dsputil_mmx.c

ffmpeg移植到symbian的全部源代码

        "paddb %%mm1, %%mm6             \n\t"

static void h263_v_loop_filter_mmx(uint8_t *src, int stride, int qscale){
    if(ENABLE_ANY_H263) {
    const int strength= ff_h263_loop_filter_strength[qscale];

    asm volatile(

        H263_LOOP_FILTER

        "movq %%mm3, %1                 \n\t"
        "movq %%mm4, %2                 \n\t"
        "movq %%mm5, %0                 \n\t"
        "movq %%mm6, %3                 \n\t"
        : "+m" (*(uint64_t*)(src - 2*stride)),
          "+m" (*(uint64_t*)(src - 1*stride)),
          "+m" (*(uint64_t*)(src + 0*stride)),
          "+m" (*(uint64_t*)(src + 1*stride))
        : "g" (2*strength), "m"(ff_pb_FC)
    );
    }
}

static inline void transpose4x4(uint8_t *dst, uint8_t *src, int dst_stride, int src_stride){
    asm volatile( //FIXME could save 1 instruction if done as 8x4 ...
        "movd  %4, %%mm0                \n\t"
        "movd  %5, %%mm1                \n\t"
        "movd  %6, %%mm2                \n\t"
        "movd  %7, %%mm3                \n\t"
        "punpcklbw %%mm1, %%mm0         \n\t"
        "punpcklbw %%mm3, %%mm2         \n\t"
        "movq %%mm0, %%mm1              \n\t"
        "punpcklwd %%mm2, %%mm0         \n\t"
        "punpckhwd %%mm2, %%mm1         \n\t"
        "movd  %%mm0, %0                \n\t"
        "punpckhdq %%mm0, %%mm0         \n\t"
        "movd  %%mm0, %1                \n\t"
        "movd  %%mm1, %2                \n\t"
        "punpckhdq %%mm1, %%mm1         \n\t"
        "movd  %%mm1, %3                \n\t"

        : "=m" (*(uint32_t*)(dst + 0*dst_stride)),
          "=m" (*(uint32_t*)(dst + 1*dst_stride)),
          "=m" (*(uint32_t*)(dst + 2*dst_stride)),
          "=m" (*(uint32_t*)(dst + 3*dst_stride))
        :  "m" (*(uint32_t*)(src + 0*src_stride)),
           "m" (*(uint32_t*)(src + 1*src_stride)),
           "m" (*(uint32_t*)(src + 2*src_stride)),
           "m" (*(uint32_t*)(src + 3*src_stride))
    );
}

static void h263_h_loop_filter_mmx(uint8_t *src, int stride, int qscale){
    if(ENABLE_ANY_H263) {
    const int strength= ff_h263_loop_filter_strength[qscale];
    DECLARE_ALIGNED(8, uint64_t, temp[4]);
    uint8_t *btemp= (uint8_t*)temp;

    src -= 2;

    transpose4x4(btemp  , src           , 8, stride);
    transpose4x4(btemp+4, src + 4*stride, 8, stride);
    asm volatile(
        H263_LOOP_FILTER // 5 3 4 6

        : "+m" (temp[0]),
          "+m" (temp[1]),
          "+m" (temp[2]),
          "+m" (temp[3])
        : "g" (2*strength), "m"(ff_pb_FC)
    );

    asm volatile(
        "movq %%mm5, %%mm1              \n\t"
        "movq %%mm4, %%mm0              \n\t"
        "punpcklbw %%mm3, %%mm5         \n\t"
        "punpcklbw %%mm6, %%mm4         \n\t"
        "punpckhbw %%mm3, %%mm1         \n\t"
        "punpckhbw %%mm6, %%mm0         \n\t"
        "movq %%mm5, %%mm3              \n\t"
        "movq %%mm1, %%mm6              \n\t"
        "punpcklwd %%mm4, %%mm5         \n\t"
        "punpcklwd %%mm0, %%mm1         \n\t"
        "punpckhwd %%mm4, %%mm3         \n\t"
        "punpckhwd %%mm0, %%mm6         \n\t"
        "movd %%mm5, (%0)               \n\t"
        "punpckhdq %%mm5, %%mm5         \n\t"
        "movd %%mm5, (%0,%2)            \n\t"
        "movd %%mm3, (%0,%2,2)          \n\t"
        "punpckhdq %%mm3, %%mm3         \n\t"
        "movd %%mm3, (%0,%3)            \n\t"
        "movd %%mm1, (%1)               \n\t"
        "punpckhdq %%mm1, %%mm1         \n\t"
        "movd %%mm1, (%1,%2)            \n\t"
        "movd %%mm6, (%1,%2,2)          \n\t"
        "punpckhdq %%mm6, %%mm6         \n\t"
        "movd %%mm6, (%1,%3)            \n\t"
        :: "r" (src),
           "r" (src + 4*stride),
           "r" ((x86_reg)   stride ),
           "r" ((x86_reg)(3*stride))
    );
    }
}

/* draw the edges of width 'w' of an image of size width, height
   this mmx version can only handle w==8 || w==16 */
static void draw_edges_mmx(uint8_t *buf, int wrap, int width, int height, int w)
{
    uint8_t *ptr, *last_line;
    int i;

    last_line = buf + (height - 1) * wrap;
    /* left and right */
    ptr = buf;
    if(w==8)
    {
        asm volatile(
                "1:                             \n\t"
                "movd (%0), %%mm0               \n\t"
                "punpcklbw %%mm0, %%mm0         \n\t"
                "punpcklwd %%mm0, %%mm0         \n\t"
                "punpckldq %%mm0, %%mm0         \n\t"
                "movq %%mm0, -8(%0)             \n\t"
                "movq -8(%0, %2), %%mm1         \n\t"
                "punpckhbw %%mm1, %%mm1         \n\t"
                "punpckhwd %%mm1, %%mm1         \n\t"
                "punpckhdq %%mm1, %%mm1         \n\t"
                "movq %%mm1, (%0, %2)           \n\t"
                "add %1, %0                     \n\t"
                "cmp %3, %0                     \n\t"
                " jb 1b                         \n\t"
                : "+r" (ptr)
                : "r" ((x86_reg)wrap), "r" ((x86_reg)width), "r" (ptr + wrap*height)
        );
    }
    else
    {
        asm volatile(
                "1:                             \n\t"
                "movd (%0), %%mm0               \n\t"
                "punpcklbw %%mm0, %%mm0         \n\t"
                "punpcklwd %%mm0, %%mm0         \n\t"
                "punpckldq %%mm0, %%mm0         \n\t"
                "movq %%mm0, -8(%0)             \n\t"
                "movq %%mm0, -16(%0)            \n\t"
                "movq -8(%0, %2), %%mm1         \n\t"
                "punpckhbw %%mm1, %%mm1         \n\t"
                "punpckhwd %%mm1, %%mm1         \n\t"
                "punpckhdq %%mm1, %%mm1         \n\t"
                "movq %%mm1, (%0, %2)           \n\t"
                "movq %%mm1, 8(%0, %2)          \n\t"
                "add %1, %0                     \n\t"
                "cmp %3, %0                     \n\t"
                " jb 1b                         \n\t"
                : "+r" (ptr)
                : "r" ((x86_reg)wrap), "r" ((x86_reg)width), "r" (ptr + wrap*height)
        );
    }

    for(i=0;i<w;i+=4) {
        /* top and bottom (and hopefully also the corners) */
        ptr= buf - (i + 1) * wrap - w;
        asm volatile(
                "1:                             \n\t"
                "movq (%1, %0), %%mm0           \n\t"
                "movq %%mm0, (%0)               \n\t"
                "movq %%mm0, (%0, %2)           \n\t"
                "movq %%mm0, (%0, %2, 2)        \n\t"
                "movq %%mm0, (%0, %3)           \n\t"
                "add $8, %0                     \n\t"
                "cmp %4, %0                     \n\t"
                " jb 1b                         \n\t"
                : "+r" (ptr)
                : "r" ((x86_reg)buf - (x86_reg)ptr - w), "r" ((x86_reg)-wrap), "r" ((x86_reg)-wrap*3), "r" (ptr+width+2*w)
        );
        ptr= last_line + (i + 1) * wrap - w;
        asm volatile(
                "1:                             \n\t"
                "movq (%1, %0), %%mm0           \n\t"
                "movq %%mm0, (%0)               \n\t"
                "movq %%mm0, (%0, %2)           \n\t"
                "movq %%mm0, (%0, %2, 2)        \n\t"
                "movq %%mm0, (%0, %3)           \n\t"
                "add $8, %0                     \n\t"
                "cmp %4, %0                     \n\t"
                " jb 1b                         \n\t"
                : "+r" (ptr)
                : "r" ((x86_reg)last_line - (x86_reg)ptr - w), "r" ((x86_reg)wrap), "r" ((x86_reg)wrap*3), "r" (ptr+width+2*w)
        );
    }
}

#define PAETH(cpu, abs3)\
void add_png_paeth_prediction_##cpu(uint8_t *dst, uint8_t *src, uint8_t *top, int w, int bpp)\
{\
    x86_reg i = -bpp;\
    x86_reg end = w-3;\
    asm volatile(\
        "pxor      %%mm7, %%mm7 \n"\
        "movd    (%1,%0), %%mm0 \n"\
        "movd    (%2,%0), %%mm1 \n"\
        "punpcklbw %%mm7, %%mm0 \n"\
        "punpcklbw %%mm7, %%mm1 \n"\
        "add       %4, %0 \n"\
        "1: \n"\
        "movq      %%mm1, %%mm2 \n"\
        "movd    (%2,%0), %%mm1 \n"\
        "movq      %%mm2, %%mm3 \n"\
        "punpcklbw %%mm7, %%mm1 \n"\
        "movq      %%mm2, %%mm4 \n"\
        "psubw     %%mm1, %%mm3 \n"\
        "psubw     %%mm0, %%mm4 \n"\
        "movq      %%mm3, %%mm5 \n"\
        "paddw     %%mm4, %%mm5 \n"\
        abs3\
        "movq      %%mm4, %%mm6 \n"\
        "pminsw    %%mm5, %%mm6 \n"\
        "pcmpgtw   %%mm6, %%mm3 \n"\
        "pcmpgtw   %%mm5, %%mm4 \n"\
        "movq      %%mm4, %%mm6 \n"\
        "pand      %%mm3, %%mm4 \n"\
        "pandn     %%mm3, %%mm6 \n"\
        "pandn     %%mm0, %%mm3 \n"\
        "movd    (%3,%0), %%mm0 \n"\
        "pand      %%mm1, %%mm6 \n"\
        "pand      %%mm4, %%mm2 \n"\
        "punpcklbw %%mm7, %%mm0 \n"\
        "movq      %6,    %%mm5 \n"\
        "paddw     %%mm6, %%mm0 \n"\
        "paddw     %%mm2, %%mm3 \n"\
        "paddw     %%mm3, %%mm0 \n"\
        "pand      %%mm5, %%mm0 \n"\
        "movq      %%mm0, %%mm3 \n"\
        "packuswb  %%mm3, %%mm3 \n"\
        "movd      %%mm3, (%1,%0) \n"\
        "add       %4, %0 \n"\
        "cmp       %5, %0 \n"\
        "jle 1b \n"\
        :"+r"(i)\
        :"r"(dst), "r"(top), "r"(src), "r"((x86_reg)bpp), "g"(end),\
         "m"(ff_pw_255)\
        :"memory"\
    );\
}

#define ABS3_MMX2\
        "psubw     %%mm5, %%mm7 \n"\
        "pmaxsw    %%mm7, %%mm5 \n"\
        "pxor      %%mm6, %%mm6 \n"\
        "pxor      %%mm7, %%mm7 \n"\
        "psubw     %%mm3, %%mm6 \n"\
        "psubw     %%mm4, %%mm7 \n"\
        "pmaxsw    %%mm6, %%mm3 \n"\
        "pmaxsw    %%mm7, %%mm4 \n"\
        "pxor      %%mm7, %%mm7 \n"

#define ABS3_SSSE3\
        "pabsw     %%mm3, %%mm3 \n"\
        "pabsw     %%mm4, %%mm4 \n"\
        "pabsw     %%mm5, %%mm5 \n"

PAETH(mmx2, ABS3_MMX2)
#ifdef HAVE_SSSE3
PAETH(ssse3, ABS3_SSSE3)
#endif

#define QPEL_V_LOW(m3,m4,m5,m6, pw_20, pw_3, rnd, in0, in1, in2, in7, out, OP)\
        "paddw " #m4 ", " #m3 "           \n\t" /* x1 */\
        "movq "MANGLE(ff_pw_20)", %%mm4   \n\t" /* 20 */\
        "pmullw " #m3 ", %%mm4            \n\t" /* 20x1 */\
        "movq "#in7", " #m3 "             \n\t" /* d */\
        "movq "#in0", %%mm5               \n\t" /* D */\
        "paddw " #m3 ", %%mm5             \n\t" /* x4 */\
        "psubw %%mm5, %%mm4               \n\t" /* 20x1 - x4 */\
        "movq "#in1", %%mm5               \n\t" /* C */\
        "movq "#in2", %%mm6               \n\t" /* B */\
        "paddw " #m6 ", %%mm5             \n\t" /* x3 */\
        "paddw " #m5 ", %%mm6             \n\t" /* x2 */\
        "paddw %%mm6, %%mm6               \n\t" /* 2x2 */\
        "psubw %%mm6, %%mm5               \n\t" /* -2x2 + x3 */\
        "pmullw "MANGLE(ff_pw_3)", %%mm5  \n\t" /* -6x2 + 3x3 */\
        "paddw " #rnd ", %%mm4            \n\t" /* x2 */\
        "paddw %%mm4, %%mm5               \n\t" /* 20x1 - 6x2 + 3x3 - x4 */\
        "psraw $5, %%mm5                  \n\t"\
        "packuswb %%mm5, %%mm5            \n\t"\
        OP(%%mm5, out, %%mm7, d)

#define QPEL_BASE(OPNAME, ROUNDER, RND, OP_MMX2, OP_3DNOW)\
static void OPNAME ## mpeg4_qpel16_h_lowpass_mmx2(uint8_t *dst, uint8_t *src, int dstStride, int srcStride, int h){\
    uint64_t temp;\
\
    asm volatile(\
        "pxor %%mm7, %%mm7                \n\t"\
        "1:                               \n\t"\
        "movq  (%0), %%mm0                \n\t" /* ABCDEFGH */\
        "movq %%mm0, %%mm1                \n\t" /* ABCDEFGH */\
        "movq %%mm0, %%mm2                \n\t" /* ABCDEFGH */\