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
|
#include "ekd_blas.h"
void print_vector(float* vector, uint32_t size) {
for (uint32_t i=0; i<size; i++) printf("%.2f ", vector[i]);
}
void print_matrix(float** matrix, uint32_t rows, uint32_t cols) {
for (uint32_t i=0; i<rows; i++) {
printf("\n");
for (uint32_t j=0; j<cols; j++) printf("%.2f ", matrix[i][j]);
}
}
void copy_vector(float* input_vector, float* resultant_vector, uint32_t size) {
for (uint32_t i=0; i<size; i++) resultant_vector[i] = input_vector[i];
}
void copy_matrix(float** input_matrix, float** resultant_matrix, uint32_t rows, uint32_t cols) {
for (uint32_t i=0; i<rows; i++){
for (uint32_t j=0; j<cols; j++) resultant_matrix[i][j] = input_matrix[i][j];
}
}
float trace_matrix(float** input_matrix, uint32_t size) {
float trace = 0.0f;
for (uint32_t i=0; i<size; i++) {
for (uint32_t j=0; j< size; j++) {
if (i == j) trace += input_matrix[i][j];
}
}
return trace;
}
void shift_matrix(float** input_matrix, uint32_t size, float trace, shift_matrix_operation_e operation) {
for (uint32_t i=0; i<size; i++) {
for (uint32_t j=0; j<size; j++) {
if (i == j) {
if (operation == SHIFT_ADD) {
input_matrix[i][j] += trace;
} else if (operation == SHIFT_SUBTRACT) {
input_matrix[i][j] -= trace;
}
}
}
}
}
void eye(float** input_matrix, uint32_t size) {
for (uint32_t i=0; i<size; i++) {
for (uint32_t j=0; j<size; j++) input_matrix[i][j] = (i == j) ? 1 : 0;
}
}
void outer_product(float* input_vector, uint32_t size, float** resultant_matrix) {
// for (uint32_t i=0; i<size; i++) {
// for (uint32_t j=0; j<size; j++) resultant_matrix[i][j] = input_vector[i] * input_vector[j];
// }
for (uint32_t i = 0; i < size; i++) {
for (uint32_t j = i; j < size; j++) {
float value = input_vector[i] * input_vector[j];
resultant_matrix[i][j] = value;
resultant_matrix[j][i] = value;
}
}
}
float inner_product(float* input_vector_1, float* input_vector_2, uint32_t size) {
float resultant_scalar = 0.0;
for (uint32_t i=0; i<size; i++) resultant_scalar += input_vector_1[i] * input_vector_2[i];
return resultant_scalar;
}
void transpose(float** input_matrix, uint32_t rows, uint32_t cols, float** transposed_matrix) {
for (uint32_t i=0; i<rows; i++){
for (uint32_t j=0; j<cols; j++) transposed_matrix[j][i] = input_matrix[i][j];
}
}
void matrix_operation(float** input_matrix_A, uint32_t rows_A, uint32_t cols_A, float** input_matrix_B, uint32_t rows_B, uint32_t cols_B, float** resultant_matrix, matrix_matrix_operation_e operation) {
if (operation == MM_ADD) {
for (uint32_t i=0; i<rows_A; i++) {
for (uint32_t j=0; j<cols_A; j++) resultant_matrix[i][j] = input_matrix_A[i][j] + input_matrix_B[i][j];
}
} else if (operation == MM_SUBTRACT) {
for (uint32_t i=0; i<rows_A; i++) {
for (uint32_t j=0; j<cols_A; j++) resultant_matrix[i][j] = input_matrix_A[i][j] - input_matrix_B[i][j];
}
} else if (operation == MM_MULTIPLY) {
for (uint32_t i=0; i<rows_A; i++) {
for (uint32_t j=0; j<cols_B; j++) {
resultant_matrix[i][j] = 0;
for (uint32_t k=0; k<cols_A; k++) resultant_matrix[i][j] += input_matrix_A[i][k] * input_matrix_B[k][j];
}
}
} else if (operation == MM_MULTIPLY_OMP) {
uint32_t a, b, c;
#pragma omp parallel for private(a, b, c) shared(input_matrix_A, input_matrix_B, resultant_matrix)
for (a=0; a<rows_A; a++) {
for (b=0; b<cols_B; b++) {
resultant_matrix[a][b] = 0;
for (c=0; c<cols_A; c++) resultant_matrix[a][b] += input_matrix_A[a][c] * input_matrix_B[c][b];
}
}
} else if (operation == MM_MULTIPLY_GEMM) {
float* temp_matrix_A = malloc(rows_A * cols_A * sizeof(float));
float* temp_matrix_B = malloc(rows_B * cols_B * sizeof(float));
float* temp_resultant_matrix = malloc(rows_A * cols_B * sizeof(float));
for (uint32_t i = 0; i < rows_A; i++) {
for (uint32_t j = 0; j < cols_A; j++) {
temp_matrix_A[i * cols_A + j] = input_matrix_A[i][j];
}
}
for (uint32_t i = 0; i < rows_B; i++) {
for (uint32_t j = 0; j < cols_B; j++) {
temp_matrix_B[i * cols_B + j] = input_matrix_B[i][j];
}
}
cblas_sgemm(CblasRowMajor, CblasNoTrans, CblasNoTrans, rows_A, cols_B, cols_A, 1.0, temp_matrix_A, cols_A, temp_matrix_B, cols_B, 0.0, temp_resultant_matrix, cols_B);
for (uint32_t i = 0; i < rows_A; i++) {
for (uint32_t j = 0; j < cols_B; j++) {
resultant_matrix[i][j] = temp_resultant_matrix[i * cols_B + j];
}
}
free(temp_matrix_A);
free(temp_matrix_B);
free(temp_resultant_matrix);
}
}
uint8_t sign(float num) {
return (num > 0) - (num < 0);
}
void vector_normalize(float* vector, uint32_t size) {
float norm = 0.0;
for (uint32_t i=0; i<size; i++) norm += vector[i] * vector[i];
norm = sqrt(norm);
for (uint32_t i=0; i<size; i++) vector[i] = (norm == 0) ? 0 : vector[i]/norm;
}
float vector_norm(float* vector, uint32_t size) {
float norm = 0.0;
for (uint32_t i=0; i<size; i++) norm += vector[i] * vector[i];
return sqrt(norm);
}
float matrix_norm(float** input_matrix_A1, float** input_matrix_A, uint32_t rows, uint32_t cols) {
float norm = 0.0;
for (uint32_t i=0; i<rows; i++) {
for (uint32_t j=0; j<cols; j++) norm += fabs(input_matrix_A1[i][j] - input_matrix_A[i][j]) * fabs(input_matrix_A1[i][j] - input_matrix_A[i][j]);
}
return sqrt(norm);
}
float cosine_similarity(float* curr_vector, float* new_vector, uint32_t size) {
float dot_product = 0.0f;
for (uint32_t i=0; i<size; i++) dot_product += curr_vector[i] * new_vector[i];
return (vector_norm(curr_vector, size) == 0 || vector_norm(new_vector, size) == 0) ? 1e-6 : dot_product / (vector_norm(curr_vector, size) * vector_norm(new_vector, size));
}
void matrix_vector_multiply(float** input_matrix, uint32_t rows, uint32_t cols, float* input_vector, float* resultant_vector, matrix_vector_operation_e operation) {
if (operation == MV_MULTIPLY) {
for (uint32_t i=0; i<rows; i++) {
resultant_vector[i] = 0;
for (uint32_t j=0; j<cols; j++) {
resultant_vector[i] += input_matrix[i][j] * input_vector[j];
}
}
} else if (operation == MV_MULTIPLY_GEMV) {
float* temp_matrix = malloc(rows * cols * sizeof(float));
for (uint32_t i = 0; i < rows; i++) {
for (uint32_t j = 0; j < cols; j++) {
temp_matrix[i * cols + j] = input_matrix[i][j];
}
}
cblas_sgemv(CblasRowMajor, CblasNoTrans, rows, cols, 1.0f, temp_matrix, cols, input_vector, 1, 0.0f, resultant_vector, 1);
free(temp_matrix);
}
}
|
