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
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
|
#include "Halide.h"
#include <iostream>
#include <limits>
#include <stdio.h>
// Disable a warning in MSVC that we know will be triggered here.
#ifdef _MSC_VER
#pragma warning(disable : 4756) // "overflow in constant arithmetic"
#endif
using namespace Halide;
using namespace Halide::ConciseCasts;
typedef Expr (*cast_maker_t)(Expr);
// Local alias for terseness. This must be used any time the
// cast could be a float->int conversion that might not fit in the destination.
template<typename DST, typename SRC>
DST safe_cast(SRC s) {
return Internal::safe_numeric_cast<DST, SRC>(s);
}
template<typename source_t, typename target_t>
void test_saturating() {
source_t source_min, source_max;
if (std::numeric_limits<source_t>::has_infinity) {
source_max = std::numeric_limits<source_t>::infinity();
source_min = -source_max;
} else {
source_min = std::numeric_limits<source_t>::lowest();
source_max = std::numeric_limits<source_t>::max();
}
target_t target_min, target_max;
if (std::numeric_limits<target_t>::has_infinity) {
target_max = std::numeric_limits<target_t>::infinity();
target_min = -target_max;
} else {
target_min = std::numeric_limits<target_t>::lowest();
target_max = std::numeric_limits<target_t>::max();
}
Buffer<source_t> in(7);
in(0) = (source_t)0;
in(1) = (source_t)1;
// This can intentionally change the value if source_t is unsigned
in(2) = (source_t)-1;
in(3) = safe_cast<source_t>(source_max);
in(4) = safe_cast<source_t>(source_min);
// These two can intentionally change the value if source_t is smaller than target_t
in(5) = safe_cast<source_t>(target_min);
in(6) = safe_cast<source_t>(target_max);
Var x;
Func f;
f(x) = saturating_cast<target_t>(in(x));
Buffer<target_t> result = f.realize({7});
for (int32_t i = 0; i < 7; i++) {
bool source_signed = std::numeric_limits<source_t>::is_signed;
bool target_signed = std::numeric_limits<target_t>::is_signed;
bool source_floating = !std::numeric_limits<source_t>::is_integer;
bool target_floating = !std::numeric_limits<target_t>::is_integer;
target_t correct_result;
if (source_floating) {
double bounded_lower = std::max((double)in(i), (double)target_min);
if (bounded_lower >= (double)target_max) {
correct_result = target_max;
} else {
correct_result = (target_t)bounded_lower;
}
} else if (target_floating) {
correct_result = (target_t)std::min((double)in(i), (double)target_max);
} else if (source_signed == target_signed) {
if (sizeof(source_t) > sizeof(target_t)) {
correct_result = (target_t)std::min(std::max(in(i),
safe_cast<source_t>(target_min)),
safe_cast<source_t>(target_max));
} else {
correct_result = (target_t)in(i);
}
} else {
if (source_signed) {
source_t val = std::max(in(i), (source_t)0);
if (sizeof(source_t) > sizeof(target_t)) {
correct_result = (target_t)std::min(val, safe_cast<source_t>(target_max));
} else {
correct_result = (target_t)val;
}
} else {
if (sizeof(source_t) >= sizeof(target_t)) {
correct_result = (target_t)std::min(in(i), safe_cast<source_t>(target_max));
} else { // dest is signed, but larger so unsigned source_t guaranteed to fit
correct_result = std::min((target_t)in(i), target_max);
}
}
}
// Do a simpler verification as well if a 64-bit int will hold everything
if (!target_floating && (sizeof(target_t) < 8 || target_signed) &&
!source_floating && (sizeof(source_t) < 8 || source_signed)) {
int64_t simpler_correct_result;
simpler_correct_result = std::min(std::max((int64_t)in(i),
(int64_t)target_min),
(int64_t)target_max);
if (simpler_correct_result != (int64_t)correct_result) {
std::cout << "Simpler verification failed for index " << i
<< " correct_result is " << correct_result
<< " correct_result casted to int64_t is " << (int64_t)correct_result
<< " simpler_correct_result is " << simpler_correct_result << "\n";
std::cout << "in(i) " << in(i)
<< " target_min " << target_min
<< " target_max " << target_max << "\n";
}
assert(simpler_correct_result == (int64_t)correct_result);
}
if (result(i) != correct_result) {
std::cout << "Match failure at index " << i
<< " got " << result(i)
<< " expected " << correct_result
<< " for input " << in(i) << "\n";
}
assert(result(i) == correct_result);
}
}
template<typename source_t, typename target_t>
void test_concise(cast_maker_t cast_maker, bool saturating) {
source_t source_min = std::numeric_limits<source_t>::min();
source_t source_max = std::numeric_limits<source_t>::max();
target_t target_min = std::numeric_limits<target_t>::min();
target_t target_max = std::numeric_limits<target_t>::max();
Buffer<source_t> in(7);
in(0) = (source_t)0;
in(1) = (source_t)1;
// This can intentionally change the value if source_t is unsigned
in(2) = (source_t)-1;
in(3) = source_max;
in(4) = source_min;
// These two can intentionally change the value if source_t is smaller than target_t
in(5) = safe_cast<source_t>(target_min);
in(6) = safe_cast<source_t>(target_max);
Var x;
Func f;
f(x) = cast_maker(in(x));
Buffer<target_t> result = f.realize({7});
for (int32_t i = 0; i < 7; i++) {
bool source_signed = std::numeric_limits<source_t>::is_signed;
bool target_signed = std::numeric_limits<target_t>::is_signed;
bool source_floating = !std::numeric_limits<source_t>::is_integer;
target_t correct_result;
if (saturating) {
if (source_floating) {
source_t bounded_lower = std::max(in(i), safe_cast<source_t>(target_min));
if (bounded_lower >= safe_cast<source_t>(target_max)) {
correct_result = target_max;
} else {
correct_result = (target_t)bounded_lower;
}
} else if (source_signed == target_signed) {
if (sizeof(source_t) > sizeof(target_t)) {
correct_result = (target_t)std::min(std::max(in(i),
safe_cast<source_t>(target_min)),
safe_cast<source_t>(target_max));
} else {
correct_result = (target_t)in(i);
}
} else {
if (source_signed) {
source_t val = std::max(in(i), (source_t)0);
if (sizeof(source_t) > sizeof(target_t)) {
correct_result = (target_t)std::min(val, safe_cast<source_t>(target_max));
} else {
correct_result = (target_t)val;
}
} else {
if (sizeof(source_t) >= sizeof(target_t)) {
correct_result = (target_t)std::min(in(i), safe_cast<source_t>(target_max));
} else { // dest is signed, but larger so unsigned source_t guaranteed to fit
correct_result = std::min((target_t)in(i), target_max);
}
}
}
// Do a simpler verification as well if a 64-bit int will hold everything
if ((sizeof(target_t) < 8 || target_signed) &&
(source_floating || (sizeof(source_t) < 8 || source_signed))) {
int64_t simpler_correct_result;
if (source_floating) {
double bounded_lower = std::max((double)in(i), (double)target_min);
if (bounded_lower >= (double)target_max) {
simpler_correct_result = target_max;
} else {
simpler_correct_result = (int64_t)bounded_lower;
}
} else {
simpler_correct_result = std::min(std::max((int64_t)in(i),
(int64_t)target_min),
(int64_t)target_max);
}
if (simpler_correct_result != (int64_t)correct_result) {
std::cout << "Simpler verification failed for index " << i
<< " correct_result is " << correct_result
<< " correct_result casted to int64_t is " << (int64_t)correct_result
<< " simpler_correct_result is " << simpler_correct_result << "\n"
<< "in(i) " << in(i)
<< " target_min " << target_min
<< " target_max " << target_max << "\n";
}
assert(simpler_correct_result == (int64_t)correct_result);
}
} else {
correct_result = (target_t)in(i);
}
if (result(i) != correct_result) {
std::cout << "Match failure at index " << i
<< " got " << result(i)
<< " expected " << correct_result
<< " for input " << in(i)
<< (saturating ? " saturating" : " nonsaturating") << "\n";
}
assert(result(i) == correct_result);
}
}
template<typename source_t>
void test_one_source_saturating() {
test_saturating<source_t, int8_t>();
test_saturating<source_t, uint8_t>();
test_saturating<source_t, int16_t>();
test_saturating<source_t, uint16_t>();
test_saturating<source_t, int32_t>();
test_saturating<source_t, uint32_t>();
test_saturating<source_t, int64_t>();
test_saturating<source_t, uint64_t>();
test_saturating<source_t, float>();
test_saturating<source_t, double>();
}
template<typename source_t>
void test_one_source_concise() {
test_concise<source_t, int8_t>(i8, false);
test_concise<source_t, uint8_t>(u8, false);
test_concise<source_t, int8_t>(i8_sat, true);
test_concise<source_t, uint8_t>(u8_sat, true);
test_concise<source_t, int16_t>(i16, false);
test_concise<source_t, uint16_t>(u16, false);
test_concise<source_t, int16_t>(i16_sat, true);
test_concise<source_t, uint16_t>(u16_sat, true);
test_concise<source_t, int32_t>(i32, false);
test_concise<source_t, uint32_t>(u32, false);
test_concise<source_t, int32_t>(i32_sat, true);
test_concise<source_t, uint32_t>(u32_sat, true);
test_concise<source_t, int64_t>(i64, false);
test_concise<source_t, uint64_t>(u64, false);
test_concise<source_t, int64_t>(i64_sat, true);
test_concise<source_t, uint64_t>(u64_sat, true);
}
template<typename source_t>
void test_one_source() {
test_one_source_saturating<source_t>();
test_one_source_concise<source_t>();
}
int main(int argc, char **argv) {
#if defined(__i386__) || defined(_M_IX86)
printf("[SKIP] Skipping test because it requires bit-exact int to float casts,\n"
"and on i386 without SSE it is hard to guarantee that the test binary won't use x87 instructions.\n");
return 0;
#endif
test_one_source<int8_t>();
test_one_source<uint8_t>();
test_one_source<int16_t>();
test_one_source<uint16_t>();
test_one_source<int32_t>();
test_one_source<uint32_t>();
test_one_source<int64_t>();
test_one_source<uint64_t>();
// Casting out of range values from floating-point
// to integer types is undefined behavior so only
// do the saturating casts.
test_one_source_saturating<float>();
test_one_source_saturating<double>();
printf("Success!\n");
return 0;
}
|
