h264dsp_mmx.c 40.6 KB
Newer Older
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49
/*
 * Copyright (c) 2004-2005 Michael Niedermayer, Loren Merritt
 *
 * This library is free software; you can redistribute it and/or
 * modify it under the terms of the GNU Lesser General Public
 * License as published by the Free Software Foundation; either
 * version 2 of the License, or (at your option) any later version.
 *
 * This library is distributed in the hope that it will be useful,
 * 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
 * License along with this library; if not, write to the Free Software
 * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
 */


/***********************************/
/* IDCT */

/* in/out: mma=mma+mmb, mmb=mmb-mma */
#define SUMSUB_BA( a, b ) \
    "paddw "#b", "#a" \n\t"\
    "paddw "#b", "#b" \n\t"\
    "psubw "#a", "#b" \n\t"

#define SUMSUB_BADC( a, b, c, d ) \
    "paddw "#b", "#a" \n\t"\
    "paddw "#d", "#c" \n\t"\
    "paddw "#b", "#b" \n\t"\
    "paddw "#d", "#d" \n\t"\
    "psubw "#a", "#b" \n\t"\
    "psubw "#c", "#d" \n\t"

#define SUMSUBD2_AB( a, b, t ) \
    "movq  "#b", "#t" \n\t"\
    "psraw  $1 , "#b" \n\t"\
    "paddw "#a", "#b" \n\t"\
    "psraw  $1 , "#a" \n\t"\
    "psubw "#t", "#a" \n\t"

#define IDCT4_1D( s02, s13, d02, d13, t ) \
    SUMSUB_BA  ( s02, d02 )\
    SUMSUBD2_AB( s13, d13, t )\
    SUMSUB_BADC( d13, s02, s13, d02 )

#define SBUTTERFLY(a,b,t,n)\
50 51 52
    "movq " #a ", " #t "                \n\t" /* abcd */\
    "punpckl" #n " " #b ", " #a "       \n\t" /* aebf */\
    "punpckh" #n " " #b ", " #t "       \n\t" /* cgdh */\
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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164

#define TRANSPOSE4(a,b,c,d,t)\
    SBUTTERFLY(a,b,t,wd) /* a=aebf t=cgdh */\
    SBUTTERFLY(c,d,b,wd) /* c=imjn b=kolp */\
    SBUTTERFLY(a,c,d,dq) /* a=aeim d=bfjn */\
    SBUTTERFLY(t,b,c,dq) /* t=cgko c=dhlp */

#define STORE_DIFF_4P( p, t, z ) \
    "psraw      $6,     "#p" \n\t"\
    "movd       (%0),   "#t" \n\t"\
    "punpcklbw "#z",    "#t" \n\t"\
    "paddsw    "#t",    "#p" \n\t"\
    "packuswb  "#z",    "#p" \n\t"\
    "movd      "#p",    (%0) \n\t"

void ff_h264_idct_add_mmx2(uint8_t *dst, int16_t *block, int stride)
{
    /* Load dct coeffs */
    asm volatile(
        "movq   (%0), %%mm0 \n\t"
        "movq  8(%0), %%mm1 \n\t"
        "movq 16(%0), %%mm2 \n\t"
        "movq 24(%0), %%mm3 \n\t"
    :: "r"(block) );

    asm volatile(
        /* mm1=s02+s13  mm2=s02-s13  mm4=d02+d13  mm0=d02-d13 */
        IDCT4_1D( %%mm2, %%mm1, %%mm0, %%mm3, %%mm4 )

        "movq      %0,    %%mm6 \n\t"
        /* in: 1,4,0,2  out: 1,2,3,0 */
        TRANSPOSE4( %%mm3, %%mm1, %%mm0, %%mm2, %%mm4 )

        "paddw     %%mm6, %%mm3 \n\t"

        /* mm2=s02+s13  mm3=s02-s13  mm4=d02+d13  mm1=d02-d13 */
        IDCT4_1D( %%mm4, %%mm2, %%mm3, %%mm0, %%mm1 )

        "pxor %%mm7, %%mm7    \n\t"
    :: "m"(ff_pw_32));

    asm volatile(
    STORE_DIFF_4P( %%mm0, %%mm1, %%mm7)
        "add %1, %0             \n\t"
    STORE_DIFF_4P( %%mm2, %%mm1, %%mm7)
        "add %1, %0             \n\t"
    STORE_DIFF_4P( %%mm3, %%mm1, %%mm7)
        "add %1, %0             \n\t"
    STORE_DIFF_4P( %%mm4, %%mm1, %%mm7)
        : "+r"(dst)
        : "r" ((long)stride)
    );
}


/***********************************/
/* deblocking */

// out: o = |x-y|>a
// clobbers: t
#define DIFF_GT_MMX(x,y,a,o,t)\
    "movq     "#y", "#t"  \n\t"\
    "movq     "#x", "#o"  \n\t"\
    "psubusb  "#x", "#t"  \n\t"\
    "psubusb  "#y", "#o"  \n\t"\
    "por      "#t", "#o"  \n\t"\
    "psubusb  "#a", "#o"  \n\t"

// in: mm0=p1 mm1=p0 mm2=q0 mm3=q1
// out: mm5=beta-1, mm7=mask
// clobbers: mm4,mm6
#define H264_DEBLOCK_MASK(alpha1, beta1) \
    "pshufw $0, "#alpha1", %%mm4 \n\t"\
    "pshufw $0, "#beta1 ", %%mm5 \n\t"\
    "packuswb  %%mm4, %%mm4      \n\t"\
    "packuswb  %%mm5, %%mm5      \n\t"\
    DIFF_GT_MMX(%%mm1, %%mm2, %%mm4, %%mm7, %%mm6) /* |p0-q0| > alpha-1 */\
    DIFF_GT_MMX(%%mm0, %%mm1, %%mm5, %%mm4, %%mm6) /* |p1-p0| > beta-1 */\
    "por       %%mm4, %%mm7      \n\t"\
    DIFF_GT_MMX(%%mm3, %%mm2, %%mm5, %%mm4, %%mm6) /* |q1-q0| > beta-1 */\
    "por       %%mm4, %%mm7      \n\t"\
    "pxor      %%mm6, %%mm6      \n\t"\
    "pcmpeqb   %%mm6, %%mm7      \n\t"

// in: mm0=p1 mm1=p0 mm2=q0 mm3=q1 mm7=(tc&mask)
// out: mm1=p0' mm2=q0'
// clobbers: mm0,3-6
#define H264_DEBLOCK_P0_Q0(pb_01, pb_3f)\
        /* a = q0^p0^((p1-q1)>>2) */\
        "movq    %%mm0, %%mm4  \n\t"\
        "psubb   %%mm3, %%mm4  \n\t"\
        "psrlw   $2,    %%mm4  \n\t"\
        "pxor    %%mm1, %%mm4  \n\t"\
        "pxor    %%mm2, %%mm4  \n\t"\
        /* b = p0^(q1>>2) */\
        "psrlw   $2,    %%mm3  \n\t"\
        "pand "#pb_3f", %%mm3  \n\t"\
        "movq    %%mm1, %%mm5  \n\t"\
        "pxor    %%mm3, %%mm5  \n\t"\
        /* c = q0^(p1>>2) */\
        "psrlw   $2,    %%mm0  \n\t"\
        "pand "#pb_3f", %%mm0  \n\t"\
        "movq    %%mm2, %%mm6  \n\t"\
        "pxor    %%mm0, %%mm6  \n\t"\
        /* d = (c^b) & ~(b^a) & 1 */\
        "pxor    %%mm5, %%mm6  \n\t"\
        "pxor    %%mm4, %%mm5  \n\t"\
        "pandn   %%mm6, %%mm5  \n\t"\
        "pand "#pb_01", %%mm5  \n\t"\
        /* delta = (avg(q0, p1>>2) + (d&a))
         *       - (avg(p0, q1>>2) + (d&~a)) */\
        "pavgb   %%mm2, %%mm0  \n\t"\
165 166
        "pand    %%mm5, %%mm4  \n\t"\
        "paddusb %%mm4, %%mm0  \n\t"\
167
        "pavgb   %%mm1, %%mm3  \n\t"\
168
        "pxor    %%mm5, %%mm4  \n\t"\
169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371
        "paddusb %%mm4, %%mm3  \n\t"\
        /* p0 += clip(delta, -tc0, tc0)
         * q0 -= clip(delta, -tc0, tc0) */\
        "movq    %%mm0, %%mm4  \n\t"\
        "psubusb %%mm3, %%mm0  \n\t"\
        "psubusb %%mm4, %%mm3  \n\t"\
        "pminub  %%mm7, %%mm0  \n\t"\
        "pminub  %%mm7, %%mm3  \n\t"\
        "paddusb %%mm0, %%mm1  \n\t"\
        "paddusb %%mm3, %%mm2  \n\t"\
        "psubusb %%mm3, %%mm1  \n\t"\
        "psubusb %%mm0, %%mm2  \n\t"

// in: mm0=p1 mm1=p0 mm2=q0 mm3=q1 mm7=(tc&mask) %8=mm_bone
// out: (q1addr) = clip( (q2+((p0+q0+1)>>1))>>1, q1-tc0, q1+tc0 )
// clobbers: q2, tmp, tc0
#define H264_DEBLOCK_Q1(p1, q2, q2addr, q1addr, tc0, tmp)\
        "movq     %%mm1,  "#tmp"   \n\t"\
        "pavgb    %%mm2,  "#tmp"   \n\t"\
        "pavgb    "#tmp", "#q2"    \n\t" /* avg(p2,avg(p0,q0)) */\
        "pxor   "q2addr", "#tmp"   \n\t"\
        "pand     %8,     "#tmp"   \n\t" /* (p2^avg(p0,q0))&1 */\
        "psubusb  "#tmp", "#q2"    \n\t" /* (p2+((p0+q0+1)>>1))>>1 */\
        "movq     "#p1",  "#tmp"   \n\t"\
        "psubusb  "#tc0", "#tmp"   \n\t"\
        "paddusb  "#p1",  "#tc0"   \n\t"\
        "pmaxub   "#tmp", "#q2"    \n\t"\
        "pminub   "#tc0", "#q2"    \n\t"\
        "movq     "#q2",  "q1addr" \n\t"

static inline void h264_loop_filter_luma_mmx2(uint8_t *pix, int stride, int alpha1, int beta1, int8_t *tc0)
{
    uint64_t tmp0;
    uint64_t tc = (uint8_t)tc0[1]*0x01010000 | (uint8_t)tc0[0]*0x0101;
    // with luma, tc0=0 doesn't mean no filtering, so we need a separate input mask
    uint32_t mask[2] = { (tc0[0]>=0)*0xffffffff, (tc0[1]>=0)*0xffffffff };

    asm volatile(
        "movq    (%1,%3), %%mm0    \n\t" //p1
        "movq    (%1,%3,2), %%mm1  \n\t" //p0
        "movq    (%2),    %%mm2    \n\t" //q0
        "movq    (%2,%3), %%mm3    \n\t" //q1
        H264_DEBLOCK_MASK(%6, %7)
        "pand     %5,     %%mm7    \n\t"
        "movq     %%mm7,  %0       \n\t"

        /* filter p1 */
        "movq     (%1),   %%mm3    \n\t" //p2
        DIFF_GT_MMX(%%mm1, %%mm3, %%mm5, %%mm6, %%mm4) // |p2-p0|>beta-1
        "pandn    %%mm7,  %%mm6    \n\t"
        "pcmpeqb  %%mm7,  %%mm6    \n\t"
        "pand     %%mm7,  %%mm6    \n\t" // mask & |p2-p0|<beta
        "pshufw  $80, %4, %%mm4    \n\t"
        "pand     %%mm7,  %%mm4    \n\t" // mask & tc0
        "movq     %8,     %%mm7    \n\t"
        "pand     %%mm6,  %%mm7    \n\t" // mask & |p2-p0|<beta & 1
        "pand     %%mm4,  %%mm6    \n\t" // mask & |p2-p0|<beta & tc0
        "paddb    %%mm4,  %%mm7    \n\t" // tc++
        H264_DEBLOCK_Q1(%%mm0, %%mm3, "(%1)", "(%1,%3)", %%mm6, %%mm4)

        /* filter q1 */
        "movq    (%2,%3,2), %%mm4  \n\t" //q2
        DIFF_GT_MMX(%%mm2, %%mm4, %%mm5, %%mm6, %%mm3) // |q2-q0|>beta-1
        "pandn    %0,     %%mm6    \n\t"
        "pcmpeqb  %0,     %%mm6    \n\t"
        "pand     %0,     %%mm6    \n\t"
        "pshufw  $80, %4, %%mm5    \n\t"
        "pand     %%mm6,  %%mm5    \n\t"
        "pand     %8,     %%mm6    \n\t"
        "paddb    %%mm6,  %%mm7    \n\t"
        "movq    (%2,%3), %%mm3    \n\t"
        H264_DEBLOCK_Q1(%%mm3, %%mm4, "(%2,%3,2)", "(%2,%3)", %%mm5, %%mm6)

        /* filter p0, q0 */
        H264_DEBLOCK_P0_Q0(%8, %9)
        "movq      %%mm1, (%1,%3,2) \n\t"
        "movq      %%mm2, (%2)      \n\t"

        : "=m"(tmp0)
        : "r"(pix-3*stride), "r"(pix), "r"((long)stride),
          "m"(tc), "m"(*(uint64_t*)mask), "m"(alpha1), "m"(beta1),
          "m"(mm_bone), "m"(ff_pb_3F)
    );
}

static void h264_v_loop_filter_luma_mmx2(uint8_t *pix, int stride, int alpha, int beta, int8_t *tc0)
{
    if((tc0[0] & tc0[1]) >= 0)
        h264_loop_filter_luma_mmx2(pix, stride, alpha-1, beta-1, tc0);
    if((tc0[2] & tc0[3]) >= 0)
        h264_loop_filter_luma_mmx2(pix+8, stride, alpha-1, beta-1, tc0+2);
}
static void h264_h_loop_filter_luma_mmx2(uint8_t *pix, int stride, int alpha, int beta, int8_t *tc0)
{
    //FIXME: could cut some load/stores by merging transpose with filter
    // also, it only needs to transpose 6x8
    uint8_t trans[8*8];
    int i;
    for(i=0; i<2; i++, pix+=8*stride, tc0+=2) {
        if((tc0[0] & tc0[1]) < 0)
            continue;
        transpose4x4(trans,       pix-4,          8, stride);
        transpose4x4(trans  +4*8, pix,            8, stride);
        transpose4x4(trans+4,     pix-4+4*stride, 8, stride);
        transpose4x4(trans+4+4*8, pix  +4*stride, 8, stride);
        h264_loop_filter_luma_mmx2(trans+4*8, 8, alpha-1, beta-1, tc0);
        transpose4x4(pix-2,          trans  +2*8, stride, 8);
        transpose4x4(pix-2+4*stride, trans+4+2*8, stride, 8);
    }
}

static inline void h264_loop_filter_chroma_mmx2(uint8_t *pix, int stride, int alpha1, int beta1, int8_t *tc0)
{
    asm volatile(
        "movq    (%0),    %%mm0     \n\t" //p1
        "movq    (%0,%2), %%mm1     \n\t" //p0
        "movq    (%1),    %%mm2     \n\t" //q0
        "movq    (%1,%2), %%mm3     \n\t" //q1
        H264_DEBLOCK_MASK(%4, %5)
        "movd      %3,    %%mm6     \n\t"
        "punpcklbw %%mm6, %%mm6     \n\t"
        "pand      %%mm6, %%mm7     \n\t" // mm7 = tc&mask
        H264_DEBLOCK_P0_Q0(%6, %7)
        "movq      %%mm1, (%0,%2)   \n\t"
        "movq      %%mm2, (%1)      \n\t"

        :: "r"(pix-2*stride), "r"(pix), "r"((long)stride),
           "r"(*(uint32_t*)tc0),
           "m"(alpha1), "m"(beta1), "m"(mm_bone), "m"(ff_pb_3F)
    );
}

static void h264_v_loop_filter_chroma_mmx2(uint8_t *pix, int stride, int alpha, int beta, int8_t *tc0)
{
    h264_loop_filter_chroma_mmx2(pix, stride, alpha-1, beta-1, tc0);
}

static void h264_h_loop_filter_chroma_mmx2(uint8_t *pix, int stride, int alpha, int beta, int8_t *tc0)
{
    //FIXME: could cut some load/stores by merging transpose with filter
    uint8_t trans[8*4];
    transpose4x4(trans, pix-2, 8, stride);
    transpose4x4(trans+4, pix-2+4*stride, 8, stride);
    h264_loop_filter_chroma_mmx2(trans+2*8, 8, alpha-1, beta-1, tc0);
    transpose4x4(pix-2, trans, stride, 8);
    transpose4x4(pix-2+4*stride, trans+4, stride, 8);
}

// p0 = (p0 + q1 + 2*p1 + 2) >> 2
#define H264_FILTER_CHROMA4(p0, p1, q1, one) \
    "movq    "#p0", %%mm4  \n\t"\
    "pxor    "#q1", %%mm4  \n\t"\
    "pand   "#one", %%mm4  \n\t" /* mm4 = (p0^q1)&1 */\
    "pavgb   "#q1", "#p0"  \n\t"\
    "psubusb %%mm4, "#p0"  \n\t"\
    "pavgb   "#p1", "#p0"  \n\t" /* dst = avg(p1, avg(p0,q1) - ((p0^q1)&1)) */\

static inline void h264_loop_filter_chroma_intra_mmx2(uint8_t *pix, int stride, int alpha1, int beta1)
{
    asm volatile(
        "movq    (%0),    %%mm0     \n\t"
        "movq    (%0,%2), %%mm1     \n\t"
        "movq    (%1),    %%mm2     \n\t"
        "movq    (%1,%2), %%mm3     \n\t"
        H264_DEBLOCK_MASK(%3, %4)
        "movq    %%mm1,   %%mm5     \n\t"
        "movq    %%mm2,   %%mm6     \n\t"
        H264_FILTER_CHROMA4(%%mm1, %%mm0, %%mm3, %5) //p0'
        H264_FILTER_CHROMA4(%%mm2, %%mm3, %%mm0, %5) //q0'
        "psubb   %%mm5,   %%mm1     \n\t"
        "psubb   %%mm6,   %%mm2     \n\t"
        "pand    %%mm7,   %%mm1     \n\t"
        "pand    %%mm7,   %%mm2     \n\t"
        "paddb   %%mm5,   %%mm1     \n\t"
        "paddb   %%mm6,   %%mm2     \n\t"
        "movq    %%mm1,   (%0,%2)   \n\t"
        "movq    %%mm2,   (%1)      \n\t"
        :: "r"(pix-2*stride), "r"(pix), "r"((long)stride),
           "m"(alpha1), "m"(beta1), "m"(mm_bone)
    );
}

static void h264_v_loop_filter_chroma_intra_mmx2(uint8_t *pix, int stride, int alpha, int beta)
{
    h264_loop_filter_chroma_intra_mmx2(pix, stride, alpha-1, beta-1);
}

static void h264_h_loop_filter_chroma_intra_mmx2(uint8_t *pix, int stride, int alpha, int beta)
{
    //FIXME: could cut some load/stores by merging transpose with filter
    uint8_t trans[8*4];
    transpose4x4(trans, pix-2, 8, stride);
    transpose4x4(trans+4, pix-2+4*stride, 8, stride);
    h264_loop_filter_chroma_intra_mmx2(trans+2*8, 8, alpha-1, beta-1);
    transpose4x4(pix-2, trans, stride, 8);
    transpose4x4(pix-2+4*stride, trans+4, stride, 8);
}


/***********************************/
/* motion compensation */

#define QPEL_H264V(A,B,C,D,E,F,OP)\
372 373 374 375 376 377 378 379 380 381 382 383 384 385
        "movd (%0), "#F"            \n\t"\
        "movq "#C", %%mm6           \n\t"\
        "paddw "#D", %%mm6          \n\t"\
        "psllw $2, %%mm6            \n\t"\
        "psubw "#B", %%mm6          \n\t"\
        "psubw "#E", %%mm6          \n\t"\
        "pmullw %4, %%mm6           \n\t"\
        "add %2, %0                 \n\t"\
        "punpcklbw %%mm7, "#F"      \n\t"\
        "paddw %5, "#A"             \n\t"\
        "paddw "#F", "#A"           \n\t"\
        "paddw "#A", %%mm6          \n\t"\
        "psraw $5, %%mm6            \n\t"\
        "packuswb %%mm6, %%mm6      \n\t"\
386
        OP(%%mm6, (%1), A, d)\
387
        "add %3, %1                 \n\t"
388 389

#define QPEL_H264HV(A,B,C,D,E,F,OF)\
390 391 392 393 394 395 396 397 398 399 400 401
        "movd (%0), "#F"            \n\t"\
        "movq "#C", %%mm6           \n\t"\
        "paddw "#D", %%mm6          \n\t"\
        "psllw $2, %%mm6            \n\t"\
        "psubw "#B", %%mm6          \n\t"\
        "psubw "#E", %%mm6          \n\t"\
        "pmullw %3, %%mm6           \n\t"\
        "add %2, %0                 \n\t"\
        "punpcklbw %%mm7, "#F"      \n\t"\
        "paddw "#F", "#A"           \n\t"\
        "paddw "#A", %%mm6          \n\t"\
        "movq %%mm6, "#OF"(%1)      \n\t"
402

403 404 405 406 407
#define QPEL_H264(OPNAME, OP, MMX)\
static void OPNAME ## h264_qpel4_h_lowpass_ ## MMX(uint8_t *dst, uint8_t *src, int dstStride, int srcStride){\
    int h=4;\
\
    asm volatile(\
408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433
        "pxor %%mm7, %%mm7          \n\t"\
        "movq %5, %%mm4             \n\t"\
        "movq %6, %%mm5             \n\t"\
        "1:                         \n\t"\
        "movd  -1(%0), %%mm1        \n\t"\
        "movd    (%0), %%mm2        \n\t"\
        "movd   1(%0), %%mm3        \n\t"\
        "movd   2(%0), %%mm0        \n\t"\
        "punpcklbw %%mm7, %%mm1     \n\t"\
        "punpcklbw %%mm7, %%mm2     \n\t"\
        "punpcklbw %%mm7, %%mm3     \n\t"\
        "punpcklbw %%mm7, %%mm0     \n\t"\
        "paddw %%mm0, %%mm1         \n\t"\
        "paddw %%mm3, %%mm2         \n\t"\
        "movd  -2(%0), %%mm0        \n\t"\
        "movd   3(%0), %%mm3        \n\t"\
        "punpcklbw %%mm7, %%mm0     \n\t"\
        "punpcklbw %%mm7, %%mm3     \n\t"\
        "paddw %%mm3, %%mm0         \n\t"\
        "psllw $2, %%mm2            \n\t"\
        "psubw %%mm1, %%mm2         \n\t"\
        "pmullw %%mm4, %%mm2        \n\t"\
        "paddw %%mm5, %%mm0         \n\t"\
        "paddw %%mm2, %%mm0         \n\t"\
        "psraw $5, %%mm0            \n\t"\
        "packuswb %%mm0, %%mm0      \n\t"\
434
        OP(%%mm0, (%1),%%mm6, d)\
435 436 437 438
        "add %3, %0                 \n\t"\
        "add %4, %1                 \n\t"\
        "decl %2                    \n\t"\
        " jnz 1b                    \n\t"\
439 440 441 442 443 444 445 446
        : "+a"(src), "+c"(dst), "+m"(h)\
        : "d"((long)srcStride), "S"((long)dstStride), "m"(ff_pw_5), "m"(ff_pw_16)\
        : "memory"\
    );\
}\
static void OPNAME ## h264_qpel4_v_lowpass_ ## MMX(uint8_t *dst, uint8_t *src, int dstStride, int srcStride){\
    src -= 2*srcStride;\
    asm volatile(\
447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462
        "pxor %%mm7, %%mm7          \n\t"\
        "movd (%0), %%mm0           \n\t"\
        "add %2, %0                 \n\t"\
        "movd (%0), %%mm1           \n\t"\
        "add %2, %0                 \n\t"\
        "movd (%0), %%mm2           \n\t"\
        "add %2, %0                 \n\t"\
        "movd (%0), %%mm3           \n\t"\
        "add %2, %0                 \n\t"\
        "movd (%0), %%mm4           \n\t"\
        "add %2, %0                 \n\t"\
        "punpcklbw %%mm7, %%mm0     \n\t"\
        "punpcklbw %%mm7, %%mm1     \n\t"\
        "punpcklbw %%mm7, %%mm2     \n\t"\
        "punpcklbw %%mm7, %%mm3     \n\t"\
        "punpcklbw %%mm7, %%mm4     \n\t"\
463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478
        QPEL_H264V(%%mm0, %%mm1, %%mm2, %%mm3, %%mm4, %%mm5, OP)\
        QPEL_H264V(%%mm1, %%mm2, %%mm3, %%mm4, %%mm5, %%mm0, OP)\
        QPEL_H264V(%%mm2, %%mm3, %%mm4, %%mm5, %%mm0, %%mm1, OP)\
        QPEL_H264V(%%mm3, %%mm4, %%mm5, %%mm0, %%mm1, %%mm2, OP)\
         \
        : "+a"(src), "+c"(dst)\
        : "S"((long)srcStride), "D"((long)dstStride), "m"(ff_pw_5), "m"(ff_pw_16)\
        : "memory"\
    );\
}\
static void OPNAME ## h264_qpel4_hv_lowpass_ ## MMX(uint8_t *dst, int16_t *tmp, uint8_t *src, int dstStride, int tmpStride, int srcStride){\
    int h=4;\
    int w=3;\
    src -= 2*srcStride+2;\
    while(w--){\
        asm volatile(\
479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494
            "pxor %%mm7, %%mm7      \n\t"\
            "movd (%0), %%mm0       \n\t"\
            "add %2, %0             \n\t"\
            "movd (%0), %%mm1       \n\t"\
            "add %2, %0             \n\t"\
            "movd (%0), %%mm2       \n\t"\
            "add %2, %0             \n\t"\
            "movd (%0), %%mm3       \n\t"\
            "add %2, %0             \n\t"\
            "movd (%0), %%mm4       \n\t"\
            "add %2, %0             \n\t"\
            "punpcklbw %%mm7, %%mm0 \n\t"\
            "punpcklbw %%mm7, %%mm1 \n\t"\
            "punpcklbw %%mm7, %%mm2 \n\t"\
            "punpcklbw %%mm7, %%mm3 \n\t"\
            "punpcklbw %%mm7, %%mm4 \n\t"\
495 496 497 498 499 500 501 502 503 504 505 506 507 508
            QPEL_H264HV(%%mm0, %%mm1, %%mm2, %%mm3, %%mm4, %%mm5, 0*8*3)\
            QPEL_H264HV(%%mm1, %%mm2, %%mm3, %%mm4, %%mm5, %%mm0, 1*8*3)\
            QPEL_H264HV(%%mm2, %%mm3, %%mm4, %%mm5, %%mm0, %%mm1, 2*8*3)\
            QPEL_H264HV(%%mm3, %%mm4, %%mm5, %%mm0, %%mm1, %%mm2, 3*8*3)\
             \
            : "+a"(src)\
            : "c"(tmp), "S"((long)srcStride), "m"(ff_pw_5)\
            : "memory"\
        );\
        tmp += 4;\
        src += 4 - 9*srcStride;\
    }\
    tmp -= 3*4;\
    asm volatile(\
509 510 511 512 513 514 515 516 517 518 519 520
        "movq %4, %%mm6             \n\t"\
        "1:                         \n\t"\
        "movq     (%0), %%mm0       \n\t"\
        "paddw  10(%0), %%mm0       \n\t"\
        "movq    2(%0), %%mm1       \n\t"\
        "paddw   8(%0), %%mm1       \n\t"\
        "movq    4(%0), %%mm2       \n\t"\
        "paddw   6(%0), %%mm2       \n\t"\
        "psubw %%mm1, %%mm0         \n\t"/*a-b   (abccba)*/\
        "psraw $2, %%mm0            \n\t"/*(a-b)/4 */\
        "psubw %%mm1, %%mm0         \n\t"/*(a-b)/4-b */\
        "paddsw %%mm2, %%mm0        \n\t"\
521
        "psraw $2, %%mm0            \n\t"/*((a-b)/4-b+c)/4 */\
522
        "paddw %%mm6, %%mm2         \n\t"\
523
        "paddw %%mm2, %%mm0         \n\t"/*(a-5*b+20*c)/16 +32 */\
524 525
        "psraw $6, %%mm0            \n\t"\
        "packuswb %%mm0, %%mm0      \n\t"\
526
        OP(%%mm0, (%1),%%mm7, d)\
527 528 529 530
        "add $24, %0                \n\t"\
        "add %3, %1                 \n\t"\
        "decl %2                    \n\t"\
        " jnz 1b                    \n\t"\
531 532 533 534 535 536 537 538 539
        : "+a"(tmp), "+c"(dst), "+m"(h)\
        : "S"((long)dstStride), "m"(ff_pw_32)\
        : "memory"\
    );\
}\
\
static void OPNAME ## h264_qpel8_h_lowpass_ ## MMX(uint8_t *dst, uint8_t *src, int dstStride, int srcStride){\
    int h=8;\
    asm volatile(\
540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582
        "pxor %%mm7, %%mm7          \n\t"\
        "movq %5, %%mm6             \n\t"\
        "1:                         \n\t"\
        "movq    (%0), %%mm0        \n\t"\
        "movq   1(%0), %%mm2        \n\t"\
        "movq %%mm0, %%mm1          \n\t"\
        "movq %%mm2, %%mm3          \n\t"\
        "punpcklbw %%mm7, %%mm0     \n\t"\
        "punpckhbw %%mm7, %%mm1     \n\t"\
        "punpcklbw %%mm7, %%mm2     \n\t"\
        "punpckhbw %%mm7, %%mm3     \n\t"\
        "paddw %%mm2, %%mm0         \n\t"\
        "paddw %%mm3, %%mm1         \n\t"\
        "psllw $2, %%mm0            \n\t"\
        "psllw $2, %%mm1            \n\t"\
        "movq   -1(%0), %%mm2       \n\t"\
        "movq    2(%0), %%mm4       \n\t"\
        "movq %%mm2, %%mm3          \n\t"\
        "movq %%mm4, %%mm5          \n\t"\
        "punpcklbw %%mm7, %%mm2     \n\t"\
        "punpckhbw %%mm7, %%mm3     \n\t"\
        "punpcklbw %%mm7, %%mm4     \n\t"\
        "punpckhbw %%mm7, %%mm5     \n\t"\
        "paddw %%mm4, %%mm2         \n\t"\
        "paddw %%mm3, %%mm5         \n\t"\
        "psubw %%mm2, %%mm0         \n\t"\
        "psubw %%mm5, %%mm1         \n\t"\
        "pmullw %%mm6, %%mm0        \n\t"\
        "pmullw %%mm6, %%mm1        \n\t"\
        "movd   -2(%0), %%mm2       \n\t"\
        "movd    7(%0), %%mm5       \n\t"\
        "punpcklbw %%mm7, %%mm2     \n\t"\
        "punpcklbw %%mm7, %%mm5     \n\t"\
        "paddw %%mm3, %%mm2         \n\t"\
        "paddw %%mm5, %%mm4         \n\t"\
        "movq %6, %%mm5             \n\t"\
        "paddw %%mm5, %%mm2         \n\t"\
        "paddw %%mm5, %%mm4         \n\t"\
        "paddw %%mm2, %%mm0         \n\t"\
        "paddw %%mm4, %%mm1         \n\t"\
        "psraw $5, %%mm0            \n\t"\
        "psraw $5, %%mm1            \n\t"\
        "packuswb %%mm1, %%mm0      \n\t"\
583
        OP(%%mm0, (%1),%%mm5, q)\
584 585 586 587
        "add %3, %0                 \n\t"\
        "add %4, %1                 \n\t"\
        "decl %2                    \n\t"\
        " jnz 1b                    \n\t"\
588 589 590 591 592 593 594 595 596 597 598 599
        : "+a"(src), "+c"(dst), "+m"(h)\
        : "d"((long)srcStride), "S"((long)dstStride), "m"(ff_pw_5), "m"(ff_pw_16)\
        : "memory"\
    );\
}\
\
static void OPNAME ## h264_qpel8_v_lowpass_ ## MMX(uint8_t *dst, uint8_t *src, int dstStride, int srcStride){\
    int h= 2;\
    src -= 2*srcStride;\
    \
    while(h--){\
      asm volatile(\
600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615
        "pxor %%mm7, %%mm7          \n\t"\
        "movd (%0), %%mm0           \n\t"\
        "add %2, %0                 \n\t"\
        "movd (%0), %%mm1           \n\t"\
        "add %2, %0                 \n\t"\
        "movd (%0), %%mm2           \n\t"\
        "add %2, %0                 \n\t"\
        "movd (%0), %%mm3           \n\t"\
        "add %2, %0                 \n\t"\
        "movd (%0), %%mm4           \n\t"\
        "add %2, %0                 \n\t"\
        "punpcklbw %%mm7, %%mm0     \n\t"\
        "punpcklbw %%mm7, %%mm1     \n\t"\
        "punpcklbw %%mm7, %%mm2     \n\t"\
        "punpcklbw %%mm7, %%mm3     \n\t"\
        "punpcklbw %%mm7, %%mm4     \n\t"\
616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638
        QPEL_H264V(%%mm0, %%mm1, %%mm2, %%mm3, %%mm4, %%mm5, OP)\
        QPEL_H264V(%%mm1, %%mm2, %%mm3, %%mm4, %%mm5, %%mm0, OP)\
        QPEL_H264V(%%mm2, %%mm3, %%mm4, %%mm5, %%mm0, %%mm1, OP)\
        QPEL_H264V(%%mm3, %%mm4, %%mm5, %%mm0, %%mm1, %%mm2, OP)\
        QPEL_H264V(%%mm4, %%mm5, %%mm0, %%mm1, %%mm2, %%mm3, OP)\
        QPEL_H264V(%%mm5, %%mm0, %%mm1, %%mm2, %%mm3, %%mm4, OP)\
        QPEL_H264V(%%mm0, %%mm1, %%mm2, %%mm3, %%mm4, %%mm5, OP)\
        QPEL_H264V(%%mm1, %%mm2, %%mm3, %%mm4, %%mm5, %%mm0, OP)\
         \
        : "+a"(src), "+c"(dst)\
        : "S"((long)srcStride), "D"((long)dstStride), "m"(ff_pw_5), "m"(ff_pw_16)\
        : "memory"\
     );\
     src += 4-13*srcStride;\
     dst +=  4-8*dstStride;\
   }\
}\
static void OPNAME ## h264_qpel8_hv_lowpass_ ## MMX(uint8_t *dst, int16_t *tmp, uint8_t *src, int dstStride, int tmpStride, int srcStride){\
    int h=8;\
    int w=4;\
    src -= 2*srcStride+2;\
    while(w--){\
        asm volatile(\
639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654
            "pxor %%mm7, %%mm7      \n\t"\
            "movd (%0), %%mm0       \n\t"\
            "add %2, %0             \n\t"\
            "movd (%0), %%mm1       \n\t"\
            "add %2, %0             \n\t"\
            "movd (%0), %%mm2       \n\t"\
            "add %2, %0             \n\t"\
            "movd (%0), %%mm3       \n\t"\
            "add %2, %0             \n\t"\
            "movd (%0), %%mm4       \n\t"\
            "add %2, %0             \n\t"\
            "punpcklbw %%mm7, %%mm0 \n\t"\
            "punpcklbw %%mm7, %%mm1 \n\t"\
            "punpcklbw %%mm7, %%mm2 \n\t"\
            "punpcklbw %%mm7, %%mm3 \n\t"\
            "punpcklbw %%mm7, %%mm4 \n\t"\
655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672
            QPEL_H264HV(%%mm0, %%mm1, %%mm2, %%mm3, %%mm4, %%mm5, 0*8*4)\
            QPEL_H264HV(%%mm1, %%mm2, %%mm3, %%mm4, %%mm5, %%mm0, 1*8*4)\
            QPEL_H264HV(%%mm2, %%mm3, %%mm4, %%mm5, %%mm0, %%mm1, 2*8*4)\
            QPEL_H264HV(%%mm3, %%mm4, %%mm5, %%mm0, %%mm1, %%mm2, 3*8*4)\
            QPEL_H264HV(%%mm4, %%mm5, %%mm0, %%mm1, %%mm2, %%mm3, 4*8*4)\
            QPEL_H264HV(%%mm5, %%mm0, %%mm1, %%mm2, %%mm3, %%mm4, 5*8*4)\
            QPEL_H264HV(%%mm0, %%mm1, %%mm2, %%mm3, %%mm4, %%mm5, 6*8*4)\
            QPEL_H264HV(%%mm1, %%mm2, %%mm3, %%mm4, %%mm5, %%mm0, 7*8*4)\
             \
            : "+a"(src)\
            : "c"(tmp), "S"((long)srcStride), "m"(ff_pw_5)\
            : "memory"\
        );\
        tmp += 4;\
        src += 4 - 13*srcStride;\
    }\
    tmp -= 4*4;\
    asm volatile(\
673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703
        "movq %4, %%mm6             \n\t"\
        "1:                         \n\t"\
        "movq     (%0), %%mm0       \n\t"\
        "movq    8(%0), %%mm3       \n\t"\
        "movq    2(%0), %%mm1       \n\t"\
        "movq   10(%0), %%mm4       \n\t"\
        "paddw   %%mm4, %%mm0       \n\t"\
        "paddw   %%mm3, %%mm1       \n\t"\
        "paddw  18(%0), %%mm3       \n\t"\
        "paddw  16(%0), %%mm4       \n\t"\
        "movq    4(%0), %%mm2       \n\t"\
        "movq   12(%0), %%mm5       \n\t"\
        "paddw   6(%0), %%mm2       \n\t"\
        "paddw  14(%0), %%mm5       \n\t"\
        "psubw %%mm1, %%mm0         \n\t"\
        "psubw %%mm4, %%mm3         \n\t"\
        "psraw $2, %%mm0            \n\t"\
        "psraw $2, %%mm3            \n\t"\
        "psubw %%mm1, %%mm0         \n\t"\
        "psubw %%mm4, %%mm3         \n\t"\
        "paddsw %%mm2, %%mm0        \n\t"\
        "paddsw %%mm5, %%mm3        \n\t"\
        "psraw $2, %%mm0            \n\t"\
        "psraw $2, %%mm3            \n\t"\
        "paddw %%mm6, %%mm2         \n\t"\
        "paddw %%mm6, %%mm5         \n\t"\
        "paddw %%mm2, %%mm0         \n\t"\
        "paddw %%mm5, %%mm3         \n\t"\
        "psraw $6, %%mm0            \n\t"\
        "psraw $6, %%mm3            \n\t"\
        "packuswb %%mm3, %%mm0      \n\t"\
704
        OP(%%mm0, (%1),%%mm7, q)\
705 706 707 708
        "add $32, %0                \n\t"\
        "add %3, %1                 \n\t"\
        "decl %2                    \n\t"\
        " jnz 1b                    \n\t"\
709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768 769 770 771 772 773 774 775 776 777 778 779 780 781 782 783 784 785 786 787 788 789 790 791 792 793 794 795 796 797 798 799 800 801 802 803 804 805 806 807 808 809 810 811 812 813 814 815 816 817 818 819 820 821 822 823 824 825 826 827 828 829 830 831 832 833 834 835 836 837 838 839 840 841 842 843 844 845 846 847 848 849 850 851 852 853 854 855 856 857 858 859 860 861 862 863 864
        : "+a"(tmp), "+c"(dst), "+m"(h)\
        : "S"((long)dstStride), "m"(ff_pw_32)\
        : "memory"\
    );\
}\
static void OPNAME ## h264_qpel16_v_lowpass_ ## MMX(uint8_t *dst, uint8_t *src, int dstStride, int srcStride){\
    OPNAME ## h264_qpel8_v_lowpass_ ## MMX(dst  , src  , dstStride, srcStride);\
    OPNAME ## h264_qpel8_v_lowpass_ ## MMX(dst+8, src+8, dstStride, srcStride);\
    src += 8*srcStride;\
    dst += 8*dstStride;\
    OPNAME ## h264_qpel8_v_lowpass_ ## MMX(dst  , src  , dstStride, srcStride);\
    OPNAME ## h264_qpel8_v_lowpass_ ## MMX(dst+8, src+8, dstStride, srcStride);\
}\
\
static void OPNAME ## h264_qpel16_h_lowpass_ ## MMX(uint8_t *dst, uint8_t *src, int dstStride, int srcStride){\
    OPNAME ## h264_qpel8_h_lowpass_ ## MMX(dst  , src  , dstStride, srcStride);\
    OPNAME ## h264_qpel8_h_lowpass_ ## MMX(dst+8, src+8, dstStride, srcStride);\
    src += 8*srcStride;\
    dst += 8*dstStride;\
    OPNAME ## h264_qpel8_h_lowpass_ ## MMX(dst  , src  , dstStride, srcStride);\
    OPNAME ## h264_qpel8_h_lowpass_ ## MMX(dst+8, src+8, dstStride, srcStride);\
}\
\
static void OPNAME ## h264_qpel16_hv_lowpass_ ## MMX(uint8_t *dst, int16_t *tmp, uint8_t *src, int dstStride, int tmpStride, int srcStride){\
    OPNAME ## h264_qpel8_hv_lowpass_ ## MMX(dst  , tmp  , src  , dstStride, tmpStride, srcStride);\
    OPNAME ## h264_qpel8_hv_lowpass_ ## MMX(dst+8, tmp  , src+8, dstStride, tmpStride, srcStride);\
    src += 8*srcStride;\
    dst += 8*dstStride;\
    OPNAME ## h264_qpel8_hv_lowpass_ ## MMX(dst  , tmp  , src  , dstStride, tmpStride, srcStride);\
    OPNAME ## h264_qpel8_hv_lowpass_ ## MMX(dst+8, tmp  , src+8, dstStride, tmpStride, srcStride);\
}\

#define H264_MC(OPNAME, SIZE, MMX) \
static void OPNAME ## h264_qpel ## SIZE ## _mc00_ ## MMX (uint8_t *dst, uint8_t *src, int stride){\
    OPNAME ## pixels ## SIZE ## _mmx(dst, src, stride, SIZE);\
}\
\
static void OPNAME ## h264_qpel ## SIZE ## _mc10_ ## MMX(uint8_t *dst, uint8_t *src, int stride){\
    uint64_t temp[SIZE*SIZE/8];\
    uint8_t * const half= (uint8_t*)temp;\
    put_h264_qpel ## SIZE ## _h_lowpass_ ## MMX(half, src, SIZE, stride);\
    OPNAME ## pixels ## SIZE ## _l2_ ## MMX(dst, src, half, stride, stride, SIZE);\
}\
\
static void OPNAME ## h264_qpel ## SIZE ## _mc20_ ## MMX(uint8_t *dst, uint8_t *src, int stride){\
    OPNAME ## h264_qpel ## SIZE ## _h_lowpass_ ## MMX(dst, src, stride, stride);\
}\
\
static void OPNAME ## h264_qpel ## SIZE ## _mc30_ ## MMX(uint8_t *dst, uint8_t *src, int stride){\
    uint64_t temp[SIZE*SIZE/8];\
    uint8_t * const half= (uint8_t*)temp;\
    put_h264_qpel ## SIZE ## _h_lowpass_ ## MMX(half, src, SIZE, stride);\
    OPNAME ## pixels ## SIZE ## _l2_ ## MMX(dst, src+1, half, stride, stride, SIZE);\
}\
\
static void OPNAME ## h264_qpel ## SIZE ## _mc01_ ## MMX(uint8_t *dst, uint8_t *src, int stride){\
    uint64_t temp[SIZE*SIZE/8];\
    uint8_t * const half= (uint8_t*)temp;\
    put_h264_qpel ## SIZE ## _v_lowpass_ ## MMX(half, src, SIZE, stride);\
    OPNAME ## pixels ## SIZE ## _l2_ ## MMX(dst, src, half, stride, stride, SIZE);\
}\
\
static void OPNAME ## h264_qpel ## SIZE ## _mc02_ ## MMX(uint8_t *dst, uint8_t *src, int stride){\
    OPNAME ## h264_qpel ## SIZE ## _v_lowpass_ ## MMX(dst, src, stride, stride);\
}\
\
static void OPNAME ## h264_qpel ## SIZE ## _mc03_ ## MMX(uint8_t *dst, uint8_t *src, int stride){\
    uint64_t temp[SIZE*SIZE/8];\
    uint8_t * const half= (uint8_t*)temp;\
    put_h264_qpel ## SIZE ## _v_lowpass_ ## MMX(half, src, SIZE, stride);\
    OPNAME ## pixels ## SIZE ## _l2_ ## MMX(dst, src+stride, half, stride, stride, SIZE);\
}\
\
static void OPNAME ## h264_qpel ## SIZE ## _mc11_ ## MMX(uint8_t *dst, uint8_t *src, int stride){\
    uint64_t temp[SIZE*SIZE/4];\
    uint8_t * const halfH= (uint8_t*)temp;\
    uint8_t * const halfV= ((uint8_t*)temp) + SIZE*SIZE;\
    put_h264_qpel ## SIZE ## _h_lowpass_ ## MMX(halfH, src, SIZE, stride);\
    put_h264_qpel ## SIZE ## _v_lowpass_ ## MMX(halfV, src, SIZE, stride);\
    OPNAME ## pixels ## SIZE ## _l2_ ## MMX(dst, halfH, halfV, stride, SIZE, SIZE);\
}\
\
static void OPNAME ## h264_qpel ## SIZE ## _mc31_ ## MMX(uint8_t *dst, uint8_t *src, int stride){\
    uint64_t temp[SIZE*SIZE/4];\
    uint8_t * const halfH= (uint8_t*)temp;\
    uint8_t * const halfV= ((uint8_t*)temp) + SIZE*SIZE;\
    put_h264_qpel ## SIZE ## _h_lowpass_ ## MMX(halfH, src, SIZE, stride);\
    put_h264_qpel ## SIZE ## _v_lowpass_ ## MMX(halfV, src+1, SIZE, stride);\
    OPNAME ## pixels ## SIZE ## _l2_ ## MMX(dst, halfH, halfV, stride, SIZE, SIZE);\
}\
\
static void OPNAME ## h264_qpel ## SIZE ## _mc13_ ## MMX(uint8_t *dst, uint8_t *src, int stride){\
    uint64_t temp[SIZE*SIZE/4];\
    uint8_t * const halfH= (uint8_t*)temp;\
    uint8_t * const halfV= ((uint8_t*)temp) + SIZE*SIZE;\
    put_h264_qpel ## SIZE ## _h_lowpass_ ## MMX(halfH, src + stride, SIZE, stride);\
    put_h264_qpel ## SIZE ## _v_lowpass_ ## MMX(halfV, src, SIZE, stride);\
    OPNAME ## pixels ## SIZE ## _l2_ ## MMX(dst, halfH, halfV, stride, SIZE, SIZE);\
}\
\
static void OPNAME ## h264_qpel ## SIZE ## _mc33_ ## MMX(uint8_t *dst, uint8_t *src, int stride){\
    uint64_t temp[SIZE*SIZE/4];\
    uint8_t * const halfH= (uint8_t*)temp;\
    uint8_t * const halfV= ((uint8_t*)temp) + SIZE*SIZE;\
    put_h264_qpel ## SIZE ## _h_lowpass_ ## MMX(halfH, src + stride, SIZE, stride);\
    put_h264_qpel ## SIZE ## _v_lowpass_ ## MMX(halfV, src+1, SIZE, stride);\
    OPNAME ## pixels ## SIZE ## _l2_ ## MMX(dst, halfH, halfV, stride, SIZE, SIZE);\
}\
\
static void OPNAME ## h264_qpel ## SIZE ## _mc22_ ## MMX(uint8_t *dst, uint8_t *src, int stride){\
    uint64_t temp[SIZE*(SIZE+8)/4];\
    int16_t * const tmp= (int16_t*)temp;\
    OPNAME ## h264_qpel ## SIZE ## _hv_lowpass_ ## MMX(dst, tmp, src, stride, SIZE, stride);\
}\
\
static void OPNAME ## h264_qpel ## SIZE ## _mc21_ ## MMX(uint8_t *dst, uint8_t *src, int stride){\
    uint64_t temp[SIZE*(SIZE+8)/4 + SIZE*SIZE/4];\
    uint8_t * const halfH= (uint8_t*)temp;\
    uint8_t * const halfHV= ((uint8_t*)temp) + SIZE*SIZE;\
    int16_t * const tmp= ((int16_t*)temp) + SIZE*SIZE;\
    put_h264_qpel ## SIZE ## _h_lowpass_ ## MMX(halfH, src, SIZE, stride);\
    put_h264_qpel ## SIZE ## _hv_lowpass_ ## MMX(halfHV, tmp, src, SIZE, SIZE, stride);\
    OPNAME ## pixels ## SIZE ## _l2_ ## MMX(dst, halfH, halfHV, stride, SIZE, SIZE);\
}\
\
static void OPNAME ## h264_qpel ## SIZE ## _mc23_ ## MMX(uint8_t *dst, uint8_t *src, int stride){\
    uint64_t temp[SIZE*(SIZE+8)/4 + SIZE*SIZE/4];\
    uint8_t * const halfH= (uint8_t*)temp;\
    uint8_t * const halfHV= ((uint8_t*)temp) + SIZE*SIZE;\
    int16_t * const tmp= ((int16_t*)temp) + SIZE*SIZE;\
    put_h264_qpel ## SIZE ## _h_lowpass_ ## MMX(halfH, src + stride, SIZE, stride);\
    put_h264_qpel ## SIZE ## _hv_lowpass_ ## MMX(halfHV, tmp, src, SIZE, SIZE, stride);\
    OPNAME ## pixels ## SIZE ## _l2_ ## MMX(dst, halfH, halfHV, stride, SIZE, SIZE);\
}\
\
static void OPNAME ## h264_qpel ## SIZE ## _mc12_ ## MMX(uint8_t *dst, uint8_t *src, int stride){\
    uint64_t temp[SIZE*(SIZE+8)/4 + SIZE*SIZE/4];\
    uint8_t * const halfV= (uint8_t*)temp;\
    uint8_t * const halfHV= ((uint8_t*)temp) + SIZE*SIZE;\
    int16_t * const tmp= ((int16_t*)temp) + SIZE*SIZE;\
    put_h264_qpel ## SIZE ## _v_lowpass_ ## MMX(halfV, src, SIZE, stride);\
    put_h264_qpel ## SIZE ## _hv_lowpass_ ## MMX(halfHV, tmp, src, SIZE, SIZE, stride);\
    OPNAME ## pixels ## SIZE ## _l2_ ## MMX(dst, halfV, halfHV, stride, SIZE, SIZE);\
}\
\
static void OPNAME ## h264_qpel ## SIZE ## _mc32_ ## MMX(uint8_t *dst, uint8_t *src, int stride){\
    uint64_t temp[SIZE*(SIZE+8)/4 + SIZE*SIZE/4];\
    uint8_t * const halfV= (uint8_t*)temp;\
    uint8_t * const halfHV= ((uint8_t*)temp) + SIZE*SIZE;\
    int16_t * const tmp= ((int16_t*)temp) + SIZE*SIZE;\
    put_h264_qpel ## SIZE ## _v_lowpass_ ## MMX(halfV, src+1, SIZE, stride);\
    put_h264_qpel ## SIZE ## _hv_lowpass_ ## MMX(halfHV, tmp, src, SIZE, SIZE, stride);\
    OPNAME ## pixels ## SIZE ## _l2_ ## MMX(dst, halfV, halfHV, stride, SIZE, SIZE);\
}\


865
#define PUT_OP(a,b,temp, size) "mov" #size " " #a ", " #b "    \n\t"
866
#define AVG_3DNOW_OP(a,b,temp, size) \
867 868 869
"mov" #size " " #b ", " #temp "   \n\t"\
"pavgusb " #temp ", " #a "        \n\t"\
"mov" #size " " #a ", " #b "      \n\t"
870
#define AVG_MMX2_OP(a,b,temp, size) \
871 872 873
"mov" #size " " #b ", " #temp "   \n\t"\
"pavgb " #temp ", " #a "          \n\t"\
"mov" #size " " #a ", " #b "      \n\t"
874 875 876 877 878 879 880 881 882 883 884 885 886 887 888 889 890 891 892 893 894

QPEL_H264(put_,       PUT_OP, 3dnow)
QPEL_H264(avg_, AVG_3DNOW_OP, 3dnow)
QPEL_H264(put_,       PUT_OP, mmx2)
QPEL_H264(avg_,  AVG_MMX2_OP, mmx2)

H264_MC(put_, 4, 3dnow)
H264_MC(put_, 8, 3dnow)
H264_MC(put_, 16,3dnow)
H264_MC(avg_, 4, 3dnow)
H264_MC(avg_, 8, 3dnow)
H264_MC(avg_, 16,3dnow)
H264_MC(put_, 4, mmx2)
H264_MC(put_, 8, mmx2)
H264_MC(put_, 16,mmx2)
H264_MC(avg_, 4, mmx2)
H264_MC(avg_, 8, mmx2)
H264_MC(avg_, 16,mmx2)


#define H264_CHROMA_OP(S,D)
895
#define H264_CHROMA_OP4(S,D,T)
896
#define H264_CHROMA_MC8_TMPL put_h264_chroma_mc8_mmx
897 898
#define H264_CHROMA_MC4_TMPL put_h264_chroma_mc4_mmx
#define H264_CHROMA_MC8_MV0 put_pixels8_mmx
899 900
#include "dsputil_h264_template_mmx.c"
#undef H264_CHROMA_OP
901
#undef H264_CHROMA_OP4
902
#undef H264_CHROMA_MC8_TMPL
903 904
#undef H264_CHROMA_MC4_TMPL
#undef H264_CHROMA_MC8_MV0
905 906

#define H264_CHROMA_OP(S,D) "pavgb " #S ", " #D " \n\t"
907 908
#define H264_CHROMA_OP4(S,D,T) "movd  " #S ", " #T " \n\t"\
                               "pavgb " #T ", " #D " \n\t"
909
#define H264_CHROMA_MC8_TMPL avg_h264_chroma_mc8_mmx2
910 911
#define H264_CHROMA_MC4_TMPL avg_h264_chroma_mc4_mmx2
#define H264_CHROMA_MC8_MV0 avg_pixels8_mmx2
912 913
#include "dsputil_h264_template_mmx.c"
#undef H264_CHROMA_OP
914
#undef H264_CHROMA_OP4
915
#undef H264_CHROMA_MC8_TMPL
916 917
#undef H264_CHROMA_MC4_TMPL
#undef H264_CHROMA_MC8_MV0
918 919

#define H264_CHROMA_OP(S,D) "pavgusb " #S ", " #D " \n\t"
920 921
#define H264_CHROMA_OP4(S,D,T) "movd " #S ", " #T " \n\t"\
                               "pavgusb " #T ", " #D " \n\t"
922
#define H264_CHROMA_MC8_TMPL avg_h264_chroma_mc8_3dnow
923 924
#define H264_CHROMA_MC4_TMPL avg_h264_chroma_mc4_3dnow
#define H264_CHROMA_MC8_MV0 avg_pixels8_3dnow
925 926
#include "dsputil_h264_template_mmx.c"
#undef H264_CHROMA_OP
927
#undef H264_CHROMA_OP4
928
#undef H264_CHROMA_MC8_TMPL
929 930
#undef H264_CHROMA_MC4_TMPL
#undef H264_CHROMA_MC8_MV0
931

932 933 934 935 936 937 938 939 940 941 942 943 944 945 946 947 948 949 950 951 952 953 954 955 956 957
/***********************************/
/* weighted prediction */

static inline void ff_h264_weight_WxH_mmx2(uint8_t *dst, int stride, int log2_denom, int weight, int offset, int w, int h)
{
    int x, y;
    offset <<= log2_denom;
    offset += (1 << log2_denom) >> 1;
    asm volatile(
        "movd    %0, %%mm4        \n\t"
        "movd    %1, %%mm5        \n\t"
        "movd    %2, %%mm6        \n\t"
        "pshufw  $0, %%mm4, %%mm4 \n\t"
        "pshufw  $0, %%mm5, %%mm5 \n\t"
        "pxor    %%mm7, %%mm7     \n\t"
        :: "g"(weight), "g"(offset), "g"(log2_denom)
    );
    for(y=0; y<h; y+=2){
        for(x=0; x<w; x+=4){
            asm volatile(
                "movd      %0,    %%mm0 \n\t"
                "movd      %1,    %%mm1 \n\t"
                "punpcklbw %%mm7, %%mm0 \n\t"
                "punpcklbw %%mm7, %%mm1 \n\t"
                "pmullw    %%mm4, %%mm0 \n\t"
                "pmullw    %%mm4, %%mm1 \n\t"
958 959
                "paddsw    %%mm5, %%mm0 \n\t"
                "paddsw    %%mm5, %%mm1 \n\t"
960 961 962 963 964 965 966 967 968 969 970 971 972 973 974 975 976 977 978 979 980 981 982 983 984 985 986 987 988 989 990 991 992 993 994 995 996 997
                "psraw     %%mm6, %%mm0 \n\t"
                "psraw     %%mm6, %%mm1 \n\t"
                "packuswb  %%mm7, %%mm0 \n\t"
                "packuswb  %%mm7, %%mm1 \n\t"
                "movd      %%mm0, %0    \n\t"
                "movd      %%mm1, %1    \n\t"
                : "+m"(*(uint32_t*)(dst+x)),
                  "+m"(*(uint32_t*)(dst+x+stride))
            );
        }
        dst += 2*stride;
    }
}

static inline void ff_h264_biweight_WxH_mmx2(uint8_t *dst, uint8_t *src, int stride, int log2_denom, int weightd, int weights, int offsetd, int offsets, int w, int h)
{
    int x, y;
    int offset = ((offsets + offsetd + 1) | 1) << log2_denom;
    asm volatile(
        "movd    %0, %%mm3        \n\t"
        "movd    %1, %%mm4        \n\t"
        "movd    %2, %%mm5        \n\t"
        "movd    %3, %%mm6        \n\t"
        "pshufw  $0, %%mm3, %%mm3 \n\t"
        "pshufw  $0, %%mm4, %%mm4 \n\t"
        "pshufw  $0, %%mm5, %%mm5 \n\t"
        "pxor    %%mm7, %%mm7     \n\t"
        :: "g"(weightd), "g"(weights), "g"(offset), "g"(log2_denom+1)
    );
    for(y=0; y<h; y++){
        for(x=0; x<w; x+=4){
            asm volatile(
                "movd      %0,    %%mm0 \n\t"
                "movd      %1,    %%mm1 \n\t"
                "punpcklbw %%mm7, %%mm0 \n\t"
                "punpcklbw %%mm7, %%mm1 \n\t"
                "pmullw    %%mm3, %%mm0 \n\t"
                "pmullw    %%mm4, %%mm1 \n\t"
998 999
                "paddsw    %%mm1, %%mm0 \n\t"
                "paddsw    %%mm5, %%mm0 \n\t"
1000 1001 1002 1003 1004 1005 1006 1007 1008 1009 1010 1011 1012 1013 1014 1015 1016 1017 1018 1019 1020 1021 1022 1023 1024 1025 1026 1027 1028
                "psraw     %%mm6, %%mm0 \n\t"
                "packuswb  %%mm0, %%mm0 \n\t"
                "movd      %%mm0, %0    \n\t"
                : "+m"(*(uint32_t*)(dst+x))
                :  "m"(*(uint32_t*)(src+x))
            );
        }
        src += stride;
        dst += stride;
    }
}

#define H264_WEIGHT(W,H) \
static void ff_h264_biweight_ ## W ## x ## H ## _mmx2(uint8_t *dst, uint8_t *src, int stride, int log2_denom, int weightd, int weights, int offsetd, int offsets){ \
    ff_h264_biweight_WxH_mmx2(dst, src, stride, log2_denom, weightd, weights, offsetd, offsets, W, H); \
} \
static void ff_h264_weight_ ## W ## x ## H ## _mmx2(uint8_t *dst, int stride, int log2_denom, int weight, int offset){ \
    ff_h264_weight_WxH_mmx2(dst, stride, log2_denom, weight, offset, W, H); \
}

H264_WEIGHT(16,16)
H264_WEIGHT(16, 8)
H264_WEIGHT( 8,16)
H264_WEIGHT( 8, 8)
H264_WEIGHT( 8, 4)
H264_WEIGHT( 4, 8)
H264_WEIGHT( 4, 4)
H264_WEIGHT( 4, 2)