-
Notifications
You must be signed in to change notification settings - Fork 0
/
toygenerator.cu
executable file
·207 lines (194 loc) · 9.5 KB
/
toygenerator.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
/* CUDA toy
* Thong Nguyen, 2020 */
#include <cstdio>
#include <cuda_runtime.h>
#include <curand_kernel.h>
#include "toygenerator.cuh"
__global__
void generate_goodness_of_fit_toys(float * dev_bkg_expected,
float * dev_obs_data,
float * dev_q_toys,
int n_bins,
unsigned long ntoys,
curandState *states,
unsigned int nStates)
{
int toy;
unsigned long tid = threadIdx.x + blockDim.x * blockIdx.x;
float sum_log_likelihood;
while (tid < ntoys)
{
curand_init((unsigned long long)clock() + tid, 0, 0, &states[tid % nStates]);
sum_log_likelihood = 0;
for (int bin = 0; bin < n_bins; bin++)
{
toy = curand_poisson(&states[tid % nStates], dev_bkg_expected[bin]);
sum_log_likelihood += chisquare(dev_bkg_expected[bin], toy);
}
dev_q_toys[tid] = sum_log_likelihood;
tid += blockDim.x * gridDim.x;
}
}
__global__
void count_extreme_goodness_of_fit(float * dev_bkg_expected,
float * dev_obs_data,
unsigned int * dev_larger_gpu,
float q_obs,
int n_bins,
unsigned long ntoys,
curandState *states,
unsigned int nStates)
{
int toy;
unsigned long tid = threadIdx.x + blockDim.x * blockIdx.x;
float sum_log_likelihood;
while (tid < ntoys)
{
curand_init((unsigned long long)clock() + tid, 0, 0, &states[tid % nStates]);
sum_log_likelihood = 0;
for (int bin = 0; bin < n_bins; bin++)
{
toy = curand_poisson(&states[tid % nStates], dev_bkg_expected[bin]);
sum_log_likelihood += chisquare(dev_bkg_expected[bin], toy);
}
if (sum_log_likelihood > q_obs) atomicAdd(dev_larger_gpu, (unsigned int) 1);
tid += blockDim.x * gridDim.x;
}
}
__global__
void generate_neyman_pearson_toys(float * dev_bkg_expected,
float * dev_sig_expected,
float * dev_obs_data,
float * dev_q_toys,
int n_bins,
unsigned long ntoys,
curandState *states,
unsigned int nStates)
{
int toy;
unsigned long tid = (threadIdx.x + blockDim.x * blockIdx.x);
float sum_log_likelihood, numerator, denominator;
while (tid < ntoys)
{
curand_init((unsigned long long)clock() + tid, 0, 0, &states[tid % nStates]);
sum_log_likelihood = 0;
for (int bin = 0; bin < n_bins; bin++)
{
toy = curand_poisson(&states[tid % nStates], dev_bkg_expected[bin]);
numerator = log_poisson(dev_bkg_expected[bin]+dev_sig_expected[bin], toy);
denominator = log_poisson(dev_bkg_expected[bin], toy);
sum_log_likelihood += 2 * (numerator-denominator);
}
dev_q_toys[tid] = sum_log_likelihood;
tid += (blockDim.x * gridDim.x);
}
}
__global__
void count_extreme_neyman_pearson(float * dev_bkg_expected,
float * dev_sig_expected,
float * dev_obs_data,
unsigned int * dev_larger_gpu,
float q_obs,
int n_bins,
unsigned long ntoys,
curandState *states,
unsigned int nStates)
{
int toy;
unsigned long tid = (threadIdx.x + blockDim.x * blockIdx.x);
float sum_log_likelihood, numerator, denominator;
while (tid < ntoys)
{
curand_init((unsigned long long)clock() + tid, 0, 0, &states[tid % nStates]);
sum_log_likelihood = 0;
for (int bin = 0; bin < n_bins; bin++)
{
toy = curand_poisson(&states[tid % nStates], dev_bkg_expected[bin]);
numerator = log_poisson(dev_bkg_expected[bin]+dev_sig_expected[bin], toy);
denominator = log_poisson(dev_bkg_expected[bin], toy);
sum_log_likelihood += 2 * (numerator-denominator);
}
if (sum_log_likelihood > q_obs) atomicAdd(dev_larger_gpu, (unsigned int) 1);
tid += (blockDim.x * gridDim.x);
}
}
void cuda_call_generate_goodness_of_fit_toys(unsigned int nBlocks,
int threadsPerBlock,
float * dev_bkg_expected,
float * dev_obs_data,
float * dev_q_toys,
int n_bins,
unsigned long ntoys,
curandState * devStates,
unsigned int nStates)
{
generate_goodness_of_fit_toys<<<nBlocks, threadsPerBlock>>>(dev_bkg_expected,
dev_obs_data,
dev_q_toys,
n_bins,
ntoys,
devStates,
nStates);
}
void cuda_call_count_extreme_goodness_of_fit(unsigned int nBlocks,
int threadsPerBlock,
float * dev_bkg_expected,
float * dev_obs_data,
unsigned int * dev_larger_gpu,
float q_obs,
int n_bins,
unsigned long ntoys,
curandState * devStates,
unsigned int nStates)
{
count_extreme_goodness_of_fit<<<nBlocks, threadsPerBlock>>>(dev_bkg_expected,
dev_obs_data,
dev_larger_gpu,
q_obs,
n_bins,
ntoys,
devStates,
nStates);
}
void cuda_call_generate_neyman_pearson_toys(unsigned int nBlocks,
int threadsPerBlock,
float * dev_bkg_expected,
float * dev_sig_expected,
float * dev_obs_data,
float * dev_q_toys,
int n_bins,
unsigned long ntoys,
curandState * devStates,
unsigned int nStates)
{
generate_neyman_pearson_toys<<<nBlocks, threadsPerBlock>>>(dev_bkg_expected,
dev_sig_expected,
dev_obs_data,
dev_q_toys,
n_bins,
ntoys,
devStates,
nStates);
}
void cuda_call_count_extreme_neyman_pearson(unsigned int nBlocks,
int threadsPerBlock,
float * dev_bkg_expected,
float * dev_sig_expected,
float * dev_obs_data,
unsigned int * dev_larger_gpu,
float q_obs,
int n_bins,
unsigned long ntoys,
curandState * devStates,
unsigned int nStates)
{
count_extreme_neyman_pearson<<<nBlocks, threadsPerBlock>>>(dev_bkg_expected,
dev_sig_expected,
dev_obs_data,
dev_larger_gpu,
q_obs,
n_bins,
ntoys,
devStates,
nStates);
}