-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathpoint.cu
358 lines (340 loc) · 14.6 KB
/
point.cu
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
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
#ifdef __cplusplus
extern "C++"{
#endif // __cplusplus
#include "point.h"
Point::Point(){
num = 0;
std::cout<<"构造函数"<<std::endl;
}
Point::Point(std::string fileName){
std::ifstream fid(fileName);
num = 300000000;
cudaHostAlloc((void**)&pos_cpu, num * 4 * sizeof(float), cudaHostAllocDefault);
std::string line;
float x;
for(int i = 0; i < num; i++) {
getline(fid, line);
std::istringstream s(line);
for(int j = 0; j < 3; j++) {
s >> x;
pos_cpu[j*num + i] = x;
}
}
cudaMalloc((void**)&pos_gpu, num * 4 * sizeof(float));
cudaMemcpy(pos_gpu, pos_cpu, num * 4 * sizeof(float), cudaMemcpyHostToDevice);
}
Point::~Point(){
num = -1;
cudaFreeHost(pos_cpu);
cudaFree(pos_gpu);
cudaFreeHost(index_cpu);
cudaFree(index_gpu);
cudaFreeHost(distance_cpu);
cudaFree(distance_gpu);
cudaFreeHost(sum_cpu);
cudaFree(sum_gpu);
cudaFreeHost(sum_out_cpu);
cudaFree(sum_out_gpu);
cudaFreeHost(tile_index_with_point_cpu);
cudaFree(tile_index_with_point_gpu);
cudaFreeHost(point_index_in_tile_cpu);
cudaFree(point_index_in_tile_gpu);
cudaFreeHost(point_index_cpu);
cudaFree(point_index_gpu);
std::cout<<"析构函数"<<std::endl;
}
__global__ void set_tile_kernel(){}
int Point::set_tile(int x_n, int y_n, int z_n){
xn = x_n;
yn = y_n;
zn = z_n;
grid_num = xn*yn*zn*23*23*23;
tile_num = xn*yn*zn;
xl = (x_end - x_start) / xn;
yl = (y_end - y_start) / yn;
zl = (z_end - z_start) / zn;
gxl = (x_end - x_start) / (xn * 23);
gyl = (y_end - y_start) / (yn * 23);
gzl = (z_end - z_start) / (zn * 23);
// set grid memory
cudaHostAlloc((void**)&index_cpu, grid_num * sizeof(uint), cudaHostAllocDefault);
cudaMalloc((void**)&index_gpu, grid_num * sizeof(uint));
cudaHostAlloc((void**)&distance_cpu, grid_num * sizeof(int), cudaHostAllocDefault);
cudaMalloc((void**)&distance_gpu, grid_num * sizeof(int));
cudaHostAlloc((void**)&sum_cpu, 4096 * sizeof(uint), cudaHostAllocDefault);
cudaMalloc((void**)&sum_gpu, 4096 * sizeof(uint));
cudaHostAlloc((void**)&sum_out_cpu, 4096 * sizeof(uint), cudaHostAllocDefault);
cudaMalloc((void**)&sum_out_gpu, 4096 * sizeof(uint));
// 记录每个点，在每个tile的顺序
cudaHostAlloc((void**)&tile_index_with_point_cpu, num * sizeof(uint), cudaHostAllocDefault);
cudaMalloc((void**)&tile_index_with_point_gpu, num * sizeof(uint));
cudaHostAlloc((void**)&point_index_in_tile_cpu, num * sizeof(uint), cudaHostAllocDefault);
cudaMalloc((void**)&point_index_in_tile_gpu, num * sizeof(uint));
cudaHostAlloc((void**)&point_index_cpu, num * sizeof(uint), cudaHostAllocDefault);
cudaMalloc((void**)&point_index_gpu, num * sizeof(uint));
for(int i = 0; i < grid_num; i++){
distance_cpu[i] = 2e9;
}
cudaMemcpy(distance_gpu, distance_cpu, grid_num * sizeof(int), cudaMemcpyHostToDevice);
for(int i = 0; i < 4096; i++){
sum_cpu[i] = 0;
//sum_out_cpu[i] = 1;
}
cudaMemcpy(sum_gpu, sum_cpu, 4096 * sizeof(uint), cudaMemcpyHostToDevice);
//cudaMemcpy(sum_out_gpu, sum_out_cpu, 4096 * sizeof(uint), cudaMemcpyHostToDevice);
return 1;
}
__global__ void sample_kernel(uint *index, int *distance, float *pos, int num,
float x_start, float y_start, float z_start,
float gxl, float gyl, float gzl){
int tid = blockIdx.x * blockDim.x + threadIdx.x;
for(int i = tid; i < num; i += gridDim.x * blockDim.x){
float x = (pos[i] - x_start) / gxl;
float y = (pos[i+num] - y_start) / gyl;
float z = (pos[i+2*num] - z_start) / gzl;
uint x_index = uint(x);
uint y_index = uint(y);
uint z_index = uint(z);
uint grid_index = x_index*360*360 + y_index*360 + z_index;
float grid_distance = (x-x_index)*(x-x_index) + (y-y_index)*(y_index) + (z-z_index)*(z-z_index);
int grid_distance_int = __float_as_int(grid_distance);
atomicMin(&distance[grid_index], grid_distance_int);
__syncthreads();
if(grid_distance_int == distance[grid_index]){
index[grid_index] = i;
}
}
}
int Point::sample(){
sample_kernel<<<1024, 1024>>>(index_gpu, distance_gpu, pos_gpu, num,
x_start, y_start, z_start, gxl, gyl, gzl);
return 1;
}
__global__ void scan(uint *g_odata, uint *g_idata, int n){
extern __shared__ uint temp[]; // allocated on invocation
int thid = threadIdx.x;
int offset = 1;
uint first_value = 0, last_value;
for(int i = 0; i < n; i += 2048){
temp[2*thid] = g_idata[2*thid + i]; // load input into shared memory
temp[2*thid+1] = g_idata[2*thid+1 + i];
if (thid == 0) {
temp[0] += first_value;
}
for (int d = 2048>>1; d > 0; d >>= 1){
__syncthreads();
if (thid < d){
int ai = offset*(2*thid+1)-1;
int bi = offset*(2*(thid+1))-1;
temp[bi] += temp[ai];
}
offset *= 2;
}
if (thid == 0) {
last_value = temp[2048 - 1];
temp[2048 - 1] = 0;
}
for (int d = 1; d < 2048; d *= 2){
offset >>= 1;
__syncthreads();
if (thid < d){
int ai = offset*(2*thid+1)-1;
int bi = offset*(2*(thid+1))-1;
uint t = temp[ai];
temp[ai] = temp[bi];
temp[bi] += t;
}
}
__syncthreads();
g_odata[2*thid + i] = temp[2*thid]; // write results to device memory
g_odata[2*thid+1 + i] = temp[2*thid+1];
if (thid == 0) {
g_odata[i] = first_value;
first_value = last_value;
}
}
}
__global__ void count_kernel(uint *index, int *distance, uint *sum_gpu,
uint *sum_out_gpu, uint *tile_index_with_point_gpu,
uint *point_index_in_tile_gpu, float *pos, int num,
float x_start, float y_start, float z_start,
float gxl, float gyl, float gzl){
int tid = blockIdx.x * blockDim.x + threadIdx.x;
for(int i = tid; i < num; i += gridDim.x * blockDim.x){
float x = (pos[i] - x_start) / (gxl*23);
float y = (pos[i+num] - y_start) / (gyl*23);
float z = (pos[i+2*num] - z_start) / (gzl*23);
uint x_index = uint(x);
uint y_index = uint(y);
uint z_index = uint(z);
uint tile_index = x_index*16*16 + y_index*16 + z_index;
uint point_index_in_tile = atomicAdd(&sum_gpu[tile_index], 1);
tile_index_with_point_gpu[i] = tile_index;
point_index_in_tile_gpu[i] = point_index_in_tile;
}
//__syncthreads();
//__threadfence();
//scan<<<1, 1024, 2048 * 4, cudaStreamTailLaunch>>>(sum_out_gpu, sum_gpu, 8192);
//if(blockIdx.x == 0){
// extern __shared__ uint temp[2048]; // allocated on invocation
// int n = 8192;
// int thid = threadIdx.x;
// int offset = 1;
// uint first_value = 0, last_value;
// for(int i2 = 0; i2 < n; i2 += 2048){
// // temp[2*thid] = 0; // load input into shared memory
// temp[2*thid] = sum_gpu[2*thid + i2]; // load input into shared memory
// temp[2*thid+1] = sum_gpu[2*thid+1 + i2];
// if (thid == 0) {
// temp[0] += first_value;
// }
// for (int d = 2048>>1; d > 0; d >>= 1){
// __syncthreads();
// if (thid < d){
// int ai = offset*(2*thid+1)-1;
// int bi = offset*(2*(thid+1))-1;
// temp[bi] += temp[ai];
// }
// offset *= 2;
// }
// if (thid == 0) {
// last_value = temp[2048 - 1];
// temp[2048 - 1] = 0;
// }
// for (int d = 1; d < 2048; d *= 2){
// offset >>= 1;
// __syncthreads();
// if (thid < d){
// int ai = offset*(2*thid+1)-1;
// int bi = offset*(2*(thid+1))-1;
// uint t = temp[ai];
// temp[ai] = temp[bi];
// temp[bi] += t;
// }
// }
// __syncthreads();
// sum_out_gpu[2*thid + i2] = temp[2*thid]; // write results to device memory
// sum_out_gpu[2*thid+1 + i2] = temp[2*thid+1];
// if (thid == 0) {
// sum_out_gpu[i2] = first_value;
// first_value = last_value;
// }
// }
//}
}
__global__ void sort_index_kernel(uint *point_index_gpu,
float *pos_gpu,
uint *tile_index_with_point_gpu,
uint *point_index_in_tile_gpu,
uint *sum_out_gpu, int num){
int tid = blockIdx.x * blockDim.x + threadIdx.x;
for(int i = tid; i < num; i += gridDim.x * blockDim.x){
point_index_gpu[sum_out_gpu[tile_index_with_point_gpu[i]] + point_index_in_tile_gpu[i]] = i;
//pos_gpu[sum_out_gpu[tile_index_with_point_gpu[i]] + point_index_in_tile_gpu[i] + 3*num] = pos_gpu[i+2*num];
//pos_gpu[sum_out_gpu[tile_index_with_point_gpu[i]] + point_index_in_tile_gpu[i] + 2*num] = pos_gpu[i+num];
//pos_gpu[sum_out_gpu[tile_index_with_point_gpu[i]] + point_index_in_tile_gpu[i] + num] = pos_gpu[i];
}
for(int i = tid; i < num; i += gridDim.x * blockDim.x){
uint index = point_index_gpu[i];
pos_gpu[i + 3*num] = pos_gpu[index+2*num];
pos_gpu[i + 2*num] = pos_gpu[index+1*num];
pos_gpu[i + 1*num] = pos_gpu[index+0*num];
}
//for(int i = tid; i < num; i += gridDim.x * blockDim.x){
// pos_gpu[sum_out_gpu[tile_index_with_point_gpu[i]] + point_index_in_tile_gpu[i] + 2*num] = pos_gpu[i+num];
//}
//for(int i = tid; i < num; i += gridDim.x * blockDim.x){
// pos_gpu[sum_out_gpu[tile_index_with_point_gpu[i]] + point_index_in_tile_gpu[i] + num] = pos_gpu[i];
//}
}
__global__ void sample_point_kernel(uint *index_gpu, uint *point_index_gpu,
uint *tile_index_with_point_gpu,
uint *point_index_in_tile_gpu,
uint *sum_gpu, uint *sum_out_gpu,
float *pos_gpu, int num,
float x_start, float y_start, float z_start,
float gxl, float gyl, float gzl){
extern __shared__ uint distance_sm[]; // allocated on invocation
for (uint i = blockIdx.x; i < 4096; i += gridDim.x){
uint point_num_in_tile = sum_gpu[i];
uint base_index = sum_out_gpu[i];
for (uint j = threadIdx.x; j < 12167; j += blockDim.x) {
distance_sm[j] = 2e9;
}
//__syncthreads();
for (uint j = threadIdx.x; j < point_num_in_tile; j += blockDim.x){
//uint index = point_index_gpu[base_index + j];
uint index = base_index + j;
float x = (pos_gpu[index] - x_start) / gxl;
float y = (pos_gpu[index+num] - y_start) / gyl;
float z = (pos_gpu[index+2*num] - z_start) / gzl;
uint x_index = uint(x) % 23;
uint y_index = uint(y) % 23;
uint z_index = uint(z) % 23;
uint grid_index = x_index*23*23 + y_index*23 + z_index;
float grid_distance = (x-x_index)*(x-x_index) + (y-y_index)*(y_index) + (z-z_index)*(z-z_index);
uint grid_distance_int = __float_as_uint(grid_distance);
atomicMin(&distance_sm[grid_index], grid_distance_int);
//tile_index_with_point_gpu[base_index + j] = grid_index;
//point_index_in_tile_gpu[base_index + j] = grid_distance_int;
//__syncthreads();
if(grid_distance_int == distance_sm[grid_index]){
index_gpu[i*12167 + grid_index] = index;
}
}
//__syncthreads();
//for (uint j = threadIdx.x; j < point_num_in_tile; j += blockDim.x){
// uint index = point_index_gpu[base_index + j];
// uint grid_index = tile_index_with_point_gpu[base_index + j];
// uint grid_distance_int = point_index_in_tile_gpu[base_index + j];
// if(grid_distance_int == distance_sm[grid_index]){
// index_gpu[i*5832 + grid_index] = index;
// }
//}
}
}
int Point::sample_2(){
count_kernel<<<1024, 1024>>>(index_gpu, distance_gpu, sum_gpu, sum_out_gpu,
tile_index_with_point_gpu, point_index_in_tile_gpu, pos_gpu, num,
x_start, y_start, z_start, gxl, gyl, gzl);
scan<<<1, 1024, 2048 * sizeof(uint)>>>(sum_out_gpu, sum_gpu, 4096);
sort_index_kernel<<<1024, 1024>>>(point_index_gpu, pos_gpu, tile_index_with_point_gpu, point_index_in_tile_gpu, sum_out_gpu, num);
//scan<<<1, 1024, 2048 * 4, cudaStreamTailLaunch>>>(sum_out_gpu, sum_gpu, 8192);
sample_point_kernel<<<1024, 1024, 12167 * sizeof(uint)>>>(index_gpu, point_index_gpu, tile_index_with_point_gpu,
point_index_in_tile_gpu, sum_gpu, sum_out_gpu,
pos_gpu+num, num, x_start, y_start, z_start,
gxl, gyl, gzl);
cudaMemcpy(sum_out_cpu, sum_out_gpu, 4096 * sizeof(uint), cudaMemcpyDeviceToHost);
for(int i = 0; i < 4096; i++){
std::cout<<i<<":"<<sum_out_cpu[i]<<" ";
}
std::cout<<std::endl;
cudaMemcpy(point_index_cpu, point_index_gpu, 4096 * sizeof(uint), cudaMemcpyDeviceToHost);
for(int i = 0; i < 4096; i++){
std::cout<<i<<":"<<point_index_cpu[i]<<" ";
}
std::cout<<std::endl;
return 1;
}
//__global__ void sample_3_kernel(uint *index, int *distance, uint *sum_gpu, uint *sum_out_gpu, float *pos, int num,
// float x_start, float y_start, float z_start,
// float gxl, float gyl, float gzl){
// extern __shared__ uint temp[2048]; // allocated on invocation
// int tid = blockIdx.x * blockDim.x + threadIdx.x;
// for(int i = tid; i < num; i += gridDim.x * blockDim.x){
// float x = (pos[i] - x_start) / (gxl*18);
// float y = (pos[i+num] - y_start) / (gyl*18);
// float z = (pos[i+2*num] - z_start) / (gzl*18);
// uint x_index = uint(x);
// uint y_index = uint(y);
// uint z_index = uint(z);
// uint tile_index = x_index*20*20 + y_index*20 + z_index;
// uint point_index_in_tile = atomicAdd(&sum_gpu[tile_index], 1);
// __syncthreads();
//}
int Point::sample_3(){
return 1;
}
#ifdef __cplusplus
};
#endif // __cplusplus