algorithm_extended.c
036e04b14503louis

algorithm_extended.c
View Options

			/*
			============================================================================
			Filename : algorithm.c
			Author : Your names go here
			SCIPER : Your SCIPER numbers

			============================================================================
			*/
			#include <math.h>

			#define INPUT(I,J) input[(I)*length+(J)]
			#define OUTPUT(I,J) output[(I)*length+(J)]

			#define CACHE_SIZE 8
			#define COORD(I,J,K) (I) = K / length; (J) = (K) - (I) * length;

			void simulate(double input, double output, int threads, int length, int iterations)
			{
			double *temp;

			omp_set_num_threads(threads);

			int i;
			int j;
			const int matsize = length*length;
			double horiz_block[3];

			for(int n=0; n < iterations; n++)
			{
			printf("Loop: %d\n", n);

			#pragma omp parallel for private(i,j, horiz_block)
			for(int k=1; k<matsize; k++) // loop j first
			{
			COORD(i,j,k)

			if ((i == 0) \|\| (i == length-1) \|\| (j == 0) \|\| (j == length-1) \|\|
			(((i == length/2-1) \|\| (i== length/2))
			&& ((j == length/2-1) \|\| (j == length/2))) ) {
			continue;
			}

			// write on one chache miss, read on three
			horiz_block[0] = INPUT(i-1,j-1) + INPUT(i-1,j) + INPUT(i-1,j+1);
			horiz_block[1] = INPUT(i,j-1) + INPUT(i,j) + INPUT(i,j+1);
			horiz_block[2] = INPUT(i+1,j-1) + INPUT(i+1,j) + INPUT(i+1,j+1);

			// write on one chache miss
			OUTPUT(i,j) = (horiz_block[0] + horiz_block[1] + horiz_block[2]) / 9;

			// Original line
			/*
			OUTPUT(i,j) = (INPUT(i-1,j-1) + INPUT(i-1,j) + INPUT(i-1,j+1) +
			INPUT(i,j-1) + INPUT(i,j) + INPUT(i,j+1) +
			INPUT(i+1,j-1) + INPUT(i+1,j) + INPUT(i+1,j+1) )/9; */
			}

			temp = input;
			input = output;
			output = temp;
			}
			}