view libmpcodecs/vf_ilpack.c @ 17588:79081ba52e00

Move the v{Y,C}CoeffsBank vectors into the SwsContext, filling them in just once when the scaler is initialized, instead of building them and freeing them over and over. This gives massive performance improvements. patch by Alan Curry, pacman*at*TheWorld*dot*com

#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include <inttypes.h>

#include "config.h"
#include "mp_msg.h"
#include "cpudetect.h"

#include "img_format.h"
#include "mp_image.h"
#include "vf.h"

#include "libvo/fastmemcpy.h"
#include "postproc/rgb2rgb.h"

typedef void (pack_func_t)(unsigned char *dst, unsigned char *y,
	unsigned char *u, unsigned char *v, int w, int us, int vs);

struct vf_priv_s {
	int mode;
	pack_func_t *pack[2];
};

static void pack_nn_C(unsigned char *dst, unsigned char *y,
	unsigned char *u, unsigned char *v, int w)
{
	int j;
	for (j = w/2; j; j--) {
		*dst++ = *y++;
		*dst++ = *u++;
		*dst++ = *y++;
		*dst++ = *v++;
	}
}

static void pack_li_0_C(unsigned char *dst, unsigned char *y,
	unsigned char *u, unsigned char *v, int w, int us, int vs)
{
	int j;
	for (j = w/2; j; j--) {
		*dst++ = *y++;
		*dst++ = (u[us+us] + 7*u[0])>>3;
		*dst++ = *y++;
		*dst++ = (v[vs+vs] + 7*v[0])>>3;
		u++; v++;
	}
}

static void pack_li_1_C(unsigned char *dst, unsigned char *y,
	unsigned char *u, unsigned char *v, int w, int us, int vs)
{
	int j;
	for (j = w/2; j; j--) {
		*dst++ = *y++;
		*dst++ = (3*u[us+us] + 5*u[0])>>3;
		*dst++ = *y++;
		*dst++ = (3*v[vs+vs] + 5*v[0])>>3;
		u++; v++;
	}
}

#ifdef HAVE_MMX
static void pack_nn_MMX(unsigned char *dst, unsigned char *y,
	unsigned char *u, unsigned char *v, int w)
{
	int j;
	asm volatile (""
		".balign 16 \n\t"
		"1: \n\t"
		"movq (%0), %%mm1 \n\t"
		"movq (%0), %%mm2 \n\t"
		"movq (%1), %%mm4 \n\t"
		"movq (%2), %%mm6 \n\t"
		"punpcklbw %%mm6, %%mm4 \n\t"
		"punpcklbw %%mm4, %%mm1 \n\t"
		"punpckhbw %%mm4, %%mm2 \n\t"
		
		"add $8, %0 \n\t"
		"add $4, %1 \n\t"
		"add $4, %2 \n\t"
		"movq %%mm1, (%3) \n\t"
		"movq %%mm2, 8(%3) \n\t"
		"add $16, %3 \n\t"
		"decl %4 \n\t"
		"jnz 1b \n\t"
		"emms \n\t"
		: 
		: "r" (y), "r" (u), "r" (v), "r" (dst), "r" (w/8)
		: "memory"
		);
	pack_nn_C(dst, y, u, v, (w&7));
}

static void pack_li_0_MMX(unsigned char *dst, unsigned char *y,
	unsigned char *u, unsigned char *v, int w, int us, int vs)
{
	asm volatile (""
		"push %%"REG_BP" \n\t"
#ifdef ARCH_X86_64
		"mov %6, %%"REG_BP" \n\t"
#else
		"movl 4(%%"REG_d"), %%"REG_BP" \n\t"
		"movl (%%"REG_d"), %%"REG_d" \n\t"
#endif
		"pxor %%mm0, %%mm0 \n\t"
		
		".balign 16 \n\t"
		".Lli0: \n\t"
		"movq (%%"REG_S"), %%mm1 \n\t"
		"movq (%%"REG_S"), %%mm2 \n\t"
		
		"movq (%%"REG_a",%%"REG_d",2), %%mm4 \n\t"
		"movq (%%"REG_b",%%"REG_BP",2), %%mm6 \n\t"
		"punpcklbw %%mm0, %%mm4 \n\t"
		"punpcklbw %%mm0, %%mm6 \n\t"
		"movq (%%"REG_a"), %%mm3 \n\t"
		"movq (%%"REG_b"), %%mm5 \n\t"
		"punpcklbw %%mm0, %%mm3 \n\t"
		"punpcklbw %%mm0, %%mm5 \n\t"
		"paddw %%mm3, %%mm4 \n\t"
		"paddw %%mm5, %%mm6 \n\t"
		"paddw %%mm3, %%mm4 \n\t"
		"paddw %%mm5, %%mm6 \n\t"
		"paddw %%mm3, %%mm4 \n\t"
		"paddw %%mm5, %%mm6 \n\t"
		"paddw %%mm3, %%mm4 \n\t"
		"paddw %%mm5, %%mm6 \n\t"
		"paddw %%mm3, %%mm4 \n\t"
		"paddw %%mm5, %%mm6 \n\t"
		"paddw %%mm3, %%mm4 \n\t"
		"paddw %%mm5, %%mm6 \n\t"
		"paddw %%mm3, %%mm4 \n\t"
		"paddw %%mm5, %%mm6 \n\t"
		"psrlw $3, %%mm4 \n\t"
		"psrlw $3, %%mm6 \n\t"
		"packuswb %%mm4, %%mm4 \n\t"
		"packuswb %%mm6, %%mm6 \n\t"
		"punpcklbw %%mm6, %%mm4 \n\t"
		"punpcklbw %%mm4, %%mm1 \n\t"
		"punpckhbw %%mm4, %%mm2 \n\t"
		
		"movq %%mm1, (%%"REG_D") \n\t"
		"movq %%mm2, 8(%%"REG_D") \n\t"
		
		"movq 8(%%"REG_S"), %%mm1 \n\t"
		"movq 8(%%"REG_S"), %%mm2 \n\t"
		
		"movq (%%"REG_a",%%"REG_d",2), %%mm4 \n\t"
		"movq (%%"REG_b",%%"REG_BP",2), %%mm6 \n\t"
		"punpckhbw %%mm0, %%mm4 \n\t"
		"punpckhbw %%mm0, %%mm6 \n\t"
		"movq (%%"REG_a"), %%mm3 \n\t"
		"movq (%%"REG_b"), %%mm5 \n\t"
		"punpckhbw %%mm0, %%mm3 \n\t"
		"punpckhbw %%mm0, %%mm5 \n\t"
		"paddw %%mm3, %%mm4 \n\t"
		"paddw %%mm5, %%mm6 \n\t"
		"paddw %%mm3, %%mm4 \n\t"
		"paddw %%mm5, %%mm6 \n\t"
		"paddw %%mm3, %%mm4 \n\t"
		"paddw %%mm5, %%mm6 \n\t"
		"paddw %%mm3, %%mm4 \n\t"
		"paddw %%mm5, %%mm6 \n\t"
		"paddw %%mm3, %%mm4 \n\t"
		"paddw %%mm5, %%mm6 \n\t"
		"paddw %%mm3, %%mm4 \n\t"
		"paddw %%mm5, %%mm6 \n\t"
		"paddw %%mm3, %%mm4 \n\t"
		"paddw %%mm5, %%mm6 \n\t"
		"psrlw $3, %%mm4 \n\t"
		"psrlw $3, %%mm6 \n\t"
		"packuswb %%mm4, %%mm4 \n\t"
		"packuswb %%mm6, %%mm6 \n\t"
		"punpcklbw %%mm6, %%mm4 \n\t"
		"punpcklbw %%mm4, %%mm1 \n\t"
		"punpckhbw %%mm4, %%mm2 \n\t"
		
		"add $16, %%"REG_S" \n\t"
		"add $8, %%"REG_a" \n\t"
		"add $8, %%"REG_b" \n\t"
		
		"movq %%mm1, 16(%%"REG_D") \n\t"
		"movq %%mm2, 24(%%"REG_D") \n\t"
		"add $32, %%"REG_D" \n\t"
		
		"decl %%ecx \n\t"
		"jnz .Lli0 \n\t"
		"emms \n\t"
		"pop %%"REG_BP" \n\t"
		: 
		: "S" (y), "D" (dst), "a" (u), "b" (v), "c" (w/16),
#ifdef ARCH_X86_64
		"d" ((long)us), "r" ((long)vs)
#else
		"d" (&us)
#endif
		: "memory"
		);
	pack_li_0_C(dst, y, u, v, (w&15), us, vs);
}

static void pack_li_1_MMX(unsigned char *dst, unsigned char *y,
	unsigned char *u, unsigned char *v, int w, int us, int vs)
{
	asm volatile (""
		"push %%"REG_BP" \n\t"
#ifdef ARCH_X86_64
		"mov %6, %%"REG_BP" \n\t"
#else
		"movl 4(%%"REG_d"), %%"REG_BP" \n\t"
		"movl (%%"REG_d"), %%"REG_d" \n\t"
#endif
		"pxor %%mm0, %%mm0 \n\t"
		
		".balign 16 \n\t"
		".Lli1: \n\t"
		"movq (%%"REG_S"), %%mm1 \n\t"
		"movq (%%"REG_S"), %%mm2 \n\t"
		
		"movq (%%"REG_a",%%"REG_d",2), %%mm4 \n\t"
		"movq (%%"REG_b",%%"REG_BP",2), %%mm6 \n\t"
		"punpcklbw %%mm0, %%mm4 \n\t"
		"punpcklbw %%mm0, %%mm6 \n\t"
		"movq (%%"REG_a"), %%mm3 \n\t"
		"movq (%%"REG_b"), %%mm5 \n\t"
		"punpcklbw %%mm0, %%mm3 \n\t"
		"punpcklbw %%mm0, %%mm5 \n\t"
		"movq %%mm4, %%mm7 \n\t"
		"paddw %%mm4, %%mm4 \n\t"
		"paddw %%mm7, %%mm4 \n\t"
		"movq %%mm6, %%mm7 \n\t"
		"paddw %%mm6, %%mm6 \n\t"
		"paddw %%mm7, %%mm6 \n\t"
		"paddw %%mm3, %%mm4 \n\t"
		"paddw %%mm5, %%mm6 \n\t"
		"paddw %%mm3, %%mm4 \n\t"
		"paddw %%mm5, %%mm6 \n\t"
		"paddw %%mm3, %%mm4 \n\t"
		"paddw %%mm5, %%mm6 \n\t"
		"paddw %%mm3, %%mm4 \n\t"
		"paddw %%mm5, %%mm6 \n\t"
		"paddw %%mm3, %%mm4 \n\t"
		"paddw %%mm5, %%mm6 \n\t"
		"psrlw $3, %%mm4 \n\t"
		"psrlw $3, %%mm6 \n\t"
		"packuswb %%mm4, %%mm4 \n\t"
		"packuswb %%mm6, %%mm6 \n\t"
		"punpcklbw %%mm6, %%mm4 \n\t"
		"punpcklbw %%mm4, %%mm1 \n\t"
		"punpckhbw %%mm4, %%mm2 \n\t"
		
		"movq %%mm1, (%%"REG_D") \n\t"
		"movq %%mm2, 8(%%"REG_D") \n\t"
		
		"movq 8(%%"REG_S"), %%mm1 \n\t"
		"movq 8(%%"REG_S"), %%mm2 \n\t"
		
		"movq (%%"REG_a",%%"REG_d",2), %%mm4 \n\t"
		"movq (%%"REG_b",%%"REG_BP",2), %%mm6 \n\t"
		"punpckhbw %%mm0, %%mm4 \n\t"
		"punpckhbw %%mm0, %%mm6 \n\t"
		"movq (%%"REG_a"), %%mm3 \n\t"
		"movq (%%"REG_b"), %%mm5 \n\t"
		"punpckhbw %%mm0, %%mm3 \n\t"
		"punpckhbw %%mm0, %%mm5 \n\t"
		"movq %%mm4, %%mm7 \n\t"
		"paddw %%mm4, %%mm4 \n\t"
		"paddw %%mm7, %%mm4 \n\t"
		"movq %%mm6, %%mm7 \n\t"
		"paddw %%mm6, %%mm6 \n\t"
		"paddw %%mm7, %%mm6 \n\t"
		"paddw %%mm3, %%mm4 \n\t"
		"paddw %%mm5, %%mm6 \n\t"
		"paddw %%mm3, %%mm4 \n\t"
		"paddw %%mm5, %%mm6 \n\t"
		"paddw %%mm3, %%mm4 \n\t"
		"paddw %%mm5, %%mm6 \n\t"
		"paddw %%mm3, %%mm4 \n\t"
		"paddw %%mm5, %%mm6 \n\t"
		"paddw %%mm3, %%mm4 \n\t"
		"paddw %%mm5, %%mm6 \n\t"
		"psrlw $3, %%mm4 \n\t"
		"psrlw $3, %%mm6 \n\t"
		"packuswb %%mm4, %%mm4 \n\t"
		"packuswb %%mm6, %%mm6 \n\t"
		"punpcklbw %%mm6, %%mm4 \n\t"
		"punpcklbw %%mm4, %%mm1 \n\t"
		"punpckhbw %%mm4, %%mm2 \n\t"
		
		"add $16, %%"REG_S" \n\t"
		"add $8, %%"REG_a" \n\t"
		"add $8, %%"REG_b" \n\t"
		
		"movq %%mm1, 16(%%"REG_D") \n\t"
		"movq %%mm2, 24(%%"REG_D") \n\t"
		"add $32, %%"REG_D" \n\t"
		
		"decl %%ecx \n\t"
		"jnz .Lli1 \n\t"
		"emms \n\t"
		"pop %%"REG_BP" \n\t"
		: 
		: "S" (y), "D" (dst), "a" (u), "b" (v), "c" (w/16),
#ifdef ARCH_X86_64
		"d" ((long)us), "r" ((long)vs)
#else
		"d" (&us)
#endif
		: "memory"
		);
	pack_li_1_C(dst, y, u, v, (w&15), us, vs);
}
#endif

static pack_func_t *pack_nn;
static pack_func_t *pack_li_0;
static pack_func_t *pack_li_1;

static void ilpack(unsigned char *dst, unsigned char *src[3],
	int dststride, int srcstride[3], int w, int h, pack_func_t *pack[2])
{
	int i;
	unsigned char *y, *u, *v;
	int ys = srcstride[0], us = srcstride[1], vs = srcstride[2];
	int a, b;

	y = src[0];
	u = src[1];
	v = src[2];

	pack_nn(dst, y, u, v, w, 0, 0);
	y += ys; dst += dststride;
	pack_nn(dst, y, u+us, v+vs, w, 0, 0);
	y += ys; dst += dststride;
	for (i=2; i<h-2; i++) {
		a = (i&2) ? 1 : -1;
		b = (i&1) ^ ((i&2)>>1);
		pack[b](dst, y, u, v, w, us*a, vs*a);
		y += ys;
		if ((i&3) == 1) {
			u -= us;
			v -= vs;
		} else {
			u += us;
			v += vs;
		}
		dst += dststride;
	}
	pack_nn(dst, y, u, v, w, 0, 0);
	y += ys; dst += dststride; u += us; v += vs;
	pack_nn(dst, y, u, v, w, 0, 0);
}


static int put_image(struct vf_instance_s* vf, mp_image_t *mpi)
{
	mp_image_t *dmpi;

	// hope we'll get DR buffer:
	dmpi=vf_get_image(vf->next, IMGFMT_YUY2,
			  MP_IMGTYPE_TEMP, MP_IMGFLAG_ACCEPT_STRIDE,
			  mpi->w, mpi->h);

	ilpack(dmpi->planes[0], mpi->planes, dmpi->stride[0], mpi->stride, mpi->w, mpi->h, vf->priv->pack);

	return vf_next_put_image(vf,dmpi);
}

static int config(struct vf_instance_s* vf,
		  int width, int height, int d_width, int d_height,
		  unsigned int flags, unsigned int outfmt)
{
	/* FIXME - also support UYVY output? */
	return vf_next_config(vf, width, height, d_width, d_height, flags, IMGFMT_YUY2);
}


static int query_format(struct vf_instance_s* vf, unsigned int fmt)
{
	/* FIXME - really any YUV 4:2:0 input format should work */
	switch (fmt) {
	case IMGFMT_YV12:
	case IMGFMT_IYUV:
	case IMGFMT_I420:
		return vf_next_query_format(vf,IMGFMT_YUY2);
	}
	return 0;
}

static int open(vf_instance_t *vf, char* args)
{
	vf->config=config;
	vf->query_format=query_format;
	vf->put_image=put_image;
	vf->priv = calloc(1, sizeof(struct vf_priv_s));
	vf->priv->mode = 1;
	if (args) sscanf(args, "%d", &vf->priv->mode);
	
	pack_nn = (pack_func_t *)pack_nn_C;
	pack_li_0 = pack_li_0_C;
	pack_li_1 = pack_li_1_C;
#ifdef HAVE_MMX
	if(gCpuCaps.hasMMX) {
		pack_nn = (pack_func_t *)pack_nn_MMX;
		pack_li_0 = pack_li_0_MMX;
		pack_li_1 = pack_li_1_MMX;
	}
#endif

	switch(vf->priv->mode) {
	case 0:
		vf->priv->pack[0] = vf->priv->pack[1] = pack_nn;
		break;
	default:
		mp_msg(MSGT_VFILTER, MSGL_WARN,
			"ilpack: unknown mode %d (fallback to linear)\n",
			vf->priv->mode);
	case 1:
		vf->priv->pack[0] = pack_li_0;
		vf->priv->pack[1] = pack_li_1;
		break;
	}
	
	return 1;
}

vf_info_t vf_info_ilpack = {
	"4:2:0 planar -> 4:2:2 packed reinterlacer",
	"ilpack",
	"Richard Felker",
	"",
	open,
	NULL
};