gpu-desktop-calculator/verb.cu at master · maya-undefined/gpu-desktop-calculator · GitHub

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
#include "verb.h"
#include "kernels.h"

Verb::Verb(std::string file1, std::string outputFile) {
       host_A_file = new FH(file1);
       outputFilePtr = new std::ofstream(outputFile);

}

Verb::Verb(std::string file1, std::string file2, std::string outputFile) {
       host_A_file = new FH(file1);
       host_B_file = new FH(file2);
       outputFilePtr =  new std::ofstream(outputFile);
}

void Verb::dispatch() {
       if (host_B_file == nullptr) {

              size_t loops = 0;
              while (!host_A_file->eof()) {
                     A_rows = host_A_file->row_len();
                     // Remember how many rows we read so far

                     std::vector<float> host_A = host_A_file->read_data_from_file();

                     A_rows = host_A_file->row_len() - A_rows;
                     // and now we can calculate how many rows we need to process in this chunk

                     if (A_cols != host_A_file->col_len()) { A_cols = host_A_file->col_len(); }

                     cudaMalloc((void **)&device_C, host_A.size() * sizeof(float));
                     cudaMalloc((void **)&device_A, host_A.size() * sizeof(float));
                     cudaMemcpy(device_A, host_A.data(), host_A.size() * sizeof(float), cudaMemcpyHostToDevice);

                     execute();

                     // Copy result back to host
                     // we only need to keep track of how many elements since we are using a flat array
                     std::vector<float> host_C(A_rows);
                     cudaMemcpy(host_C.data(), device_C, A_rows * sizeof(float), cudaMemcpyDeviceToHost);

                     for (float value : host_C) {
                            *outputFilePtr << std::fixed << std::setprecision(6) << value << "\n";
                     }
                     outputFilePtr->flush();
                     loops++;
                     cudaFree(device_A);
                     cudaFree(device_C);
              } // while
       } else {

              size_t loops = 0;
              while (!host_A_file->eof()) {
                     A_rows = host_A_file->row_len();
                     B_rows = host_B_file->row_len();
                     // Remember how many rows we read so far

                     std::vector<float> host_A = host_A_file->read_data_from_file();
                     std::vector<float> host_B = host_B_file->read_data_from_file();

                     A_rows = host_A_file->row_len() - A_rows;
                     B_rows = host_B_file->row_len() - B_rows;
                     // and now we can calculate how many rows we need to process in this chunk

                     if (B_cols != host_B_file->col_len()) { B_cols = host_B_file->col_len(); }
                     if (A_cols != host_A_file->col_len()) { A_cols = host_A_file->col_len(); }

                     cudaMalloc((void **)&device_C, host_A.size() * sizeof(float));
                     cudaMalloc((void **)&device_A, host_A.size() * sizeof(float));
                     cudaMalloc((void **)&device_B, host_B.size() * sizeof(float));
                     cudaMemcpy(device_A, host_A.data(), host_A.size() * sizeof(float), cudaMemcpyHostToDevice);
                     cudaMemcpy(device_B, host_B.data(), host_B.size() * sizeof(float), cudaMemcpyHostToDevice);

                     execute();

                     // Copy result back to host
                     // we only need to keep track of how many elements since we are using a flat array
                     size_t ele_to_read = max(A_rows, B_rows);
                     std::vector<float> host_C(ele_to_read);
                     cudaMemcpy(host_C.data(), device_C, ele_to_read * sizeof(float), cudaMemcpyDeviceToHost);

                     for (float value : host_C) {
                            *outputFilePtr << std::fixed << std::setprecision(6) << value << "\n";
                     }
                     outputFilePtr->flush();
                     loops++;
                     cudaFree(device_A);
                     cudaFree(device_B);
                     cudaFree(device_C);
              } // while
       } // if 1 file or 2 file operations
}

void Add::execute() {
       dim3 blockSize(256);
       dim3 gridSize((A_rows + blockSize.x - 1) / blockSize.x);
       addMultipleArrays<<<gridSize, blockSize>>>(
              device_A, device_B, device_C,
              A_rows, B_rows, // rows
              A_cols, B_cols // columns
       );

}

void Exp::execute() {
       dim3 blockSize(256);
       dim3 gridSize((A_rows + blockSize.x - 1) / blockSize.x);
       expArrays<<<gridSize, blockSize>>>(
              device_A, device_C,
              A_rows // rows
       );

}

void Div::execute() {
       dim3 blockSize(256);
       dim3 gridSize((A_rows + blockSize.x - 1) / blockSize.x);
       divArrays<<<gridSize, blockSize>>>(
              device_A, device_B, device_C,
              A_rows, B_rows // rows
       );
}

void Mul::execute() {
       dim3 blockSize(256);
       dim3 gridSize((A_rows + blockSize.x - 1) / blockSize.x);
       mulMultipleArrays<<<gridSize, blockSize>>>(
              device_A, device_B, device_C,
              A_rows, B_rows, // rows
              A_cols, B_cols // columns
       );
}

void Sin::execute() {

}

void Cos::execute() {

}

void Tan::execute() {

}