h264dsp_mmx.c.svn-base

mediastreamer2是开源的网络传输媒体流的库

        "packuswb %%mm1, %%mm0          \n\t"\
        "packuswb %%mm3, %%mm2          \n\t"\
        PAVGB"     (%0), %%mm0          \n\t"\
        PAVGB"  (%0,%3), %%mm2          \n\t"\
        OP(%%mm0, (%2), %%mm5, q)\
        OP(%%mm2, (%2,%4), %%mm5, q)\
        ::"a"(src8), "c"(src16), "d"(dst),\
          "r"((long)src8Stride), "r"((long)dstStride)\
        :"memory");\
        src8 += 2L*src8Stride;\
        src16 += 48;\
        dst += 2L*dstStride;\
    }while(h-=2);\
}\
static void OPNAME ## pixels16_l2_shift5_ ## MMX(uint8_t *dst, int16_t *src16, uint8_t *src8, int dstStride, int src8Stride, int h)\
{\
    OPNAME ## pixels8_l2_shift5_ ## MMX(dst  , src16  , src8  , dstStride, src8Stride, h);\
    OPNAME ## pixels8_l2_shift5_ ## MMX(dst+8, src16+8, src8+8, dstStride, src8Stride, h);\
}\


#ifdef ARCH_X86_64
#define QPEL_H264_H16_XMM(OPNAME, OP, MMX)\
static av_noinline void OPNAME ## h264_qpel16_h_lowpass_l2_ ## MMX(uint8_t *dst, uint8_t *src, uint8_t *src2, int dstStride, int src2Stride){\
    int h=16;\
    asm volatile(\
        "pxor %%xmm15, %%xmm15      \n\t"\
        "movdqa %6, %%xmm14         \n\t"\
        "movdqa %7, %%xmm13         \n\t"\
        "1:                         \n\t"\
        "lddqu    3(%0), %%xmm1     \n\t"\
        "lddqu   -5(%0), %%xmm7     \n\t"\
        "movdqa  %%xmm1, %%xmm0     \n\t"\
        "punpckhbw %%xmm15, %%xmm1  \n\t"\
        "punpcklbw %%xmm15, %%xmm0  \n\t"\
        "punpcklbw %%xmm15, %%xmm7  \n\t"\
        "movdqa  %%xmm1, %%xmm2     \n\t"\
        "movdqa  %%xmm0, %%xmm6     \n\t"\
        "movdqa  %%xmm1, %%xmm3     \n\t"\
        "movdqa  %%xmm0, %%xmm8     \n\t"\
        "movdqa  %%xmm1, %%xmm4     \n\t"\
        "movdqa  %%xmm0, %%xmm9     \n\t"\
        "movdqa  %%xmm1, %%xmm5     \n\t"\
        "movdqa  %%xmm0, %%xmm10    \n\t"\
        "palignr $6, %%xmm0, %%xmm5 \n\t"\
        "palignr $6, %%xmm7, %%xmm10\n\t"\
        "palignr $8, %%xmm0, %%xmm4 \n\t"\
        "palignr $8, %%xmm7, %%xmm9 \n\t"\
        "palignr $10,%%xmm0, %%xmm3 \n\t"\
        "palignr $10,%%xmm7, %%xmm8 \n\t"\
        "paddw   %%xmm1, %%xmm5     \n\t"\
        "paddw   %%xmm0, %%xmm10    \n\t"\
        "palignr $12,%%xmm0, %%xmm2 \n\t"\
        "palignr $12,%%xmm7, %%xmm6 \n\t"\
        "palignr $14,%%xmm0, %%xmm1 \n\t"\
        "palignr $14,%%xmm7, %%xmm0 \n\t"\
        "paddw   %%xmm3, %%xmm2     \n\t"\
        "paddw   %%xmm8, %%xmm6     \n\t"\
        "paddw   %%xmm4, %%xmm1     \n\t"\
        "paddw   %%xmm9, %%xmm0     \n\t"\
        "psllw   $2,     %%xmm2     \n\t"\
        "psllw   $2,     %%xmm6     \n\t"\
        "psubw   %%xmm1, %%xmm2     \n\t"\
        "psubw   %%xmm0, %%xmm6     \n\t"\
        "paddw   %%xmm13,%%xmm5     \n\t"\
        "paddw   %%xmm13,%%xmm10    \n\t"\
        "pmullw  %%xmm14,%%xmm2     \n\t"\
        "pmullw  %%xmm14,%%xmm6     \n\t"\
        "lddqu   (%2),   %%xmm3     \n\t"\
        "paddw   %%xmm5, %%xmm2     \n\t"\
        "paddw   %%xmm10,%%xmm6     \n\t"\
        "psraw   $5,     %%xmm2     \n\t"\
        "psraw   $5,     %%xmm6     \n\t"\
        "packuswb %%xmm2,%%xmm6     \n\t"\
        "pavgb   %%xmm3, %%xmm6     \n\t"\
        OP(%%xmm6, (%1), %%xmm4, dqa)\
        "add %5, %0                 \n\t"\
        "add %5, %1                 \n\t"\
        "add %4, %2                 \n\t"\
        "decl %3                    \n\t"\
        "jg 1b                      \n\t"\
        : "+a"(src), "+c"(dst), "+d"(src2), "+g"(h)\
        : "D"((long)src2Stride), "S"((long)dstStride),\
          "m"(ff_pw_5), "m"(ff_pw_16)\
        : "memory"\
    );\
}
#else // ARCH_X86_64
#define QPEL_H264_H16_XMM(OPNAME, OP, MMX)\
static av_noinline void OPNAME ## h264_qpel16_h_lowpass_l2_ ## MMX(uint8_t *dst, uint8_t *src, uint8_t *src2, int dstStride, int src2Stride){\
    OPNAME ## h264_qpel8_h_lowpass_l2_ ## MMX(dst  , src  , src2  , dstStride, src2Stride);\
    OPNAME ## h264_qpel8_h_lowpass_l2_ ## MMX(dst+8, src+8, src2+8, dstStride, src2Stride);\
    src += 8*dstStride;\
    dst += 8*dstStride;\
    src2 += 8*src2Stride;\
    OPNAME ## h264_qpel8_h_lowpass_l2_ ## MMX(dst  , src  , src2  , dstStride, src2Stride);\
    OPNAME ## h264_qpel8_h_lowpass_l2_ ## MMX(dst+8, src+8, src2+8, dstStride, src2Stride);\
}
#endif // ARCH_X86_64

#define QPEL_H264_H_XMM(OPNAME, OP, MMX)\
static av_noinline void OPNAME ## h264_qpel8_h_lowpass_l2_ ## MMX(uint8_t *dst, uint8_t *src, uint8_t *src2, int dstStride, int src2Stride){\
    int h=8;\
    asm volatile(\
        "pxor %%xmm7, %%xmm7        \n\t"\
        "movdqa %0, %%xmm6          \n\t"\
        :: "m"(ff_pw_5)\
    );\
    do{\
    asm volatile(\
        "lddqu   -5(%0), %%xmm1     \n\t"\
        "movdqa  %%xmm1, %%xmm0     \n\t"\
        "punpckhbw %%xmm7, %%xmm1   \n\t"\
        "punpcklbw %%xmm7, %%xmm0   \n\t"\
        "movdqa  %%xmm1, %%xmm2     \n\t"\
        "movdqa  %%xmm1, %%xmm3     \n\t"\
        "movdqa  %%xmm1, %%xmm4     \n\t"\
        "movdqa  %%xmm1, %%xmm5     \n\t"\
        "palignr $6, %%xmm0, %%xmm5 \n\t"\
        "palignr $8, %%xmm0, %%xmm4 \n\t"\
        "palignr $10,%%xmm0, %%xmm3 \n\t"\
        "paddw   %%xmm1, %%xmm5     \n\t"\
        "palignr $12,%%xmm0, %%xmm2 \n\t"\
        "palignr $14,%%xmm0, %%xmm1 \n\t"\
        "paddw   %%xmm3, %%xmm2     \n\t"\
        "paddw   %%xmm4, %%xmm1     \n\t"\
        "psllw   $2,     %%xmm2     \n\t"\
        "movq    (%2),   %%xmm3     \n\t"\
        "psubw   %%xmm1, %%xmm2     \n\t"\
        "paddw   %5,     %%xmm5     \n\t"\
        "pmullw  %%xmm6, %%xmm2     \n\t"\
        "paddw   %%xmm5, %%xmm2     \n\t"\
        "psraw   $5,     %%xmm2     \n\t"\
        "packuswb %%xmm2, %%xmm2    \n\t"\
        "pavgb   %%xmm3, %%xmm2     \n\t"\
        OP(%%xmm2, (%1), %%xmm4, q)\
        "add %4, %0                 \n\t"\
        "add %4, %1                 \n\t"\
        "add %3, %2                 \n\t"\
        : "+a"(src), "+c"(dst), "+d"(src2)\
        : "D"((long)src2Stride), "S"((long)dstStride),\
          "m"(ff_pw_16)\
        : "memory"\
    );\
    }while(--h);\
}\
QPEL_H264_H16_XMM(OPNAME, OP, MMX)\
\
static av_noinline void OPNAME ## h264_qpel8_h_lowpass_ ## MMX(uint8_t *dst, uint8_t *src, int dstStride, int srcStride){\
    int h=8;\
    asm volatile(\
        "pxor %%xmm7, %%xmm7        \n\t"\
        "movdqa %5, %%xmm6          \n\t"\
        "1:                         \n\t"\
        "lddqu   -5(%0), %%xmm1     \n\t"\
        "movdqa  %%xmm1, %%xmm0     \n\t"\
        "punpckhbw %%xmm7, %%xmm1   \n\t"\
        "punpcklbw %%xmm7, %%xmm0   \n\t"\
        "movdqa  %%xmm1, %%xmm2     \n\t"\
        "movdqa  %%xmm1, %%xmm3     \n\t"\
        "movdqa  %%xmm1, %%xmm4     \n\t"\
        "movdqa  %%xmm1, %%xmm5     \n\t"\
        "palignr $6, %%xmm0, %%xmm5 \n\t"\
        "palignr $8, %%xmm0, %%xmm4 \n\t"\
        "palignr $10,%%xmm0, %%xmm3 \n\t"\
        "paddw   %%xmm1, %%xmm5     \n\t"\
        "palignr $12,%%xmm0, %%xmm2 \n\t"\
        "palignr $14,%%xmm0, %%xmm1 \n\t"\
        "paddw   %%xmm3, %%xmm2     \n\t"\
        "paddw   %%xmm4, %%xmm1     \n\t"\
        "psllw   $2,     %%xmm2     \n\t"\
        "psubw   %%xmm1, %%xmm2     \n\t"\
        "paddw   %6,     %%xmm5     \n\t"\
        "pmullw  %%xmm6, %%xmm2     \n\t"\
        "paddw   %%xmm5, %%xmm2     \n\t"\
        "psraw   $5,     %%xmm2     \n\t"\
        "packuswb %%xmm2, %%xmm2    \n\t"\
        OP(%%xmm2, (%1), %%xmm4, q)\
        "add %3, %0                 \n\t"\
        "add %4, %1                 \n\t"\
        "decl %2                    \n\t"\
        " jnz 1b                    \n\t"\
        : "+a"(src), "+c"(dst), "+g"(h)\
        : "D"((long)srcStride), "S"((long)dstStride),\
          "m"(ff_pw_5), "m"(ff_pw_16)\
        : "memory"\
    );\
}\
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);\
}\

#define QPEL_H264_V_XMM(OPNAME, OP, MMX)\
static av_noinline void OPNAME ## h264_qpel8or16_v_lowpass_ ## MMX(uint8_t *dst, uint8_t *src, int dstStride, int srcStride, int h){\
    src -= 2*srcStride;\
    \
    asm volatile(\
        "pxor %%xmm7, %%xmm7        \n\t"\
        "movq (%0), %%xmm0          \n\t"\
        "add %2, %0                 \n\t"\
        "movq (%0), %%xmm1          \n\t"\
        "add %2, %0                 \n\t"\
        "movq (%0), %%xmm2          \n\t"\
        "add %2, %0                 \n\t"\
        "movq (%0), %%xmm3          \n\t"\
        "add %2, %0                 \n\t"\
        "movq (%0), %%xmm4          \n\t"\
        "add %2, %0                 \n\t"\
        "punpcklbw %%xmm7, %%xmm0   \n\t"\
        "punpcklbw %%xmm7, %%xmm1   \n\t"\
        "punpcklbw %%xmm7, %%xmm2   \n\t"\
        "punpcklbw %%xmm7, %%xmm3   \n\t"\
        "punpcklbw %%xmm7, %%xmm4   \n\t"\
        QPEL_H264V_XMM(%%xmm0, %%xmm1, %%xmm2, %%xmm3, %%xmm4, %%xmm5, OP)\
        QPEL_H264V_XMM(%%xmm1, %%xmm2, %%xmm3, %%xmm4, %%xmm5, %%xmm0, OP)\
        QPEL_H264V_XMM(%%xmm2, %%xmm3, %%xmm4, %%xmm5, %%xmm0, %%xmm1, OP)\
        QPEL_H264V_XMM(%%xmm3, %%xmm4, %%xmm5, %%xmm0, %%xmm1, %%xmm2, OP)\
        QPEL_H264V_XMM(%%xmm4, %%xmm5, %%xmm0, %%xmm1, %%xmm2, %%xmm3, OP)\
        QPEL_H264V_XMM(%%xmm5, %%xmm0, %%xmm1, %%xmm2, %%xmm3, %%xmm4, OP)\
        QPEL_H264V_XMM(%%xmm0, %%xmm1, %%xmm2, %%xmm3, %%xmm4, %%xmm5, OP)\
        QPEL_H264V_XMM(%%xmm1, %%xmm2, %%xmm3, %%xmm4, %%xmm5, %%xmm0, OP)\
         \
        : "+a"(src), "+c"(dst)\
        : "S"((long)srcStride), "D"((long)dstStride), "m"(ff_pw_5), "m"(ff_pw_16)\
        : "memory"\
    );\
    if(h==16){\
        asm volatile(\
            QPEL_H264V_XMM(%%xmm2, %%xmm3, %%xmm4, %%xmm5, %%xmm0, %%xmm1, OP)\
            QPEL_H264V_XMM(%%xmm3, %%xmm4, %%xmm5, %%xmm0, %%xmm1, %%xmm2, OP)\
            QPEL_H264V_XMM(%%xmm4, %%xmm5, %%xmm0, %%xmm1, %%xmm2, %%xmm3, OP)\
            QPEL_H264V_XMM(%%xmm5, %%xmm0, %%xmm1, %%xmm2, %%xmm3, %%xmm4, OP)\
            QPEL_H264V_XMM(%%xmm0, %%xmm1, %%xmm2, %%xmm3, %%xmm4, %%xmm5, OP)\
            QPEL_H264V_XMM(%%xmm1, %%xmm2, %%xmm3, %%xmm4, %%xmm5, %%xmm0, OP)\
            QPEL_H264V_XMM(%%xmm2, %%xmm3, %%xmm4, %%xmm5, %%xmm0, %%xmm1, OP)\
            QPEL_H264V_XMM(%%xmm3, %%xmm4, %%xmm5, %%xmm0, %%xmm1, %%xmm2, OP)\
            \
            : "+a"(src), "+c"(dst)\
            : "S"((long)srcStride), "D"((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){\
    OPNAME ## h264_qpel8or16_v_lowpass_ ## MMX(dst  , src  , dstStride, srcStride, 8);\
}\
static av_noinline void OPNAME ## h264_qpel16_v_lowpass_ ## MMX(uint8_t *dst, uint8_t *src, int dstStride, int srcStride){\
    OPNAME ## h264_qpel8or16_v_lowpass_ ## MMX(dst  , src  , dstStride, srcStride, 16);\
    OPNAME ## h264_qpel8or16_v_lowpass_ ## MMX(dst+8, src+8, dstStride, srcStride, 16);\
}

static av_always_inline void put_h264_qpel8or16_hv1_lowpass_sse2(int16_t *tmp, uint8_t *src, int tmpStride, int srcStride, int size){
    int w = (size+8)>>3;
    src -= 2*srcStride+2;
    while(w--){
        asm volatile(
            "pxor %%xmm7, %%xmm7        \n\t"
            "movq (%0), %%xmm0          \n\t"
            "add %2, %0                 \n\t"
            "movq (%0), %%xmm1          \n\t"
            "add %2, %0                 \n\t"
            "movq (%0), %%xmm2          \n\t"
            "add %2, %0                 \n\t"
            "movq (%0), %%xmm3          \n\t"
            "add %2, %0                 \n\t"
            "movq (%0), %%xmm4          \n\t"
            "add %2, %0                 \n\t"
            "punpcklbw %%xmm7, %%xmm0   \n\t"
            "punpcklbw %%xmm7, %%xmm1   \n\t"
            "punpcklbw %%xmm7, %%xmm2   \n\t"
            "punpcklbw %%xmm7, %%xmm3   \n\t"
            "punpcklbw %%xmm7, %%xmm4   \n\t"
            QPEL_H264HV_XMM(%%xmm0, %%xmm1, %%xmm2, %%xmm3, %%xmm4, %%xmm5, 0*48)
            QPEL_H264HV_XMM(%%xmm1, %%xmm2, %%xmm3, %%xmm4, %%xmm5, %%xmm0, 1*48)
            QPEL_H264HV_XMM(%%xmm2, %%xmm3, %%xmm4, %%xmm5, %%xmm0, %%xmm1, 2*48)
            QPEL_H264HV_XMM(%%xmm3, %%xmm4, %%xmm5, %%xmm0, %%xmm1, %%xmm2, 3*48)
            QPEL_H264HV_XMM(%%xmm4, %%xmm5, %%xmm0, %%xmm1, %%xmm2, %%xmm3, 4*48)
            QPEL_H264HV_XMM(%%xmm5, %%xmm0, %%xmm1, %%xmm2, %%xmm3, %%xmm4, 5*48)
            QPEL_H264HV_XMM(%%xmm0, %%xmm1, %%xmm2, %%xmm3, %%xmm4, %%xmm5, 6*48)
            QPEL_H264HV_XMM(%%xmm1, %%xmm2, %%xmm3, %%xmm4, %%xmm5, %%xmm0, 7*48)
            : "+a"(src)
            : "c"(tmp), "S"((long)srcStride), "m"(ff_pw_5), "m"(ff_pw_16)
            : "memory"
        );
        if(size==16){
            asm volatile(
                QPEL_H264HV_XMM(%%xmm2, %%xmm3, %%xmm4, %%xmm5, %%xmm0, %%xmm1,  8*48)
                QPEL_H264HV_XMM(%%xmm3, %%xmm4, %%xmm5, %%xmm0, %%xmm1, %%xmm2,  9*48)
                QPEL_H264HV_XMM(%%xmm4, %%xmm5, %%xmm0, %%xmm1, %%xmm2, %%xmm3, 10*48)
                QPEL_H264HV_XMM(%%xmm5, %%xmm0, %%xmm1, %%xmm2, %%xmm3, %%xmm4, 11*48)
                QPEL_H264HV_XMM(%%xmm0, %%xmm1, %%xmm2, %%xmm3, %%xmm4, %%xmm5, 12*48)
                QPEL_H264HV_XMM(%%xmm1, %%xmm2, %%xmm3, %%xmm4, %%xmm5, %%xmm0, 13*48)
                QPEL_H264HV_XMM(%%xmm2, %%xmm3, %%xmm4, %%xmm5, %%xmm0, %%xmm1, 14*48)
                QPEL_H264HV_XMM(%%xmm3, %%xmm4, %%xmm5, %%xmm0, %%xmm1, %%xmm2, 15*48)
                : "+a"(src)
                : "c"(tmp), "S"((long)srcStride), "m"(ff_pw_5), "m"(ff_pw_16)
                : "memory"
            );
        }
        tmp += 8;
        src += 8 - (size+5)*srcStride;
    }
}

#define QPEL_H264_HV2_XMM(OPNAME, OP, MMX)\
static av_always_inline void OPNAME ## h264_qpel8or16_hv2_lowpass_ ## MMX(uint8_t *dst, int16_t *tmp, int dstStride, int tmpStride, int size){\
    int h = size;\
    if(size == 16){\
        asm volatile(\
            "1:                         \n\t"\
            "movdqa 32(%0), %%xmm4      \n\t"\
            "movdqa 16(%0), %%xmm5      \n\t"\
            "movdqa   (%0), %%xmm7      \n\t"\
            "movdqa %%xmm4, %%xmm3      \n\t"\
            "movdqa %%xmm4, %%xmm2      \n\t"\
            "movdqa %%xmm4, %%xmm1      \n\t"\
            "movdqa %%xmm4, %%xmm0      \n\t"\