interpolate8x8.h

TMS320C6713Xvid视频压缩算法源代码.rar

	dst = cur + y * stride + x;

	switch((y_frac << 2) | (x_frac)) {

	case 0:
		transfer8x8_copy(dst, src, stride);
		break;

	case 1:
		interpolate8x8_lowpass_h(halfpel_h, src, stride, rounding);
		interpolate8x8_avg2(dst, src, halfpel_h, stride, rounding, 8);
  		break;

	case 2:
	    interpolate8x8_lowpass_h(dst, src, stride, rounding);
  		break;

	case 3:
		interpolate8x8_lowpass_h(halfpel_h, src, stride, rounding);
		interpolate8x8_avg2(dst, src + 1, halfpel_h, stride, rounding, 8);
  		break;

	case 4:
		interpolate8x8_lowpass_v(halfpel_v, src, stride, rounding);
		interpolate8x8_avg2(dst, src, halfpel_v, stride, rounding, 8);
 		break;

	case 5:
		interpolate8x8_lowpass_h(halfpel_h, src, stride, rounding);
		interpolate8x8_avg2(halfpel_v, src, halfpel_h, stride, rounding, 9);
		interpolate8x8_lowpass_v(halfpel_hv, halfpel_v, stride, rounding);
		interpolate8x8_avg2(dst, halfpel_v, halfpel_hv, stride, rounding, 8);
 		break;

	case 6:
		interpolate8x8_lowpass_hv(halfpel_hv, halfpel_h, src, stride, rounding);
		interpolate8x8_avg2(dst, halfpel_h, halfpel_hv, stride, rounding, 8);
 		break;

	case 7:
		interpolate8x8_lowpass_h(halfpel_h, src, stride, rounding);
		interpolate8x8_avg2(halfpel_v, src + 1, halfpel_h, stride, rounding, 9);
		interpolate8x8_lowpass_v(halfpel_hv, halfpel_v, stride, rounding);
		interpolate8x8_avg2(dst, halfpel_v, halfpel_hv, stride, rounding, 8);
 		break;

	case 8:
	    interpolate8x8_lowpass_v(dst, src, stride, rounding);
		break;

	case 9:
		interpolate8x8_lowpass_h(halfpel_h, src, stride, rounding);
		interpolate8x8_avg2(halfpel_v, src, halfpel_h, stride, rounding, 9);
		interpolate8x8_lowpass_v(dst, halfpel_v, stride, rounding);
  		break;

	case 10:
		interpolate8x8_lowpass_hv(dst, halfpel_h, src, stride, rounding);
		break;

	case 11:
		interpolate8x8_lowpass_h(halfpel_h, src, stride, rounding);
		interpolate8x8_avg2(halfpel_v, src + 1, halfpel_h, stride, rounding, 9);
		interpolate8x8_lowpass_v(dst, halfpel_v, stride, rounding);
   		break;

	case 12:
		interpolate8x8_lowpass_v(halfpel_v, src, stride, rounding);
		interpolate8x8_avg2(dst, src+stride, halfpel_v, stride, rounding, 8);
   		break;

	case 13:
		interpolate8x8_lowpass_h(halfpel_h, src, stride, rounding);
		interpolate8x8_avg2(halfpel_v, src, halfpel_h, stride, rounding, 9);
		interpolate8x8_lowpass_v(halfpel_hv, halfpel_v, stride, rounding);
		interpolate8x8_avg2(dst, halfpel_v+stride, halfpel_hv, stride, rounding, 8);
 		break;

	case 14:
		interpolate8x8_lowpass_hv(halfpel_hv, halfpel_h, src, stride, rounding);
		interpolate8x8_avg2(dst, halfpel_h+stride, halfpel_hv, stride, rounding, 8);
 		break;

	case 15:
		interpolate8x8_lowpass_h(halfpel_h, src, stride, rounding);
		interpolate8x8_avg2(halfpel_v, src + 1, halfpel_h, stride, rounding, 9);
		interpolate8x8_lowpass_v(halfpel_hv, halfpel_v, stride, rounding);
		interpolate8x8_avg2(dst, halfpel_hv, halfpel_v + stride, stride, rounding, 8);
		break;
	}
}

static __inline void interpolate16x16_quarterpel(uint8_t * const cur,
				     uint8_t * const refn,
					 uint8_t * const refh,
					 uint8_t * const refv,
					 uint8_t * const refhv,
				     const uint32_t x, const uint32_t y,
					 const int32_t dx,  const int dy,
					 const uint32_t stride,
					 const uint32_t rounding)
{
	const int32_t xRef = x*4 + dx;
	const int32_t yRef = y*4 + dy;

	uint8_t *src, *dst;
	uint8_t *halfpel_h, *halfpel_v, *halfpel_hv;
	int32_t x_int, y_int, x_frac, y_frac;

	x_int = xRef/4;
	if (xRef < 0 && xRef % 4)
		x_int--;

	x_frac = xRef - (4*x_int);

	y_int  = yRef/4;
	if (yRef < 0 && yRef % 4)
		y_int--;

	y_frac = yRef - (4*y_int);

	src = refn + y_int * stride + x_int;
	halfpel_h = refh;
	halfpel_v = refv;
	halfpel_hv = refhv;

	dst = cur + y * stride + x;

	switch((y_frac << 2) | (x_frac)) {

	case 0:
		transfer16x16_copy(dst, src, stride);
		break;

	case 1:
		interpolate16x16_lowpass_h(halfpel_h, src, stride, rounding);
		interpolate8x8_avg2(dst, src, halfpel_h, stride, rounding, 8);
		interpolate8x8_avg2(dst+8, src+8, halfpel_h+8, stride, rounding, 8);
		interpolate8x8_avg2(dst+8*stride, src+8*stride, halfpel_h+8*stride, stride, rounding, 8);
		interpolate8x8_avg2(dst+8*stride+8, src+8*stride+8, halfpel_h+8*stride+8, stride, rounding, 8);
		break;

	case 2:
	    interpolate16x16_lowpass_h(dst, src, stride, rounding);
  		break;

	case 3:
		interpolate16x16_lowpass_h(halfpel_h, src, stride, rounding);
		interpolate8x8_avg2(dst, src + 1, halfpel_h, stride, rounding, 8);
		interpolate8x8_avg2(dst+8, src + 8 + 1, halfpel_h+8, stride, rounding, 8);
		interpolate8x8_avg2(dst+8*stride, src + 8*stride + 1, halfpel_h+8*stride, stride, rounding, 8);
		interpolate8x8_avg2(dst+8*stride+8, src+8*stride+8 + 1, halfpel_h+8*stride+8, stride, rounding, 8);
		break;

	case 4:
		interpolate16x16_lowpass_v(halfpel_v, src, stride, rounding);
		interpolate8x8_avg2(dst, src, halfpel_v, stride, rounding, 8);
		interpolate8x8_avg2(dst+8, src+8, halfpel_v+8, stride, rounding, 8);
		interpolate8x8_avg2(dst+8*stride, src+8*stride, halfpel_v+8*stride, stride, rounding, 8);
		interpolate8x8_avg2(dst+8*stride+8, src+8*stride+8, halfpel_v+8*stride+8, stride, rounding, 8);
		break;

	case 5:
		interpolate16x16_lowpass_h(halfpel_h, src, stride, rounding);
		interpolate8x8_avg2(halfpel_v, src, halfpel_h, stride, rounding, 9);
		interpolate8x8_avg2(halfpel_v+8, src + 8, halfpel_h+8, stride, rounding, 9);
		interpolate8x8_avg2(halfpel_v+8*stride, src + 8*stride, halfpel_h+8*stride, stride, rounding, 9);
		interpolate8x8_avg2(halfpel_v+8*stride+8, src+8*stride+8, halfpel_h+8*stride+8, stride, rounding, 9);

		interpolate16x16_lowpass_v(halfpel_hv, halfpel_v, stride, rounding);
		interpolate8x8_avg2(dst, halfpel_hv, halfpel_v, stride, rounding, 8);
		interpolate8x8_avg2(dst+8, halfpel_hv+8, halfpel_v+8, stride, rounding, 8);
		interpolate8x8_avg2(dst+8*stride, halfpel_hv+8*stride, halfpel_v+8*stride, stride, rounding, 8);
		interpolate8x8_avg2(dst+8*stride+8, halfpel_hv+8*stride+8, halfpel_v+8*stride+8, stride, rounding, 8);
		break;

	case 6:
		interpolate16x16_lowpass_hv(halfpel_hv, halfpel_h, src, stride, rounding);
		interpolate8x8_avg2(dst, halfpel_h, halfpel_hv, stride, rounding, 8);
		interpolate8x8_avg2(dst+8, halfpel_h+8, halfpel_hv+8, stride, rounding, 8);
		interpolate8x8_avg2(dst+8*stride, halfpel_h+8*stride, halfpel_hv+8*stride, stride, rounding, 8);
		interpolate8x8_avg2(dst+8*stride+8, halfpel_h+8*stride+8, halfpel_hv+8*stride+8, stride, rounding, 8);
		break;

	case 7:
		interpolate16x16_lowpass_h(halfpel_h, src, stride, rounding);
		interpolate8x8_avg2(halfpel_v, src+1, halfpel_h, stride, rounding, 9);
		interpolate8x8_avg2(halfpel_v+8, src+1 + 8, halfpel_h+8, stride, rounding, 9);
		interpolate8x8_avg2(halfpel_v+8*stride, src+1 + 8*stride, halfpel_h+8*stride, stride, rounding, 9);
		interpolate8x8_avg2(halfpel_v+8*stride+8, src+1+8*stride+8, halfpel_h+8*stride+8, stride, rounding, 9);

		interpolate16x16_lowpass_v(halfpel_hv, halfpel_v, stride, rounding);
		interpolate8x8_avg2(dst, halfpel_hv, halfpel_v, stride, rounding, 8);
		interpolate8x8_avg2(dst+8, halfpel_hv+8, halfpel_v+8, stride, rounding, 8);
		interpolate8x8_avg2(dst+8*stride, halfpel_hv+8*stride, halfpel_v+8*stride, stride, rounding, 8);
		interpolate8x8_avg2(dst+8*stride+8, halfpel_hv+8*stride+8, halfpel_v+8*stride+8, stride, rounding, 8);
		break;

	case 8:
	    interpolate16x16_lowpass_v(dst, src, stride, rounding);
		break;

	case 9:
		interpolate16x16_lowpass_h(halfpel_h, src, stride, rounding);
		interpolate8x8_avg2(halfpel_v, src, halfpel_h, stride, rounding, 9);
		interpolate8x8_avg2(halfpel_v+8, src + 8, halfpel_h+8, stride, rounding, 9);
		interpolate8x8_avg2(halfpel_v+8*stride, src + 8*stride, halfpel_h+8*stride, stride, rounding, 9);
		interpolate8x8_avg2(halfpel_v+8*stride+8, src+8*stride+8, halfpel_h+8*stride+8, stride, rounding, 9);
		interpolate16x16_lowpass_v(dst, halfpel_v, stride, rounding);
		break;

	case 10:
		interpolate16x16_lowpass_hv(dst, halfpel_h, src, stride, rounding);
		break;

	case 11:
		interpolate16x16_lowpass_h(halfpel_h, src, stride, rounding);
		interpolate8x8_avg2(halfpel_v, src+1, halfpel_h, stride, rounding, 9);
		interpolate8x8_avg2(halfpel_v+8, src+1 + 8, halfpel_h+8, stride, rounding, 9);
		interpolate8x8_avg2(halfpel_v+8*stride, src+1 + 8*stride, halfpel_h+8*stride, stride, rounding, 9);
		interpolate8x8_avg2(halfpel_v+8*stride+8, src+1+8*stride+8, halfpel_h+8*stride+8, stride, rounding, 9);
		interpolate16x16_lowpass_v(dst, halfpel_v, stride, rounding);
		break;

	case 12:
		interpolate16x16_lowpass_v(halfpel_v, src, stride, rounding);
		interpolate8x8_avg2(dst, src+stride, halfpel_v, stride, rounding, 8);
		interpolate8x8_avg2(dst+8, src+stride+8, halfpel_v+8, stride, rounding, 8);
		interpolate8x8_avg2(dst+8*stride, src+stride+8*stride, halfpel_v+8*stride, stride, rounding, 8);
		interpolate8x8_avg2(dst+8*stride+8, src+stride+8*stride+8, halfpel_v+8*stride+8, stride, rounding, 8);
		break;

	case 13:
		interpolate16x16_lowpass_h(halfpel_h, src, stride, rounding);
		interpolate8x8_avg2(halfpel_v, src, halfpel_h, stride, rounding, 9);
		interpolate8x8_avg2(halfpel_v+8, src + 8, halfpel_h+8, stride, rounding, 9);
		interpolate8x8_avg2(halfpel_v+8*stride, src + 8*stride, halfpel_h+8*stride, stride, rounding, 9);
		interpolate8x8_avg2(halfpel_v+8*stride+8, src+8*stride+8, halfpel_h+8*stride+8, stride, rounding, 9);

		interpolate16x16_lowpass_v(halfpel_hv, halfpel_v, stride, rounding);
		interpolate8x8_avg2(dst, halfpel_hv, halfpel_v+stride, stride, rounding, 8);
		interpolate8x8_avg2(dst+8, halfpel_hv+8, halfpel_v+stride+8, stride, rounding, 8);
		interpolate8x8_avg2(dst+8*stride, halfpel_hv+8*stride, halfpel_v+stride+8*stride, stride, rounding, 8);
		interpolate8x8_avg2(dst+8*stride+8, halfpel_hv+8*stride+8, halfpel_v+stride+8*stride+8, stride, rounding, 8);
		break;

	case 14:
		interpolate16x16_lowpass_hv(halfpel_hv, halfpel_h, src, stride, rounding);
		interpolate8x8_avg2(dst, halfpel_h+stride, halfpel_hv, stride, rounding, 8);
		interpolate8x8_avg2(dst+8, halfpel_h+stride+8, halfpel_hv+8, stride, rounding, 8);
		interpolate8x8_avg2(dst+8*stride, halfpel_h+stride+8*stride, halfpel_hv+8*stride, stride, rounding, 8);
		interpolate8x8_avg2(dst+8*stride+8, halfpel_h+stride+8*stride+8, halfpel_hv+8*stride+8, stride, rounding, 8);
		break;

	case 15:
		interpolate16x16_lowpass_h(halfpel_h, src, stride, rounding);
		interpolate8x8_avg2(halfpel_v, src+1, halfpel_h, stride, rounding, 9);
		interpolate8x8_avg2(halfpel_v+8, src+1 + 8, halfpel_h+8, stride, rounding, 9);
		interpolate8x8_avg2(halfpel_v+8*stride, src+1 + 8*stride, halfpel_h+8*stride, stride, rounding, 9);
		interpolate8x8_avg2(halfpel_v+8*stride+8, src+1+8*stride+8, halfpel_h+8*stride+8, stride, rounding, 9);

		interpolate16x16_lowpass_v(halfpel_hv, halfpel_v, stride, rounding);
		interpolate8x8_avg2(dst, halfpel_hv, halfpel_v+stride, stride, rounding, 8);
		interpolate8x8_avg2(dst+8, halfpel_hv+8, halfpel_v+stride+8, stride, rounding, 8);
		interpolate8x8_avg2(dst+8*stride, halfpel_hv+8*stride, halfpel_v+stride+8*stride, stride, rounding, 8);
		interpolate8x8_avg2(dst+8*stride+8, halfpel_hv+8*stride+8, halfpel_v+stride+8*stride+8, stride, rounding, 8);
		break;
	}
}

#endif