2 * VC-1 and WMV3 decoder - DSP functions
3 * Copyright (c) 2006 Konstantin Shishkov
5 * This library is free software; you can redistribute it and/or
6 * modify it under the terms of the GNU Lesser General Public
7 * License as published by the Free Software Foundation; either
8 * version 2 of the License, or (at your option) any later version.
10 * This library is distributed in the hope that it will be useful,
11 * but WITHOUT ANY WARRANTY; without even the implied warranty of
12 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
13 * Lesser General Public License for more details.
15 * You should have received a copy of the GNU Lesser General Public
16 * License along with this library; if not, write to the Free Software
17 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
23 * VC-1 and WMV3 decoder
30 /** Apply overlap transform to vertical edge
32 static void vc1_v_overlap_c(uint8_t* src, int stride, int rnd)
36 for(i = 0; i < 8; i++) {
42 src[-2*stride] = clip_uint8((7*a + d + 4 - rnd) >> 3);
43 src[-stride] = clip_uint8((-a + 7*b + c + d + 3 + rnd) >> 3);
44 src[0] = clip_uint8((a + b + 7*c - d + 4 - rnd) >> 3);
45 src[stride] = clip_uint8((a + 7*d + 3 + rnd) >> 3);
50 /** Apply overlap transform to horizontal edge
52 static void vc1_h_overlap_c(uint8_t* src, int stride, int rnd)
56 for(i = 0; i < 8; i++) {
62 src[-2] = clip_uint8((7*a + d + 4 - rnd) >> 3);
63 src[-1] = clip_uint8((-a + 7*b + c + d + 3 + rnd) >> 3);
64 src[0] = clip_uint8((a + b + 7*c - d + 4 - rnd) >> 3);
65 src[1] = clip_uint8((a + 7*d + 3 + rnd) >> 3);
71 /** Do inverse transform on 8x8 block
73 static void vc1_inv_trans_8x8_c(DCTELEM block[64])
76 register int t1,t2,t3,t4,t5,t6,t7,t8;
81 for(i = 0; i < 8; i++){
82 t1 = 12 * (src[0] + src[4]);
83 t2 = 12 * (src[0] - src[4]);
84 t3 = 16 * src[2] + 6 * src[6];
85 t4 = 6 * src[2] - 16 * src[6];
92 t1 = 16 * src[1] + 15 * src[3] + 9 * src[5] + 4 * src[7];
93 t2 = 15 * src[1] - 4 * src[3] - 16 * src[5] - 9 * src[7];
94 t3 = 9 * src[1] - 16 * src[3] + 4 * src[5] + 15 * src[7];
95 t4 = 4 * src[1] - 9 * src[3] + 15 * src[5] - 16 * src[7];
97 dst[0] = (t5 + t1 + 4) >> 3;
98 dst[1] = (t6 + t2 + 4) >> 3;
99 dst[2] = (t7 + t3 + 4) >> 3;
100 dst[3] = (t8 + t4 + 4) >> 3;
101 dst[4] = (t8 - t4 + 4) >> 3;
102 dst[5] = (t7 - t3 + 4) >> 3;
103 dst[6] = (t6 - t2 + 4) >> 3;
104 dst[7] = (t5 - t1 + 4) >> 3;
112 for(i = 0; i < 8; i++){
113 t1 = 12 * (src[ 0] + src[32]);
114 t2 = 12 * (src[ 0] - src[32]);
115 t3 = 16 * src[16] + 6 * src[48];
116 t4 = 6 * src[16] - 16 * src[48];
123 t1 = 16 * src[ 8] + 15 * src[24] + 9 * src[40] + 4 * src[56];
124 t2 = 15 * src[ 8] - 4 * src[24] - 16 * src[40] - 9 * src[56];
125 t3 = 9 * src[ 8] - 16 * src[24] + 4 * src[40] + 15 * src[56];
126 t4 = 4 * src[ 8] - 9 * src[24] + 15 * src[40] - 16 * src[56];
128 dst[ 0] = (t5 + t1 + 64) >> 7;
129 dst[ 8] = (t6 + t2 + 64) >> 7;
130 dst[16] = (t7 + t3 + 64) >> 7;
131 dst[24] = (t8 + t4 + 64) >> 7;
132 dst[32] = (t8 - t4 + 64 + 1) >> 7;
133 dst[40] = (t7 - t3 + 64 + 1) >> 7;
134 dst[48] = (t6 - t2 + 64 + 1) >> 7;
135 dst[56] = (t5 - t1 + 64 + 1) >> 7;
142 /** Do inverse transform on 8x4 part of block
144 static void vc1_inv_trans_8x4_c(DCTELEM block[64], int n)
147 register int t1,t2,t3,t4,t5,t6,t7,t8;
154 for(i = 0; i < 4; i++){
155 t1 = 12 * (src[0] + src[4]);
156 t2 = 12 * (src[0] - src[4]);
157 t3 = 16 * src[2] + 6 * src[6];
158 t4 = 6 * src[2] - 16 * src[6];
165 t1 = 16 * src[1] + 15 * src[3] + 9 * src[5] + 4 * src[7];
166 t2 = 15 * src[1] - 4 * src[3] - 16 * src[5] - 9 * src[7];
167 t3 = 9 * src[1] - 16 * src[3] + 4 * src[5] + 15 * src[7];
168 t4 = 4 * src[1] - 9 * src[3] + 15 * src[5] - 16 * src[7];
170 dst[0] = (t5 + t1 + 4) >> 3;
171 dst[1] = (t6 + t2 + 4) >> 3;
172 dst[2] = (t7 + t3 + 4) >> 3;
173 dst[3] = (t8 + t4 + 4) >> 3;
174 dst[4] = (t8 - t4 + 4) >> 3;
175 dst[5] = (t7 - t3 + 4) >> 3;
176 dst[6] = (t6 - t2 + 4) >> 3;
177 dst[7] = (t5 - t1 + 4) >> 3;
185 for(i = 0; i < 8; i++){
186 t1 = 17 * (src[ 0] + src[16]);
187 t2 = 17 * (src[ 0] - src[16]);
193 dst[ 0] = (t1 + t3 + t6 + 64) >> 7;
194 dst[ 8] = (t2 - t4 + t5 + 64) >> 7;
195 dst[16] = (t2 + t4 - t5 + 64) >> 7;
196 dst[24] = (t1 - t3 - t6 + 64) >> 7;
203 /** Do inverse transform on 4x8 parts of block
205 static void vc1_inv_trans_4x8_c(DCTELEM block[64], int n)
208 register int t1,t2,t3,t4,t5,t6,t7,t8;
215 for(i = 0; i < 8; i++){
216 t1 = 17 * (src[0] + src[2]);
217 t2 = 17 * (src[0] - src[2]);
223 dst[0] = (t1 + t3 + t6 + 4) >> 3;
224 dst[1] = (t2 - t4 + t5 + 4) >> 3;
225 dst[2] = (t2 + t4 - t5 + 4) >> 3;
226 dst[3] = (t1 - t3 - t6 + 4) >> 3;
234 for(i = 0; i < 4; i++){
235 t1 = 12 * (src[ 0] + src[32]);
236 t2 = 12 * (src[ 0] - src[32]);
237 t3 = 16 * src[16] + 6 * src[48];
238 t4 = 6 * src[16] - 16 * src[48];
245 t1 = 16 * src[ 8] + 15 * src[24] + 9 * src[40] + 4 * src[56];
246 t2 = 15 * src[ 8] - 4 * src[24] - 16 * src[40] - 9 * src[56];
247 t3 = 9 * src[ 8] - 16 * src[24] + 4 * src[40] + 15 * src[56];
248 t4 = 4 * src[ 8] - 9 * src[24] + 15 * src[40] - 16 * src[56];
250 dst[ 0] = (t5 + t1 + 64) >> 7;
251 dst[ 8] = (t6 + t2 + 64) >> 7;
252 dst[16] = (t7 + t3 + 64) >> 7;
253 dst[24] = (t8 + t4 + 64) >> 7;
254 dst[32] = (t8 - t4 + 64 + 1) >> 7;
255 dst[40] = (t7 - t3 + 64 + 1) >> 7;
256 dst[48] = (t6 - t2 + 64 + 1) >> 7;
257 dst[56] = (t5 - t1 + 64 + 1) >> 7;
264 /** Do inverse transform on 4x4 part of block
266 static void vc1_inv_trans_4x4_c(DCTELEM block[64], int n)
269 register int t1,t2,t3,t4,t5,t6;
273 off = (n&1) * 4 + (n&2) * 16;
276 for(i = 0; i < 4; i++){
277 t1 = 17 * (src[0] + src[2]);
278 t2 = 17 * (src[0] - src[2]);
284 dst[0] = (t1 + t3 + t6 + 4) >> 3;
285 dst[1] = (t2 - t4 + t5 + 4) >> 3;
286 dst[2] = (t2 + t4 - t5 + 4) >> 3;
287 dst[3] = (t1 - t3 - t6 + 4) >> 3;
295 for(i = 0; i < 4; i++){
296 t1 = 17 * (src[ 0] + src[16]);
297 t2 = 17 * (src[ 0] - src[16]);
303 dst[ 0] = (t1 + t3 + t6 + 64) >> 7;
304 dst[ 8] = (t2 - t4 + t5 + 64) >> 7;
305 dst[16] = (t2 + t4 - t5 + 64) >> 7;
306 dst[24] = (t1 - t3 - t6 + 64) >> 7;
313 /* motion compensation functions */
315 /** Filter used to interpolate fractional pel values
317 static always_inline int vc1_mspel_filter(const uint8_t *src, int stride, int mode, int r)
323 return (-4*src[-stride] + 53*src[0] + 18*src[stride] - 3*src[stride*2] + 32 - r) >> 6;
325 return (-src[-stride] + 9*src[0] + 9*src[stride] - src[stride*2] + 8 - r) >> 4;
327 return (-3*src[-stride] + 18*src[0] + 53*src[stride] - 4*src[stride*2] + 32 - r) >> 6;
329 return 0; //should not occur
332 /** Function used to do motion compensation with bicubic interpolation
334 static void vc1_mspel_mc(uint8_t *dst, const uint8_t *src, int stride, int mode, int rnd)
337 uint8_t tmp[8*11], *tptr;
344 for(j = 0; j < 11; j++) {
345 for(i = 0; i < 8; i++)
346 tptr[i] = clip_uint8(vc1_mspel_filter(src + i, 1, m, r));
354 for(j = 0; j < 8; j++) {
355 for(i = 0; i < 8; i++)
356 dst[i] = clip_uint8(vc1_mspel_filter(tptr + i, 8, m, r));
362 /* pixel functions - really are entry points to vc1_mspel_mc */
364 /* this one is defined in dsputil.c */
365 void ff_put_vc1_mspel_mc00_c(uint8_t *dst, const uint8_t *src, int stride, int rnd);
367 static void ff_put_vc1_mspel_mc10_c(uint8_t *dst, const uint8_t *src, int stride, int rnd) {
368 vc1_mspel_mc(dst, src, stride, 0x1, rnd);
371 static void ff_put_vc1_mspel_mc20_c(uint8_t *dst, const uint8_t *src, int stride, int rnd) {
372 vc1_mspel_mc(dst, src, stride, 0x2, rnd);
375 static void ff_put_vc1_mspel_mc30_c(uint8_t *dst, const uint8_t *src, int stride, int rnd) {
376 vc1_mspel_mc(dst, src, stride, 0x3, rnd);
379 static void ff_put_vc1_mspel_mc01_c(uint8_t *dst, const uint8_t *src, int stride, int rnd) {
380 vc1_mspel_mc(dst, src, stride, 0x4, rnd);
383 static void ff_put_vc1_mspel_mc11_c(uint8_t *dst, const uint8_t *src, int stride, int rnd) {
384 vc1_mspel_mc(dst, src, stride, 0x5, rnd);
387 static void ff_put_vc1_mspel_mc21_c(uint8_t *dst, const uint8_t *src, int stride, int rnd) {
388 vc1_mspel_mc(dst, src, stride, 0x6, rnd);
391 static void ff_put_vc1_mspel_mc31_c(uint8_t *dst, const uint8_t *src, int stride, int rnd) {
392 vc1_mspel_mc(dst, src, stride, 0x7, rnd);
395 static void ff_put_vc1_mspel_mc02_c(uint8_t *dst, const uint8_t *src, int stride, int rnd) {
396 vc1_mspel_mc(dst, src, stride, 0x8, rnd);
399 static void ff_put_vc1_mspel_mc12_c(uint8_t *dst, const uint8_t *src, int stride, int rnd) {
400 vc1_mspel_mc(dst, src, stride, 0x9, rnd);
403 static void ff_put_vc1_mspel_mc22_c(uint8_t *dst, const uint8_t *src, int stride, int rnd) {
404 vc1_mspel_mc(dst, src, stride, 0xA, rnd);
407 static void ff_put_vc1_mspel_mc32_c(uint8_t *dst, const uint8_t *src, int stride, int rnd) {
408 vc1_mspel_mc(dst, src, stride, 0xB, rnd);
411 static void ff_put_vc1_mspel_mc03_c(uint8_t *dst, const uint8_t *src, int stride, int rnd) {
412 vc1_mspel_mc(dst, src, stride, 0xC, rnd);
415 static void ff_put_vc1_mspel_mc13_c(uint8_t *dst, const uint8_t *src, int stride, int rnd) {
416 vc1_mspel_mc(dst, src, stride, 0xD, rnd);
419 static void ff_put_vc1_mspel_mc23_c(uint8_t *dst, const uint8_t *src, int stride, int rnd) {
420 vc1_mspel_mc(dst, src, stride, 0xE, rnd);
423 static void ff_put_vc1_mspel_mc33_c(uint8_t *dst, const uint8_t *src, int stride, int rnd) {
424 vc1_mspel_mc(dst, src, stride, 0xF, rnd);
427 void ff_vc1dsp_init(DSPContext* dsp, AVCodecContext *avctx) {
428 dsp->vc1_inv_trans_8x8 = vc1_inv_trans_8x8_c;
429 dsp->vc1_inv_trans_4x8 = vc1_inv_trans_4x8_c;
430 dsp->vc1_inv_trans_8x4 = vc1_inv_trans_8x4_c;
431 dsp->vc1_inv_trans_4x4 = vc1_inv_trans_4x4_c;
432 dsp->vc1_h_overlap = vc1_h_overlap_c;
433 dsp->vc1_v_overlap = vc1_v_overlap_c;
435 dsp->put_vc1_mspel_pixels_tab[ 0] = ff_put_vc1_mspel_mc00_c;
436 dsp->put_vc1_mspel_pixels_tab[ 1] = ff_put_vc1_mspel_mc10_c;
437 dsp->put_vc1_mspel_pixels_tab[ 2] = ff_put_vc1_mspel_mc20_c;
438 dsp->put_vc1_mspel_pixels_tab[ 3] = ff_put_vc1_mspel_mc30_c;
439 dsp->put_vc1_mspel_pixels_tab[ 4] = ff_put_vc1_mspel_mc01_c;
440 dsp->put_vc1_mspel_pixels_tab[ 5] = ff_put_vc1_mspel_mc11_c;
441 dsp->put_vc1_mspel_pixels_tab[ 6] = ff_put_vc1_mspel_mc21_c;
442 dsp->put_vc1_mspel_pixels_tab[ 7] = ff_put_vc1_mspel_mc31_c;
443 dsp->put_vc1_mspel_pixels_tab[ 8] = ff_put_vc1_mspel_mc02_c;
444 dsp->put_vc1_mspel_pixels_tab[ 9] = ff_put_vc1_mspel_mc12_c;
445 dsp->put_vc1_mspel_pixels_tab[10] = ff_put_vc1_mspel_mc22_c;
446 dsp->put_vc1_mspel_pixels_tab[11] = ff_put_vc1_mspel_mc32_c;
447 dsp->put_vc1_mspel_pixels_tab[12] = ff_put_vc1_mspel_mc03_c;
448 dsp->put_vc1_mspel_pixels_tab[13] = ff_put_vc1_mspel_mc13_c;
449 dsp->put_vc1_mspel_pixels_tab[14] = ff_put_vc1_mspel_mc23_c;
450 dsp->put_vc1_mspel_pixels_tab[15] = ff_put_vc1_mspel_mc33_c;