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
|
// Copyright (C) 2022 Advanced Micro Devices, Inc. All rights reserved.
//
// Permission is hereby granted, free of charge, to any person obtaining a copy
// of this software and associated documentation files (the "Software"), to deal
// in the Software without restriction, including without limitation the rights
// to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
// copies of the Software, and to permit persons to whom the Software is
// furnished to do so, subject to the following conditions:
//
// The above copyright notice and this permission notice shall be included in
// all copies or substantial portions of the Software.
//
// THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
// IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
// FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
// AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
// LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
// OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
// THE SOFTWARE.
#include "accuracy_test.h"
extern bool fftw_compare;
inline auto param_checkstride()
{
// checkstride requires us to copy data back to the host for
// checking, which we only do when comparing against FFTW.
if(!fftw_compare)
return std::vector<fft_params>{};
// tuples of length,stride,nbatch,dist to test. strides are arranged so
// there's space either between elements on the fastest dim, or
// between dims, or both.
std::vector<std::tuple<std::vector<size_t>, std::vector<size_t>, size_t, size_t>> sizes = {
// 1D single kernel non-unit stride
{{64}, {2}, 2, 140},
// 1D single kernel unit stride but non-contiguous batch
{{64}, {1}, 2, 80},
// 1D odd length (to test odd-length R2C/C2R)
{{15}, {2}, 2, 40},
// 1D SBCC+SBRC
{{8192}, {2}, 2, 17000},
// 1D TRTRT
{{24000}, {2}, 2, 50000},
// 2D_RTRT
{{20, 30}, {80, 2}, 2, 1700},
{{40, 30}, {80, 2}, 2, 3600},
// 2D_RTRT unit stride along fast dim
{{20, 30}, {40, 1}, 2, 1000},
{{40, 30}, {40, 1}, 2, 2000},
// 2D_RC
{{64, 64}, {130, 2}, 2, 8400},
// 3D_RC
{{64, 64, 64}, {8400, 130, 2}, 2, 540000},
// 3D_RTRTRT
{{2, 3, 4}, {40, 10, 2}, 2, 100},
// bigger 3D_RTRTRT
{{30, 40, 50}, {3000, 60, 1}, 2, 100000},
};
std::vector<fft_params> params;
for(const auto trans_type : trans_type_range)
{
for(const auto& s : sizes)
{
for(const auto precision : precision_range_sp_dp)
{
for(const auto& types :
generate_types(trans_type, {fft_placement_notinplace}, true))
{
for(bool callback : {true, false})
{
// callbacks don't work for planar
bool is_planar = std::get<2>(types) == fft_array_type_complex_planar
|| std::get<2>(types) == fft_array_type_hermitian_planar
|| std::get<3>(types) == fft_array_type_complex_planar
|| std::get<3>(types) == fft_array_type_hermitian_planar;
if(callback && is_planar)
continue;
fft_params param;
param.length = std::get<0>(s);
param.istride = std::get<1>(s);
param.ostride = std::get<1>(s);
param.nbatch = std::get<2>(s);
param.precision = precision;
param.idist = std::get<3>(s);
param.odist = std::get<3>(s);
param.transform_type = std::get<0>(types);
param.placement = std::get<1>(types);
param.itype = std::get<2>(types);
param.otype = std::get<3>(types);
param.run_callbacks = callback;
param.check_output_strides = true;
params.push_back(param);
}
}
}
}
}
return params;
}
INSTANTIATE_TEST_SUITE_P(checkstride,
accuracy_test,
::testing::ValuesIn(param_checkstride()),
accuracy_test::TestName);
|
