idct_mmi.c
Go to the documentation of this file.
1 /*
2  * Originally provided by Intel at Application Note AP-922.
3  *
4  * Column code adapted from Peter Gubanov.
5  * Copyright (c) 2000-2001 Peter Gubanov <peter@elecard.net.ru>
6  * http://www.elecard.com/peter/idct.shtml
7  * rounding trick copyright (c) 2000 Michel Lespinasse <walken@zoy.org>
8  *
9  * MMI port and (c) 2002 by Leon van Stuivenberg
10  *
11  * This file is part of Libav.
12  *
13  * Libav is free software; you can redistribute it and/or
14  * modify it under the terms of the GNU Lesser General Public
15  * License as published by the Free Software Foundation; either
16  * version 2.1 of the License, or (at your option) any later version.
17  *
18  * Libav is distributed in the hope that it will be useful,
19  * but WITHOUT ANY WARRANTY; without even the implied warranty of
20  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
21  * Lesser General Public License for more details.
22  *
23  * You should have received a copy of the GNU Lesser General Public
24  * License along with Libav; if not, write to the Free Software
25  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
26  */
27 
28 #include "libavutil/common.h"
29 #include "libavcodec/dsputil.h"
30 #include "mmi.h"
31 
32 #define BITS_INV_ACC 5 // 4 or 5 for IEEE
33 #define SHIFT_INV_ROW (16 - BITS_INV_ACC)
34 #define SHIFT_INV_COL (1 + BITS_INV_ACC)
35 
36 #define TG1 6518
37 #define TG2 13573
38 #define TG3 21895
39 #define CS4 23170
40 
41 #define ROUNDER_0 0
42 #define ROUNDER_1 16
43 
44 #define TAB_i_04 (32+0)
45 #define TAB_i_17 (32+64)
46 #define TAB_i_26 (32+128)
47 #define TAB_i_35 (32+192)
48 
49 #define TG_1_16 (32+256+0)
50 #define TG_2_16 (32+256+16)
51 #define TG_3_16 (32+256+32)
52 #define COS_4_16 (32+256+48)
53 
54 #define CLIPMAX (32+256+64+0)
55 
56 static short consttable[] align16 = {
57 /* rounder 0*/ // assume SHIFT_INV_ROW == 11
58  0x3ff, 1, 0x3ff, 1, 0x3ff, 1, 0x3ff, 1,
59 /* rounder 1*/
60  0x3ff, 0, 0x3ff, 0, 0x3ff, 0, 0x3ff, 0,
61 /* row 0/4*/
62  16384, 21407, -16384, -21407, 22725, 19266, -22725, -12873,
63  8867, 16384, 8867, 16384, 4520, 12873, -4520, 19266,
64  16384, -8867, 16384, -8867, 12873, -22725, 19266, -22725,
65  21407, -16384, -21407, 16384, 19266, 4520, -12873, 4520,
66 /* row 1/7*/
67  22725, 29692, -22725, -29692, 31521, 26722, -31521, -17855,
68  12299, 22725, 12299, 22725, 6270, 17855, -6270, 26722,
69  22725, -12299, 22725, -12299, 17855, -31521, 26722, -31521,
70  29692, -22725, -29692, 22725, 26722, 6270, -17855, 6270,
71 /* row 2/6*/
72  21407, 27969, -21407, -27969, 29692, 25172, -29692, -16819,
73  11585, 21407, 11585, 21407, 5906, 16819, -5906, 25172,
74  21407, -11585, 21407, -11585, 16819, -29692, 25172, -29692,
75  27969, -21407, -27969, 21407, 25172, 5906, -16819, 5906,
76 /*row 3/5*/
77  19266, 25172, -19266, -25172, 26722, 22654, -26722, -15137,
78  10426, 19266, 10426, 19266, 5315, 15137, -5315, 22654,
79  19266, -10426, 19266, -10426, 15137, -26722, 22654, -26722,
80  25172, -19266, -25172, 19266, 22654, 5315, -15137, 5315,
81 /*column constants*/
82  TG1, TG1, TG1, TG1, TG1, TG1, TG1, TG1,
83  TG2, TG2, TG2, TG2, TG2, TG2, TG2, TG2,
84  TG3, TG3, TG3, TG3, TG3, TG3, TG3, TG3,
85  CS4, CS4, CS4, CS4, CS4, CS4, CS4, CS4,
86 /* clamp */
87  255, 255, 255, 255, 255, 255, 255, 255
88 };
89 
90 
91 #define DCT_8_INV_ROW1(blk, rowoff, taboff, rnd, outreg) { \
92  lq(blk, rowoff, $16); /* r16 = x7 x5 x3 x1 x6 x4 x2 x0 */ \
93  /*slot*/ \
94  lq($24, 0+taboff, $17); /* r17 = w */ \
95  /*delay slot $16*/ \
96  lq($24, 16+taboff, $18);/* r18 = w */ \
97  prevh($16, $2); /* r2 = x1 x3 x5 x7 x0 x2 x4 x6 */ \
98  lq($24, 32+taboff, $19);/* r19 = w */ \
99  phmadh($17, $16, $17); /* r17 = b1"b0'a1"a0' */ \
100  lq($24, 48+taboff, $20);/* r20 = w */ \
101  phmadh($18, $2, $18); /* r18 = b1'b0"a1'a0" */ \
102  phmadh($19, $16, $19); /* r19 = b3"b2'a3"a2' */ \
103  phmadh($20, $2, $20); /* r20 = b3'b2"a3'a2" */ \
104  paddw($17, $18, $17); /* r17 = (b1)(b0)(a1)(a0) */ \
105  paddw($19, $20, $19); /* r19 = (b3)(b2)(a3)(a2) */ \
106  pcpyld($19, $17, $18); /* r18 = (a3)(a2)(a1)(a0) */ \
107  pcpyud($17, $19, $20); /* r20 = (b3)(b2)(b1)(b0) */ \
108  paddw($18, rnd, $18); /* r18 = (a3)(a2)(a1)(a0) */\
109  paddw($18, $20, $17); /* r17 = ()()()(a0+b0) */ \
110  psubw($18, $20, $20); /* r20 = ()()()(a0-b0) */ \
111  psraw($17, SHIFT_INV_ROW, $17); /* r17 = (y3 y2 y1 y0) */ \
112  psraw($20, SHIFT_INV_ROW, $20); /* r20 = (y4 y5 y6 y7) */ \
113  ppach($20, $17, outreg);/* out = y4 y5 y6 y7 y3 y2 y1 y0 Note order */ \
114 \
115  prevh(outreg, $2); \
116  pcpyud($2, $2, $2); \
117  pcpyld($2, outreg, outreg); \
118 }
119 
120 
121 #define DCT_8_INV_COL8() \
122 \
123  lq($24, TG_3_16, $2); /* r2 = tn3 */ \
124 \
125  pmulth($11, $2, $17); /* r17 = x3 * tn3 (6420) */ \
126  psraw($17, 15, $17); \
127  pmfhl_uw($3); /* r3 = 7531 */ \
128  psraw($3, 15, $3); \
129  pinteh($3, $17, $17); /* r17 = x3 * tn3 */ \
130  psubh($17, $13, $17); /* r17 = tm35 */ \
131 \
132  pmulth($13, $2, $18); /* r18 = x5 * tn3 (6420) */ \
133  psraw($18, 15, $18); \
134  pmfhl_uw($3); /* r3 = 7531 */ \
135  psraw($3, 15, $3); \
136  pinteh($3, $18, $18); /* r18 = x5 * tn3 */ \
137  paddh($18, $11, $18); /* r18 = tp35 */ \
138 \
139  lq($24, TG_1_16, $2); /* r2 = tn1 */ \
140 \
141  pmulth($15, $2, $19); /* r19 = x7 * tn1 (6420) */ \
142  psraw($19, 15, $19); \
143  pmfhl_uw($3); /* r3 = 7531 */ \
144  psraw($3, 15, $3); \
145  pinteh($3, $19, $19); /* r19 = x7 * tn1 */ \
146  paddh($19, $9, $19); /* r19 = tp17 */ \
147 \
148  pmulth($9, $2, $20); /* r20 = x1 * tn1 (6420) */ \
149  psraw($20, 15, $20); \
150  pmfhl_uw($3); /* r3 = 7531 */ \
151  psraw($3, 15, $3); \
152  pinteh($3, $20, $20); /* r20 = x1 * tn1 */ \
153  psubh($20, $15, $20); /* r20 = tm17 */ \
154 \
155  psubh($19, $18, $3); /* r3 = t1 */ \
156  paddh($20, $17, $16); /* r16 = t2 */ \
157  psubh($20, $17, $23); /* r23 = b3 */ \
158  paddh($19, $18, $20); /* r20 = b0 */ \
159 \
160  lq($24, COS_4_16, $2); /* r2 = cs4 */ \
161 \
162  paddh($3, $16, $21); /* r21 = t1+t2 */ \
163  psubh($3, $16, $22); /* r22 = t1-t2 */ \
164 \
165  pmulth($21, $2, $21); /* r21 = cs4 * (t1+t2) 6420 */ \
166  psraw($21, 15, $21); \
167  pmfhl_uw($3); /* r3 = 7531 */ \
168  psraw($3, 15, $3); \
169  pinteh($3, $21, $21); /* r21 = b1 */ \
170 \
171  pmulth($22, $2, $22); /* r22 = cs4 * (t1-t2) 6420 */ \
172  psraw($22, 15, $22); \
173  pmfhl_uw($3); /* r3 = 7531 */ \
174  psraw($3, 15, $3); \
175  pinteh($3, $22, $22); /* r22 = b2 */ \
176 \
177  lq($24, TG_2_16, $2); /* r2 = tn2 */ \
178 \
179  pmulth($10, $2, $17); /* r17 = x2 * tn2 (6420) */ \
180  psraw($17, 15, $17); \
181  pmfhl_uw($3); /* r3 = 7531 */ \
182  psraw($3, 15, $3); \
183  pinteh($3, $17, $17); /* r17 = x3 * tn3 */ \
184  psubh($17, $14, $17); /* r17 = tm26 */ \
185 \
186  pmulth($14, $2, $18); /* r18 = x6 * tn2 (6420) */ \
187  psraw($18, 15, $18); \
188  pmfhl_uw($3); /* r3 = 7531 */ \
189  psraw($3, 15, $3); \
190  pinteh($3, $18, $18); /* r18 = x6 * tn2 */ \
191  paddh($18, $10, $18); /* r18 = tp26 */ \
192 \
193  paddh($8, $12, $2); /* r2 = tp04 */ \
194  psubh($8, $12, $3); /* r3 = tm04 */ \
195 \
196  paddh($2, $18, $16); /* r16 = a0 */ \
197  psubh($2, $18, $19); /* r19 = a3 */ \
198  psubh($3, $17, $18); /* r18 = a2 */ \
199  paddh($3, $17, $17); /* r17 = a1 */
200 
201 
202 #define DCT_8_INV_COL8_STORE(blk) \
203 \
204  paddh($16, $20, $2); /* y0 a0+b0 */ \
205  psubh($16, $20, $16); /* y7 a0-b0 */ \
206  psrah($2, SHIFT_INV_COL, $2); \
207  psrah($16, SHIFT_INV_COL, $16); \
208  sq($2, 0, blk); \
209  sq($16, 112, blk); \
210 \
211  paddh($17, $21, $3); /* y1 a1+b1 */ \
212  psubh($17, $21, $17); /* y6 a1-b1 */ \
213  psrah($3, SHIFT_INV_COL, $3); \
214  psrah($17, SHIFT_INV_COL, $17); \
215  sq($3, 16, blk); \
216  sq($17, 96, blk); \
217 \
218  paddh($18, $22, $2); /* y2 a2+b2 */ \
219  psubh($18, $22, $18); /* y5 a2-b2 */ \
220  psrah($2, SHIFT_INV_COL, $2); \
221  psrah($18, SHIFT_INV_COL, $18); \
222  sq($2, 32, blk); \
223  sq($18, 80, blk); \
224 \
225  paddh($19, $23, $3); /* y3 a3+b3 */ \
226  psubh($19, $23, $19); /* y4 a3-b3 */ \
227  psrah($3, SHIFT_INV_COL, $3); \
228  psrah($19, SHIFT_INV_COL, $19); \
229  sq($3, 48, blk); \
230  sq($19, 64, blk);
231 
232 
233 
234 #define DCT_8_INV_COL8_PMS() \
235  paddh($16, $20, $2); /* y0 a0+b0 */ \
236  psubh($16, $20, $20); /* y7 a0-b0 */ \
237  psrah($2, SHIFT_INV_COL, $16); \
238  psrah($20, SHIFT_INV_COL, $20); \
239 \
240  paddh($17, $21, $3); /* y1 a1+b1 */ \
241  psubh($17, $21, $21); /* y6 a1-b1 */ \
242  psrah($3, SHIFT_INV_COL, $17); \
243  psrah($21, SHIFT_INV_COL, $21); \
244 \
245  paddh($18, $22, $2); /* y2 a2+b2 */ \
246  psubh($18, $22, $22); /* y5 a2-b2 */ \
247  psrah($2, SHIFT_INV_COL, $18); \
248  psrah($22, SHIFT_INV_COL, $22); \
249 \
250  paddh($19, $23, $3); /* y3 a3+b3 */ \
251  psubh($19, $23, $23); /* y4 a3-b3 */ \
252  psrah($3, SHIFT_INV_COL, $19); \
253  psrah($23, SHIFT_INV_COL, $23);
254 
255 #define PUT(rs) \
256  pminh(rs, $11, $2); \
257  pmaxh($2, $0, $2); \
258  ppacb($0, $2, $2); \
259  sd3(2, 0, 4); \
260  __asm__ volatile ("add $4, $5, $4");
261 
262 #define DCT_8_INV_COL8_PUT() \
263  PUT($16); \
264  PUT($17); \
265  PUT($18); \
266  PUT($19); \
267  PUT($23); \
268  PUT($22); \
269  PUT($21); \
270  PUT($20);
271 
272 #define ADD(rs) \
273  ld3(4, 0, 2); \
274  pextlb($0, $2, $2); \
275  paddh($2, rs, $2); \
276  pminh($2, $11, $2); \
277  pmaxh($2, $0, $2); \
278  ppacb($0, $2, $2); \
279  sd3(2, 0, 4); \
280  __asm__ volatile ("add $4, $5, $4");
281 
282 /*fixme: schedule*/
283 #define DCT_8_INV_COL8_ADD() \
284  ADD($16); \
285  ADD($17); \
286  ADD($18); \
287  ADD($19); \
288  ADD($23); \
289  ADD($22); \
290  ADD($21); \
291  ADD($20);
292 
293 
294 void ff_mmi_idct(int16_t * block)
295 {
296  /* $4 = block */
297  __asm__ volatile("la $24, %0"::"m"(consttable[0]));
298  lq($24, ROUNDER_0, $8);
299  lq($24, ROUNDER_1, $7);
300  DCT_8_INV_ROW1($4, 0, TAB_i_04, $8, $8);
301  DCT_8_INV_ROW1($4, 16, TAB_i_17, $7, $9);
302  DCT_8_INV_ROW1($4, 32, TAB_i_26, $7, $10);
303  DCT_8_INV_ROW1($4, 48, TAB_i_35, $7, $11);
304  DCT_8_INV_ROW1($4, 64, TAB_i_04, $7, $12);
305  DCT_8_INV_ROW1($4, 80, TAB_i_35, $7, $13);
306  DCT_8_INV_ROW1($4, 96, TAB_i_26, $7, $14);
307  DCT_8_INV_ROW1($4, 112, TAB_i_17, $7, $15);
308  DCT_8_INV_COL8();
310 
311  //let savedtemp regs be saved
312  __asm__ volatile(" ":::"$16", "$17", "$18", "$19", "$20", "$21", "$22", "$23");
313 }
314 
315 
316 void ff_mmi_idct_put(uint8_t *dest, int line_size, DCTELEM *block)
317 {
318  /* $4 = dest, $5 = line_size, $6 = block */
319  __asm__ volatile("la $24, %0"::"m"(consttable[0]));
320  lq($24, ROUNDER_0, $8);
321  lq($24, ROUNDER_1, $7);
322  DCT_8_INV_ROW1($6, 0, TAB_i_04, $8, $8);
323  DCT_8_INV_ROW1($6, 16, TAB_i_17, $7, $9);
324  DCT_8_INV_ROW1($6, 32, TAB_i_26, $7, $10);
325  DCT_8_INV_ROW1($6, 48, TAB_i_35, $7, $11);
326  DCT_8_INV_ROW1($6, 64, TAB_i_04, $7, $12);
327  DCT_8_INV_ROW1($6, 80, TAB_i_35, $7, $13);
328  DCT_8_INV_ROW1($6, 96, TAB_i_26, $7, $14);
329  DCT_8_INV_ROW1($6, 112, TAB_i_17, $7, $15);
330  DCT_8_INV_COL8();
331  lq($24, CLIPMAX, $11);
334 
335  //let savedtemp regs be saved
336  __asm__ volatile(" ":::"$16", "$17", "$18", "$19", "$20", "$21", "$22", "$23");
337 }
338 
339 
340 void ff_mmi_idct_add(uint8_t *dest, int line_size, DCTELEM *block)
341 {
342  /* $4 = dest, $5 = line_size, $6 = block */
343  __asm__ volatile("la $24, %0"::"m"(consttable[0]));
344  lq($24, ROUNDER_0, $8);
345  lq($24, ROUNDER_1, $7);
346  DCT_8_INV_ROW1($6, 0, TAB_i_04, $8, $8);
347  DCT_8_INV_ROW1($6, 16, TAB_i_17, $7, $9);
348  DCT_8_INV_ROW1($6, 32, TAB_i_26, $7, $10);
349  DCT_8_INV_ROW1($6, 48, TAB_i_35, $7, $11);
350  DCT_8_INV_ROW1($6, 64, TAB_i_04, $7, $12);
351  DCT_8_INV_ROW1($6, 80, TAB_i_35, $7, $13);
352  DCT_8_INV_ROW1($6, 96, TAB_i_26, $7, $14);
353  DCT_8_INV_ROW1($6, 112, TAB_i_17, $7, $15);
354  DCT_8_INV_COL8();
355  lq($24, CLIPMAX, $11);
358 
359  //let savedtemp regs be saved
360  __asm__ volatile(" ":::"$16", "$17", "$18", "$19", "$20", "$21", "$22", "$23");
361 }
362