  2.334, 8.65, 4.556,   5.776, 7.565, 2.545, 8.65, 5.786,
                         4.556,   5.776, 7.565, 2.545, 8.65, 1.44,   0.5665, 4.566, 7.666, 4.66, 4.556,   5.776, 7.565, 2.545, 8.65, 4.556,   5.776, 7.565, 2.545, 8.65, 6.5656,
                         5.56,    3.455, 6.76,  2.334, 8.65, 5.56, 3.455, 6.76,  2.334, 8.65, 4.556,   5.776, 7.565, 2.545, 8.65, 5.4343, 9.6556, 5,  34.75, 7.56, 7.4554,
                         1.44,   0.5665, 4.566, 7.666, 4.66 , 4.556,   5.776, 7.565, 2.545, 8.65, 1.44,   0.5665, 4.566, 7.666, 4.66, 5.56,    3.455, 6.76,  2.334, 8.65, 7.344};   



unsigned t0, t1;
int counter = 0;
void processor(void *arg   ){
   printf(" Dentro de la funcion\n");
   float matriz[1280]; 
   float *buf_out = &matriz[0];
  // variable local
   int var_local =0;
   var_local = var_local +10;
  // variable global
   counter = counter +10;

   Mat mat_out = Mat(32,40, CV_32FC1); //output          
   Mat mat_flip_kernel;                // se cargara la matriz rotada 180°
   Mat mat_kernel( Size(21, 21), CV_32FC1, buf_kernel );       // kernel
   Mat mat_in( Size(40, 32),CV_32FC1, buf_in);                // input
   flip(mat_kernel, mat_flip_kernel,-1);   // rotando la matriz 180° para la convolucion, porque filter2d solo realiza la correlacion 
   Point anchor( -1,-1);                   // para que este centrado en el centro del kerneL
   filter2D(mat_in, mat_out, -1, mat_flip_kernel, anchor, 0, BORDER_DEFAULT); 

    for (int i=0; i<32; i++){
               float *ptr_out = mat_out.ptr<float>(i); // ptr_out puntero a float, to 1st element in row i  
               for (int j=0 ;j<40 ;j++){ 
                    buf_out[i*40+j] =  ptr_out[j];
                    cout<<buf_out[i*40 + j]<<"; ";
             }
   }



  // variable local
   var_local++;
  // variable global
   counter++;
  printf(" variable local: %d\n",var_local);
  printf(" variable global : %d\n",counter);
  return ;
}



int main(void){
    int i = 0;

    t0 = clock();
    processor(NULL);
    processor(NULL);
    processor(NULL);
    processor(NULL);
    t1 = clock();
    double time = (double(t1-t0)/CLOCKS_PER_SEC);
    cout << "Execution Time: " << time << endl;



}

As 4 convolutions are processed independently, choose to use multithreading in order to reduce processing time, but when I implement it I am surprised that the progamma block where multithreading is performed requires a processing time of 600msec, causing Now that the total program takes a processing time of 900msec, that is to say, it gets worse, as it is possible that implementing multithreading is slower than using a program with a thread.

This is the previous simple program of using two-dimensional convolution with the filter2D command of openCV but using multithreading, but each bidimensional convolution in each core or thread, that is, they are processed in parallel, I imagine that it should be executed in parallel. for this specific simple example program, the display program" Execution time : 0.029157sec ", which is a bit longer than the processing time using a single thread, obviously it is wrong, because this time should be a quarter of the execution time using a single thread, not it's like that.

#include <stdio.h>
#include <stdlib.h>
#include <pthread.h>
#include <opencv2/core/core.hpp>
#include <opencv2/highgui/highgui.hpp>
#include "opencv2/imgproc/imgproc.hpp"
#include <iostream>
#include <ctime>
#include <opencv2/opencv.hpp>
#include <math.h>

using namespace cv;
using namespace std;
pthread_mutex_t mutex1 = PTHREAD_MUTEX_INITIALIZER;
pthread_t thread_id[4];

    float buf_in[1280]  = { for reasons of space limit I do not place the data, but it is the same as the example program above};    

float buf_kernel[441] ={for reasons of space limit I do not place the data, but it is the same as the example program above};   


unsigned t0, t1;
int counter = 0;
void *thread_processor(void *arg   ){
   printf(" Dentro de Thread\n");


   float matriz[1280]; 
    float *buf_out = &matriz[0];



  // variable local
   pthread_mutex_lock( &mutex1 );
   int var_local =0;
   var_local = var_local +10;
   pthread_mutex_unlock( &mutex1 );
 // variable global
   pthread_mutex_lock( &mutex1 );
   counter = counter +10;
   pthread_mutex_unlock( &mutex1 );

   Mat mat_out = Mat(32,40, CV_32FC1); //output          
   Mat mat_flip_kernel;// = Mat(21,21, CV_32FC1); // se cargara la matriz rotada 180°
   Mat mat_kernel( Size(21, 21), CV_32FC1, buf_kernel );       // kernel
   Mat mat_in( Size(40, 32),CV_32FC1, buf_in);    // input
   flip(mat_kernel, mat_flip_kernel,-1);   // rotando la matriz 180° para la convolucion, porque filter2d solo realiza la correlacion 


   Point anchor( -1,-1);                   // para que este centrado en el centro del kerneL
   filter2D(mat_in, mat_out, -1, mat_flip_kernel, anchor, 0, BORDER_DEFAULT); 
 //  pthread_mutex_lock( &mutex1 );
 //  cout<<"mat_out :"<<mat_out<<endl;
 //  pthread_mutex_unlock( &mutex1 );


// Pero al usar multithread,por eso cambie a otro tipo de declaracion de puntero que si funciono
    pthread_mutex_lock( &mutex1 );   
    for (int i=0; i<32; i++){
               float *ptr_out = mat_out.ptr<float>(i); // ptr_out puntero a float, to 1st element in row 0  
                for (int j=0 ;j<40 ;j++){ 
               // pthread_mutex_lock( &mutex1 );     
               buf_out[i*40+j] =  ptr_out[j];
               // cout<<" Elemento ["<<i<<"] ["<<j<<"] :"<<buf_out[i*40 + j];
                cout<<buf_out[i*40 + j]<<"; ";
               // pthread_mutex_unlock( &mutex1 );
        }

     }  
     pthread_mutex_unlock( &mutex1);



  // variable local
   pthread_mutex_lock( &mutex1 );
   var_local++;
   pthread_mutex_unlock( &mutex1 );
  // variable global
   pthread_mutex_lock( &mutex1 );
   counter++;
   pthread_mutex_unlock( &mutex1 );
   printf(" variable local: %d\n",var_local);
   printf(" variable global : %d\n",counter);
   printf(" Fuera de Thread\n");

  return NULL;
}



int main(void){
  int i = 0;
  int error;

  t0 = clock();

  while(i < 4) {
    error = pthread_create(&(thread_id[i]), NULL, &thread_processor,NULL);
    if (error != 0){  
        printf("\nthere was a problem creating thread:");
    }
    else{
       printf("\n Thread number %d created.\n", i);
    }

   i++;
 }

 for ( i =0 ; i< 4 ; i++){
   pthread_join( thread_id[i], NULL);


}

   t1 = clock();
   double time = (double(t1-t0)/CLOCKS_PER_SEC);
   cout << "Execution Time: " << time << endl;



}

Can somebody help me find what might slow the program down when this is ran with multiple threads and maybe a solution? Thank you for all help and advice in advance!

ALBERTO ROSELLO ALVAREZ
  • 1
  • 1
  • May help, https://stackoverflow.com/questions/39495136/why-settting-cpu-affinity-make-threads-run-slower/39499787#39499787 – Matthew Fisher Sep 16 '18 at 03:14
  • 6
    Yes, some algorithms perform worse with multiple threads than with one execution thread. Multithreading is not a magical button that makes everything run faster. One must understand how multithreading works, and implement whatever algorithm being used so that it can take advantage of multithreading. In the shown algorithm, for example, it appears that the main calculation loop is mutex-protected. Which means that you sure ain't going to have multiple threads executing in parallel during the main part of your code. Plus, now you also suffer mutex locking overhead! – Sam Varshavchik Sep 16 '18 at 03:23
  • 2
    clock() measures cpu time. Measure "wall time" instead. – Micka Sep 16 '18 at 06:34

1 Answers1

0

You should definitely read some good tutorials or books about concurrency, because there are some grave issues in your program code:

  • Quoting the comment from Sam Varshavchik "Multithreading is not a magical button that makes everything run faster": You are creating four threads and executing the same function on all threads. This does not automatically split up the computations among the four threads, but just leads to the same function being executed 4 times, so you're not getting any benefit. You need to split up the data and distribute it to the different threads by hand.
  • You are protecting EVERYTHING with a mutex, though you should only protect variables that are shared between the different threads. So for example int val_local is a local variable and every thread has its own, no need for a mutex here. And the same applies to all of the other variables, except counter. The for-loop does not need a mutex neither as it does not use any global variables. Mutexes lead to blocking and preemption, which causes the performance to drop significantly, so you should only use them when needed.

  • I don't know what happens in the following code section, but you are using the global variables here. When they are written concurrently it might lead to race conditions and would require synchronization through a mutex:

    Mat mat_out = Mat(32,40, CV_32FC1); //output Mat mat_flip_kernel;// = Mat(21,21, CV_32FC1); // se cargara la matriz rotada 180° Mat mat_kernel( Size(21, 21), CV_32FC1, buf_kernel ); // kernel Mat mat_in( Size(40, 32),CV_32FC1, buf_in); // input flip(mat_kernel, mat_flip_kernel,-1); // rotando la matriz 180° para la convolucion, porque filter2d solo realiza la correlacion

Mike van Dyke
  • 2,724
  • 3
  • 16
  • 31