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
|
// BSD 3-Clause License; see https://github.com/scikit-hep/awkward-1.0/blob/main/LICENSE
#define FILENAME(line) \
FILENAME_FOR_EXCEPTIONS_CUDA("src/cuda-kernels/awkward_reduce_argmax.cu", \
line)
#include "standard_parallel_algorithms.h"
#include "awkward/kernels.h"
template <typename OUT, typename IN>
__global__ void
awkward_reduce_argmax_kernel(OUT* toptr,
const IN* fromptr,
const int64_t* parents,
int64_t lenparents) {
int64_t thread_id = blockIdx.x * blockDim.x + threadIdx.x;
if (thread_id < lenparents) {
int64_t parent = parents[thread_id];
if (toptr[parent] == -1 || (fromptr[thread_id] > (fromptr[toptr[parent]]))) {
toptr[parent] = thread_id;
}
}
}
template <typename OUT, typename IN>
ERROR
awkward_reduce_argmax(OUT* toptr,
const IN* fromptr,
const int64_t* parents,
int64_t lenparents,
int64_t outlength) {
HANDLE_ERROR(cudaMemset(toptr, -1, sizeof(OUT) * outlength));
dim3 blocks_per_grid = blocks(lenparents);
dim3 threads_per_block = threads(lenparents);
awkward_reduce_argmax_kernel<<<blocks_per_grid, threads_per_block>>>(
toptr, fromptr, parents, lenparents);
return success();
}
ERROR
awkward_reduce_argmax_int8_64(int64_t* toptr,
const int8_t* fromptr,
const int64_t* parents,
int64_t lenparents,
int64_t outlength) {
return awkward_reduce_argmax<int64_t, int8_t>(
toptr, fromptr, parents, lenparents, outlength);
}
ERROR
awkward_reduce_argmax_uint8_64(int64_t* toptr,
const uint8_t* fromptr,
const int64_t* parents,
int64_t lenparents,
int64_t outlength) {
return awkward_reduce_argmax<int64_t, uint8_t>(
toptr, fromptr, parents, lenparents, outlength);
}
ERROR
awkward_reduce_argmax_int16_64(int64_t* toptr,
const int16_t* fromptr,
const int64_t* parents,
int64_t lenparents,
int64_t outlength) {
return awkward_reduce_argmax<int64_t, int16_t>(
toptr, fromptr, parents, lenparents, outlength);
}
ERROR
awkward_reduce_argmax_uint16_64(int64_t* toptr,
const uint16_t* fromptr,
const int64_t* parents,
int64_t lenparents,
int64_t outlength) {
return awkward_reduce_argmax<int64_t, uint16_t>(
toptr, fromptr, parents, lenparents, outlength);
}
ERROR
awkward_reduce_argmax_int32_64(int64_t* toptr,
const int32_t* fromptr,
const int64_t* parents,
int64_t lenparents,
int64_t outlength) {
return awkward_reduce_argmax<int64_t, int32_t>(
toptr, fromptr, parents, lenparents, outlength);
}
ERROR
awkward_reduce_argmax_uint32_64(int64_t* toptr,
const uint32_t* fromptr,
const int64_t* parents,
int64_t lenparents,
int64_t outlength) {
return awkward_reduce_argmax<int64_t, uint32_t>(
toptr, fromptr, parents, lenparents, outlength);
}
ERROR
awkward_reduce_argmax_int64_64(int64_t* toptr,
const int64_t* fromptr,
const int64_t* parents,
int64_t lenparents,
int64_t outlength) {
return awkward_reduce_argmax<int64_t, int64_t>(
toptr, fromptr, parents, lenparents, outlength);
}
ERROR
awkward_reduce_argmax_uint64_64(int64_t* toptr,
const uint64_t* fromptr,
const int64_t* parents,
int64_t lenparents,
int64_t outlength) {
return awkward_reduce_argmax<int64_t, uint64_t>(
toptr, fromptr, parents, lenparents, outlength);
}
ERROR
awkward_reduce_argmax_float32_64(int64_t* toptr,
const float* fromptr,
const int64_t* parents,
int64_t lenparents,
int64_t outlength) {
return awkward_reduce_argmax<int64_t, float>(
toptr, fromptr, parents, lenparents, outlength);
}
ERROR
awkward_reduce_argmax_float64_64(int64_t* toptr,
const double* fromptr,
const int64_t* parents,
int64_t lenparents,
int64_t outlength) {
return awkward_reduce_argmax<int64_t, double>(
toptr, fromptr, parents, lenparents, outlength);
}
|
