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
|
// This file is part of Eigen, a lightweight C++ template library
// for linear algebra.
//
// Copyright (C) 2014-2015 Benoit Steiner <benoit.steiner.goog@gmail.com>
//
// This Source Code Form is subject to the terms of the Mozilla
// Public License v. 2.0. If a copy of the MPL was not distributed
// with this file, You can obtain one at http://mozilla.org/MPL/2.0/.
#include "main.h"
#include <Eigen/CXX11/Tensor>
void test_signed_32bit()
{
// Divide by one
const Eigen::internal::TensorIntDivisor<int32_t, false> div_by_one(1);
for (int32_t j = 0; j < 25000; ++j) {
const int32_t fast_div = j / div_by_one;
const int32_t slow_div = j / 1;
VERIFY_IS_EQUAL(fast_div, slow_div);
}
// Standard divide by 2 or more
for (int32_t i = 2; i < 25000; ++i) {
const Eigen::internal::TensorIntDivisor<int32_t, false> div(i);
for (int32_t j = 0; j < 25000; ++j) {
const int32_t fast_div = j / div;
const int32_t slow_div = j / i;
VERIFY_IS_EQUAL(fast_div, slow_div);
}
}
// Optimized divide by 2 or more
for (int32_t i = 2; i < 25000; ++i) {
const Eigen::internal::TensorIntDivisor<int32_t, true> div(i);
for (int32_t j = 0; j < 25000; ++j) {
const int32_t fast_div = j / div;
const int32_t slow_div = j / i;
VERIFY_IS_EQUAL(fast_div, slow_div);
}
}
}
void test_unsigned_32bit()
{
for (uint32_t i = 1; i < 25000; ++i) {
const Eigen::internal::TensorIntDivisor<uint32_t> div(i);
for (uint32_t j = 0; j < 25000; ++j) {
const uint32_t fast_div = j / div;
const uint32_t slow_div = j / i;
VERIFY_IS_EQUAL(fast_div, slow_div);
}
}
}
void test_signed_64bit()
{
for (int64_t i = 1; i < 25000; ++i) {
const Eigen::internal::TensorIntDivisor<int64_t> div(i);
for (int64_t j = 0; j < 25000; ++j) {
const int64_t fast_div = j / div;
const int64_t slow_div = j / i;
VERIFY_IS_EQUAL(fast_div, slow_div);
}
}
}
void test_unsigned_64bit()
{
for (uint64_t i = 1; i < 25000; ++i) {
const Eigen::internal::TensorIntDivisor<uint64_t> div(i);
for (uint64_t j = 0; j < 25000; ++j) {
const uint64_t fast_div = j / div;
const uint64_t slow_div = j / i;
VERIFY_IS_EQUAL(fast_div, slow_div);
}
}
}
void test_powers_32bit() {
for (int expon = 1; expon < 31; expon++) {
int32_t div = (1 << expon);
for (int num_expon = 0; num_expon < 32; num_expon++) {
int32_t start_num = (1 << num_expon) - 100;
int32_t end_num = (1 << num_expon) + 100;
if (start_num < 0)
start_num = 0;
for (int32_t num = start_num; num < end_num; num++) {
Eigen::internal::TensorIntDivisor<int32_t> divider =
Eigen::internal::TensorIntDivisor<int32_t>(div);
int32_t result = num/div;
int32_t result_op = divider.divide(num);
VERIFY_IS_EQUAL(result_op, result);
}
}
}
}
void test_powers_64bit() {
for (int expon = 0; expon < 63; expon++) {
int64_t div = (1ull << expon);
for (int num_expon = 0; num_expon < 63; num_expon++) {
int64_t start_num = (1ull << num_expon) - 10;
int64_t end_num = (1ull << num_expon) + 10;
if (start_num < 0)
start_num = 0;
for (int64_t num = start_num; num < end_num; num++) {
Eigen::internal::TensorIntDivisor<int64_t> divider(div);
int64_t result = num/div;
int64_t result_op = divider.divide(num);
VERIFY_IS_EQUAL(result_op, result);
}
}
}
}
void test_specific() {
// A particular combination that was previously failing
int64_t div = 209715200;
int64_t num = 3238002688ll;
Eigen::internal::TensorIntDivisor<int64_t> divider(div);
int64_t result = num/div;
int64_t result_op = divider.divide(num);
VERIFY_IS_EQUAL(result, result_op);
}
void test_cxx11_tensor_intdiv()
{
CALL_SUBTEST_1(test_signed_32bit());
CALL_SUBTEST_2(test_unsigned_32bit());
CALL_SUBTEST_3(test_signed_64bit());
CALL_SUBTEST_4(test_unsigned_64bit());
CALL_SUBTEST_5(test_powers_32bit());
CALL_SUBTEST_6(test_powers_64bit());
CALL_SUBTEST_7(test_specific());
}
|
