vc1dsp_altivec.c 11.8 KB
Newer Older
1 2 3 4
/*
 * VC-1 and WMV3 decoder - DSP functions AltiVec-optimized
 * Copyright (c) 2006 Konstantin Shishkov
 *
5 6 7
 * This file is part of FFmpeg.
 *
 * FFmpeg is free software; you can redistribute it and/or
8 9
 * modify it under the terms of the GNU Lesser General Public
 * License as published by the Free Software Foundation; either
10
 * version 2.1 of the License, or (at your option) any later version.
11
 *
12
 * FFmpeg is distributed in the hope that it will be useful,
13 14 15 16 17
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
 * Lesser General Public License for more details.
 *
 * You should have received a copy of the GNU Lesser General Public
18
 * License along with FFmpeg; if not, write to the Free Software
19 20 21
 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
 */

22
#include "config.h"
23
#include "libavutil/attributes.h"
24
#include "libavutil/cpu.h"
25
#include "libavutil/ppc/cpu.h"
26 27
#include "libavutil/ppc/types_altivec.h"
#include "libavutil/ppc/util_altivec.h"
Ronald S. Bultje's avatar
Ronald S. Bultje committed
28
#include "libavcodec/vc1dsp.h"
29

30 31
#if HAVE_ALTIVEC

32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136
// main steps of 8x8 transform
#define STEP8(s0, s1, s2, s3, s4, s5, s6, s7, vec_rnd) \
do { \
    t0 = vec_sl(vec_add(s0, s4), vec_2); \
    t0 = vec_add(vec_sl(t0, vec_1), t0); \
    t0 = vec_add(t0, vec_rnd); \
    t1 = vec_sl(vec_sub(s0, s4), vec_2); \
    t1 = vec_add(vec_sl(t1, vec_1), t1); \
    t1 = vec_add(t1, vec_rnd); \
    t2 = vec_add(vec_sl(s6, vec_2), vec_sl(s6, vec_1)); \
    t2 = vec_add(t2, vec_sl(s2, vec_4)); \
    t3 = vec_add(vec_sl(s2, vec_2), vec_sl(s2, vec_1)); \
    t3 = vec_sub(t3, vec_sl(s6, vec_4)); \
    t4 = vec_add(t0, t2); \
    t5 = vec_add(t1, t3); \
    t6 = vec_sub(t1, t3); \
    t7 = vec_sub(t0, t2); \
\
    t0 = vec_sl(vec_add(s1, s3), vec_4); \
    t0 = vec_add(t0, vec_sl(s5, vec_3)); \
    t0 = vec_add(t0, vec_sl(s7, vec_2)); \
    t0 = vec_add(t0, vec_sub(s5, s3)); \
\
    t1 = vec_sl(vec_sub(s1, s5), vec_4); \
    t1 = vec_sub(t1, vec_sl(s7, vec_3)); \
    t1 = vec_sub(t1, vec_sl(s3, vec_2)); \
    t1 = vec_sub(t1, vec_add(s1, s7)); \
\
    t2 = vec_sl(vec_sub(s7, s3), vec_4); \
    t2 = vec_add(t2, vec_sl(s1, vec_3)); \
    t2 = vec_add(t2, vec_sl(s5, vec_2)); \
    t2 = vec_add(t2, vec_sub(s1, s7)); \
\
    t3 = vec_sl(vec_sub(s5, s7), vec_4); \
    t3 = vec_sub(t3, vec_sl(s3, vec_3)); \
    t3 = vec_add(t3, vec_sl(s1, vec_2)); \
    t3 = vec_sub(t3, vec_add(s3, s5)); \
\
    s0 = vec_add(t4, t0); \
    s1 = vec_add(t5, t1); \
    s2 = vec_add(t6, t2); \
    s3 = vec_add(t7, t3); \
    s4 = vec_sub(t7, t3); \
    s5 = vec_sub(t6, t2); \
    s6 = vec_sub(t5, t1); \
    s7 = vec_sub(t4, t0); \
}while(0)

#define SHIFT_HOR8(s0, s1, s2, s3, s4, s5, s6, s7) \
do { \
    s0 = vec_sra(s0, vec_3); \
    s1 = vec_sra(s1, vec_3); \
    s2 = vec_sra(s2, vec_3); \
    s3 = vec_sra(s3, vec_3); \
    s4 = vec_sra(s4, vec_3); \
    s5 = vec_sra(s5, vec_3); \
    s6 = vec_sra(s6, vec_3); \
    s7 = vec_sra(s7, vec_3); \
}while(0)

#define SHIFT_VERT8(s0, s1, s2, s3, s4, s5, s6, s7) \
do { \
    s0 = vec_sra(s0, vec_7); \
    s1 = vec_sra(s1, vec_7); \
    s2 = vec_sra(s2, vec_7); \
    s3 = vec_sra(s3, vec_7); \
    s4 = vec_sra(vec_add(s4, vec_1s), vec_7); \
    s5 = vec_sra(vec_add(s5, vec_1s), vec_7); \
    s6 = vec_sra(vec_add(s6, vec_1s), vec_7); \
    s7 = vec_sra(vec_add(s7, vec_1s), vec_7); \
}while(0)

/* main steps of 4x4 transform */
#define STEP4(s0, s1, s2, s3, vec_rnd) \
do { \
    t1 = vec_add(vec_sl(s0, vec_4), s0); \
    t1 = vec_add(t1, vec_rnd); \
    t2 = vec_add(vec_sl(s2, vec_4), s2); \
    t0 = vec_add(t1, t2); \
    t1 = vec_sub(t1, t2); \
    t3 = vec_sl(vec_sub(s3, s1), vec_1); \
    t3 = vec_add(t3, vec_sl(t3, vec_2)); \
    t2 = vec_add(t3, vec_sl(s1, vec_5)); \
    t3 = vec_add(t3, vec_sl(s3, vec_3)); \
    t3 = vec_add(t3, vec_sl(s3, vec_2)); \
    s0 = vec_add(t0, t2); \
    s1 = vec_sub(t1, t3); \
    s2 = vec_add(t1, t3); \
    s3 = vec_sub(t0, t2); \
}while (0)

#define SHIFT_HOR4(s0, s1, s2, s3) \
    s0 = vec_sra(s0, vec_3); \
    s1 = vec_sra(s1, vec_3); \
    s2 = vec_sra(s2, vec_3); \
    s3 = vec_sra(s3, vec_3);

#define SHIFT_VERT4(s0, s1, s2, s3) \
    s0 = vec_sra(s0, vec_7); \
    s1 = vec_sra(s1, vec_7); \
    s2 = vec_sra(s2, vec_7); \
    s3 = vec_sra(s3, vec_7);

/** Do inverse transform on 8x8 block
*/
Diego Biurrun's avatar
Diego Biurrun committed
137
static void vc1_inv_trans_8x8_altivec(int16_t block[64])
138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231
{
    vector signed short src0, src1, src2, src3, src4, src5, src6, src7;
    vector signed int s0, s1, s2, s3, s4, s5, s6, s7;
    vector signed int s8, s9, sA, sB, sC, sD, sE, sF;
    vector signed int t0, t1, t2, t3, t4, t5, t6, t7;
    const vector signed int vec_64 = vec_sl(vec_splat_s32(4), vec_splat_u32(4));
    const vector unsigned int vec_7 = vec_splat_u32(7);
    const vector unsigned int vec_4 = vec_splat_u32(4);
    const vector  signed int vec_4s = vec_splat_s32(4);
    const vector unsigned int vec_3 = vec_splat_u32(3);
    const vector unsigned int vec_2 = vec_splat_u32(2);
    const vector  signed int vec_1s = vec_splat_s32(1);
    const vector unsigned int vec_1 = vec_splat_u32(1);

    src0 = vec_ld(  0, block);
    src1 = vec_ld( 16, block);
    src2 = vec_ld( 32, block);
    src3 = vec_ld( 48, block);
    src4 = vec_ld( 64, block);
    src5 = vec_ld( 80, block);
    src6 = vec_ld( 96, block);
    src7 = vec_ld(112, block);

    s0 = vec_unpackl(src0);
    s1 = vec_unpackl(src1);
    s2 = vec_unpackl(src2);
    s3 = vec_unpackl(src3);
    s4 = vec_unpackl(src4);
    s5 = vec_unpackl(src5);
    s6 = vec_unpackl(src6);
    s7 = vec_unpackl(src7);
    s8 = vec_unpackh(src0);
    s9 = vec_unpackh(src1);
    sA = vec_unpackh(src2);
    sB = vec_unpackh(src3);
    sC = vec_unpackh(src4);
    sD = vec_unpackh(src5);
    sE = vec_unpackh(src6);
    sF = vec_unpackh(src7);
    STEP8(s0, s1, s2, s3, s4, s5, s6, s7, vec_4s);
    SHIFT_HOR8(s0, s1, s2, s3, s4, s5, s6, s7);
    STEP8(s8, s9, sA, sB, sC, sD, sE, sF, vec_4s);
    SHIFT_HOR8(s8, s9, sA, sB, sC, sD, sE, sF);
    src0 = vec_pack(s8, s0);
    src1 = vec_pack(s9, s1);
    src2 = vec_pack(sA, s2);
    src3 = vec_pack(sB, s3);
    src4 = vec_pack(sC, s4);
    src5 = vec_pack(sD, s5);
    src6 = vec_pack(sE, s6);
    src7 = vec_pack(sF, s7);
    TRANSPOSE8(src0, src1, src2, src3, src4, src5, src6, src7);

    s0 = vec_unpackl(src0);
    s1 = vec_unpackl(src1);
    s2 = vec_unpackl(src2);
    s3 = vec_unpackl(src3);
    s4 = vec_unpackl(src4);
    s5 = vec_unpackl(src5);
    s6 = vec_unpackl(src6);
    s7 = vec_unpackl(src7);
    s8 = vec_unpackh(src0);
    s9 = vec_unpackh(src1);
    sA = vec_unpackh(src2);
    sB = vec_unpackh(src3);
    sC = vec_unpackh(src4);
    sD = vec_unpackh(src5);
    sE = vec_unpackh(src6);
    sF = vec_unpackh(src7);
    STEP8(s0, s1, s2, s3, s4, s5, s6, s7, vec_64);
    SHIFT_VERT8(s0, s1, s2, s3, s4, s5, s6, s7);
    STEP8(s8, s9, sA, sB, sC, sD, sE, sF, vec_64);
    SHIFT_VERT8(s8, s9, sA, sB, sC, sD, sE, sF);
    src0 = vec_pack(s8, s0);
    src1 = vec_pack(s9, s1);
    src2 = vec_pack(sA, s2);
    src3 = vec_pack(sB, s3);
    src4 = vec_pack(sC, s4);
    src5 = vec_pack(sD, s5);
    src6 = vec_pack(sE, s6);
    src7 = vec_pack(sF, s7);

    vec_st(src0,  0, block);
    vec_st(src1, 16, block);
    vec_st(src2, 32, block);
    vec_st(src3, 48, block);
    vec_st(src4, 64, block);
    vec_st(src5, 80, block);
    vec_st(src6, 96, block);
    vec_st(src7,112, block);
}

/** Do inverse transform on 8x4 part of block
*/
Diego Biurrun's avatar
Diego Biurrun committed
232
static void vc1_inv_trans_8x4_altivec(uint8_t *dest, int stride, int16_t *block)
233 234 235 236 237 238 239 240 241 242 243 244 245
{
    vector signed short src0, src1, src2, src3, src4, src5, src6, src7;
    vector signed int s0, s1, s2, s3, s4, s5, s6, s7;
    vector signed int s8, s9, sA, sB, sC, sD, sE, sF;
    vector signed int t0, t1, t2, t3, t4, t5, t6, t7;
    const vector signed int vec_64 = vec_sl(vec_splat_s32(4), vec_splat_u32(4));
    const vector unsigned int vec_7 = vec_splat_u32(7);
    const vector unsigned int vec_5 = vec_splat_u32(5);
    const vector unsigned int vec_4 = vec_splat_u32(4);
    const vector  signed int vec_4s = vec_splat_s32(4);
    const vector unsigned int vec_3 = vec_splat_u32(3);
    const vector unsigned int vec_2 = vec_splat_u32(2);
    const vector unsigned int vec_1 = vec_splat_u32(1);
246 247 248
    vector unsigned char tmp;
    vector signed short tmp2, tmp3;
    vector unsigned char perm0, perm1, p0, p1, p;
249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289

    src0 = vec_ld(  0, block);
    src1 = vec_ld( 16, block);
    src2 = vec_ld( 32, block);
    src3 = vec_ld( 48, block);
    src4 = vec_ld( 64, block);
    src5 = vec_ld( 80, block);
    src6 = vec_ld( 96, block);
    src7 = vec_ld(112, block);

    TRANSPOSE8(src0, src1, src2, src3, src4, src5, src6, src7);
    s0 = vec_unpackl(src0);
    s1 = vec_unpackl(src1);
    s2 = vec_unpackl(src2);
    s3 = vec_unpackl(src3);
    s4 = vec_unpackl(src4);
    s5 = vec_unpackl(src5);
    s6 = vec_unpackl(src6);
    s7 = vec_unpackl(src7);
    s8 = vec_unpackh(src0);
    s9 = vec_unpackh(src1);
    sA = vec_unpackh(src2);
    sB = vec_unpackh(src3);
    sC = vec_unpackh(src4);
    sD = vec_unpackh(src5);
    sE = vec_unpackh(src6);
    sF = vec_unpackh(src7);
    STEP8(s0, s1, s2, s3, s4, s5, s6, s7, vec_4s);
    SHIFT_HOR8(s0, s1, s2, s3, s4, s5, s6, s7);
    STEP8(s8, s9, sA, sB, sC, sD, sE, sF, vec_4s);
    SHIFT_HOR8(s8, s9, sA, sB, sC, sD, sE, sF);
    src0 = vec_pack(s8, s0);
    src1 = vec_pack(s9, s1);
    src2 = vec_pack(sA, s2);
    src3 = vec_pack(sB, s3);
    src4 = vec_pack(sC, s4);
    src5 = vec_pack(sD, s5);
    src6 = vec_pack(sE, s6);
    src7 = vec_pack(sF, s7);
    TRANSPOSE8(src0, src1, src2, src3, src4, src5, src6, src7);

290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305
    s0 = vec_unpackh(src0);
    s1 = vec_unpackh(src1);
    s2 = vec_unpackh(src2);
    s3 = vec_unpackh(src3);
    s8 = vec_unpackl(src0);
    s9 = vec_unpackl(src1);
    sA = vec_unpackl(src2);
    sB = vec_unpackl(src3);
    STEP4(s0, s1, s2, s3, vec_64);
    SHIFT_VERT4(s0, s1, s2, s3);
    STEP4(s8, s9, sA, sB, vec_64);
    SHIFT_VERT4(s8, s9, sA, sB);
    src0 = vec_pack(s0, s8);
    src1 = vec_pack(s1, s9);
    src2 = vec_pack(s2, sA);
    src3 = vec_pack(s3, sB);
306

307 308 309 310 311 312 313 314 315
    p0 = vec_lvsl (0, dest);
    p1 = vec_lvsl (stride, dest);
    p = vec_splat_u8 (-1);
    perm0 = vec_mergeh (p, p0);
    perm1 = vec_mergeh (p, p1);

#define ADD(dest,src,perm)                                              \
    /* *(uint64_t *)&tmp = *(uint64_t *)dest; */                        \
    tmp = vec_ld (0, dest);                                             \
316
    tmp2 = (vector signed short)vec_perm (tmp, vec_splat_u8(0), perm);  \
317 318 319 320
    tmp3 = vec_adds (tmp2, src);                                        \
    tmp = vec_packsu (tmp3, tmp3);                                      \
    vec_ste ((vector unsigned int)tmp, 0, (unsigned int *)dest);        \
    vec_ste ((vector unsigned int)tmp, 4, (unsigned int *)dest);
321

322 323 324 325
    ADD (dest, src0, perm0)      dest += stride;
    ADD (dest, src1, perm1)      dest += stride;
    ADD (dest, src2, perm0)      dest += stride;
    ADD (dest, src3, perm1)
326 327
}

Ronald S. Bultje's avatar
Ronald S. Bultje committed
328 329 330 331 332
#define PUT_OP_U8_ALTIVEC(d, s, dst) d = s
#define AVG_OP_U8_ALTIVEC(d, s, dst) d = vec_avg(dst, s)

#define OP_U8_ALTIVEC                          PUT_OP_U8_ALTIVEC
#define PREFIX_no_rnd_vc1_chroma_mc8_altivec   put_no_rnd_vc1_chroma_mc8_altivec
333
#include "h264chroma_template.c"
Ronald S. Bultje's avatar
Ronald S. Bultje committed
334 335 336 337 338
#undef OP_U8_ALTIVEC
#undef PREFIX_no_rnd_vc1_chroma_mc8_altivec

#define OP_U8_ALTIVEC                          AVG_OP_U8_ALTIVEC
#define PREFIX_no_rnd_vc1_chroma_mc8_altivec   avg_no_rnd_vc1_chroma_mc8_altivec
339
#include "h264chroma_template.c"
Ronald S. Bultje's avatar
Ronald S. Bultje committed
340 341
#undef OP_U8_ALTIVEC
#undef PREFIX_no_rnd_vc1_chroma_mc8_altivec
342

343 344 345
#endif /* HAVE_ALTIVEC */

av_cold void ff_vc1dsp_init_ppc(VC1DSPContext *dsp)
346
{
347
#if HAVE_ALTIVEC
348
    if (!PPC_ALTIVEC(av_get_cpu_flags()))
349 350
        return;

351
    dsp->vc1_inv_trans_8x8 = vc1_inv_trans_8x8_altivec;
352
    dsp->vc1_inv_trans_8x4 = vc1_inv_trans_8x4_altivec;
Ronald S. Bultje's avatar
Ronald S. Bultje committed
353 354
    dsp->put_no_rnd_vc1_chroma_pixels_tab[0] = put_no_rnd_vc1_chroma_mc8_altivec;
    dsp->avg_no_rnd_vc1_chroma_pixels_tab[0] = avg_no_rnd_vc1_chroma_mc8_altivec;
355
#endif /* HAVE_ALTIVEC */
356
}