mpegvideo_mmx.c

mediastreamer2是开源的网络传输媒体流的库

                "por %%mm7, %%mm0               \n\t"
                "por %%mm7, %%mm1               \n\t"
                "pxor %%mm2, %%mm0              \n\t"
                "pxor %%mm3, %%mm1              \n\t"
                "psubw %%mm2, %%mm0             \n\t"
                "psubw %%mm3, %%mm1             \n\t"
                "pandn %%mm0, %%mm4             \n\t"
                "pandn %%mm1, %%mm5             \n\t"
                "movq %%mm4, (%0, %%"REG_a")    \n\t"
                "movq %%mm5, 8(%0, %%"REG_a")   \n\t"

                "add $16, %%"REG_a"             \n\t"
                "js 1b                          \n\t"
                ::"r" (block+nCoeffs), "r"(quant_matrix+nCoeffs), "g" (qscale), "g" (-2*nCoeffs)
                : "%"REG_a, "memory"
        );
    block[0]= block0;
}

static void dct_unquantize_mpeg1_inter_mmx(MpegEncContext *s,
                                     DCTELEM *block, int n, int qscale)
{
    long nCoeffs;
    const uint16_t *quant_matrix;

    assert(s->block_last_index[n]>=0);

    nCoeffs= s->intra_scantable.raster_end[ s->block_last_index[n] ]+1;

        quant_matrix = s->inter_matrix;
asm volatile(
                "pcmpeqw %%mm7, %%mm7           \n\t"
                "psrlw $15, %%mm7               \n\t"
                "movd %2, %%mm6                 \n\t"
                "packssdw %%mm6, %%mm6          \n\t"
                "packssdw %%mm6, %%mm6          \n\t"
                "mov %3, %%"REG_a"              \n\t"
                ASMALIGN(4)
                "1:                             \n\t"
                "movq (%0, %%"REG_a"), %%mm0    \n\t"
                "movq 8(%0, %%"REG_a"), %%mm1   \n\t"
                "movq (%1, %%"REG_a"), %%mm4    \n\t"
                "movq 8(%1, %%"REG_a"), %%mm5   \n\t"
                "pmullw %%mm6, %%mm4            \n\t" // q=qscale*quant_matrix[i]
                "pmullw %%mm6, %%mm5            \n\t" // q=qscale*quant_matrix[i]
                "pxor %%mm2, %%mm2              \n\t"
                "pxor %%mm3, %%mm3              \n\t"
                "pcmpgtw %%mm0, %%mm2           \n\t" // block[i] < 0 ? -1 : 0
                "pcmpgtw %%mm1, %%mm3           \n\t" // block[i] < 0 ? -1 : 0
                "pxor %%mm2, %%mm0              \n\t"
                "pxor %%mm3, %%mm1              \n\t"
                "psubw %%mm2, %%mm0             \n\t" // abs(block[i])
                "psubw %%mm3, %%mm1             \n\t" // abs(block[i])
                "paddw %%mm0, %%mm0             \n\t" // abs(block[i])*2
                "paddw %%mm1, %%mm1             \n\t" // abs(block[i])*2
                "paddw %%mm7, %%mm0             \n\t" // abs(block[i])*2 + 1
                "paddw %%mm7, %%mm1             \n\t" // abs(block[i])*2 + 1
                "pmullw %%mm4, %%mm0            \n\t" // (abs(block[i])*2 + 1)*q
                "pmullw %%mm5, %%mm1            \n\t" // (abs(block[i])*2 + 1)*q
                "pxor %%mm4, %%mm4              \n\t"
                "pxor %%mm5, %%mm5              \n\t" // FIXME slow
                "pcmpeqw (%0, %%"REG_a"), %%mm4 \n\t" // block[i] == 0 ? -1 : 0
                "pcmpeqw 8(%0, %%"REG_a"), %%mm5\n\t" // block[i] == 0 ? -1 : 0
                "psraw $4, %%mm0                \n\t"
                "psraw $4, %%mm1                \n\t"
                "psubw %%mm7, %%mm0             \n\t"
                "psubw %%mm7, %%mm1             \n\t"
                "por %%mm7, %%mm0               \n\t"
                "por %%mm7, %%mm1               \n\t"
                "pxor %%mm2, %%mm0              \n\t"
                "pxor %%mm3, %%mm1              \n\t"
                "psubw %%mm2, %%mm0             \n\t"
                "psubw %%mm3, %%mm1             \n\t"
                "pandn %%mm0, %%mm4             \n\t"
                "pandn %%mm1, %%mm5             \n\t"
                "movq %%mm4, (%0, %%"REG_a")    \n\t"
                "movq %%mm5, 8(%0, %%"REG_a")   \n\t"

                "add $16, %%"REG_a"             \n\t"
                "js 1b                          \n\t"
                ::"r" (block+nCoeffs), "r"(quant_matrix+nCoeffs), "g" (qscale), "g" (-2*nCoeffs)
                : "%"REG_a, "memory"
        );
}

static void dct_unquantize_mpeg2_intra_mmx(MpegEncContext *s,
                                     DCTELEM *block, int n, int qscale)
{
    long nCoeffs;
    const uint16_t *quant_matrix;
    int block0;

    assert(s->block_last_index[n]>=0);

    if(s->alternate_scan) nCoeffs= 63; //FIXME
    else nCoeffs= s->intra_scantable.raster_end[ s->block_last_index[n] ];

    if (n < 4)
        block0 = block[0] * s->y_dc_scale;
    else
        block0 = block[0] * s->c_dc_scale;
    quant_matrix = s->intra_matrix;
asm volatile(
                "pcmpeqw %%mm7, %%mm7           \n\t"
                "psrlw $15, %%mm7               \n\t"
                "movd %2, %%mm6                 \n\t"
                "packssdw %%mm6, %%mm6          \n\t"
                "packssdw %%mm6, %%mm6          \n\t"
                "mov %3, %%"REG_a"              \n\t"
                ASMALIGN(4)
                "1:                             \n\t"
                "movq (%0, %%"REG_a"), %%mm0    \n\t"
                "movq 8(%0, %%"REG_a"), %%mm1   \n\t"
                "movq (%1, %%"REG_a"), %%mm4    \n\t"
                "movq 8(%1, %%"REG_a"), %%mm5   \n\t"
                "pmullw %%mm6, %%mm4            \n\t" // q=qscale*quant_matrix[i]
                "pmullw %%mm6, %%mm5            \n\t" // q=qscale*quant_matrix[i]
                "pxor %%mm2, %%mm2              \n\t"
                "pxor %%mm3, %%mm3              \n\t"
                "pcmpgtw %%mm0, %%mm2           \n\t" // block[i] < 0 ? -1 : 0
                "pcmpgtw %%mm1, %%mm3           \n\t" // block[i] < 0 ? -1 : 0
                "pxor %%mm2, %%mm0              \n\t"
                "pxor %%mm3, %%mm1              \n\t"
                "psubw %%mm2, %%mm0             \n\t" // abs(block[i])
                "psubw %%mm3, %%mm1             \n\t" // abs(block[i])
                "pmullw %%mm4, %%mm0            \n\t" // abs(block[i])*q
                "pmullw %%mm5, %%mm1            \n\t" // abs(block[i])*q
                "pxor %%mm4, %%mm4              \n\t"
                "pxor %%mm5, %%mm5              \n\t" // FIXME slow
                "pcmpeqw (%0, %%"REG_a"), %%mm4 \n\t" // block[i] == 0 ? -1 : 0
                "pcmpeqw 8(%0, %%"REG_a"), %%mm5\n\t" // block[i] == 0 ? -1 : 0
                "psraw $3, %%mm0                \n\t"
                "psraw $3, %%mm1                \n\t"
                "pxor %%mm2, %%mm0              \n\t"
                "pxor %%mm3, %%mm1              \n\t"
                "psubw %%mm2, %%mm0             \n\t"
                "psubw %%mm3, %%mm1             \n\t"
                "pandn %%mm0, %%mm4             \n\t"
                "pandn %%mm1, %%mm5             \n\t"
                "movq %%mm4, (%0, %%"REG_a")    \n\t"
                "movq %%mm5, 8(%0, %%"REG_a")   \n\t"

                "add $16, %%"REG_a"             \n\t"
                "jng 1b                         \n\t"
                ::"r" (block+nCoeffs), "r"(quant_matrix+nCoeffs), "g" (qscale), "g" (-2*nCoeffs)
                : "%"REG_a, "memory"
        );
    block[0]= block0;
        //Note, we do not do mismatch control for intra as errors cannot accumulate
}

static void dct_unquantize_mpeg2_inter_mmx(MpegEncContext *s,
                                     DCTELEM *block, int n, int qscale)
{
    long nCoeffs;
    const uint16_t *quant_matrix;

    assert(s->block_last_index[n]>=0);

    if(s->alternate_scan) nCoeffs= 63; //FIXME
    else nCoeffs= s->intra_scantable.raster_end[ s->block_last_index[n] ];

        quant_matrix = s->inter_matrix;
asm volatile(
                "pcmpeqw %%mm7, %%mm7           \n\t"
                "psrlq $48, %%mm7               \n\t"
                "movd %2, %%mm6                 \n\t"
                "packssdw %%mm6, %%mm6          \n\t"
                "packssdw %%mm6, %%mm6          \n\t"
                "mov %3, %%"REG_a"              \n\t"
                ASMALIGN(4)
                "1:                             \n\t"
                "movq (%0, %%"REG_a"), %%mm0    \n\t"
                "movq 8(%0, %%"REG_a"), %%mm1   \n\t"
                "movq (%1, %%"REG_a"), %%mm4    \n\t"
                "movq 8(%1, %%"REG_a"), %%mm5   \n\t"
                "pmullw %%mm6, %%mm4            \n\t" // q=qscale*quant_matrix[i]
                "pmullw %%mm6, %%mm5            \n\t" // q=qscale*quant_matrix[i]
                "pxor %%mm2, %%mm2              \n\t"
                "pxor %%mm3, %%mm3              \n\t"
                "pcmpgtw %%mm0, %%mm2           \n\t" // block[i] < 0 ? -1 : 0
                "pcmpgtw %%mm1, %%mm3           \n\t" // block[i] < 0 ? -1 : 0
                "pxor %%mm2, %%mm0              \n\t"
                "pxor %%mm3, %%mm1              \n\t"
                "psubw %%mm2, %%mm0             \n\t" // abs(block[i])
                "psubw %%mm3, %%mm1             \n\t" // abs(block[i])
                "paddw %%mm0, %%mm0             \n\t" // abs(block[i])*2
                "paddw %%mm1, %%mm1             \n\t" // abs(block[i])*2
                "pmullw %%mm4, %%mm0            \n\t" // abs(block[i])*2*q
                "pmullw %%mm5, %%mm1            \n\t" // abs(block[i])*2*q
                "paddw %%mm4, %%mm0             \n\t" // (abs(block[i])*2 + 1)*q
                "paddw %%mm5, %%mm1             \n\t" // (abs(block[i])*2 + 1)*q
                "pxor %%mm4, %%mm4              \n\t"
                "pxor %%mm5, %%mm5              \n\t" // FIXME slow
                "pcmpeqw (%0, %%"REG_a"), %%mm4 \n\t" // block[i] == 0 ? -1 : 0
                "pcmpeqw 8(%0, %%"REG_a"), %%mm5\n\t" // block[i] == 0 ? -1 : 0
                "psrlw $4, %%mm0                \n\t"
                "psrlw $4, %%mm1                \n\t"
                "pxor %%mm2, %%mm0              \n\t"
                "pxor %%mm3, %%mm1              \n\t"
                "psubw %%mm2, %%mm0             \n\t"
                "psubw %%mm3, %%mm1             \n\t"
                "pandn %%mm0, %%mm4             \n\t"
                "pandn %%mm1, %%mm5             \n\t"
                "pxor %%mm4, %%mm7              \n\t"
                "pxor %%mm5, %%mm7              \n\t"
                "movq %%mm4, (%0, %%"REG_a")    \n\t"
                "movq %%mm5, 8(%0, %%"REG_a")   \n\t"

                "add $16, %%"REG_a"             \n\t"
                "jng 1b                         \n\t"
                "movd 124(%0, %3), %%mm0        \n\t"
                "movq %%mm7, %%mm6              \n\t"
                "psrlq $32, %%mm7               \n\t"
                "pxor %%mm6, %%mm7              \n\t"
                "movq %%mm7, %%mm6              \n\t"
                "psrlq $16, %%mm7               \n\t"
                "pxor %%mm6, %%mm7              \n\t"
                "pslld $31, %%mm7               \n\t"
                "psrlq $15, %%mm7               \n\t"
                "pxor %%mm7, %%mm0              \n\t"
                "movd %%mm0, 124(%0, %3)        \n\t"

                ::"r" (block+nCoeffs), "r"(quant_matrix+nCoeffs), "g" (qscale), "r" (-2*nCoeffs)
                : "%"REG_a, "memory"
        );
}

/* 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"