yading@10: /*
yading@10:  * H.26L/H.264/AVC/JVT/14496-10/... encoder/decoder
yading@10:  * Copyright (c) 2003-2011 Michael Niedermayer <michaelni@gmx.at>
yading@10:  *
yading@10:  * This file is part of FFmpeg.
yading@10:  *
yading@10:  * FFmpeg is free software; you can redistribute it and/or
yading@10:  * modify it under the terms of the GNU Lesser General Public
yading@10:  * License as published by the Free Software Foundation; either
yading@10:  * version 2.1 of the License, or (at your option) any later version.
yading@10:  *
yading@10:  * FFmpeg is distributed in the hope that it will be useful,
yading@10:  * but WITHOUT ANY WARRANTY; without even the implied warranty of
yading@10:  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
yading@10:  * Lesser General Public License for more details.
yading@10:  *
yading@10:  * You should have received a copy of the GNU Lesser General Public
yading@10:  * License along with FFmpeg; if not, write to the Free Software
yading@10:  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
yading@10:  */
yading@10: 
yading@10: /**
yading@10:  * @file
yading@10:  * H.264 / AVC / MPEG4 part10 DSP functions.
yading@10:  * @author Michael Niedermayer <michaelni@gmx.at>
yading@10:  */
yading@10: 
yading@10: #include "bit_depth_template.c"
yading@10: 
yading@10: #define op_scale1(x)  block[x] = av_clip_pixel( (block[x]*weight + offset) >> log2_denom )
yading@10: #define op_scale2(x)  dst[x] = av_clip_pixel( (src[x]*weights + dst[x]*weightd + offset) >> (log2_denom+1))
yading@10: #define H264_WEIGHT(W) \
yading@10: static void FUNCC(weight_h264_pixels ## W)(uint8_t *_block, int stride, int height, \
yading@10:                                            int log2_denom, int weight, int offset) \
yading@10: { \
yading@10:     int y; \
yading@10:     pixel *block = (pixel*)_block; \
yading@10:     stride >>= sizeof(pixel)-1; \
yading@10:     offset <<= (log2_denom + (BIT_DEPTH-8)); \
yading@10:     if(log2_denom) offset += 1<<(log2_denom-1); \
yading@10:     for (y = 0; y < height; y++, block += stride) { \
yading@10:         op_scale1(0); \
yading@10:         op_scale1(1); \
yading@10:         if(W==2) continue; \
yading@10:         op_scale1(2); \
yading@10:         op_scale1(3); \
yading@10:         if(W==4) continue; \
yading@10:         op_scale1(4); \
yading@10:         op_scale1(5); \
yading@10:         op_scale1(6); \
yading@10:         op_scale1(7); \
yading@10:         if(W==8) continue; \
yading@10:         op_scale1(8); \
yading@10:         op_scale1(9); \
yading@10:         op_scale1(10); \
yading@10:         op_scale1(11); \
yading@10:         op_scale1(12); \
yading@10:         op_scale1(13); \
yading@10:         op_scale1(14); \
yading@10:         op_scale1(15); \
yading@10:     } \
yading@10: } \
yading@10: static void FUNCC(biweight_h264_pixels ## W)(uint8_t *_dst, uint8_t *_src, int stride, int height, \
yading@10:                                              int log2_denom, int weightd, int weights, int offset) \
yading@10: { \
yading@10:     int y; \
yading@10:     pixel *dst = (pixel*)_dst; \
yading@10:     pixel *src = (pixel*)_src; \
yading@10:     stride >>= sizeof(pixel)-1; \
yading@10:     offset <<= (BIT_DEPTH-8); \
yading@10:     offset = ((offset + 1) | 1) << log2_denom; \
yading@10:     for (y = 0; y < height; y++, dst += stride, src += stride) { \
yading@10:         op_scale2(0); \
yading@10:         op_scale2(1); \
yading@10:         if(W==2) continue; \
yading@10:         op_scale2(2); \
yading@10:         op_scale2(3); \
yading@10:         if(W==4) continue; \
yading@10:         op_scale2(4); \
yading@10:         op_scale2(5); \
yading@10:         op_scale2(6); \
yading@10:         op_scale2(7); \
yading@10:         if(W==8) continue; \
yading@10:         op_scale2(8); \
yading@10:         op_scale2(9); \
yading@10:         op_scale2(10); \
yading@10:         op_scale2(11); \
yading@10:         op_scale2(12); \
yading@10:         op_scale2(13); \
yading@10:         op_scale2(14); \
yading@10:         op_scale2(15); \
yading@10:     } \
yading@10: }
yading@10: 
yading@10: H264_WEIGHT(16)
yading@10: H264_WEIGHT(8)
yading@10: H264_WEIGHT(4)
yading@10: H264_WEIGHT(2)
yading@10: 
yading@10: #undef op_scale1
yading@10: #undef op_scale2
yading@10: #undef H264_WEIGHT
yading@10: 
yading@10: static av_always_inline av_flatten void FUNCC(h264_loop_filter_luma)(uint8_t *p_pix, int xstride, int ystride, int inner_iters, int alpha, int beta, int8_t *tc0)
yading@10: {
yading@10:     pixel *pix = (pixel*)p_pix;
yading@10:     int i, d;
yading@10:     xstride >>= sizeof(pixel)-1;
yading@10:     ystride >>= sizeof(pixel)-1;
yading@10:     alpha <<= BIT_DEPTH - 8;
yading@10:     beta  <<= BIT_DEPTH - 8;
yading@10:     for( i = 0; i < 4; i++ ) {
yading@10:         const int tc_orig = tc0[i] << (BIT_DEPTH - 8);
yading@10:         if( tc_orig < 0 ) {
yading@10:             pix += inner_iters*ystride;
yading@10:             continue;
yading@10:         }
yading@10:         for( d = 0; d < inner_iters; d++ ) {
yading@10:             const int p0 = pix[-1*xstride];
yading@10:             const int p1 = pix[-2*xstride];
yading@10:             const int p2 = pix[-3*xstride];
yading@10:             const int q0 = pix[0];
yading@10:             const int q1 = pix[1*xstride];
yading@10:             const int q2 = pix[2*xstride];
yading@10: 
yading@10:             if( FFABS( p0 - q0 ) < alpha &&
yading@10:                 FFABS( p1 - p0 ) < beta &&
yading@10:                 FFABS( q1 - q0 ) < beta ) {
yading@10: 
yading@10:                 int tc = tc_orig;
yading@10:                 int i_delta;
yading@10: 
yading@10:                 if( FFABS( p2 - p0 ) < beta ) {
yading@10:                     if(tc_orig)
yading@10:                     pix[-2*xstride] = p1 + av_clip( (( p2 + ( ( p0 + q0 + 1 ) >> 1 ) ) >> 1) - p1, -tc_orig, tc_orig );
yading@10:                     tc++;
yading@10:                 }
yading@10:                 if( FFABS( q2 - q0 ) < beta ) {
yading@10:                     if(tc_orig)
yading@10:                     pix[   xstride] = q1 + av_clip( (( q2 + ( ( p0 + q0 + 1 ) >> 1 ) ) >> 1) - q1, -tc_orig, tc_orig );
yading@10:                     tc++;
yading@10:                 }
yading@10: 
yading@10:                 i_delta = av_clip( (((q0 - p0 ) << 2) + (p1 - q1) + 4) >> 3, -tc, tc );
yading@10:                 pix[-xstride] = av_clip_pixel( p0 + i_delta );    /* p0' */
yading@10:                 pix[0]        = av_clip_pixel( q0 - i_delta );    /* q0' */
yading@10:             }
yading@10:             pix += ystride;
yading@10:         }
yading@10:     }
yading@10: }
yading@10: static void FUNCC(h264_v_loop_filter_luma)(uint8_t *pix, int stride, int alpha, int beta, int8_t *tc0)
yading@10: {
yading@10:     FUNCC(h264_loop_filter_luma)(pix, stride, sizeof(pixel), 4, alpha, beta, tc0);
yading@10: }
yading@10: static void FUNCC(h264_h_loop_filter_luma)(uint8_t *pix, int stride, int alpha, int beta, int8_t *tc0)
yading@10: {
yading@10:     FUNCC(h264_loop_filter_luma)(pix, sizeof(pixel), stride, 4, alpha, beta, tc0);
yading@10: }
yading@10: static void FUNCC(h264_h_loop_filter_luma_mbaff)(uint8_t *pix, int stride, int alpha, int beta, int8_t *tc0)
yading@10: {
yading@10:     FUNCC(h264_loop_filter_luma)(pix, sizeof(pixel), stride, 2, alpha, beta, tc0);
yading@10: }
yading@10: 
yading@10: static av_always_inline av_flatten void FUNCC(h264_loop_filter_luma_intra)(uint8_t *p_pix, int xstride, int ystride, int inner_iters, int alpha, int beta)
yading@10: {
yading@10:     pixel *pix = (pixel*)p_pix;
yading@10:     int d;
yading@10:     xstride >>= sizeof(pixel)-1;
yading@10:     ystride >>= sizeof(pixel)-1;
yading@10:     alpha <<= BIT_DEPTH - 8;
yading@10:     beta  <<= BIT_DEPTH - 8;
yading@10:     for( d = 0; d < 4 * inner_iters; d++ ) {
yading@10:         const int p2 = pix[-3*xstride];
yading@10:         const int p1 = pix[-2*xstride];
yading@10:         const int p0 = pix[-1*xstride];
yading@10: 
yading@10:         const int q0 = pix[ 0*xstride];
yading@10:         const int q1 = pix[ 1*xstride];
yading@10:         const int q2 = pix[ 2*xstride];
yading@10: 
yading@10:         if( FFABS( p0 - q0 ) < alpha &&
yading@10:             FFABS( p1 - p0 ) < beta &&
yading@10:             FFABS( q1 - q0 ) < beta ) {
yading@10: 
yading@10:             if(FFABS( p0 - q0 ) < (( alpha >> 2 ) + 2 )){
yading@10:                 if( FFABS( p2 - p0 ) < beta)
yading@10:                 {
yading@10:                     const int p3 = pix[-4*xstride];
yading@10:                     /* p0', p1', p2' */
yading@10:                     pix[-1*xstride] = ( p2 + 2*p1 + 2*p0 + 2*q0 + q1 + 4 ) >> 3;
yading@10:                     pix[-2*xstride] = ( p2 + p1 + p0 + q0 + 2 ) >> 2;
yading@10:                     pix[-3*xstride] = ( 2*p3 + 3*p2 + p1 + p0 + q0 + 4 ) >> 3;
yading@10:                 } else {
yading@10:                     /* p0' */
yading@10:                     pix[-1*xstride] = ( 2*p1 + p0 + q1 + 2 ) >> 2;
yading@10:                 }
yading@10:                 if( FFABS( q2 - q0 ) < beta)
yading@10:                 {
yading@10:                     const int q3 = pix[3*xstride];
yading@10:                     /* q0', q1', q2' */
yading@10:                     pix[0*xstride] = ( p1 + 2*p0 + 2*q0 + 2*q1 + q2 + 4 ) >> 3;
yading@10:                     pix[1*xstride] = ( p0 + q0 + q1 + q2 + 2 ) >> 2;
yading@10:                     pix[2*xstride] = ( 2*q3 + 3*q2 + q1 + q0 + p0 + 4 ) >> 3;
yading@10:                 } else {
yading@10:                     /* q0' */
yading@10:                     pix[0*xstride] = ( 2*q1 + q0 + p1 + 2 ) >> 2;
yading@10:                 }
yading@10:             }else{
yading@10:                 /* p0', q0' */
yading@10:                 pix[-1*xstride] = ( 2*p1 + p0 + q1 + 2 ) >> 2;
yading@10:                 pix[ 0*xstride] = ( 2*q1 + q0 + p1 + 2 ) >> 2;
yading@10:             }
yading@10:         }
yading@10:         pix += ystride;
yading@10:     }
yading@10: }
yading@10: static void FUNCC(h264_v_loop_filter_luma_intra)(uint8_t *pix, int stride, int alpha, int beta)
yading@10: {
yading@10:     FUNCC(h264_loop_filter_luma_intra)(pix, stride, sizeof(pixel), 4, alpha, beta);
yading@10: }
yading@10: static void FUNCC(h264_h_loop_filter_luma_intra)(uint8_t *pix, int stride, int alpha, int beta)
yading@10: {
yading@10:     FUNCC(h264_loop_filter_luma_intra)(pix, sizeof(pixel), stride, 4, alpha, beta);
yading@10: }
yading@10: static void FUNCC(h264_h_loop_filter_luma_mbaff_intra)(uint8_t *pix, int stride, int alpha, int beta)
yading@10: {
yading@10:     FUNCC(h264_loop_filter_luma_intra)(pix, sizeof(pixel), stride, 2, alpha, beta);
yading@10: }
yading@10: 
yading@10: static av_always_inline av_flatten void FUNCC(h264_loop_filter_chroma)(uint8_t *p_pix, int xstride, int ystride, int inner_iters, int alpha, int beta, int8_t *tc0)
yading@10: {
yading@10:     pixel *pix = (pixel*)p_pix;
yading@10:     int i, d;
yading@10:     alpha <<= BIT_DEPTH - 8;
yading@10:     beta  <<= BIT_DEPTH - 8;
yading@10:     xstride >>= sizeof(pixel)-1;
yading@10:     ystride >>= sizeof(pixel)-1;
yading@10:     for( i = 0; i < 4; i++ ) {
yading@10:         const int tc = ((tc0[i] - 1) << (BIT_DEPTH - 8)) + 1;
yading@10:         if( tc <= 0 ) {
yading@10:             pix += inner_iters*ystride;
yading@10:             continue;
yading@10:         }
yading@10:         for( d = 0; d < inner_iters; d++ ) {
yading@10:             const int p0 = pix[-1*xstride];
yading@10:             const int p1 = pix[-2*xstride];
yading@10:             const int q0 = pix[0];
yading@10:             const int q1 = pix[1*xstride];
yading@10: 
yading@10:             if( FFABS( p0 - q0 ) < alpha &&
yading@10:                 FFABS( p1 - p0 ) < beta &&
yading@10:                 FFABS( q1 - q0 ) < beta ) {
yading@10: 
yading@10:                 int delta = av_clip( (((q0 - p0 ) << 2) + (p1 - q1) + 4) >> 3, -tc, tc );
yading@10: 
yading@10:                 pix[-xstride] = av_clip_pixel( p0 + delta );    /* p0' */
yading@10:                 pix[0]        = av_clip_pixel( q0 - delta );    /* q0' */
yading@10:             }
yading@10:             pix += ystride;
yading@10:         }
yading@10:     }
yading@10: }
yading@10: static void FUNCC(h264_v_loop_filter_chroma)(uint8_t *pix, int stride, int alpha, int beta, int8_t *tc0)
yading@10: {
yading@10:     FUNCC(h264_loop_filter_chroma)(pix, stride, sizeof(pixel), 2, alpha, beta, tc0);
yading@10: }
yading@10: static void FUNCC(h264_h_loop_filter_chroma)(uint8_t *pix, int stride, int alpha, int beta, int8_t *tc0)
yading@10: {
yading@10:     FUNCC(h264_loop_filter_chroma)(pix, sizeof(pixel), stride, 2, alpha, beta, tc0);
yading@10: }
yading@10: static void FUNCC(h264_h_loop_filter_chroma_mbaff)(uint8_t *pix, int stride, int alpha, int beta, int8_t *tc0)
yading@10: {
yading@10:     FUNCC(h264_loop_filter_chroma)(pix, sizeof(pixel), stride, 1, alpha, beta, tc0);
yading@10: }
yading@10: static void FUNCC(h264_h_loop_filter_chroma422)(uint8_t *pix, int stride, int alpha, int beta, int8_t *tc0)
yading@10: {
yading@10:     FUNCC(h264_loop_filter_chroma)(pix, sizeof(pixel), stride, 4, alpha, beta, tc0);
yading@10: }
yading@10: static void FUNCC(h264_h_loop_filter_chroma422_mbaff)(uint8_t *pix, int stride, int alpha, int beta, int8_t *tc0)
yading@10: {
yading@10:     FUNCC(h264_loop_filter_chroma)(pix, sizeof(pixel), stride, 2, alpha, beta, tc0);
yading@10: }
yading@10: 
yading@10: static av_always_inline av_flatten void FUNCC(h264_loop_filter_chroma_intra)(uint8_t *p_pix, int xstride, int ystride, int inner_iters, int alpha, int beta)
yading@10: {
yading@10:     pixel *pix = (pixel*)p_pix;
yading@10:     int d;
yading@10:     xstride >>= sizeof(pixel)-1;
yading@10:     ystride >>= sizeof(pixel)-1;
yading@10:     alpha <<= BIT_DEPTH - 8;
yading@10:     beta  <<= BIT_DEPTH - 8;
yading@10:     for( d = 0; d < 4 * inner_iters; d++ ) {
yading@10:         const int p0 = pix[-1*xstride];
yading@10:         const int p1 = pix[-2*xstride];
yading@10:         const int q0 = pix[0];
yading@10:         const int q1 = pix[1*xstride];
yading@10: 
yading@10:         if( FFABS( p0 - q0 ) < alpha &&
yading@10:             FFABS( p1 - p0 ) < beta &&
yading@10:             FFABS( q1 - q0 ) < beta ) {
yading@10: 
yading@10:             pix[-xstride] = ( 2*p1 + p0 + q1 + 2 ) >> 2;   /* p0' */
yading@10:             pix[0]        = ( 2*q1 + q0 + p1 + 2 ) >> 2;   /* q0' */
yading@10:         }
yading@10:         pix += ystride;
yading@10:     }
yading@10: }
yading@10: static void FUNCC(h264_v_loop_filter_chroma_intra)(uint8_t *pix, int stride, int alpha, int beta)
yading@10: {
yading@10:     FUNCC(h264_loop_filter_chroma_intra)(pix, stride, sizeof(pixel), 2, alpha, beta);
yading@10: }
yading@10: static void FUNCC(h264_h_loop_filter_chroma_intra)(uint8_t *pix, int stride, int alpha, int beta)
yading@10: {
yading@10:     FUNCC(h264_loop_filter_chroma_intra)(pix, sizeof(pixel), stride, 2, alpha, beta);
yading@10: }
yading@10: static void FUNCC(h264_h_loop_filter_chroma_mbaff_intra)(uint8_t *pix, int stride, int alpha, int beta)
yading@10: {
yading@10:     FUNCC(h264_loop_filter_chroma_intra)(pix, sizeof(pixel), stride, 1, alpha, beta);
yading@10: }
yading@10: static void FUNCC(h264_h_loop_filter_chroma422_intra)(uint8_t *pix, int stride, int alpha, int beta)
yading@10: {
yading@10:     FUNCC(h264_loop_filter_chroma_intra)(pix, sizeof(pixel), stride, 4, alpha, beta);
yading@10: }
yading@10: static void FUNCC(h264_h_loop_filter_chroma422_mbaff_intra)(uint8_t *pix, int stride, int alpha, int beta)
yading@10: {
yading@10:     FUNCC(h264_loop_filter_chroma_intra)(pix, sizeof(pixel), stride, 2, alpha, beta);
yading@10: }