Implement filter-based adaptive algorithm, Add bigger datasets

.gitignore

@@ -53,3 +53,5 @@ dkms.conf
 
 #Output files
 *_output
+
+adj_matrix

datasets/list2matrix.c

@@ -0,0 +1,80 @@
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+
+
+int main(int argc, char **argv){
+
+FILE *fnodes;
+char nodes_file[1000];
+FILE *flist;
+char  list_file[1000];
+FILE *fmatrix;
+char  matrix_file[1000];
+int i,j;
+int **E;
+char *path;
+int N; // number of nodes
+
+ /*** 
+	The argument for the program is the directory name 
+      	of the query for which we want to create the adjacency matrix
+ ***/
+
+ if (argc != 2){ 
+	printf("list2matrix query_directory\n");
+	exit(1);
+ }
+
+ path = strdup(argv[1]);
+
+
+/*** open the nodes file to obtain the number of nodes ***/
+
+ sprintf(nodes_file,"%s/graph/nodes",path);
+ fnodes = fopen(nodes_file,"r");
+ if (fnodes == NULL){
+     printf("ERROR: Cant open file %s\n",nodes_file);
+     exit(1);
+ }
+ fscanf(fnodes,"%d",&N);
+
+ fclose(fnodes);
+
+
+/**** Read List and Construct the adjacency matrix E ****/
+
+ E = (int **)malloc(N*sizeof(int *));
+ for (i = 0; i < N; i ++){
+	  E[i] = (int *)malloc(N*sizeof(int));
+	  for (j = 0; j < N; j ++){
+			E[i][j] = 0;
+	  }
+ }
+
+ sprintf(list_file,"%s/graph/adj_list",path);
+ flist = fopen(list_file,"r");
+ for (i = 0; i < N; i ++){
+     fscanf(flist,"%*d: %d",&j);
+     while (j != -1){
+			E[i][j] = 1;
+			fscanf(flist,"%d",&j);
+	  }
+ }
+
+ fclose(flist);
+
+
+ /*** print the adjacency matrix ***/
+
+ sprintf(matrix_file,"%s/graph/adj_matrix",path);
+ fmatrix = fopen(matrix_file,"w");
+ for (i = 0; i < N; i ++){
+        for (j = 0; j < N; j ++){
+             fprintf(fmatrix,"%d ", E[i][j]);
+        }
+	fprintf(fmatrix,"\n");
+ }
+ fclose(fmatrix);
+}
+

serial/serial_gs_pagerank_functions.c

@@ -16,7 +16,12 @@ int pagerank(double ***transitionMatrix, double **pagerankVector, Parameters par
 	int iterations = 0;
 	double delta,
 	*vectorDifference = (double *) malloc(parameters.numberOfPages * sizeof(double)),
-	*previousPagerankVector = (double *) malloc(parameters.numberOfPages * sizeof(double));
+	*previousPagerankVector = (double *) malloc(parameters.numberOfPages * sizeof(double)),
+	*convergedPagerankVector = (double *) malloc(parameters.numberOfPages * sizeof(double));
+
+	for (int i=0; i<parameters.numberOfPages; ++i) {
+		convergedPagerankVector[i] = 0;
+	}
 
 	if (parameters.verbose) {
 		printf("\n----- Starting iterations -----\n");
@@ -25,7 +30,7 @@ int pagerank(double ***transitionMatrix, double **pagerankVector, Parameters par
 	do {
 		memcpy(previousPagerankVector, *pagerankVector, parameters.numberOfPages * sizeof(double));
 
-		matrixVectorMultiplication(transitionMatrix, previousPagerankVector,
+		matrixVectorMultiplication(transitionMatrix, previousPagerankVector, convergedPagerankVector,
 			pagerankVector, parameters.numberOfPages, parameters.dampingFactor);
 
 		if (parameters.history) {
@@ -38,6 +43,18 @@ int pagerank(double ***transitionMatrix, double **pagerankVector, Parameters par
 		}
 		delta = vectorNorm(vectorDifference, parameters.numberOfPages);
 
+		if (!iterations % 5) {
+			for (int i=0; i<parameters.numberOfPages; ++i) {
+				double temp = fabs((*pagerankVector)[i] - previousPagerankVector[i]) / fabs(previousPagerankVector[i]);
+				if (temp < parameters.convergenceCriterion){
+					convergedPagerankVector[i] = (*pagerankVector)[i];
+					for (int j=0; j<parameters.numberOfPages; ++j){
+						(*transitionMatrix)[i][j] = 0;
+					}
+				}
+			}
+		}
+
 		++iterations;
 		printf("Iteration %d: delta = %f\n", iterations, delta);
 	} while (delta > parameters.convergenceCriterion &&
@@ -141,7 +158,8 @@ void generateNormalizedTransitionMatrix(double ***transitionMatrix,
  * between a matrix and the a vector in a cheap way.
 */
 void matrixVectorMultiplication(double ***matrix, double *vector,
-	double **product, int vectorSize, double dampingFactor) {
+	double *convergedPagerankVector, double **product, int vectorSize,
+	double dampingFactor) {
 	double webUniformProbability = 1. / vectorSize;
 
 	for (int i=0; i<vectorSize; ++i) {
@@ -150,7 +168,7 @@ void matrixVectorMultiplication(double ***matrix, double *vector,
 		for (int j=0; j<vectorSize; ++j) {
 			sum += (*matrix)[i][j] * vector[j];
 		}
-		(*product)[i] = dampingFactor * sum;
+		(*product)[i] = dampingFactor * sum + convergedPagerankVector[i];
 	}
 
 	double normDifference = vectorNorm(vector, vectorSize) -

serial/serial_gs_pagerank_functions.h

@@ -75,7 +75,8 @@ double vectorNorm(double *vector, int vectorSize);
 // Function matrixVectorMultiplication calculates the product of the
 // multiplication between a matrix and the a vector.
 void matrixVectorMultiplication(double ***matrix, double *vector,
-	double **product, int vectorSize, double dampingFactor);
+	double *convergedPagerankVector, double **product, int vectorSize,
+	double dampingFactor);
 
 // Function pagerank iteratively calculates the pagerank of each page until
 // either the convergence criterion is met or the maximum number of iterations