Residual !!

Hallo,

I have two sparse matrix vector multiplication one is with jagged diagonal storage format another normal matrix multiplication. That is problem : "Compare the results by computing the norm of the difference of both result vectors". Result must be 10^12 - 10^15. How can I calculate residual ?

My code is: It does not work correctly. It give me always 0.

1
2
3
4
5
6
7
8
  double* k = new double[M];
	        double accum = 0.0;
    for (int i = 0; i < M; ++i) {
    	  k[i]=pow(fabs(resultVector2[i]-resultVector[i]),2);
          accum+=k[i];
    }
    double norm = sqrt(accum);
	std::cout<<"Residual : "<<norm;
In,

(int i = 0; i < M; ++i)

Maybe is should be i++ rather than ++i

There is not enough code provided there for me to test run.
Last edited on
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
#include <iostream>
#include <fstream>
#include <string> 
#include <cstdlib> 
#include <math.h>

void randomVector(int N, double* vector){
	
	for(int i = 0; i < N; i++) {
        vector[i] = ((double) rand()) / RAND_MAX;
       
    }
}

void sortZeile(int M, int* nonZeros, int* perm){
	
	for(int i = 0; i < M; i++){
		int maxIndex = i;
		for(int j = i + 1; j < M; j++){
			if(nonZeros[maxIndex] < nonZeros[j]){
				maxIndex = j;
			}
		}
		std::swap(perm[maxIndex], perm[i]);
		std::swap(nonZeros[maxIndex], nonZeros[i]);
	}
}

void fullenMatrix(int M,int L, int* nonZeros,int* perm, int* rowIndex, double* jdiag,double* matrix,int* col_ind, int* columnIndex, int* begin){
	
	int nonZerosCol = 0;
	int jdiagIndex = 0;
	for(int i = 0, l = 0, h = 0; h < nonZeros[0]; ++h, i+=M){
		for(int k = 0, m = 0; k < M; k++){
			for(int j = 0; j < L; j++){
				if(perm[k] == rowIndex[j] && m++ == l){
					jdiag[jdiagIndex] = matrix[j];
					col_ind[jdiagIndex++] = columnIndex[j];
					nonZerosCol++;
					break;
				}
			}
			m = 0;
		}
		begin[l + 1] = nonZerosCol + 1;
		l++;
	}
}

void jdsMult(int M, int* nonZeros,int* begin, double* jdiag, double* vector, int* col_ind, int* perm, double* resultVector ){
	
	for(int i = 0; i < M; i++){
		int offset = i;
		double sum = 0;
		for(int j = 0; j < nonZeros[i]; j++){			
			offset = begin[j] - 1 + i;	
			sum += jdiag[offset] * vector[col_ind[offset] - 1];
		}
		resultVector[perm[i] - 1] = sum;
	}
}

void normMult(int M, int N, double* vector, double* matrixWithZeros, double* normResult){
	
	for(int i = 0; i < M; i++){
		double sum = 0;
		for(int j = 0; j < N; j++){
			sum += vector[j] * matrixWithZeros[i*N+j];
		}
		normResult[i] = sum;
	}
}

void residual(int M, double* resultVector2, double* resultVector){

    double* k = new double[M];
	        double accum = 0.0;
    for (int i = 0; i < M; ++i) {
    	  k[i]=pow(fabs(resultVector2[i]-resultVector[i]),2);
          accum+=k[i];
    }
    double norm = sqrt(accum);
	std::cout<<"Residual : "<<norm;

}

int main() {
	
	 //Open the file:
	std::ifstream fin("mcfe.txt");
	
	// Declare variables:
	int M, N, L;
	
	// Ignore headers and comments:
	while (fin.peek() == '%') fin.ignore(2048, '\n');
	
	// Read defining parameters:
	fin >> M >> N >> L;
	
	// Create your matrix:
	double* matrix;	
	double* matrixWithZeros;
			     
	matrix = new double[L];	     
	matrixWithZeros = new double[M*N];
	std::fill(matrix, matrix + L, 0.); 
	//for(int i=0;i<L;i++){
//		std::cout<<matrix[i];
//	}
    std::cout<<std::endl;
	for(int i=0;i<M;i++){
		for(int j=0;j<N;j++){
	     matrixWithZeros[i*N+j]=0.;
		//std::cout<<matrixWithZeros[i*N+j];
		}
	}
	
	int* rowIndex = new int[L];
	int* columnIndex = new int[L];
	std::cout<<std::endl;
	
	// Read the data
	for (int l = 0; l < L; l++)
	{
		int m, n;
		double data;
		fin >> m >> n >> data;
		matrix[l] = data;
		//std::cout<<matrix[l];
		//std::cout<<std::endl;
		rowIndex[l] = m;
		columnIndex[l] = n;
		matrixWithZeros[((rowIndex[l]-1)*N)+(columnIndex[l]-1)] = matrix[l];
	//	cout<<matrixWithZeros[];
	}
	std::cout<<std::endl;
	//for(int i=0;i<M;i++){
//		for(int j=0;j<N;j++){
//			matrixWithZeros[i*N+j]=matrix[i];
//		}
//	}
	
//	for(int i=0;i<M;i++){
//		for(int j=0;j<N;j++){
//		std::cout<<matrixWithZeros[i*N+j];
//	}
//	}
		
	fin.close();

  std::cout<<std::endl;
	
	double* vector = new double[N];
	// For counting non Zero elements in Row
	int * nonZeros = new int[M]{}; 
	
	// Count non Zero elements in Rows
	for(int i = 0; i < L; i++){
		nonZeros[rowIndex[i] - 1]++;
	}
	// Perm array for keeping order after sorting
	int* perm = new int[M];
	// Initialization of perm array with values from 1..M
	for(int i = 0; i < M; i++){
		perm[i] = i + 1;
	}
	double* jdiag = new double[L];
	int* col_ind = new int[L];
	sortZeile(M,nonZeros,perm);
	// Initialization of begin array with first value 1
	int* begin = new int[nonZeros[0]];
	// NonZerosCol is used to calculate the number of non Zeros in Column

	begin[0] = 1;
    double* jdsResult = new double[M];
    double* normResult = new double[M];
	
    fullenMatrix(M,L,nonZeros,perm,rowIndex,jdiag,matrix,col_ind,columnIndex,begin);
	
 /*	std::cout << "\nJdiag:\n";
	for(int i = 0; i < L; i++){
		std::cout<<jdiag[i] << " ";
	}
	std::cout<< "\nCol_ind:\n";
	for(int i = 0; i < L; i++){
		std::cout<<col_ind[i] << " ";
	}
	std::cout<< "\nPerm:\n";
	for(int i = 0; i < M; i++){
		std::cout<< perm[i] << " ";
	}	
	std::cout<< "\nBegin:\n";
	for(int i = 0; i < nonZeros[0]; i++){
		std::cout<<begin[i] << " ";
	}*/
	
	//random vector
	randomVector(N, vector);
	
	//Matrix Vector multiplication with jds
	jdsMult(M, nonZeros,begin,jdiag,vector,col_ind,perm,jdsResult);

	// This is the output of the result vector
	//std::cout<< "\njdsResult:\n";
//	for(int i = 0; i < M; i++){
//		std::cout << jdsResult[i] << " ";
//	}

	//Normal Matrix Vector multiplication 
	normMult(M, N, vector, matrixWithZeros, normResult);
	// This is the output of the result vector
	//std::cout<< "\nnormalMatrixResult:\n";
//	for(int i = 0; i < M; i++){
//		std::cout << normResult[i] << " ";
//	}
   residual(M, normResult, jdsResult);	

    return 0;
}


There is not enough space for mcfe.txt
its look like so :
%%MatrixMarket matrix coordinate real general
765 765 24382
1 1 1.5718649900000e+03
2 1 3.8365468800000e+04
3 1 1.3600124000000e+04
4 1 3.8541311000000e+03
5 1 3.1747673300000e+03
6 1 2.5003071300000e+03
7 1 9.0686396500000e+03
8 1 1.8564418900000e+03
9 1 1.7582593800000e+04

for all values you can download from here :
http://math.nist.gov/MatrixMarket/data/Harwell-Boeing/astroph/mcfe.html
Topic archived. No new replies allowed.