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
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
|
/*
* Copyright 2012-15 Advanced Micro Devices, Inc.
*
* 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 COPYRIGHT HOLDER(S) OR AUTHOR(S) 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.
*
* Authors: AMD
*
*/
#include "dm_services.h"
#include "include/fixed31_32.h"
static const struct fixed31_32 dc_fixpt_two_pi = { 26986075409LL };
static const struct fixed31_32 dc_fixpt_ln2 = { 2977044471LL };
static const struct fixed31_32 dc_fixpt_ln2_div_2 = { 1488522236LL };
static inline unsigned long long abs_i64(
long long arg)
{
if (arg > 0)
return (unsigned long long)arg;
else
return (unsigned long long)(-arg);
}
/*
* @brief
* result = dividend / divisor
* *remainder = dividend % divisor
*/
static inline unsigned long long complete_integer_division_u64(
unsigned long long dividend,
unsigned long long divisor,
unsigned long long *remainder)
{
unsigned long long result;
ASSERT(divisor);
result = div64_u64_rem(dividend, divisor, remainder);
return result;
}
#define FRACTIONAL_PART_MASK \
((1ULL << FIXED31_32_BITS_PER_FRACTIONAL_PART) - 1)
#define GET_INTEGER_PART(x) \
((x) >> FIXED31_32_BITS_PER_FRACTIONAL_PART)
#define GET_FRACTIONAL_PART(x) \
(FRACTIONAL_PART_MASK & (x))
struct fixed31_32 dc_fixpt_from_fraction(long long numerator, long long denominator)
{
struct fixed31_32 res;
bool arg1_negative = numerator < 0;
bool arg2_negative = denominator < 0;
unsigned long long arg1_value = arg1_negative ? -numerator : numerator;
unsigned long long arg2_value = arg2_negative ? -denominator : denominator;
unsigned long long remainder;
/* determine integer part */
unsigned long long res_value = complete_integer_division_u64(
arg1_value, arg2_value, &remainder);
ASSERT(res_value <= LONG_MAX);
/* determine fractional part */
{
unsigned int i = FIXED31_32_BITS_PER_FRACTIONAL_PART;
do {
remainder <<= 1;
res_value <<= 1;
if (remainder >= arg2_value) {
res_value |= 1;
remainder -= arg2_value;
}
} while (--i != 0);
}
/* round up LSB */
{
unsigned long long summand = (remainder << 1) >= arg2_value;
ASSERT(res_value <= LLONG_MAX - summand);
res_value += summand;
}
res.value = (long long)res_value;
if (arg1_negative ^ arg2_negative)
res.value = -res.value;
return res;
}
struct fixed31_32 dc_fixpt_mul(struct fixed31_32 arg1, struct fixed31_32 arg2)
{
struct fixed31_32 res;
bool arg1_negative = arg1.value < 0;
bool arg2_negative = arg2.value < 0;
unsigned long long arg1_value = arg1_negative ? -arg1.value : arg1.value;
unsigned long long arg2_value = arg2_negative ? -arg2.value : arg2.value;
unsigned long long arg1_int = GET_INTEGER_PART(arg1_value);
unsigned long long arg2_int = GET_INTEGER_PART(arg2_value);
unsigned long long arg1_fra = GET_FRACTIONAL_PART(arg1_value);
unsigned long long arg2_fra = GET_FRACTIONAL_PART(arg2_value);
unsigned long long tmp;
res.value = arg1_int * arg2_int;
res.value <<= FIXED31_32_BITS_PER_FRACTIONAL_PART;
tmp = arg1_int * arg2_fra;
ASSERT(tmp <= (unsigned long long)(LLONG_MAX - res.value));
res.value += tmp;
tmp = arg2_int * arg1_fra;
ASSERT(tmp <= (unsigned long long)(LLONG_MAX - res.value));
res.value += tmp;
tmp = arg1_fra * arg2_fra;
tmp = (tmp >> FIXED31_32_BITS_PER_FRACTIONAL_PART) +
(tmp >= (unsigned long long)dc_fixpt_half.value);
ASSERT(tmp <= (unsigned long long)(LLONG_MAX - res.value));
res.value += tmp;
if (arg1_negative ^ arg2_negative)
res.value = -res.value;
return res;
}
struct fixed31_32 dc_fixpt_sqr(struct fixed31_32 arg)
{
struct fixed31_32 res;
unsigned long long arg_value = abs_i64(arg.value);
unsigned long long arg_int = GET_INTEGER_PART(arg_value);
unsigned long long arg_fra = GET_FRACTIONAL_PART(arg_value);
unsigned long long tmp;
res.value = arg_int * arg_int;
res.value <<= FIXED31_32_BITS_PER_FRACTIONAL_PART;
tmp = arg_int * arg_fra;
ASSERT(tmp <= (unsigned long long)(LLONG_MAX - res.value));
res.value += tmp;
ASSERT(tmp <= (unsigned long long)(LLONG_MAX - res.value));
res.value += tmp;
tmp = arg_fra * arg_fra;
tmp = (tmp >> FIXED31_32_BITS_PER_FRACTIONAL_PART) +
(tmp >= (unsigned long long)dc_fixpt_half.value);
ASSERT(tmp <= (unsigned long long)(LLONG_MAX - res.value));
res.value += tmp;
return res;
}
struct fixed31_32 dc_fixpt_recip(struct fixed31_32 arg)
{
/*
* @note
* Good idea to use Newton's method
*/
ASSERT(arg.value);
return dc_fixpt_from_fraction(
dc_fixpt_one.value,
arg.value);
}
struct fixed31_32 dc_fixpt_sinc(struct fixed31_32 arg)
{
struct fixed31_32 square;
struct fixed31_32 res = dc_fixpt_one;
int n = 27;
struct fixed31_32 arg_norm = arg;
if (dc_fixpt_le(
dc_fixpt_two_pi,
dc_fixpt_abs(arg))) {
arg_norm = dc_fixpt_sub(
arg_norm,
dc_fixpt_mul_int(
dc_fixpt_two_pi,
(int)div64_s64(
arg_norm.value,
dc_fixpt_two_pi.value)));
}
square = dc_fixpt_sqr(arg_norm);
do {
res = dc_fixpt_sub(
dc_fixpt_one,
dc_fixpt_div_int(
dc_fixpt_mul(
square,
res),
n * (n - 1)));
n -= 2;
} while (n > 2);
if (arg.value != arg_norm.value)
res = dc_fixpt_div(
dc_fixpt_mul(res, arg_norm),
arg);
return res;
}
struct fixed31_32 dc_fixpt_sin(struct fixed31_32 arg)
{
return dc_fixpt_mul(
arg,
dc_fixpt_sinc(arg));
}
struct fixed31_32 dc_fixpt_cos(struct fixed31_32 arg)
{
/* TODO implement argument normalization */
const struct fixed31_32 square = dc_fixpt_sqr(arg);
struct fixed31_32 res = dc_fixpt_one;
int n = 26;
do {
res = dc_fixpt_sub(
dc_fixpt_one,
dc_fixpt_div_int(
dc_fixpt_mul(
square,
res),
n * (n - 1)));
n -= 2;
} while (n != 0);
return res;
}
/*
* @brief
* result = exp(arg),
* where abs(arg) < 1
*
* Calculated as Taylor series.
*/
static struct fixed31_32 fixed31_32_exp_from_taylor_series(struct fixed31_32 arg)
{
unsigned int n = 9;
struct fixed31_32 res = dc_fixpt_from_fraction(
n + 2,
n + 1);
/* TODO find correct res */
ASSERT(dc_fixpt_lt(arg, dc_fixpt_one));
do
res = dc_fixpt_add(
dc_fixpt_one,
dc_fixpt_div_int(
dc_fixpt_mul(
arg,
res),
n));
while (--n != 1);
return dc_fixpt_add(
dc_fixpt_one,
dc_fixpt_mul(
arg,
res));
}
struct fixed31_32 dc_fixpt_exp(struct fixed31_32 arg)
{
/*
* @brief
* Main equation is:
* exp(x) = exp(r + m * ln(2)) = (1 << m) * exp(r),
* where m = round(x / ln(2)), r = x - m * ln(2)
*/
if (dc_fixpt_le(
dc_fixpt_ln2_div_2,
dc_fixpt_abs(arg))) {
int m = dc_fixpt_round(
dc_fixpt_div(
arg,
dc_fixpt_ln2));
struct fixed31_32 r = dc_fixpt_sub(
arg,
dc_fixpt_mul_int(
dc_fixpt_ln2,
m));
ASSERT(m != 0);
ASSERT(dc_fixpt_lt(
dc_fixpt_abs(r),
dc_fixpt_one));
if (m > 0)
return dc_fixpt_shl(
fixed31_32_exp_from_taylor_series(r),
(unsigned char)m);
else
return dc_fixpt_div_int(
fixed31_32_exp_from_taylor_series(r),
1LL << -m);
} else if (arg.value != 0)
return fixed31_32_exp_from_taylor_series(arg);
else
return dc_fixpt_one;
}
struct fixed31_32 dc_fixpt_log(struct fixed31_32 arg)
{
struct fixed31_32 res = dc_fixpt_neg(dc_fixpt_one);
/* TODO improve 1st estimation */
struct fixed31_32 error;
ASSERT(arg.value > 0);
/* TODO if arg is negative, return NaN */
/* TODO if arg is zero, return -INF */
do {
struct fixed31_32 res1 = dc_fixpt_add(
dc_fixpt_sub(
res,
dc_fixpt_one),
dc_fixpt_div(
arg,
dc_fixpt_exp(res)));
error = dc_fixpt_sub(
res,
res1);
res = res1;
/* TODO determine max_allowed_error based on quality of exp() */
} while (abs_i64(error.value) > 100ULL);
return res;
}
/* this function is a generic helper to translate fixed point value to
* specified integer format that will consist of integer_bits integer part and
* fractional_bits fractional part. For example it is used in
* dc_fixpt_u2d19 to receive 2 bits integer part and 19 bits fractional
* part in 32 bits. It is used in hw programming (scaler)
*/
static inline unsigned int ux_dy(
long long value,
unsigned int integer_bits,
unsigned int fractional_bits)
{
/* 1. create mask of integer part */
unsigned int result = (1 << integer_bits) - 1;
/* 2. mask out fractional part */
unsigned int fractional_part = FRACTIONAL_PART_MASK & value;
/* 3. shrink fixed point integer part to be of integer_bits width*/
result &= GET_INTEGER_PART(value);
/* 4. make space for fractional part to be filled in after integer */
result <<= fractional_bits;
/* 5. shrink fixed point fractional part to of fractional_bits width*/
fractional_part >>= FIXED31_32_BITS_PER_FRACTIONAL_PART - fractional_bits;
/* 6. merge the result */
return result | fractional_part;
}
static inline unsigned int clamp_ux_dy(
long long value,
unsigned int integer_bits,
unsigned int fractional_bits,
unsigned int min_clamp)
{
unsigned int truncated_val = ux_dy(value, integer_bits, fractional_bits);
if (value >= (1LL << (integer_bits + FIXED31_32_BITS_PER_FRACTIONAL_PART)))
return (1 << (integer_bits + fractional_bits)) - 1;
else if (truncated_val > min_clamp)
return truncated_val;
else
return min_clamp;
}
unsigned int dc_fixpt_u4d19(struct fixed31_32 arg)
{
return ux_dy(arg.value, 4, 19);
}
unsigned int dc_fixpt_u3d19(struct fixed31_32 arg)
{
return ux_dy(arg.value, 3, 19);
}
unsigned int dc_fixpt_u2d19(struct fixed31_32 arg)
{
return ux_dy(arg.value, 2, 19);
}
unsigned int dc_fixpt_u0d19(struct fixed31_32 arg)
{
return ux_dy(arg.value, 0, 19);
}
unsigned int dc_fixpt_clamp_u0d14(struct fixed31_32 arg)
{
return clamp_ux_dy(arg.value, 0, 14, 1);
}
unsigned int dc_fixpt_clamp_u0d10(struct fixed31_32 arg)
{
return clamp_ux_dy(arg.value, 0, 10, 1);
}
int dc_fixpt_s4d19(struct fixed31_32 arg)
{
if (arg.value < 0)
return -(int)ux_dy(dc_fixpt_abs(arg).value, 4, 19);
else
return ux_dy(arg.value, 4, 19);
}
|