Add support for VDPAU video out, including hardware decoding.
[mplayer/glamo.git] / libmpeg2 / idct_alpha.c
blobe5886ed3ff1bff7a6e4db3e71980c22efe74e2e4
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
25 #include "config.h"
27 #if ARCH_ALPHA
29 #include <stdlib.h>
30 #include <inttypes.h>
32 #include "mpeg2.h"
33 #include "attributes.h"
34 #include "mpeg2_internal.h"
35 #include "alpha_asm.h"
37 #define W1 2841 /* 2048 * sqrt (2) * cos (1 * pi / 16) */
38 #define W2 2676 /* 2048 * sqrt (2) * cos (2 * pi / 16) */
39 #define W3 2408 /* 2048 * sqrt (2) * cos (3 * pi / 16) */
40 #define W5 1609 /* 2048 * sqrt (2) * cos (5 * pi / 16) */
41 #define W6 1108 /* 2048 * sqrt (2) * cos (6 * pi / 16) */
42 #define W7 565 /* 2048 * sqrt (2) * cos (7 * pi / 16) */
44 extern uint8_t mpeg2_clip[3840 * 2 + 256];
45 #define CLIP(i) ((mpeg2_clip + 3840)[i])
47 #if 0
48 #define BUTTERFLY(t0,t1,W0,W1,d0,d1) \
49 do { \
50 t0 = W0 * d0 + W1 * d1; \
51 t1 = W0 * d1 - W1 * d0; \
52 } while (0)
53 #else
54 #define BUTTERFLY(t0,t1,W0,W1,d0,d1) \
55 do { \
56 int_fast32_t tmp = W0 * (d0 + d1); \
57 t0 = tmp + (W1 - W0) * d1; \
58 t1 = tmp - (W1 + W0) * d0; \
59 } while (0)
60 #endif
62 static inline void idct_row (int16_t * const block)
64 uint64_t l, r;
65 int_fast32_t d0, d1, d2, d3;
66 int_fast32_t a0, a1, a2, a3, b0, b1, b2, b3;
67 int_fast32_t t0, t1, t2, t3;
69 l = ldq (block);
70 r = ldq (block + 4);
72 /* shortcut */
73 if (likely (!((l & ~0xffffUL) | r))) {
74 uint64_t tmp = (uint16_t) (l >> 1);
75 tmp |= tmp << 16;
76 tmp |= tmp << 32;
77 ((int32_t *)block)[0] = tmp;
78 ((int32_t *)block)[1] = tmp;
79 ((int32_t *)block)[2] = tmp;
80 ((int32_t *)block)[3] = tmp;
81 return;
84 d0 = (sextw (l) << 11) + 2048;
85 d1 = sextw (extwl (l, 2));
86 d2 = sextw (extwl (l, 4)) << 11;
87 d3 = sextw (extwl (l, 6));
88 t0 = d0 + d2;
89 t1 = d0 - d2;
90 BUTTERFLY (t2, t3, W6, W2, d3, d1);
91 a0 = t0 + t2;
92 a1 = t1 + t3;
93 a2 = t1 - t3;
94 a3 = t0 - t2;
96 d0 = sextw (r);
97 d1 = sextw (extwl (r, 2));
98 d2 = sextw (extwl (r, 4));
99 d3 = sextw (extwl (r, 6));
100 BUTTERFLY (t0, t1, W7, W1, d3, d0);
101 BUTTERFLY (t2, t3, W3, W5, d1, d2);
102 b0 = t0 + t2;
103 b3 = t1 + t3;
104 t0 -= t2;
105 t1 -= t3;
106 b1 = ((t0 + t1) >> 8) * 181;
107 b2 = ((t0 - t1) >> 8) * 181;
109 block[0] = (a0 + b0) >> 12;
110 block[1] = (a1 + b1) >> 12;
111 block[2] = (a2 + b2) >> 12;
112 block[3] = (a3 + b3) >> 12;
113 block[4] = (a3 - b3) >> 12;
114 block[5] = (a2 - b2) >> 12;
115 block[6] = (a1 - b1) >> 12;
116 block[7] = (a0 - b0) >> 12;
119 static inline void idct_col (int16_t * const block)
121 int_fast32_t d0, d1, d2, d3;
122 int_fast32_t a0, a1, a2, a3, b0, b1, b2, b3;
123 int_fast32_t t0, t1, t2, t3;
125 d0 = (block[8*0] << 11) + 65536;
126 d1 = block[8*1];
127 d2 = block[8*2] << 11;
128 d3 = block[8*3];
129 t0 = d0 + d2;
130 t1 = d0 - d2;
131 BUTTERFLY (t2, t3, W6, W2, d3, d1);
132 a0 = t0 + t2;
133 a1 = t1 + t3;
134 a2 = t1 - t3;
135 a3 = t0 - t2;
137 d0 = block[8*4];
138 d1 = block[8*5];
139 d2 = block[8*6];
140 d3 = block[8*7];
141 BUTTERFLY (t0, t1, W7, W1, d3, d0);
142 BUTTERFLY (t2, t3, W3, W5, d1, d2);
143 b0 = t0 + t2;
144 b3 = t1 + t3;
145 t0 -= t2;
146 t1 -= t3;
147 b1 = ((t0 + t1) >> 8) * 181;
148 b2 = ((t0 - t1) >> 8) * 181;
150 block[8*0] = (a0 + b0) >> 17;
151 block[8*1] = (a1 + b1) >> 17;
152 block[8*2] = (a2 + b2) >> 17;
153 block[8*3] = (a3 + b3) >> 17;
154 block[8*4] = (a3 - b3) >> 17;
155 block[8*5] = (a2 - b2) >> 17;
156 block[8*6] = (a1 - b1) >> 17;
157 block[8*7] = (a0 - b0) >> 17;
160 void mpeg2_idct_copy_mvi (int16_t * block, uint8_t * dest, const int stride)
162 uint64_t clampmask;
163 int i;
165 for (i = 0; i < 8; i++)
166 idct_row (block + 8 * i);
168 for (i = 0; i < 8; i++)
169 idct_col (block + i);
171 clampmask = zap (-1, 0xaa); /* 0x00ff00ff00ff00ff */
172 do {
173 uint64_t shorts0, shorts1;
175 shorts0 = ldq (block);
176 shorts0 = maxsw4 (shorts0, 0);
177 shorts0 = minsw4 (shorts0, clampmask);
178 stl (pkwb (shorts0), dest);
180 shorts1 = ldq (block + 4);
181 shorts1 = maxsw4 (shorts1, 0);
182 shorts1 = minsw4 (shorts1, clampmask);
183 stl (pkwb (shorts1), dest + 4);
185 stq (0, block);
186 stq (0, block + 4);
188 dest += stride;
189 block += 8;
190 } while (--i);
193 void mpeg2_idct_add_mvi (const int last, int16_t * block,
194 uint8_t * dest, const int stride)
196 uint64_t clampmask;
197 uint64_t signmask;
198 int i;
200 if (last != 129 || (block[0] & (7 << 4)) == (4 << 4)) {
201 for (i = 0; i < 8; i++)
202 idct_row (block + 8 * i);
203 for (i = 0; i < 8; i++)
204 idct_col (block + i);
205 clampmask = zap (-1, 0xaa); /* 0x00ff00ff00ff00ff */
206 signmask = zap (-1, 0x33);
207 signmask ^= signmask >> 1; /* 0x8000800080008000 */
209 do {
210 uint64_t shorts0, pix0, signs0;
211 uint64_t shorts1, pix1, signs1;
213 shorts0 = ldq (block);
214 shorts1 = ldq (block + 4);
216 pix0 = unpkbw (ldl (dest));
217 /* signed subword add (MMX paddw). */
218 signs0 = shorts0 & signmask;
219 shorts0 &= ~signmask;
220 shorts0 += pix0;
221 shorts0 ^= signs0;
222 /* clamp. */
223 shorts0 = maxsw4 (shorts0, 0);
224 shorts0 = minsw4 (shorts0, clampmask);
226 /* next 4. */
227 pix1 = unpkbw (ldl (dest + 4));
228 signs1 = shorts1 & signmask;
229 shorts1 &= ~signmask;
230 shorts1 += pix1;
231 shorts1 ^= signs1;
232 shorts1 = maxsw4 (shorts1, 0);
233 shorts1 = minsw4 (shorts1, clampmask);
235 stl (pkwb (shorts0), dest);
236 stl (pkwb (shorts1), dest + 4);
237 stq (0, block);
238 stq (0, block + 4);
240 dest += stride;
241 block += 8;
242 } while (--i);
243 } else {
244 int DC;
245 uint64_t p0, p1, p2, p3, p4, p5, p6, p7;
246 uint64_t DCs;
248 DC = (block[0] + 64) >> 7;
249 block[0] = block[63] = 0;
251 p0 = ldq (dest + 0 * stride);
252 p1 = ldq (dest + 1 * stride);
253 p2 = ldq (dest + 2 * stride);
254 p3 = ldq (dest + 3 * stride);
255 p4 = ldq (dest + 4 * stride);
256 p5 = ldq (dest + 5 * stride);
257 p6 = ldq (dest + 6 * stride);
258 p7 = ldq (dest + 7 * stride);
260 if (DC > 0) {
261 DCs = BYTE_VEC (likely (DC <= 255) ? DC : 255);
262 p0 += minub8 (DCs, ~p0);
263 p1 += minub8 (DCs, ~p1);
264 p2 += minub8 (DCs, ~p2);
265 p3 += minub8 (DCs, ~p3);
266 p4 += minub8 (DCs, ~p4);
267 p5 += minub8 (DCs, ~p5);
268 p6 += minub8 (DCs, ~p6);
269 p7 += minub8 (DCs, ~p7);
270 } else {
271 DCs = BYTE_VEC (likely (-DC <= 255) ? -DC : 255);
272 p0 -= minub8 (DCs, p0);
273 p1 -= minub8 (DCs, p1);
274 p2 -= minub8 (DCs, p2);
275 p3 -= minub8 (DCs, p3);
276 p4 -= minub8 (DCs, p4);
277 p5 -= minub8 (DCs, p5);
278 p6 -= minub8 (DCs, p6);
279 p7 -= minub8 (DCs, p7);
282 stq (p0, dest + 0 * stride);
283 stq (p1, dest + 1 * stride);
284 stq (p2, dest + 2 * stride);
285 stq (p3, dest + 3 * stride);
286 stq (p4, dest + 4 * stride);
287 stq (p5, dest + 5 * stride);
288 stq (p6, dest + 6 * stride);
289 stq (p7, dest + 7 * stride);
293 void mpeg2_idct_copy_alpha (int16_t * block, uint8_t * dest, const int stride)
295 int i;
297 for (i = 0; i < 8; i++)
298 idct_row (block + 8 * i);
299 for (i = 0; i < 8; i++)
300 idct_col (block + i);
301 do {
302 dest[0] = CLIP (block[0]);
303 dest[1] = CLIP (block[1]);
304 dest[2] = CLIP (block[2]);
305 dest[3] = CLIP (block[3]);
306 dest[4] = CLIP (block[4]);
307 dest[5] = CLIP (block[5]);
308 dest[6] = CLIP (block[6]);
309 dest[7] = CLIP (block[7]);
311 stq(0, block);
312 stq(0, block + 4);
314 dest += stride;
315 block += 8;
316 } while (--i);
319 void mpeg2_idct_add_alpha (const int last, int16_t * block,
320 uint8_t * dest, const int stride)
322 int i;
324 if (last != 129 || (block[0] & (7 << 4)) == (4 << 4)) {
325 for (i = 0; i < 8; i++)
326 idct_row (block + 8 * i);
327 for (i = 0; i < 8; i++)
328 idct_col (block + i);
329 do {
330 dest[0] = CLIP (block[0] + dest[0]);
331 dest[1] = CLIP (block[1] + dest[1]);
332 dest[2] = CLIP (block[2] + dest[2]);
333 dest[3] = CLIP (block[3] + dest[3]);
334 dest[4] = CLIP (block[4] + dest[4]);
335 dest[5] = CLIP (block[5] + dest[5]);
336 dest[6] = CLIP (block[6] + dest[6]);
337 dest[7] = CLIP (block[7] + dest[7]);
339 stq(0, block);
340 stq(0, block + 4);
342 dest += stride;
343 block += 8;
344 } while (--i);
345 } else {
346 int DC;
348 DC = (block[0] + 64) >> 7;
349 block[0] = block[63] = 0;
350 i = 8;
351 do {
352 dest[0] = CLIP (DC + dest[0]);
353 dest[1] = CLIP (DC + dest[1]);
354 dest[2] = CLIP (DC + dest[2]);
355 dest[3] = CLIP (DC + dest[3]);
356 dest[4] = CLIP (DC + dest[4]);
357 dest[5] = CLIP (DC + dest[5]);
358 dest[6] = CLIP (DC + dest[6]);
359 dest[7] = CLIP (DC + dest[7]);
360 dest += stride;
361 } while (--i);
365 void mpeg2_idct_alpha_init (void)
367 int i, j;
369 for (i = 0; i < 64; i++) {
370 j = mpeg2_scan_norm[i];
371 mpeg2_scan_norm[i] = ((j & 0x36) >> 1) | ((j & 0x09) << 2);
372 j = mpeg2_scan_alt[i];
373 mpeg2_scan_alt[i] = ((j & 0x36) >> 1) | ((j & 0x09) << 2);
377 #endif /* ARCH_ALPHA */