--- interpolate8x8.c 2003/01/04 06:14:33 1.4.2.6 +++ interpolate8x8.c 2005/09/13 12:12:15 1.15 @@ -1,44 +1,46 @@ -/************************************************************************** +/***************************************************************************** * * XVID MPEG-4 VIDEO CODEC - * 8x8 block-based halfpel interpolation + * - 8x8 block-based halfpel interpolation - * - * This program is free software; you can redistribute it and/or modify - * it under the terms of the GNU General Public License as published by - * the Free Software Foundation; either version 2 of the License, or - * (at your option) any later version. + * Copyright(C) 2001-2003 Peter Ross * - * This program is distributed in the hope that it will be useful, - * but WITHOUT ANY WARRANTY; without even the implied warranty of - * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the - * GNU General Public License for more details. + * This program is free software ; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation ; either version 2 of the License, or + * (at your option) any later version. * - * You should have received a copy of the GNU General Public License - * along with this program; if not, write to the Free Software - * Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA. + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY ; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + * GNU General Public License for more details. * - *************************************************************************/ - -/************************************************************************** - * - * History: + * You should have received a copy of the GNU General Public License + * along with this program ; if not, write to the Free Software + * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA * - * 05.10.2002 new bilinear and qpel interpolation code - Isibaar - * 27.12.2001 modified "compensate_halfpel" - * 05.11.2001 initial version; (c)2001 peter ross + * $Id: interpolate8x8.c,v 1.15 2005/09/13 12:12:15 suxen_drol Exp $ * - *************************************************************************/ - + ****************************************************************************/ #include "../portab.h" #include "../global.h" #include "interpolate8x8.h" -// function pointers +/* function pointers */ INTERPOLATE8X8_PTR interpolate8x8_halfpel_h; INTERPOLATE8X8_PTR interpolate8x8_halfpel_v; INTERPOLATE8X8_PTR interpolate8x8_halfpel_hv; +INTERPOLATE8X8_PTR interpolate8x4_halfpel_h; +INTERPOLATE8X8_PTR interpolate8x4_halfpel_v; +INTERPOLATE8X8_PTR interpolate8x4_halfpel_hv; + +INTERPOLATE8X8_PTR interpolate8x8_halfpel_add; +INTERPOLATE8X8_PTR interpolate8x8_halfpel_h_add; +INTERPOLATE8X8_PTR interpolate8x8_halfpel_v_add; +INTERPOLATE8X8_PTR interpolate8x8_halfpel_hv_add; + INTERPOLATE8X8_AVG2_PTR interpolate8x8_avg2; INTERPOLATE8X8_AVG4_PTR interpolate8x8_avg4; @@ -54,13 +56,13 @@ INTERPOLATE8X8_6TAP_LOWPASS_PTR interpolate8x8_6tap_lowpass_h; INTERPOLATE8X8_6TAP_LOWPASS_PTR interpolate8x8_6tap_lowpass_v; -void interpolate8x8_avg2_c(uint8_t *dst, const uint8_t *src1, const uint8_t *src2, const uint32_t stride, const uint32_t rounding, const uint32_t height) +void +interpolate8x8_avg2_c(uint8_t * dst, const uint8_t * src1, const uint8_t *src2, const uint32_t stride, const uint32_t rounding, const uint32_t height) { uint32_t i; const int32_t round = 1 - rounding; - for(i = 0; i < height; i++) - { + for(i = 0; i < height; i++) { dst[0] = (src1[0] + src2[0] + round) >> 1; dst[1] = (src1[1] + src2[1] + round) >> 1; dst[2] = (src1[2] + src2[2] + round) >> 1; @@ -76,13 +78,18 @@ } } +void +interpolate8x8_halfpel_add_c(uint8_t * const dst, const uint8_t * const src, const uint32_t stride, const uint32_t rounding) +{ + interpolate8x8_avg2_c(dst, dst, src, stride, 0, 8); +} + void interpolate8x8_avg4_c(uint8_t *dst, const uint8_t *src1, const uint8_t *src2, const uint8_t *src3, const uint8_t *src4, const uint32_t stride, const uint32_t rounding) { int32_t i; const int32_t round = 2 - rounding; - for(i = 0; i < 8; i++) - { + for(i = 0; i < 8; i++) { dst[0] = (src1[0] + src2[0] + src3[0] + src4[0] + round) >> 2; dst[1] = (src1[1] + src2[1] + src3[1] + src4[1] + round) >> 2; dst[2] = (src1[2] + src2[2] + src3[2] + src4[2] + round) >> 2; @@ -91,7 +98,7 @@ dst[5] = (src1[5] + src2[5] + src3[5] + src4[5] + round) >> 2; dst[6] = (src1[6] + src2[6] + src3[6] + src4[6] + round) >> 2; dst[7] = (src1[7] + src2[7] + src3[7] + src4[7] + round) >> 2; - + dst += stride; src1 += stride; src2 += stride; @@ -100,7 +107,7 @@ } } -// dst = interpolate(src) +/* dst = interpolate(src) */ void interpolate8x8_halfpel_h_c(uint8_t * const dst, @@ -108,22 +115,104 @@ const uint32_t stride, const uint32_t rounding) { - uint32_t i, j; + uintptr_t j; + + if (rounding) { + for (j = 0; j < 8*stride; j+=stride) { + dst[j + 0] = (uint8_t)((src[j + 0] + src[j + 1] )>>1); + dst[j + 1] = (uint8_t)((src[j + 1] + src[j + 2] )>>1); + dst[j + 2] = (uint8_t)((src[j + 2] + src[j + 3] )>>1); + dst[j + 3] = (uint8_t)((src[j + 3] + src[j + 4] )>>1); + dst[j + 4] = (uint8_t)((src[j + 4] + src[j + 5] )>>1); + dst[j + 5] = (uint8_t)((src[j + 5] + src[j + 6] )>>1); + dst[j + 6] = (uint8_t)((src[j + 6] + src[j + 7] )>>1); + dst[j + 7] = (uint8_t)((src[j + 7] + src[j + 8] )>>1); + } + } else { + for (j = 0; j < 8*stride; j+=stride) { + dst[j + 0] = (uint8_t)((src[j + 0] + src[j + 1] + 1)>>1); + dst[j + 1] = (uint8_t)((src[j + 1] + src[j + 2] + 1)>>1); + dst[j + 2] = (uint8_t)((src[j + 2] + src[j + 3] + 1)>>1); + dst[j + 3] = (uint8_t)((src[j + 3] + src[j + 4] + 1)>>1); + dst[j + 4] = (uint8_t)((src[j + 4] + src[j + 5] + 1)>>1); + dst[j + 5] = (uint8_t)((src[j + 5] + src[j + 6] + 1)>>1); + dst[j + 6] = (uint8_t)((src[j + 6] + src[j + 7] + 1)>>1); + dst[j + 7] = (uint8_t)((src[j + 7] + src[j + 8] + 1)>>1); + } + } +} - for (j = 0; j < 8; j++) { - for (i = 0; i < 8; i++) { +/* dst = interpolate(src) */ - int16_t tot = - (int32_t) src[j * stride + i] + (int32_t) src[j * stride + i + - 1]; +void +interpolate8x4_halfpel_h_c(uint8_t * const dst, + const uint8_t * const src, + const uint32_t stride, + const uint32_t rounding) +{ + uintptr_t j; - tot = (int32_t) ((tot + 1 - rounding) >> 1); - dst[j * stride + i] = (uint8_t) tot; + if (rounding) { + for (j = 0; j < 4*stride; j+=stride) { + dst[j + 0] = (uint8_t)((src[j + 0] + src[j + 1] )>>1); + dst[j + 1] = (uint8_t)((src[j + 1] + src[j + 2] )>>1); + dst[j + 2] = (uint8_t)((src[j + 2] + src[j + 3] )>>1); + dst[j + 3] = (uint8_t)((src[j + 3] + src[j + 4] )>>1); + dst[j + 4] = (uint8_t)((src[j + 4] + src[j + 5] )>>1); + dst[j + 5] = (uint8_t)((src[j + 5] + src[j + 6] )>>1); + dst[j + 6] = (uint8_t)((src[j + 6] + src[j + 7] )>>1); + dst[j + 7] = (uint8_t)((src[j + 7] + src[j + 8] )>>1); + } + } else { + for (j = 0; j < 4*stride; j+=stride) { + dst[j + 0] = (uint8_t)((src[j + 0] + src[j + 1] + 1)>>1); + dst[j + 1] = (uint8_t)((src[j + 1] + src[j + 2] + 1)>>1); + dst[j + 2] = (uint8_t)((src[j + 2] + src[j + 3] + 1)>>1); + dst[j + 3] = (uint8_t)((src[j + 3] + src[j + 4] + 1)>>1); + dst[j + 4] = (uint8_t)((src[j + 4] + src[j + 5] + 1)>>1); + dst[j + 5] = (uint8_t)((src[j + 5] + src[j + 6] + 1)>>1); + dst[j + 6] = (uint8_t)((src[j + 6] + src[j + 7] + 1)>>1); + dst[j + 7] = (uint8_t)((src[j + 7] + src[j + 8] + 1)>>1); } } } +/* dst = (dst + interpolate(src)/2 */ + +void +interpolate8x8_halfpel_h_add_c(uint8_t * const dst, + const uint8_t * const src, + const uint32_t stride, + const uint32_t rounding) +{ + uintptr_t j; + + if (rounding) { + for (j = 0; j < 8*stride; j+=stride) { + dst[j + 0] = (uint8_t)((((src[j + 0] + src[j + 1] )>>1) + dst[j+0] + 1)>>1); + dst[j + 1] = (uint8_t)((((src[j + 1] + src[j + 2] )>>1) + dst[j+1] + 1)>>1); + dst[j + 2] = (uint8_t)((((src[j + 2] + src[j + 3] )>>1) + dst[j+2] + 1)>>1); + dst[j + 3] = (uint8_t)((((src[j + 3] + src[j + 4] )>>1) + dst[j+3] + 1)>>1); + dst[j + 4] = (uint8_t)((((src[j + 4] + src[j + 5] )>>1) + dst[j+4] + 1)>>1); + dst[j + 5] = (uint8_t)((((src[j + 5] + src[j + 6] )>>1) + dst[j+5] + 1)>>1); + dst[j + 6] = (uint8_t)((((src[j + 6] + src[j + 7] )>>1) + dst[j+6] + 1)>>1); + dst[j + 7] = (uint8_t)((((src[j + 7] + src[j + 8] )>>1) + dst[j+7] + 1)>>1); + } + } else { + for (j = 0; j < 8*stride; j+=stride) { + dst[j + 0] = (uint8_t)((((src[j + 0] + src[j + 1] + 1)>>1) + dst[j+0] + 1)>>1); + dst[j + 1] = (uint8_t)((((src[j + 1] + src[j + 2] + 1)>>1) + dst[j+1] + 1)>>1); + dst[j + 2] = (uint8_t)((((src[j + 2] + src[j + 3] + 1)>>1) + dst[j+2] + 1)>>1); + dst[j + 3] = (uint8_t)((((src[j + 3] + src[j + 4] + 1)>>1) + dst[j+3] + 1)>>1); + dst[j + 4] = (uint8_t)((((src[j + 4] + src[j + 5] + 1)>>1) + dst[j+4] + 1)>>1); + dst[j + 5] = (uint8_t)((((src[j + 5] + src[j + 6] + 1)>>1) + dst[j+5] + 1)>>1); + dst[j + 6] = (uint8_t)((((src[j + 6] + src[j + 7] + 1)>>1) + dst[j+6] + 1)>>1); + dst[j + 7] = (uint8_t)((((src[j + 7] + src[j + 8] + 1)>>1) + dst[j+7] + 1)>>1); + } + } +} +/* dst = interpolate(src) */ void interpolate8x8_halfpel_v_c(uint8_t * const dst, @@ -131,18 +220,107 @@ const uint32_t stride, const uint32_t rounding) { - uint32_t i, j; + uintptr_t j; - for (j = 0; j < 8; j++) { - for (i = 0; i < 8; i++) { - int16_t tot = src[j * stride + i] + src[j * stride + i + stride]; - tot = ((tot + 1 - rounding) >> 1); - dst[j * stride + i] = (uint8_t) tot; + if (rounding) { + for (j = 0; j < 8*stride; j+=stride) { + dst[j + 0] = (uint8_t)((src[j + 0] + src[j + stride + 0] )>>1); + dst[j + 1] = (uint8_t)((src[j + 1] + src[j + stride + 1] )>>1); + dst[j + 2] = (uint8_t)((src[j + 2] + src[j + stride + 2] )>>1); + dst[j + 3] = (uint8_t)((src[j + 3] + src[j + stride + 3] )>>1); + dst[j + 4] = (uint8_t)((src[j + 4] + src[j + stride + 4] )>>1); + dst[j + 5] = (uint8_t)((src[j + 5] + src[j + stride + 5] )>>1); + dst[j + 6] = (uint8_t)((src[j + 6] + src[j + stride + 6] )>>1); + dst[j + 7] = (uint8_t)((src[j + 7] + src[j + stride + 7] )>>1); + } + } else { + for (j = 0; j < 8*stride; j+=stride) { + dst[j + 0] = (uint8_t)((src[j + 0] + src[j + stride + 0] + 1)>>1); + dst[j + 1] = (uint8_t)((src[j + 1] + src[j + stride + 1] + 1)>>1); + dst[j + 2] = (uint8_t)((src[j + 2] + src[j + stride + 2] + 1)>>1); + dst[j + 3] = (uint8_t)((src[j + 3] + src[j + stride + 3] + 1)>>1); + dst[j + 4] = (uint8_t)((src[j + 4] + src[j + stride + 4] + 1)>>1); + dst[j + 5] = (uint8_t)((src[j + 5] + src[j + stride + 5] + 1)>>1); + dst[j + 6] = (uint8_t)((src[j + 6] + src[j + stride + 6] + 1)>>1); + dst[j + 7] = (uint8_t)((src[j + 7] + src[j + stride + 7] + 1)>>1); } } } +/* dst = interpolate(src) */ + +void +interpolate8x4_halfpel_v_c(uint8_t * const dst, + const uint8_t * const src, + const uint32_t stride, + const uint32_t rounding) +{ + uintptr_t j; + + + if (rounding) { + for (j = 0; j < 4*stride; j+=stride) { + dst[j + 0] = (uint8_t)((src[j + 0] + src[j + stride + 0] )>>1); + dst[j + 1] = (uint8_t)((src[j + 1] + src[j + stride + 1] )>>1); + dst[j + 2] = (uint8_t)((src[j + 2] + src[j + stride + 2] )>>1); + dst[j + 3] = (uint8_t)((src[j + 3] + src[j + stride + 3] )>>1); + dst[j + 4] = (uint8_t)((src[j + 4] + src[j + stride + 4] )>>1); + dst[j + 5] = (uint8_t)((src[j + 5] + src[j + stride + 5] )>>1); + dst[j + 6] = (uint8_t)((src[j + 6] + src[j + stride + 6] )>>1); + dst[j + 7] = (uint8_t)((src[j + 7] + src[j + stride + 7] )>>1); + } + } else { + for (j = 0; j < 4*stride; j+=stride) { + dst[j + 0] = (uint8_t)((src[j + 0] + src[j + stride + 0] + 1)>>1); + dst[j + 1] = (uint8_t)((src[j + 1] + src[j + stride + 1] + 1)>>1); + dst[j + 2] = (uint8_t)((src[j + 2] + src[j + stride + 2] + 1)>>1); + dst[j + 3] = (uint8_t)((src[j + 3] + src[j + stride + 3] + 1)>>1); + dst[j + 4] = (uint8_t)((src[j + 4] + src[j + stride + 4] + 1)>>1); + dst[j + 5] = (uint8_t)((src[j + 5] + src[j + stride + 5] + 1)>>1); + dst[j + 6] = (uint8_t)((src[j + 6] + src[j + stride + 6] + 1)>>1); + dst[j + 7] = (uint8_t)((src[j + 7] + src[j + stride + 7] + 1)>>1); + } + } +} + +/* dst = (dst + interpolate(src))/2 */ + +void +interpolate8x8_halfpel_v_add_c(uint8_t * const dst, + const uint8_t * const src, + const uint32_t stride, + const uint32_t rounding) +{ + uintptr_t j; + + + if (rounding) { + for (j = 0; j < 8*stride; j+=stride) { + dst[j + 0] = (uint8_t)((((src[j + 0] + src[j + stride + 0] )>>1) + dst[j+0] + 1)>>1); + dst[j + 1] = (uint8_t)((((src[j + 1] + src[j + stride + 1] )>>1) + dst[j+1] + 1)>>1); + dst[j + 2] = (uint8_t)((((src[j + 2] + src[j + stride + 2] )>>1) + dst[j+2] + 1)>>1); + dst[j + 3] = (uint8_t)((((src[j + 3] + src[j + stride + 3] )>>1) + dst[j+3] + 1)>>1); + dst[j + 4] = (uint8_t)((((src[j + 4] + src[j + stride + 4] )>>1) + dst[j+4] + 1)>>1); + dst[j + 5] = (uint8_t)((((src[j + 5] + src[j + stride + 5] )>>1) + dst[j+5] + 1)>>1); + dst[j + 6] = (uint8_t)((((src[j + 6] + src[j + stride + 6] )>>1) + dst[j+6] + 1)>>1); + dst[j + 7] = (uint8_t)((((src[j + 7] + src[j + stride + 7] )>>1) + dst[j+7] + 1)>>1); + } + } else { + for (j = 0; j < 8*stride; j+=stride) { + dst[j + 0] = (uint8_t)((((src[j + 0] + src[j + stride + 0] + 1)>>1) + dst[j+0] + 1)>>1); + dst[j + 1] = (uint8_t)((((src[j + 1] + src[j + stride + 1] + 1)>>1) + dst[j+1] + 1)>>1); + dst[j + 2] = (uint8_t)((((src[j + 2] + src[j + stride + 2] + 1)>>1) + dst[j+2] + 1)>>1); + dst[j + 3] = (uint8_t)((((src[j + 3] + src[j + stride + 3] + 1)>>1) + dst[j+3] + 1)>>1); + dst[j + 4] = (uint8_t)((((src[j + 4] + src[j + stride + 4] + 1)>>1) + dst[j+4] + 1)>>1); + dst[j + 5] = (uint8_t)((((src[j + 5] + src[j + stride + 5] + 1)>>1) + dst[j+5] + 1)>>1); + dst[j + 6] = (uint8_t)((((src[j + 6] + src[j + stride + 6] + 1)>>1) + dst[j+6] + 1)>>1); + dst[j + 7] = (uint8_t)((((src[j + 7] + src[j + stride + 7] + 1)>>1) + dst[j+7] + 1)>>1); + } + } +} + +/* dst = interpolate(src) */ void interpolate8x8_halfpel_hv_c(uint8_t * const dst, @@ -150,16 +328,99 @@ const uint32_t stride, const uint32_t rounding) { - uint32_t i, j; + uintptr_t j; + + if (rounding) { + for (j = 0; j < 8*stride; j+=stride) { + dst[j + 0] = (uint8_t)((src[j+0] + src[j+1] + src[j+stride+0] + src[j+stride+1] +1)>>2); + dst[j + 1] = (uint8_t)((src[j+1] + src[j+2] + src[j+stride+1] + src[j+stride+2] +1)>>2); + dst[j + 2] = (uint8_t)((src[j+2] + src[j+3] + src[j+stride+2] + src[j+stride+3] +1)>>2); + dst[j + 3] = (uint8_t)((src[j+3] + src[j+4] + src[j+stride+3] + src[j+stride+4] +1)>>2); + dst[j + 4] = (uint8_t)((src[j+4] + src[j+5] + src[j+stride+4] + src[j+stride+5] +1)>>2); + dst[j + 5] = (uint8_t)((src[j+5] + src[j+6] + src[j+stride+5] + src[j+stride+6] +1)>>2); + dst[j + 6] = (uint8_t)((src[j+6] + src[j+7] + src[j+stride+6] + src[j+stride+7] +1)>>2); + dst[j + 7] = (uint8_t)((src[j+7] + src[j+8] + src[j+stride+7] + src[j+stride+8] +1)>>2); + } + } else { + for (j = 0; j < 8*stride; j+=stride) { + dst[j + 0] = (uint8_t)((src[j+0] + src[j+1] + src[j+stride+0] + src[j+stride+1] +2)>>2); + dst[j + 1] = (uint8_t)((src[j+1] + src[j+2] + src[j+stride+1] + src[j+stride+2] +2)>>2); + dst[j + 2] = (uint8_t)((src[j+2] + src[j+3] + src[j+stride+2] + src[j+stride+3] +2)>>2); + dst[j + 3] = (uint8_t)((src[j+3] + src[j+4] + src[j+stride+3] + src[j+stride+4] +2)>>2); + dst[j + 4] = (uint8_t)((src[j+4] + src[j+5] + src[j+stride+4] + src[j+stride+5] +2)>>2); + dst[j + 5] = (uint8_t)((src[j+5] + src[j+6] + src[j+stride+5] + src[j+stride+6] +2)>>2); + dst[j + 6] = (uint8_t)((src[j+6] + src[j+7] + src[j+stride+6] + src[j+stride+7] +2)>>2); + dst[j + 7] = (uint8_t)((src[j+7] + src[j+8] + src[j+stride+7] + src[j+stride+8] +2)>>2); + } + } +} + +/* dst = interpolate(src) */ + +void +interpolate8x4_halfpel_hv_c(uint8_t * const dst, + const uint8_t * const src, + const uint32_t stride, + const uint32_t rounding) +{ + uintptr_t j; + + if (rounding) { + for (j = 0; j < 4*stride; j+=stride) { + dst[j + 0] = (uint8_t)((src[j+0] + src[j+1] + src[j+stride+0] + src[j+stride+1] +1)>>2); + dst[j + 1] = (uint8_t)((src[j+1] + src[j+2] + src[j+stride+1] + src[j+stride+2] +1)>>2); + dst[j + 2] = (uint8_t)((src[j+2] + src[j+3] + src[j+stride+2] + src[j+stride+3] +1)>>2); + dst[j + 3] = (uint8_t)((src[j+3] + src[j+4] + src[j+stride+3] + src[j+stride+4] +1)>>2); + dst[j + 4] = (uint8_t)((src[j+4] + src[j+5] + src[j+stride+4] + src[j+stride+5] +1)>>2); + dst[j + 5] = (uint8_t)((src[j+5] + src[j+6] + src[j+stride+5] + src[j+stride+6] +1)>>2); + dst[j + 6] = (uint8_t)((src[j+6] + src[j+7] + src[j+stride+6] + src[j+stride+7] +1)>>2); + dst[j + 7] = (uint8_t)((src[j+7] + src[j+8] + src[j+stride+7] + src[j+stride+8] +1)>>2); + } + } else { + for (j = 0; j < 4*stride; j+=stride) { + dst[j + 0] = (uint8_t)((src[j+0] + src[j+1] + src[j+stride+0] + src[j+stride+1] +2)>>2); + dst[j + 1] = (uint8_t)((src[j+1] + src[j+2] + src[j+stride+1] + src[j+stride+2] +2)>>2); + dst[j + 2] = (uint8_t)((src[j+2] + src[j+3] + src[j+stride+2] + src[j+stride+3] +2)>>2); + dst[j + 3] = (uint8_t)((src[j+3] + src[j+4] + src[j+stride+3] + src[j+stride+4] +2)>>2); + dst[j + 4] = (uint8_t)((src[j+4] + src[j+5] + src[j+stride+4] + src[j+stride+5] +2)>>2); + dst[j + 5] = (uint8_t)((src[j+5] + src[j+6] + src[j+stride+5] + src[j+stride+6] +2)>>2); + dst[j + 6] = (uint8_t)((src[j+6] + src[j+7] + src[j+stride+6] + src[j+stride+7] +2)>>2); + dst[j + 7] = (uint8_t)((src[j+7] + src[j+8] + src[j+stride+7] + src[j+stride+8] +2)>>2); + } + } +} + +/* dst = (interpolate(src) + dst)/2 */ + +void +interpolate8x8_halfpel_hv_add_c(uint8_t * const dst, + const uint8_t * const src, + const uint32_t stride, + const uint32_t rounding) +{ + uintptr_t j; - for (j = 0; j < 8; j++) { - for (i = 0; i < 8; i++) { - int16_t tot = - src[j * stride + i] + src[j * stride + i + 1] + - src[j * stride + i + stride] + src[j * stride + i + stride + - 1]; - tot = ((tot + 2 - rounding) >> 2); - dst[j * stride + i] = (uint8_t) tot; + if (rounding) { + for (j = 0; j < 8*stride; j+=stride) { + dst[j + 0] = (uint8_t)((((src[j+0] + src[j+1] + src[j+stride+0] + src[j+stride+1] +1)>>2) + dst[j+0])>>1); + dst[j + 1] = (uint8_t)((((src[j+1] + src[j+2] + src[j+stride+1] + src[j+stride+2] +1)>>2) + dst[j+1])>>1); + dst[j + 2] = (uint8_t)((((src[j+2] + src[j+3] + src[j+stride+2] + src[j+stride+3] +1)>>2) + dst[j+2])>>1); + dst[j + 3] = (uint8_t)((((src[j+3] + src[j+4] + src[j+stride+3] + src[j+stride+4] +1)>>2) + dst[j+3])>>1); + dst[j + 4] = (uint8_t)((((src[j+4] + src[j+5] + src[j+stride+4] + src[j+stride+5] +1)>>2) + dst[j+4])>>1); + dst[j + 5] = (uint8_t)((((src[j+5] + src[j+6] + src[j+stride+5] + src[j+stride+6] +1)>>2) + dst[j+5])>>1); + dst[j + 6] = (uint8_t)((((src[j+6] + src[j+7] + src[j+stride+6] + src[j+stride+7] +1)>>2) + dst[j+6])>>1); + dst[j + 7] = (uint8_t)((((src[j+7] + src[j+8] + src[j+stride+7] + src[j+stride+8] +1)>>2) + dst[j+7])>>1); + } + } else { + for (j = 0; j < 8*stride; j+=stride) { + dst[j + 0] = (uint8_t)((((src[j+0] + src[j+1] + src[j+stride+0] + src[j+stride+1] +2)>>2) + dst[j+0] + 1)>>1); + dst[j + 1] = (uint8_t)((((src[j+1] + src[j+2] + src[j+stride+1] + src[j+stride+2] +2)>>2) + dst[j+1] + 1)>>1); + dst[j + 2] = (uint8_t)((((src[j+2] + src[j+3] + src[j+stride+2] + src[j+stride+3] +2)>>2) + dst[j+2] + 1)>>1); + dst[j + 3] = (uint8_t)((((src[j+3] + src[j+4] + src[j+stride+3] + src[j+stride+4] +2)>>2) + dst[j+3] + 1)>>1); + dst[j + 4] = (uint8_t)((((src[j+4] + src[j+5] + src[j+stride+4] + src[j+stride+5] +2)>>2) + dst[j+4] + 1)>>1); + dst[j + 5] = (uint8_t)((((src[j+5] + src[j+6] + src[j+stride+5] + src[j+stride+6] +2)>>2) + dst[j+5] + 1)>>1); + dst[j + 6] = (uint8_t)((((src[j+6] + src[j+7] + src[j+stride+6] + src[j+stride+7] +2)>>2) + dst[j+6] + 1)>>1); + dst[j + 7] = (uint8_t)((((src[j+7] + src[j+8] + src[j+stride+7] + src[j+stride+8] +2)>>2) + dst[j+7] + 1)>>1); } } } @@ -304,7 +565,7 @@ int32_t src15 = src[15 * stride]; int32_t src16 = src[16 * stride]; - + dst[0] = CLIP(((7 * ((src0<<1) - src2) + 23 * src1 + 3 * src3 - src4 + round_add) >> 5), 0, 255); dst[stride] = CLIP(((19 * src1 + 20 * src2 - src5 + 3 * (src4 - src0 - (src3<<1)) + round_add) >> 5), 0, 255); dst[2*stride] = CLIP(((20 * (src2 + src3) + (src0<<1) + 3 * (src5 - ((src1 + src4)<<1)) - src6 + round_add) >> 5), 0, 255); @@ -345,7 +606,7 @@ int32_t src6 = src[6 * stride]; int32_t src7 = src[7 * stride]; int32_t src8 = src[8 * stride]; - + dst[0] = CLIP(((7 * ((src0<<1) - src2) + 23 * src1 + 3 * src3 - src4 + round_add) >> 5), 0, 255); dst[stride] = CLIP(((19 * src1 + 20 * src2 - src5 + 3 * (src4 - src0 - (src3 << 1)) + round_add) >> 5), 0, 255); dst[2 * stride] = CLIP(((20 * (src2 + src3) + (src0<<1) + 3 * (src5 - ((src1 + src4) <<1 )) - src6 + round_add) >> 5), 0, 255);