mpegvideo_mmx.c

Trolltech公司发布的图形界面操作系统。可在qt-embedded-2.3.7平台上编译为嵌入式图形界面操作系统。

            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"
                "movl %3, %%eax			\n\t"
		".balign 16\n\t"
		"1:				\n\t"
		"movq (%0, %%eax), %%mm0	\n\t"
		"movq 8(%0, %%eax), %%mm1	\n\t"
		"movq (%1, %%eax), %%mm4	\n\t"
		"movq 8(%1, %%eax), %%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, %%eax), %%mm4	\n\t" // block[i] == 0 ? -1 : 0
		"pcmpeqw 8(%0, %%eax), %%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, %%eax)	\n\t"
		"movq %%mm5, 8(%0, %%eax)	\n\t"

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

    } else {
        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"
                "movl %3, %%eax			\n\t"
		".balign 16\n\t"
		"1:				\n\t"
		"movq (%0, %%eax), %%mm0	\n\t"
		"movq 8(%0, %%eax), %%mm1	\n\t"
		"movq (%1, %%eax), %%mm4	\n\t"
		"movq 8(%1, %%eax), %%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, %%eax), %%mm4	\n\t" // block[i] == 0 ? -1 : 0
		"pcmpeqw 8(%0, %%eax), %%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, %%eax)	\n\t"
		"movq %%mm5, 8(%0, %%eax)	\n\t"

		"addl $16, %%eax		\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)
		: "%eax", "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 *buf, int wrap, int width, int height, int w)
{
    UINT8 *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"
		"addl %1, %0			\n\t"
		"cmpl %3, %0			\n\t"
		" jb 1b				\n\t"
		: "+r" (ptr)
		: "r" (wrap), "r" (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"
		"addl %1, %0			\n\t"
		"cmpl %3, %0			\n\t"
		" jb 1b				\n\t"		
		: "+r" (ptr)
		: "r" (wrap), "r" (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"
		"addl $8, %0			\n\t"
		"cmpl %4, %0			\n\t"
		" jb 1b				\n\t"
		: "+r" (ptr)
		: "r" ((int)buf - (int)ptr - w), "r" (-wrap), "r" (-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"
		"addl $8, %0			\n\t"
		"cmpl %4, %0			\n\t"
		" jb 1b				\n\t"
		: "+r" (ptr)
		: "r" ((int)last_line - (int)ptr - w), "r" (wrap), "r" (wrap*3), "r" (ptr+width+2*w)
	);
    }
}

#undef HAVE_MMX2
#define RENAME(a) a ## _MMX
#include "mpegvideo_mmx_template.c"

#define HAVE_MMX2
#undef RENAME
#define RENAME(a) a ## _MMX2
#include "mpegvideo_mmx_template.c"

/* external functions, from idct_mmx.c */
void ff_mmx_idct(DCTELEM *block);
void ff_mmxext_idct(DCTELEM *block);

/* XXX: those functions should be suppressed ASAP when all IDCTs are
   converted */
static void ff_libmpeg2mmx_idct_put(UINT8 *dest, int line_size, DCTELEM *block)
{
    ff_mmx_idct (block);
    put_pixels_clamped_mmx(block, dest, line_size);
}
static void ff_libmpeg2mmx_idct_add(UINT8 *dest, int line_size, DCTELEM *block)
{
    ff_mmx_idct (block);
    add_pixels_clamped_mmx(block, dest, line_size);
}
static void ff_libmpeg2mmx2_idct_put(UINT8 *dest, int line_size, DCTELEM *block)
{
    ff_mmxext_idct (block);
    put_pixels_clamped_mmx(block, dest, line_size);
}
static void ff_libmpeg2mmx2_idct_add(UINT8 *dest, int line_size, DCTELEM *block)
{
    ff_mmxext_idct (block);
    add_pixels_clamped_mmx(block, dest, line_size);
}

void MPV_common_init_mmx(MpegEncContext *s)
{
    if (mm_flags & MM_MMX) {
        int i;
        const int dct_algo = s->avctx->dct_algo;
        const int idct_algo= s->avctx->idct_algo;
        
        s->dct_unquantize_h263 = dct_unquantize_h263_mmx;
        s->dct_unquantize_mpeg1 = dct_unquantize_mpeg1_mmx;
        s->dct_unquantize_mpeg2 = dct_unquantize_mpeg2_mmx;

        draw_edges = draw_edges_mmx;

        if(dct_algo==FF_DCT_AUTO || dct_algo==FF_DCT_MMX){
            s->fdct = ff_fdct_mmx;

            if(mm_flags & MM_MMXEXT){
                s->dct_quantize= dct_quantize_MMX2;
            } else {
                s->dct_quantize= dct_quantize_MMX;
            }
        }

        if(idct_algo==FF_IDCT_AUTO || idct_algo==FF_IDCT_SIMPLEMMX){
            s->idct_put= ff_simple_idct_put_mmx;
            s->idct_add= ff_simple_idct_add_mmx;
            s->idct_permutation_type= FF_SIMPLE_IDCT_PERM;
        }else if(idct_algo==FF_IDCT_LIBMPEG2MMX){
            if(mm_flags & MM_MMXEXT){
                s->idct_put= ff_libmpeg2mmx2_idct_put;
                s->idct_add= ff_libmpeg2mmx2_idct_add;
            }else{
                s->idct_put= ff_libmpeg2mmx_idct_put;
                s->idct_add= ff_libmpeg2mmx_idct_add;
            }
            s->idct_permutation_type= FF_LIBMPEG2_IDCT_PERM;
        }
    }
}