comparison libmpcodecs/vf_spp.c @ 11305:d1585f7d5366

10l

11304:f89e212dd5b9

		*(uint32_t*)&dst[4 + y*stride]+= *(uint32_t*)&block[4 + y*8];
		*(uint32_t*)&dst[6 + y*stride]+= *(uint32_t*)&block[6 + y*8];
	}
}

static void store_slice_c(uint8_t *dst, int16_t *src, int dst_stride, int src_stride, int width, int log2_scale){
	int y, x;

#define STORE(pos) \
	temp= ((src[x + y*src_stride + pos]<<log2_scale) + d[pos])>>6;\
	if(temp & 0x100) temp= ~(temp>>31);\
	dst[x + y*dst_stride + pos]= temp;

	for(y=0; y<8; y++){
		uint8_t *d= dither[y];
		for(x=0; x<width; x+=8){
			int temp;
			STORE(0);
			STORE(1);

		}
	}
}

#ifdef HAVE_MMX
static void store_slice_mmx(uint8_t *dst, int16_t *src, int dst_stride, int src_stride, int width, int log2_scale){
	int y;

	for(y=0; y<8; y++){
		uint8_t *dst1= dst;
		int16_t *src1= src;
		asm volatile(
			"movq (%3), %%mm3	\n\t"
			"movq (%3), %%mm4	\n\t"

//	if(width != mmxw)
//		store_slice_c(dst + mmxw, src + mmxw, dst_stride, src_stride, width - mmxw, log2_scale);
}
#endif

static void (*store_slice)(uint8_t *dst, int16_t *src, int dst_stride, int src_stride, int width, int log2_scale)= store_slice_c;

static void (*requantize)(DCTELEM dst[64], DCTELEM src[64], int qp, uint8_t *permutation)= requantize_c;

static void filter(struct vf_priv_s *p, uint8_t *dst, uint8_t *src, int dst_stride, int src_stride, int width, int height, uint8_t *qp_store, int qp_stride, int is_luma){
	int x, y, i;

				p->dsp.idct(block2);
				add_block(p->temp + index, stride, block2);
			}
		}
		if(y)
			store_slice(dst + (y-8)*dst_stride, p->temp + 8 + y*stride, dst_stride, stride, width, 6-p->log2_count);
	}
#if 0
	for(y=0; y<height; y++){
		for(x=0; x<width; x++){
			if((((x>>6) ^ (y>>6)) & 1) == 0)

}

static int config(struct vf_instance_s* vf,
        int width, int height, int d_width, int d_height,
	unsigned int flags, unsigned int outfmt){

	vf->priv->temp_stride= (width+16+15)&(~15);
        vf->priv->temp= malloc(vf->priv->temp_stride*(height+16)*sizeof(int16_t));
        vf->priv->src = malloc(vf->priv->temp_stride*(height+16)*sizeof(uint8_t));
        
	return vf_next_config(vf,width,height,d_width,d_height,flags,outfmt);
}

static void get_image(struct vf_instance_s* vf, mp_image_t *mpi){

11305:d1585f7d5366

		*(uint32_t*)&dst[4 + y*stride]+= *(uint32_t*)&block[4 + y*8];
		*(uint32_t*)&dst[6 + y*stride]+= *(uint32_t*)&block[6 + y*8];
	}
}

static void store_slice_c(uint8_t *dst, int16_t *src, int dst_stride, int src_stride, int width, int height, int log2_scale){
	int y, x;

#define STORE(pos) \
	temp= ((src[x + y*src_stride + pos]<<log2_scale) + d[pos])>>6;\
	if(temp & 0x100) temp= ~(temp>>31);\
	dst[x + y*dst_stride + pos]= temp;

	for(y=0; y<height; y++){
		uint8_t *d= dither[y];
		for(x=0; x<width; x+=8){
			int temp;
			STORE(0);
			STORE(1);

		}
	}
}

#ifdef HAVE_MMX
static void store_slice_mmx(uint8_t *dst, int16_t *src, int dst_stride, int src_stride, int width, int height, int log2_scale){
	int y;

	for(y=0; y<height; y++){
		uint8_t *dst1= dst;
		int16_t *src1= src;
		asm volatile(
			"movq (%3), %%mm3	\n\t"
			"movq (%3), %%mm4	\n\t"

//	if(width != mmxw)
//		store_slice_c(dst + mmxw, src + mmxw, dst_stride, src_stride, width - mmxw, log2_scale);
}
#endif

static void (*store_slice)(uint8_t *dst, int16_t *src, int dst_stride, int src_stride, int width, int height, int log2_scale)= store_slice_c;

static void (*requantize)(DCTELEM dst[64], DCTELEM src[64], int qp, uint8_t *permutation)= requantize_c;

static void filter(struct vf_priv_s *p, uint8_t *dst, uint8_t *src, int dst_stride, int src_stride, int width, int height, uint8_t *qp_store, int qp_stride, int is_luma){
	int x, y, i;

				p->dsp.idct(block2);
				add_block(p->temp + index, stride, block2);
			}
		}
		if(y)
			store_slice(dst + (y-8)*dst_stride, p->temp + 8 + y*stride, dst_stride, stride, width, XMIN(8, height+8-y), 6-p->log2_count);
	}
#if 0
	for(y=0; y<height; y++){
		for(x=0; x<width; x++){
			if((((x>>6) ^ (y>>6)) & 1) == 0)

}

static int config(struct vf_instance_s* vf,
        int width, int height, int d_width, int d_height,
	unsigned int flags, unsigned int outfmt){
	int h= (height+16+15)&(~15);

	vf->priv->temp_stride= (width+16+15)&(~15);
        vf->priv->temp= malloc(vf->priv->temp_stride*h*sizeof(int16_t));
        vf->priv->src = malloc(vf->priv->temp_stride*h*sizeof(uint8_t));
        
	return vf_next_config(vf,width,height,d_width,d_height,flags,outfmt);
}

static void get_image(struct vf_instance_s* vf, mp_image_t *mpi){