nengel@2: /*
nengel@2:  * H.26L/H.264/AVC/JVT/14496-10/... encoder/decoder
nengel@2:  * Copyright (c) 2003-2010 Michael Niedermayer <michaelni@gmx.at>
nengel@2:  *
nengel@2:  * This file is part of FFmpeg.
nengel@2:  *
nengel@2:  * FFmpeg is free software; you can redistribute it and/or
nengel@2:  * modify it under the terms of the GNU Lesser General Public
nengel@2:  * License as published by the Free Software Foundation; either
nengel@2:  * version 2.1 of the License, or (at your option) any later version.
nengel@2:  *
nengel@2:  * FFmpeg is distributed in the hope that it will be useful,
nengel@2:  * but WITHOUT ANY WARRANTY; without even the implied warranty of
nengel@2:  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
nengel@2:  * Lesser General Public License for more details.
nengel@2:  *
nengel@2:  * You should have received a copy of the GNU Lesser General Public
nengel@2:  * License along with FFmpeg; if not, write to the Free Software
nengel@2:  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
nengel@2:  */
nengel@2: 
nengel@2: /**
nengel@2:  * @file
nengel@2:  * H.264 / AVC / MPEG4 part10 DSP functions.
nengel@2:  * @author Michael Niedermayer <michaelni@gmx.at>
nengel@2:  */
nengel@2: 
nengel@2: #include <stdint.h>
nengel@2: #include "avcodec.h"
nengel@2: #include "h264_dsp.h"
nengel@2: 
nengel@2: #define op_scale1(x)  block[x] = av_clip_uint8( (block[x]*weight + offset) >> log2_denom )
nengel@2: #define op_scale2(x)  dst[x] = av_clip_uint8( (src[x]*weights + dst[x]*weightd + offset) >> (log2_denom+1))
nengel@2: #define H264_WEIGHT(W,H) \
nengel@2: static void weight_h264_pixels ## W ## x ## H ## _c(uint8_t *block, int stride, int log2_denom, int weight, int offset){ \
nengel@2:     int y; \
nengel@2:     offset <<= log2_denom; \
nengel@2:     if(log2_denom) offset += 1<<(log2_denom-1); \
nengel@2:     for(y=0; y<H; y++, block += stride){ \
nengel@2:         op_scale1(0); \
nengel@2:         op_scale1(1); \
nengel@2:         if(W==2) continue; \
nengel@2:         op_scale1(2); \
nengel@2:         op_scale1(3); \
nengel@2:         if(W==4) continue; \
nengel@2:         op_scale1(4); \
nengel@2:         op_scale1(5); \
nengel@2:         op_scale1(6); \
nengel@2:         op_scale1(7); \
nengel@2:         if(W==8) continue; \
nengel@2:         op_scale1(8); \
nengel@2:         op_scale1(9); \
nengel@2:         op_scale1(10); \
nengel@2:         op_scale1(11); \
nengel@2:         op_scale1(12); \
nengel@2:         op_scale1(13); \
nengel@2:         op_scale1(14); \
nengel@2:         op_scale1(15); \
nengel@2:     } \
nengel@2: } \
nengel@2: static void biweight_h264_pixels ## W ## x ## H ## _c(uint8_t *dst, uint8_t *src, int stride, int log2_denom, int weightd, int weights, int offset){ \
nengel@2:     int y; \
nengel@2:     offset = ((offset + 1) | 1) << log2_denom; \
nengel@2:     for(y=0; y<H; y++, dst += stride, src += stride){ \
nengel@2:         op_scale2(0); \
nengel@2:         op_scale2(1); \
nengel@2:         if(W==2) continue; \
nengel@2:         op_scale2(2); \
nengel@2:         op_scale2(3); \
nengel@2:         if(W==4) continue; \
nengel@2:         op_scale2(4); \
nengel@2:         op_scale2(5); \
nengel@2:         op_scale2(6); \
nengel@2:         op_scale2(7); \
nengel@2:         if(W==8) continue; \
nengel@2:         op_scale2(8); \
nengel@2:         op_scale2(9); \
nengel@2:         op_scale2(10); \
nengel@2:         op_scale2(11); \
nengel@2:         op_scale2(12); \
nengel@2:         op_scale2(13); \
nengel@2:         op_scale2(14); \
nengel@2:         op_scale2(15); \
nengel@2:     } \
nengel@2: }
nengel@2: 
nengel@2: H264_WEIGHT(16,16)
nengel@2: H264_WEIGHT(16,8)
nengel@2: H264_WEIGHT(8,16)
nengel@2: H264_WEIGHT(8,8)
nengel@2: H264_WEIGHT(8,4)
nengel@2: H264_WEIGHT(4,8)
nengel@2: H264_WEIGHT(4,4)
nengel@2: H264_WEIGHT(4,2)
nengel@2: H264_WEIGHT(2,4)
nengel@2: H264_WEIGHT(2,2)
nengel@2: 
nengel@2: #undef op_scale1
nengel@2: #undef op_scale2
nengel@2: #undef H264_WEIGHT
nengel@2: 
nengel@2: static av_always_inline void h264_loop_filter_luma_c(uint8_t *pix, int xstride, int ystride, int alpha, int beta, int8_t *tc0)
nengel@2: {
nengel@2:     int i, d;
nengel@2:     for( i = 0; i < 4; i++ ) {
nengel@2:         if( tc0[i] < 0 ) {
nengel@2:             pix += 4*ystride;
nengel@2:             continue;
nengel@2:         }
nengel@2:         for( d = 0; d < 4; d++ ) {
nengel@2:             const int p0 = pix[-1*xstride];
nengel@2:             const int p1 = pix[-2*xstride];
nengel@2:             const int p2 = pix[-3*xstride];
nengel@2:             const int q0 = pix[0];
nengel@2:             const int q1 = pix[1*xstride];
nengel@2:             const int q2 = pix[2*xstride];
nengel@2: 
nengel@2:             if( FFABS( p0 - q0 ) < alpha &&
nengel@2:                 FFABS( p1 - p0 ) < beta &&
nengel@2:                 FFABS( q1 - q0 ) < beta ) {
nengel@2: 
nengel@2:                 int tc = tc0[i];
nengel@2:                 int i_delta;
nengel@2: 
nengel@2:                 if( FFABS( p2 - p0 ) < beta ) {
nengel@2:                     if(tc0[i])
nengel@2:                     pix[-2*xstride] = p1 + av_clip( (( p2 + ( ( p0 + q0 + 1 ) >> 1 ) ) >> 1) - p1, -tc0[i], tc0[i] );
nengel@2:                     tc++;
nengel@2:                 }
nengel@2:                 if( FFABS( q2 - q0 ) < beta ) {
nengel@2:                     if(tc0[i])
nengel@2:                     pix[   xstride] = q1 + av_clip( (( q2 + ( ( p0 + q0 + 1 ) >> 1 ) ) >> 1) - q1, -tc0[i], tc0[i] );
nengel@2:                     tc++;
nengel@2:                 }
nengel@2: 
nengel@2:                 i_delta = av_clip( (((q0 - p0 ) << 2) + (p1 - q1) + 4) >> 3, -tc, tc );
nengel@2:                 pix[-xstride] = av_clip_uint8( p0 + i_delta );    /* p0' */
nengel@2:                 pix[0]        = av_clip_uint8( q0 - i_delta );    /* q0' */
nengel@2:             }
nengel@2:             pix += ystride;
nengel@2:         }
nengel@2:     }
nengel@2: }
nengel@2: static void h264_v_loop_filter_luma_c(uint8_t *pix, int stride, int alpha, int beta, int8_t *tc0)
nengel@2: {
nengel@2:     h264_loop_filter_luma_c(pix, stride, 1, alpha, beta, tc0);
nengel@2: }
nengel@2: static void h264_h_loop_filter_luma_c(uint8_t *pix, int stride, int alpha, int beta, int8_t *tc0)
nengel@2: {
nengel@2:     h264_loop_filter_luma_c(pix, 1, stride, alpha, beta, tc0);
nengel@2: }
nengel@2: 
nengel@2: static av_always_inline void h264_loop_filter_luma_intra_c(uint8_t *pix, int xstride, int ystride, int alpha, int beta)
nengel@2: {
nengel@2:     int d;
nengel@2:     for( d = 0; d < 16; d++ ) {
nengel@2:         const int p2 = pix[-3*xstride];
nengel@2:         const int p1 = pix[-2*xstride];
nengel@2:         const int p0 = pix[-1*xstride];
nengel@2: 
nengel@2:         const int q0 = pix[ 0*xstride];
nengel@2:         const int q1 = pix[ 1*xstride];
nengel@2:         const int q2 = pix[ 2*xstride];
nengel@2: 
nengel@2:         if( FFABS( p0 - q0 ) < alpha &&
nengel@2:             FFABS( p1 - p0 ) < beta &&
nengel@2:             FFABS( q1 - q0 ) < beta ) {
nengel@2: 
nengel@2:             if(FFABS( p0 - q0 ) < (( alpha >> 2 ) + 2 )){
nengel@2:                 if( FFABS( p2 - p0 ) < beta)
nengel@2:                 {
nengel@2:                     const int p3 = pix[-4*xstride];
nengel@2:                     /* p0', p1', p2' */
nengel@2:                     pix[-1*xstride] = ( p2 + 2*p1 + 2*p0 + 2*q0 + q1 + 4 ) >> 3;
nengel@2:                     pix[-2*xstride] = ( p2 + p1 + p0 + q0 + 2 ) >> 2;
nengel@2:                     pix[-3*xstride] = ( 2*p3 + 3*p2 + p1 + p0 + q0 + 4 ) >> 3;
nengel@2:                 } else {
nengel@2:                     /* p0' */
nengel@2:                     pix[-1*xstride] = ( 2*p1 + p0 + q1 + 2 ) >> 2;
nengel@2:                 }
nengel@2:                 if( FFABS( q2 - q0 ) < beta)
nengel@2:                 {
nengel@2:                     const int q3 = pix[3*xstride];
nengel@2:                     /* q0', q1', q2' */
nengel@2:                     pix[0*xstride] = ( p1 + 2*p0 + 2*q0 + 2*q1 + q2 + 4 ) >> 3;
nengel@2:                     pix[1*xstride] = ( p0 + q0 + q1 + q2 + 2 ) >> 2;
nengel@2:                     pix[2*xstride] = ( 2*q3 + 3*q2 + q1 + q0 + p0 + 4 ) >> 3;
nengel@2:                 } else {
nengel@2:                     /* q0' */
nengel@2:                     pix[0*xstride] = ( 2*q1 + q0 + p1 + 2 ) >> 2;
nengel@2:                 }
nengel@2:             }else{
nengel@2:                 /* p0', q0' */
nengel@2:                 pix[-1*xstride] = ( 2*p1 + p0 + q1 + 2 ) >> 2;
nengel@2:                 pix[ 0*xstride] = ( 2*q1 + q0 + p1 + 2 ) >> 2;
nengel@2:             }
nengel@2:         }
nengel@2:         pix += ystride;
nengel@2:     }
nengel@2: }
nengel@2: static void h264_v_loop_filter_luma_intra_c(uint8_t *pix, int stride, int alpha, int beta)
nengel@2: {
nengel@2:     h264_loop_filter_luma_intra_c(pix, stride, 1, alpha, beta);
nengel@2: }
nengel@2: static void h264_h_loop_filter_luma_intra_c(uint8_t *pix, int stride, int alpha, int beta)
nengel@2: {
nengel@2:     h264_loop_filter_luma_intra_c(pix, 1, stride, alpha, beta);
nengel@2: }
nengel@2: 
nengel@2: static av_always_inline void h264_loop_filter_chroma_c(uint8_t *pix, int xstride, int ystride, int alpha, int beta, int8_t *tc0)
nengel@2: {
nengel@2:     int i, d;
nengel@2:     for( i = 0; i < 4; i++ ) {
nengel@2:         const int tc = tc0[i];
nengel@2:         if( tc <= 0 ) {
nengel@2:             pix += 2*ystride;
nengel@2:             continue;
nengel@2:         }
nengel@2:         for( d = 0; d < 2; d++ ) {
nengel@2:             const int p0 = pix[-1*xstride];
nengel@2:             const int p1 = pix[-2*xstride];
nengel@2:             const int q0 = pix[0];
nengel@2:             const int q1 = pix[1*xstride];
nengel@2: 
nengel@2:             if( FFABS( p0 - q0 ) < alpha &&
nengel@2:                 FFABS( p1 - p0 ) < beta &&
nengel@2:                 FFABS( q1 - q0 ) < beta ) {
nengel@2: 
nengel@2:                 int delta = av_clip( (((q0 - p0 ) << 2) + (p1 - q1) + 4) >> 3, -tc, tc );
nengel@2: 
nengel@2:                 pix[-xstride] = av_clip_uint8( p0 + delta );    /* p0' */
nengel@2:                 pix[0]        = av_clip_uint8( q0 - delta );    /* q0' */
nengel@2:             }
nengel@2:             pix += ystride;
nengel@2:         }
nengel@2:     }
nengel@2: }
nengel@2: static void h264_v_loop_filter_chroma_c(uint8_t *pix, int stride, int alpha, int beta, int8_t *tc0)
nengel@2: {
nengel@2:     h264_loop_filter_chroma_c(pix, stride, 1, alpha, beta, tc0);
nengel@2: }
nengel@2: static void h264_h_loop_filter_chroma_c(uint8_t *pix, int stride, int alpha, int beta, int8_t *tc0)
nengel@2: {
nengel@2:     h264_loop_filter_chroma_c(pix, 1, stride, alpha, beta, tc0);
nengel@2: }
nengel@2: 
nengel@2: static av_always_inline void h264_loop_filter_chroma_intra_c(uint8_t *pix, int xstride, int ystride, int alpha, int beta)
nengel@2: {
nengel@2:     int d;
nengel@2:     for( d = 0; d < 8; d++ ) {
nengel@2:         const int p0 = pix[-1*xstride];
nengel@2:         const int p1 = pix[-2*xstride];
nengel@2:         const int q0 = pix[0];
nengel@2:         const int q1 = pix[1*xstride];
nengel@2: 
nengel@2:         if( FFABS( p0 - q0 ) < alpha &&
nengel@2:             FFABS( p1 - p0 ) < beta &&
nengel@2:             FFABS( q1 - q0 ) < beta ) {
nengel@2: 
nengel@2:             pix[-xstride] = ( 2*p1 + p0 + q1 + 2 ) >> 2;   /* p0' */
nengel@2:             pix[0]        = ( 2*q1 + q0 + p1 + 2 ) >> 2;   /* q0' */
nengel@2:         }
nengel@2:         pix += ystride;
nengel@2:     }
nengel@2: }
nengel@2: static void h264_v_loop_filter_chroma_intra_c(uint8_t *pix, int stride, int alpha, int beta)
nengel@2: {
nengel@2:     h264_loop_filter_chroma_intra_c(pix, stride, 1, alpha, beta);
nengel@2: }
nengel@2: static void h264_h_loop_filter_chroma_intra_c(uint8_t *pix, int stride, int alpha, int beta)
nengel@2: {
nengel@2:     h264_loop_filter_chroma_intra_c(pix, 1, stride, alpha, beta);
nengel@2: }
nengel@2: 
nengel@2: void ff_h264dsp_init(H264DSPContext *c)
nengel@2: {
nengel@2:     c->h264_idct_add= ff_h264_idct_add_c;
nengel@2:     c->h264_idct8_add= ff_h264_idct8_add_c;
nengel@2:     c->h264_idct_dc_add= ff_h264_idct_dc_add_c;
nengel@2:     c->h264_idct8_dc_add= ff_h264_idct8_dc_add_c;
nengel@2:     c->h264_idct_add16     = ff_h264_idct_add16_c;
nengel@2:     c->h264_idct8_add4     = ff_h264_idct8_add4_c;
nengel@2:     c->h264_idct_add8      = ff_h264_idct_add8_c;
nengel@2:     c->h264_idct_add16intra= ff_h264_idct_add16intra_c;
nengel@2: 
nengel@2:     c->weight_h264_pixels_tab[0]= weight_h264_pixels16x16_c;
nengel@2:     c->weight_h264_pixels_tab[1]= weight_h264_pixels16x8_c;
nengel@2:     c->weight_h264_pixels_tab[2]= weight_h264_pixels8x16_c;
nengel@2:     c->weight_h264_pixels_tab[3]= weight_h264_pixels8x8_c;
nengel@2:     c->weight_h264_pixels_tab[4]= weight_h264_pixels8x4_c;
nengel@2:     c->weight_h264_pixels_tab[5]= weight_h264_pixels4x8_c;
nengel@2:     c->weight_h264_pixels_tab[6]= weight_h264_pixels4x4_c;
nengel@2:     c->weight_h264_pixels_tab[7]= weight_h264_pixels4x2_c;
nengel@2:     c->weight_h264_pixels_tab[8]= weight_h264_pixels2x4_c;
nengel@2:     c->weight_h264_pixels_tab[9]= weight_h264_pixels2x2_c;
nengel@2:     c->biweight_h264_pixels_tab[0]= biweight_h264_pixels16x16_c;
nengel@2:     c->biweight_h264_pixels_tab[1]= biweight_h264_pixels16x8_c;
nengel@2:     c->biweight_h264_pixels_tab[2]= biweight_h264_pixels8x16_c;
nengel@2:     c->biweight_h264_pixels_tab[3]= biweight_h264_pixels8x8_c;
nengel@2:     c->biweight_h264_pixels_tab[4]= biweight_h264_pixels8x4_c;
nengel@2:     c->biweight_h264_pixels_tab[5]= biweight_h264_pixels4x8_c;
nengel@2:     c->biweight_h264_pixels_tab[6]= biweight_h264_pixels4x4_c;
nengel@2:     c->biweight_h264_pixels_tab[7]= biweight_h264_pixels4x2_c;
nengel@2:     c->biweight_h264_pixels_tab[8]= biweight_h264_pixels2x4_c;
nengel@2:     c->biweight_h264_pixels_tab[9]= biweight_h264_pixels2x2_c;
nengel@2: 
nengel@2:     c->h264_v_loop_filter_luma= h264_v_loop_filter_luma_c;
nengel@2:     c->h264_h_loop_filter_luma= h264_h_loop_filter_luma_c;
nengel@2:     c->h264_v_loop_filter_luma_intra= h264_v_loop_filter_luma_intra_c;
nengel@2:     c->h264_h_loop_filter_luma_intra= h264_h_loop_filter_luma_intra_c;
nengel@2:     c->h264_v_loop_filter_chroma= h264_v_loop_filter_chroma_c;
nengel@2:     c->h264_h_loop_filter_chroma= h264_h_loop_filter_chroma_c;
nengel@2:     c->h264_v_loop_filter_chroma_intra= h264_v_loop_filter_chroma_intra_c;
nengel@2:     c->h264_h_loop_filter_chroma_intra= h264_h_loop_filter_chroma_intra_c;
nengel@2:     c->h264_loop_filter_strength= NULL;
nengel@2: 
nengel@2:     if (ARCH_ARM) ff_h264dsp_init_arm(c);
nengel@2:     if (HAVE_ALTIVEC) ff_h264dsp_init_ppc(c);
nengel@2:     if (HAVE_MMX) ff_h264dsp_init_x86(c);
nengel@2: }