Fixes
[mplayer/glamo.git] / libmpeg2 / idct_alpha.c
blob74ebedf43d84bf877f7193394001b3464c9f5ca2
1 /*
2 * idct_alpha.c
3 * Copyright (C) 2002-2003 Falk Hueffner <falk@debian.org>
4 * Copyright (C) 2000-2003 Michel Lespinasse <walken@zoy.org>
5 * Copyright (C) 1999-2000 Aaron Holtzman <aholtzma@ess.engr.uvic.ca>
7 * This file is part of mpeg2dec, a free MPEG-2 video stream decoder.
8 * See http://libmpeg2.sourceforge.net/ for updates.
10 * mpeg2dec is free software; you can redistribute it and/or modify
11 * it under the terms of the GNU General Public License as published by
12 * the Free Software Foundation; either version 2 of the License, or
13 * (at your option) any later version.
15 * mpeg2dec is distributed in the hope that it will be useful,
16 * but WITHOUT ANY WARRANTY; without even the implied warranty of
17 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
18 * GNU General Public License for more details.
20 * You should have received a copy of the GNU General Public License
21 * along with this program; if not, write to the Free Software
22 * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
24 * Modified for use with MPlayer, see libmpeg-0.4.0.diff for the exact changes.
25 * detailed changelog at http://svn.mplayerhq.hu/mplayer/trunk/
26 * $Id$
29 #include "config.h"
31 #ifdef ARCH_ALPHA
33 #include <stdlib.h>
34 #include <inttypes.h>
36 #include "mpeg2.h"
37 #include "attributes.h"
38 #include "mpeg2_internal.h"
39 #include "alpha_asm.h"
41 #define W1 2841 /* 2048 * sqrt (2) * cos (1 * pi / 16) */
42 #define W2 2676 /* 2048 * sqrt (2) * cos (2 * pi / 16) */
43 #define W3 2408 /* 2048 * sqrt (2) * cos (3 * pi / 16) */
44 #define W5 1609 /* 2048 * sqrt (2) * cos (5 * pi / 16) */
45 #define W6 1108 /* 2048 * sqrt (2) * cos (6 * pi / 16) */
46 #define W7 565 /* 2048 * sqrt (2) * cos (7 * pi / 16) */
48 extern uint8_t mpeg2_clip[3840 * 2 + 256];
49 #define CLIP(i) ((mpeg2_clip + 3840)[i])
51 #if 0
52 #define BUTTERFLY(t0,t1,W0,W1,d0,d1) \
53 do { \
54 t0 = W0 * d0 + W1 * d1; \
55 t1 = W0 * d1 - W1 * d0; \
56 } while (0)
57 #else
58 #define BUTTERFLY(t0,t1,W0,W1,d0,d1) \
59 do { \
60 int_fast32_t tmp = W0 * (d0 + d1); \
61 t0 = tmp + (W1 - W0) * d1; \
62 t1 = tmp - (W1 + W0) * d0; \
63 } while (0)
64 #endif
66 static inline void idct_row (int16_t * const block)
68 uint64_t l, r;
69 int_fast32_t d0, d1, d2, d3;
70 int_fast32_t a0, a1, a2, a3, b0, b1, b2, b3;
71 int_fast32_t t0, t1, t2, t3;
73 l = ldq (block);
74 r = ldq (block + 4);
76 /* shortcut */
77 if (likely (!((l & ~0xffffUL) | r))) {
78 uint64_t tmp = (uint16_t) (l >> 1);
79 tmp |= tmp << 16;
80 tmp |= tmp << 32;
81 ((int32_t *)block)[0] = tmp;
82 ((int32_t *)block)[1] = tmp;
83 ((int32_t *)block)[2] = tmp;
84 ((int32_t *)block)[3] = tmp;
85 return;
88 d0 = (sextw (l) << 11) + 2048;
89 d1 = sextw (extwl (l, 2));
90 d2 = sextw (extwl (l, 4)) << 11;
91 d3 = sextw (extwl (l, 6));
92 t0 = d0 + d2;
93 t1 = d0 - d2;
94 BUTTERFLY (t2, t3, W6, W2, d3, d1);
95 a0 = t0 + t2;
96 a1 = t1 + t3;
97 a2 = t1 - t3;
98 a3 = t0 - t2;
100 d0 = sextw (r);
101 d1 = sextw (extwl (r, 2));
102 d2 = sextw (extwl (r, 4));
103 d3 = sextw (extwl (r, 6));
104 BUTTERFLY (t0, t1, W7, W1, d3, d0);
105 BUTTERFLY (t2, t3, W3, W5, d1, d2);
106 b0 = t0 + t2;
107 b3 = t1 + t3;
108 t0 -= t2;
109 t1 -= t3;
110 b1 = ((t0 + t1) >> 8) * 181;
111 b2 = ((t0 - t1) >> 8) * 181;
113 block[0] = (a0 + b0) >> 12;
114 block[1] = (a1 + b1) >> 12;
115 block[2] = (a2 + b2) >> 12;
116 block[3] = (a3 + b3) >> 12;
117 block[4] = (a3 - b3) >> 12;
118 block[5] = (a2 - b2) >> 12;
119 block[6] = (a1 - b1) >> 12;
120 block[7] = (a0 - b0) >> 12;
123 static inline void idct_col (int16_t * const block)
125 int_fast32_t d0, d1, d2, d3;
126 int_fast32_t a0, a1, a2, a3, b0, b1, b2, b3;
127 int_fast32_t t0, t1, t2, t3;
129 d0 = (block[8*0] << 11) + 65536;
130 d1 = block[8*1];
131 d2 = block[8*2] << 11;
132 d3 = block[8*3];
133 t0 = d0 + d2;
134 t1 = d0 - d2;
135 BUTTERFLY (t2, t3, W6, W2, d3, d1);
136 a0 = t0 + t2;
137 a1 = t1 + t3;
138 a2 = t1 - t3;
139 a3 = t0 - t2;
141 d0 = block[8*4];
142 d1 = block[8*5];
143 d2 = block[8*6];
144 d3 = block[8*7];
145 BUTTERFLY (t0, t1, W7, W1, d3, d0);
146 BUTTERFLY (t2, t3, W3, W5, d1, d2);
147 b0 = t0 + t2;
148 b3 = t1 + t3;
149 t0 -= t2;
150 t1 -= t3;
151 b1 = ((t0 + t1) >> 8) * 181;
152 b2 = ((t0 - t1) >> 8) * 181;
154 block[8*0] = (a0 + b0) >> 17;
155 block[8*1] = (a1 + b1) >> 17;
156 block[8*2] = (a2 + b2) >> 17;
157 block[8*3] = (a3 + b3) >> 17;
158 block[8*4] = (a3 - b3) >> 17;
159 block[8*5] = (a2 - b2) >> 17;
160 block[8*6] = (a1 - b1) >> 17;
161 block[8*7] = (a0 - b0) >> 17;
164 #ifdef CAN_COMPILE_ALPHA_MVI
165 void mpeg2_idct_copy_mvi (int16_t * block, uint8_t * dest, const int stride)
167 uint64_t clampmask;
168 int i;
170 for (i = 0; i < 8; i++)
171 idct_row (block + 8 * i);
173 for (i = 0; i < 8; i++)
174 idct_col (block + i);
176 clampmask = zap (-1, 0xaa); /* 0x00ff00ff00ff00ff */
177 do {
178 uint64_t shorts0, shorts1;
180 shorts0 = ldq (block);
181 shorts0 = maxsw4 (shorts0, 0);
182 shorts0 = minsw4 (shorts0, clampmask);
183 stl (pkwb (shorts0), dest);
185 shorts1 = ldq (block + 4);
186 shorts1 = maxsw4 (shorts1, 0);
187 shorts1 = minsw4 (shorts1, clampmask);
188 stl (pkwb (shorts1), dest + 4);
190 stq (0, block);
191 stq (0, block + 4);
193 dest += stride;
194 block += 8;
195 } while (--i);
198 void mpeg2_idct_add_mvi (const int last, int16_t * block,
199 uint8_t * dest, const int stride)
201 uint64_t clampmask;
202 uint64_t signmask;
203 int i;
205 if (last != 129 || (block[0] & (7 << 4)) == (4 << 4)) {
206 for (i = 0; i < 8; i++)
207 idct_row (block + 8 * i);
208 for (i = 0; i < 8; i++)
209 idct_col (block + i);
210 clampmask = zap (-1, 0xaa); /* 0x00ff00ff00ff00ff */
211 signmask = zap (-1, 0x33);
212 signmask ^= signmask >> 1; /* 0x8000800080008000 */
214 do {
215 uint64_t shorts0, pix0, signs0;
216 uint64_t shorts1, pix1, signs1;
218 shorts0 = ldq (block);
219 shorts1 = ldq (block + 4);
221 pix0 = unpkbw (ldl (dest));
222 /* signed subword add (MMX paddw). */
223 signs0 = shorts0 & signmask;
224 shorts0 &= ~signmask;
225 shorts0 += pix0;
226 shorts0 ^= signs0;
227 /* clamp. */
228 shorts0 = maxsw4 (shorts0, 0);
229 shorts0 = minsw4 (shorts0, clampmask);
231 /* next 4. */
232 pix1 = unpkbw (ldl (dest + 4));
233 signs1 = shorts1 & signmask;
234 shorts1 &= ~signmask;
235 shorts1 += pix1;
236 shorts1 ^= signs1;
237 shorts1 = maxsw4 (shorts1, 0);
238 shorts1 = minsw4 (shorts1, clampmask);
240 stl (pkwb (shorts0), dest);
241 stl (pkwb (shorts1), dest + 4);
242 stq (0, block);
243 stq (0, block + 4);
245 dest += stride;
246 block += 8;
247 } while (--i);
248 } else {
249 int DC;
250 uint64_t p0, p1, p2, p3, p4, p5, p6, p7;
251 uint64_t DCs;
253 DC = (block[0] + 64) >> 7;
254 block[0] = block[63] = 0;
256 p0 = ldq (dest + 0 * stride);
257 p1 = ldq (dest + 1 * stride);
258 p2 = ldq (dest + 2 * stride);
259 p3 = ldq (dest + 3 * stride);
260 p4 = ldq (dest + 4 * stride);
261 p5 = ldq (dest + 5 * stride);
262 p6 = ldq (dest + 6 * stride);
263 p7 = ldq (dest + 7 * stride);
265 if (DC > 0) {
266 DCs = BYTE_VEC (likely (DC <= 255) ? DC : 255);
267 p0 += minub8 (DCs, ~p0);
268 p1 += minub8 (DCs, ~p1);
269 p2 += minub8 (DCs, ~p2);
270 p3 += minub8 (DCs, ~p3);
271 p4 += minub8 (DCs, ~p4);
272 p5 += minub8 (DCs, ~p5);
273 p6 += minub8 (DCs, ~p6);
274 p7 += minub8 (DCs, ~p7);
275 } else {
276 DCs = BYTE_VEC (likely (-DC <= 255) ? -DC : 255);
277 p0 -= minub8 (DCs, p0);
278 p1 -= minub8 (DCs, p1);
279 p2 -= minub8 (DCs, p2);
280 p3 -= minub8 (DCs, p3);
281 p4 -= minub8 (DCs, p4);
282 p5 -= minub8 (DCs, p5);
283 p6 -= minub8 (DCs, p6);
284 p7 -= minub8 (DCs, p7);
287 stq (p0, dest + 0 * stride);
288 stq (p1, dest + 1 * stride);
289 stq (p2, dest + 2 * stride);
290 stq (p3, dest + 3 * stride);
291 stq (p4, dest + 4 * stride);
292 stq (p5, dest + 5 * stride);
293 stq (p6, dest + 6 * stride);
294 stq (p7, dest + 7 * stride);
297 #endif
299 void mpeg2_idct_copy_alpha (int16_t * block, uint8_t * dest, const int stride)
301 int i;
303 for (i = 0; i < 8; i++)
304 idct_row (block + 8 * i);
305 for (i = 0; i < 8; i++)
306 idct_col (block + i);
307 do {
308 dest[0] = CLIP (block[0]);
309 dest[1] = CLIP (block[1]);
310 dest[2] = CLIP (block[2]);
311 dest[3] = CLIP (block[3]);
312 dest[4] = CLIP (block[4]);
313 dest[5] = CLIP (block[5]);
314 dest[6] = CLIP (block[6]);
315 dest[7] = CLIP (block[7]);
317 stq(0, block);
318 stq(0, block + 4);
320 dest += stride;
321 block += 8;
322 } while (--i);
325 void mpeg2_idct_add_alpha (const int last, int16_t * block,
326 uint8_t * dest, const int stride)
328 int i;
330 if (last != 129 || (block[0] & (7 << 4)) == (4 << 4)) {
331 for (i = 0; i < 8; i++)
332 idct_row (block + 8 * i);
333 for (i = 0; i < 8; i++)
334 idct_col (block + i);
335 do {
336 dest[0] = CLIP (block[0] + dest[0]);
337 dest[1] = CLIP (block[1] + dest[1]);
338 dest[2] = CLIP (block[2] + dest[2]);
339 dest[3] = CLIP (block[3] + dest[3]);
340 dest[4] = CLIP (block[4] + dest[4]);
341 dest[5] = CLIP (block[5] + dest[5]);
342 dest[6] = CLIP (block[6] + dest[6]);
343 dest[7] = CLIP (block[7] + dest[7]);
345 stq(0, block);
346 stq(0, block + 4);
348 dest += stride;
349 block += 8;
350 } while (--i);
351 } else {
352 int DC;
354 DC = (block[0] + 64) >> 7;
355 block[0] = block[63] = 0;
356 i = 8;
357 do {
358 dest[0] = CLIP (DC + dest[0]);
359 dest[1] = CLIP (DC + dest[1]);
360 dest[2] = CLIP (DC + dest[2]);
361 dest[3] = CLIP (DC + dest[3]);
362 dest[4] = CLIP (DC + dest[4]);
363 dest[5] = CLIP (DC + dest[5]);
364 dest[6] = CLIP (DC + dest[6]);
365 dest[7] = CLIP (DC + dest[7]);
366 dest += stride;
367 } while (--i);
371 void mpeg2_idct_alpha_init (void)
373 extern uint8_t mpeg2_scan_norm[64];
374 extern uint8_t mpeg2_scan_alt[64];
375 int i, j;
377 for (i = 0; i < 64; i++) {
378 j = mpeg2_scan_norm[i];
379 mpeg2_scan_norm[i] = ((j & 0x36) >> 1) | ((j & 0x09) << 2);
380 j = mpeg2_scan_alt[i];
381 mpeg2_scan_alt[i] = ((j & 0x36) >> 1) | ((j & 0x09) << 2);
385 #endif /* ARCH_ALPHA */