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
|
#define FILENAME(line) FILENAME_FOR_EXCEPTIONS_CUDA("src/cuda-kernels/manual_awkward_UnionArray_project.cu", line)
#include "awkward/kernels.h"
#include "standard_parallel_algorithms.h"
template <typename C>
__global__ void
awkward_UnionArray_project_filter_mask(
const C* fromtags,
int64_t which,
int8_t* filtered_mask,
int64_t length) {
int64_t thread_id = blockIdx.x * blockDim.x + threadIdx.x;
if(thread_id < length) {
if (fromtags[thread_id] == which) {
filtered_mask[thread_id] = 1;
}
}
}
template <typename T, typename C, typename I>
__global__ void
awkward_UnionArray_project_kernel(
int64_t* prefixedsum_mask,
T* tocarry,
const C* fromtags,
const I* fromindex,
int64_t length,
int64_t which) {
int64_t thread_id = blockIdx.x * blockDim.x + threadIdx.x;
if(thread_id < length) {
if (fromtags[thread_id] == which) {
tocarry[prefixedsum_mask[thread_id] - 1] = fromindex[thread_id];
}
}
}
template <typename T, typename C, typename I>
ERROR awkward_UnionArray_project(
int64_t* lenout,
T* tocarry,
const C* fromtags,
const I* fromindex,
int64_t length,
int64_t which) {
dim3 blocks_per_grid = blocks(length);
dim3 threads_per_block = threads(length);
int8_t* filtered_mask;
int64_t* res_temp;
HANDLE_ERROR(cudaMalloc((void**)&filtered_mask, sizeof(int8_t) * length));
HANDLE_ERROR(cudaMalloc((void**)&res_temp, sizeof(int64_t) * length));
HANDLE_ERROR(cudaMemset(filtered_mask, 0, sizeof(int8_t) * length));
awkward_UnionArray_project_filter_mask<C><<<blocks_per_grid, threads_per_block>>>(
fromtags,
which,
filtered_mask,
length);
exclusive_scan(res_temp, filtered_mask, length);
HANDLE_ERROR(cudaMemcpy(lenout, res_temp + (length - 1), sizeof(int64_t), cudaMemcpyDeviceToDevice));
awkward_UnionArray_project_kernel<T, C, I><<<blocks_per_grid, threads_per_block>>>(
res_temp,
tocarry,
fromtags,
fromindex,
length,
which);
cudaDeviceSynchronize();
return success();
}
ERROR awkward_UnionArray8_32_project_64(
int64_t* lenout,
int64_t* tocarry,
const int8_t* fromtags,
const int32_t* fromindex,
int64_t length,
int64_t which) {
return awkward_UnionArray_project<int64_t, int8_t, int32_t>(
lenout,
tocarry,
fromtags,
fromindex,
length,
which);
}
ERROR awkward_UnionArray8_U32_project_64(
int64_t* lenout,
int64_t* tocarry,
const int8_t* fromtags,
const uint32_t* fromindex,
int64_t length,
int64_t which) {
return awkward_UnionArray_project<int64_t, int8_t, uint32_t>(
lenout,
tocarry,
fromtags,
fromindex,
length,
which);
}
ERROR awkward_UnionArray8_64_project_64(
int64_t* lenout,
int64_t* tocarry,
const int8_t* fromtags,
const int64_t* fromindex,
int64_t length,
int64_t which) {
return awkward_UnionArray_project<int64_t, int8_t, int64_t>(
lenout,
tocarry,
fromtags,
fromindex,
length,
which);
}
|
