4
* Copyright (C) 1991-1996, Thomas G. Lane.
5
4
* This file is part of the Independent JPEG Group's software.
6
* For conditions of distribution and use, see the accompanying README file.
6
* The authors make NO WARRANTY or representation, either express or implied,
7
* with respect to this software, its quality, accuracy, merchantability, or
8
* fitness for a particular purpose. This software is provided "AS IS", and
9
* you, its user, assume the entire risk as to its quality and accuracy.
11
* This software is copyright (C) 1991-1996, Thomas G. Lane.
12
* All Rights Reserved except as specified below.
14
* Permission is hereby granted to use, copy, modify, and distribute this
15
* software (or portions thereof) for any purpose, without fee, subject to
17
* (1) If any part of the source code for this software is distributed, then
18
* this README file must be included, with this copyright and no-warranty
19
* notice unaltered; and any additions, deletions, or changes to the original
20
* files must be clearly indicated in accompanying documentation.
21
* (2) If only executable code is distributed, then the accompanying
22
* documentation must state that "this software is based in part on the work
23
* of the Independent JPEG Group".
24
* (3) Permission for use of this software is granted only if the user accepts
25
* full responsibility for any undesirable consequences; the authors accept
26
* NO LIABILITY for damages of any kind.
28
* These conditions apply to any software derived from or based on the IJG
29
* code, not just to the unmodified library. If you use our work, you ought
32
* Permission is NOT granted for the use of any IJG author's name or company
33
* name in advertising or publicity relating to this software or products
34
* derived from it. This software may be referred to only as "the Independent
35
* JPEG Group's software".
37
* We specifically permit and encourage the use of this software as the basis
38
* of commercial products, provided that all warranty or liability claims are
39
* assumed by the product vendor.
8
41
* This file contains a slow-but-accurate integer implementation of the
9
42
* forward DCT (Discrete Cosine Transform).
108
141
#if CONST_BITS == 13
109
#define FIX_0_298631336 ((int32_t) 2446) /* FIX(0.298631336) */
110
#define FIX_0_390180644 ((int32_t) 3196) /* FIX(0.390180644) */
111
#define FIX_0_541196100 ((int32_t) 4433) /* FIX(0.541196100) */
112
#define FIX_0_765366865 ((int32_t) 6270) /* FIX(0.765366865) */
113
#define FIX_0_899976223 ((int32_t) 7373) /* FIX(0.899976223) */
114
#define FIX_1_175875602 ((int32_t) 9633) /* FIX(1.175875602) */
115
#define FIX_1_501321110 ((int32_t) 12299) /* FIX(1.501321110) */
116
#define FIX_1_847759065 ((int32_t) 15137) /* FIX(1.847759065) */
117
#define FIX_1_961570560 ((int32_t) 16069) /* FIX(1.961570560) */
118
#define FIX_2_053119869 ((int32_t) 16819) /* FIX(2.053119869) */
119
#define FIX_2_562915447 ((int32_t) 20995) /* FIX(2.562915447) */
120
#define FIX_3_072711026 ((int32_t) 25172) /* FIX(3.072711026) */
142
#define FIX_0_298631336 ((int32_t) 2446) /* FIX(0.298631336) */
143
#define FIX_0_390180644 ((int32_t) 3196) /* FIX(0.390180644) */
144
#define FIX_0_541196100 ((int32_t) 4433) /* FIX(0.541196100) */
145
#define FIX_0_765366865 ((int32_t) 6270) /* FIX(0.765366865) */
146
#define FIX_0_899976223 ((int32_t) 7373) /* FIX(0.899976223) */
147
#define FIX_1_175875602 ((int32_t) 9633) /* FIX(1.175875602) */
148
#define FIX_1_501321110 ((int32_t) 12299) /* FIX(1.501321110) */
149
#define FIX_1_847759065 ((int32_t) 15137) /* FIX(1.847759065) */
150
#define FIX_1_961570560 ((int32_t) 16069) /* FIX(1.961570560) */
151
#define FIX_2_053119869 ((int32_t) 16819) /* FIX(2.053119869) */
152
#define FIX_2_562915447 ((int32_t) 20995) /* FIX(2.562915447) */
153
#define FIX_3_072711026 ((int32_t) 25172) /* FIX(3.072711026) */
122
155
#define FIX_0_298631336 FIX(0.298631336)
123
156
#define FIX_0_390180644 FIX(0.390180644)
152
* Perform the forward DCT on one block of samples.
156
ff_jpeg_fdct_islow (DCTELEM * data)
158
int32_t tmp0, tmp1, tmp2, tmp3, tmp4, tmp5, tmp6, tmp7;
159
int32_t tmp10, tmp11, tmp12, tmp13;
160
int32_t z1, z2, z3, z4, z5;
184
static av_always_inline void row_fdct(DCTELEM * data){
185
int_fast32_t tmp0, tmp1, tmp2, tmp3, tmp4, tmp5, tmp6, tmp7;
186
int_fast32_t tmp10, tmp11, tmp12, tmp13;
187
int_fast32_t z1, z2, z3, z4, z5;
161
188
DCTELEM *dataptr;
176
203
tmp5 = dataptr[2] - dataptr[5];
177
204
tmp3 = dataptr[3] + dataptr[4];
178
205
tmp4 = dataptr[3] - dataptr[4];
180
207
/* Even part per LL&M figure 1 --- note that published figure is faulty;
181
208
* rotator "sqrt(2)*c1" should be "sqrt(2)*c6".
184
211
tmp10 = tmp0 + tmp3;
185
212
tmp13 = tmp0 - tmp3;
186
213
tmp11 = tmp1 + tmp2;
187
214
tmp12 = tmp1 - tmp2;
189
216
dataptr[0] = (DCTELEM) ((tmp10 + tmp11) << PASS1_BITS);
190
217
dataptr[4] = (DCTELEM) ((tmp10 - tmp11) << PASS1_BITS);
192
219
z1 = MULTIPLY(tmp12 + tmp13, FIX_0_541196100);
193
220
dataptr[2] = (DCTELEM) DESCALE(z1 + MULTIPLY(tmp13, FIX_0_765366865),
194
CONST_BITS-PASS1_BITS);
221
CONST_BITS-PASS1_BITS);
195
222
dataptr[6] = (DCTELEM) DESCALE(z1 + MULTIPLY(tmp12, - FIX_1_847759065),
196
CONST_BITS-PASS1_BITS);
223
CONST_BITS-PASS1_BITS);
198
225
/* Odd part per figure 8 --- note paper omits factor of sqrt(2).
199
226
* cK represents cos(K*pi/16).
200
227
* i0..i3 in the paper are tmp4..tmp7 here.
203
230
z1 = tmp4 + tmp7;
204
231
z2 = tmp5 + tmp6;
205
232
z3 = tmp4 + tmp6;
206
233
z4 = tmp5 + tmp7;
207
234
z5 = MULTIPLY(z3 + z4, FIX_1_175875602); /* sqrt(2) * c3 */
209
236
tmp4 = MULTIPLY(tmp4, FIX_0_298631336); /* sqrt(2) * (-c1+c3+c5-c7) */
210
237
tmp5 = MULTIPLY(tmp5, FIX_2_053119869); /* sqrt(2) * ( c1+c3-c5+c7) */
211
238
tmp6 = MULTIPLY(tmp6, FIX_3_072711026); /* sqrt(2) * ( c1+c3+c5-c7) */
214
241
z2 = MULTIPLY(z2, - FIX_2_562915447); /* sqrt(2) * (-c1-c3) */
215
242
z3 = MULTIPLY(z3, - FIX_1_961570560); /* sqrt(2) * (-c3-c5) */
216
243
z4 = MULTIPLY(z4, - FIX_0_390180644); /* sqrt(2) * (c5-c3) */
221
248
dataptr[7] = (DCTELEM) DESCALE(tmp4 + z1 + z3, CONST_BITS-PASS1_BITS);
222
249
dataptr[5] = (DCTELEM) DESCALE(tmp5 + z2 + z4, CONST_BITS-PASS1_BITS);
223
250
dataptr[3] = (DCTELEM) DESCALE(tmp6 + z2 + z3, CONST_BITS-PASS1_BITS);
224
251
dataptr[1] = (DCTELEM) DESCALE(tmp7 + z1 + z4, CONST_BITS-PASS1_BITS);
226
dataptr += DCTSIZE; /* advance pointer to next row */
253
dataptr += DCTSIZE; /* advance pointer to next row */
258
* Perform the forward DCT on one block of samples.
262
ff_jpeg_fdct_islow (DCTELEM * data)
264
int_fast32_t tmp0, tmp1, tmp2, tmp3, tmp4, tmp5, tmp6, tmp7;
265
int_fast32_t tmp10, tmp11, tmp12, tmp13;
266
int_fast32_t z1, z2, z3, z4, z5;
229
273
/* Pass 2: process columns.
230
274
* We remove the PASS1_BITS scaling, but leave the results scaled up
241
285
tmp5 = dataptr[DCTSIZE*2] - dataptr[DCTSIZE*5];
242
286
tmp3 = dataptr[DCTSIZE*3] + dataptr[DCTSIZE*4];
243
287
tmp4 = dataptr[DCTSIZE*3] - dataptr[DCTSIZE*4];
245
289
/* Even part per LL&M figure 1 --- note that published figure is faulty;
246
290
* rotator "sqrt(2)*c1" should be "sqrt(2)*c6".
249
293
tmp10 = tmp0 + tmp3;
250
294
tmp13 = tmp0 - tmp3;
251
295
tmp11 = tmp1 + tmp2;
252
296
tmp12 = tmp1 - tmp2;
254
298
dataptr[DCTSIZE*0] = (DCTELEM) DESCALE(tmp10 + tmp11, PASS1_BITS);
255
299
dataptr[DCTSIZE*4] = (DCTELEM) DESCALE(tmp10 - tmp11, PASS1_BITS);
257
301
z1 = MULTIPLY(tmp12 + tmp13, FIX_0_541196100);
258
302
dataptr[DCTSIZE*2] = (DCTELEM) DESCALE(z1 + MULTIPLY(tmp13, FIX_0_765366865),
259
CONST_BITS+PASS1_BITS);
303
CONST_BITS+PASS1_BITS);
260
304
dataptr[DCTSIZE*6] = (DCTELEM) DESCALE(z1 + MULTIPLY(tmp12, - FIX_1_847759065),
261
CONST_BITS+PASS1_BITS);
305
CONST_BITS+PASS1_BITS);
263
307
/* Odd part per figure 8 --- note paper omits factor of sqrt(2).
264
308
* cK represents cos(K*pi/16).
265
309
* i0..i3 in the paper are tmp4..tmp7 here.
268
312
z1 = tmp4 + tmp7;
269
313
z2 = tmp5 + tmp6;
270
314
z3 = tmp4 + tmp6;
271
315
z4 = tmp5 + tmp7;
272
316
z5 = MULTIPLY(z3 + z4, FIX_1_175875602); /* sqrt(2) * c3 */
274
318
tmp4 = MULTIPLY(tmp4, FIX_0_298631336); /* sqrt(2) * (-c1+c3+c5-c7) */
275
319
tmp5 = MULTIPLY(tmp5, FIX_2_053119869); /* sqrt(2) * ( c1+c3-c5+c7) */
276
320
tmp6 = MULTIPLY(tmp6, FIX_3_072711026); /* sqrt(2) * ( c1+c3+c5-c7) */
279
323
z2 = MULTIPLY(z2, - FIX_2_562915447); /* sqrt(2) * (-c1-c3) */
280
324
z3 = MULTIPLY(z3, - FIX_1_961570560); /* sqrt(2) * (-c3-c5) */
281
325
z4 = MULTIPLY(z4, - FIX_0_390180644); /* sqrt(2) * (c5-c3) */
286
330
dataptr[DCTSIZE*7] = (DCTELEM) DESCALE(tmp4 + z1 + z3,
287
CONST_BITS+PASS1_BITS);
331
CONST_BITS+PASS1_BITS);
288
332
dataptr[DCTSIZE*5] = (DCTELEM) DESCALE(tmp5 + z2 + z4,
289
CONST_BITS+PASS1_BITS);
333
CONST_BITS+PASS1_BITS);
290
334
dataptr[DCTSIZE*3] = (DCTELEM) DESCALE(tmp6 + z2 + z3,
291
CONST_BITS+PASS1_BITS);
335
CONST_BITS+PASS1_BITS);
292
336
dataptr[DCTSIZE*1] = (DCTELEM) DESCALE(tmp7 + z1 + z4,
293
CONST_BITS+PASS1_BITS);
295
dataptr++; /* advance pointer to next column */
337
CONST_BITS+PASS1_BITS);
339
dataptr++; /* advance pointer to next column */
344
* The secret of DCT2-4-8 is really simple -- you do the usual 1-DCT
345
* on the rows and then, instead of doing even and odd, part on the colums
346
* you do even part two times.
349
ff_fdct248_islow (DCTELEM * data)
351
int_fast32_t tmp0, tmp1, tmp2, tmp3, tmp4, tmp5, tmp6, tmp7;
352
int_fast32_t tmp10, tmp11, tmp12, tmp13;
360
/* Pass 2: process columns.
361
* We remove the PASS1_BITS scaling, but leave the results scaled up
362
* by an overall factor of 8.
366
for (ctr = DCTSIZE-1; ctr >= 0; ctr--) {
367
tmp0 = dataptr[DCTSIZE*0] + dataptr[DCTSIZE*1];
368
tmp1 = dataptr[DCTSIZE*2] + dataptr[DCTSIZE*3];
369
tmp2 = dataptr[DCTSIZE*4] + dataptr[DCTSIZE*5];
370
tmp3 = dataptr[DCTSIZE*6] + dataptr[DCTSIZE*7];
371
tmp4 = dataptr[DCTSIZE*0] - dataptr[DCTSIZE*1];
372
tmp5 = dataptr[DCTSIZE*2] - dataptr[DCTSIZE*3];
373
tmp6 = dataptr[DCTSIZE*4] - dataptr[DCTSIZE*5];
374
tmp7 = dataptr[DCTSIZE*6] - dataptr[DCTSIZE*7];
381
dataptr[DCTSIZE*0] = (DCTELEM) DESCALE(tmp10 + tmp11, PASS1_BITS);
382
dataptr[DCTSIZE*4] = (DCTELEM) DESCALE(tmp10 - tmp11, PASS1_BITS);
384
z1 = MULTIPLY(tmp12 + tmp13, FIX_0_541196100);
385
dataptr[DCTSIZE*2] = (DCTELEM) DESCALE(z1 + MULTIPLY(tmp13, FIX_0_765366865),
386
CONST_BITS+PASS1_BITS);
387
dataptr[DCTSIZE*6] = (DCTELEM) DESCALE(z1 + MULTIPLY(tmp12, - FIX_1_847759065),
388
CONST_BITS+PASS1_BITS);
395
dataptr[DCTSIZE*1] = (DCTELEM) DESCALE(tmp10 + tmp11, PASS1_BITS);
396
dataptr[DCTSIZE*5] = (DCTELEM) DESCALE(tmp10 - tmp11, PASS1_BITS);
398
z1 = MULTIPLY(tmp12 + tmp13, FIX_0_541196100);
399
dataptr[DCTSIZE*3] = (DCTELEM) DESCALE(z1 + MULTIPLY(tmp13, FIX_0_765366865),
400
CONST_BITS+PASS1_BITS);
401
dataptr[DCTSIZE*7] = (DCTELEM) DESCALE(z1 + MULTIPLY(tmp12, - FIX_1_847759065),
402
CONST_BITS+PASS1_BITS);
404
dataptr++; /* advance pointer to next column */