2 * H.26L/H.264/AVC/JVT/14496-10/... encoder/decoder
3 * Copyright (c) 2003-2010 Michael Niedermayer <michaelni@gmx.at>
5 * This file is part of Libav.
7 * Libav is free software; you can redistribute it and/or
8 * modify it under the terms of the GNU Lesser General Public
9 * License as published by the Free Software Foundation; either
10 * version 2.1 of the License, or (at your option) any later version.
12 * Libav is distributed in the hope that it will be useful,
13 * but WITHOUT ANY WARRANTY; without even the implied warranty of
14 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
15 * Lesser General Public License for more details.
17 * You should have received a copy of the GNU Lesser General Public
18 * License along with Libav; if not, write to the Free Software
19 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
24 * H.264 / AVC / MPEG4 part10 DSP functions.
25 * @author Michael Niedermayer <michaelni@gmx.at>
32 #define op_scale1(x) block[x] = av_clip_uint8( (block[x]*weight + offset) >> log2_denom )
33 #define op_scale2(x) dst[x] = av_clip_uint8( (src[x]*weights + dst[x]*weightd + offset) >> (log2_denom+1))
34 #define H264_WEIGHT(W,H) \
35 static void weight_h264_pixels ## W ## x ## H ## _c(uint8_t *block, int stride, int log2_denom, int weight, int offset){ \
37 offset <<= log2_denom; \
38 if(log2_denom) offset += 1<<(log2_denom-1); \
39 for(y=0; y<H; y++, block += stride){ \
61 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){ \
63 offset = ((offset + 1) | 1) << log2_denom; \
64 for(y=0; y<H; y++, dst += stride, src += stride){ \
102 static av_always_inline av_flatten void h264_loop_filter_luma_c(uint8_t *pix, int xstride, int ystride, int inner_iters, int alpha, int beta, int8_t *tc0)
105 for( i = 0; i < 4; i++ ) {
107 pix += inner_iters*ystride;
110 for( d = 0; d < inner_iters; d++ ) {
111 const int p0 = pix[-1*xstride];
112 const int p1 = pix[-2*xstride];
113 const int p2 = pix[-3*xstride];
114 const int q0 = pix[0];
115 const int q1 = pix[1*xstride];
116 const int q2 = pix[2*xstride];
118 if( FFABS( p0 - q0 ) < alpha &&
119 FFABS( p1 - p0 ) < beta &&
120 FFABS( q1 - q0 ) < beta ) {
125 if( FFABS( p2 - p0 ) < beta ) {
127 pix[-2*xstride] = p1 + av_clip( (( p2 + ( ( p0 + q0 + 1 ) >> 1 ) ) >> 1) - p1, -tc0[i], tc0[i] );
130 if( FFABS( q2 - q0 ) < beta ) {
132 pix[ xstride] = q1 + av_clip( (( q2 + ( ( p0 + q0 + 1 ) >> 1 ) ) >> 1) - q1, -tc0[i], tc0[i] );
136 i_delta = av_clip( (((q0 - p0 ) << 2) + (p1 - q1) + 4) >> 3, -tc, tc );
137 pix[-xstride] = av_clip_uint8( p0 + i_delta ); /* p0' */
138 pix[0] = av_clip_uint8( q0 - i_delta ); /* q0' */
144 static void h264_v_loop_filter_luma_c(uint8_t *pix, int stride, int alpha, int beta, int8_t *tc0)
146 h264_loop_filter_luma_c(pix, stride, 1, 4, alpha, beta, tc0);
148 static void h264_h_loop_filter_luma_c(uint8_t *pix, int stride, int alpha, int beta, int8_t *tc0)
150 h264_loop_filter_luma_c(pix, 1, stride, 4, alpha, beta, tc0);
152 static void h264_h_loop_filter_luma_mbaff_c(uint8_t *pix, int stride, int alpha, int beta, int8_t *tc0)
154 h264_loop_filter_luma_c(pix, 1, stride, 2, alpha, beta, tc0);
157 static av_always_inline av_flatten void h264_loop_filter_luma_intra_c(uint8_t *pix, int xstride, int ystride, int inner_iters, int alpha, int beta)
160 for( d = 0; d < 4 * inner_iters; d++ ) {
161 const int p2 = pix[-3*xstride];
162 const int p1 = pix[-2*xstride];
163 const int p0 = pix[-1*xstride];
165 const int q0 = pix[ 0*xstride];
166 const int q1 = pix[ 1*xstride];
167 const int q2 = pix[ 2*xstride];
169 if( FFABS( p0 - q0 ) < alpha &&
170 FFABS( p1 - p0 ) < beta &&
171 FFABS( q1 - q0 ) < beta ) {
173 if(FFABS( p0 - q0 ) < (( alpha >> 2 ) + 2 )){
174 if( FFABS( p2 - p0 ) < beta)
176 const int p3 = pix[-4*xstride];
178 pix[-1*xstride] = ( p2 + 2*p1 + 2*p0 + 2*q0 + q1 + 4 ) >> 3;
179 pix[-2*xstride] = ( p2 + p1 + p0 + q0 + 2 ) >> 2;
180 pix[-3*xstride] = ( 2*p3 + 3*p2 + p1 + p0 + q0 + 4 ) >> 3;
183 pix[-1*xstride] = ( 2*p1 + p0 + q1 + 2 ) >> 2;
185 if( FFABS( q2 - q0 ) < beta)
187 const int q3 = pix[3*xstride];
189 pix[0*xstride] = ( p1 + 2*p0 + 2*q0 + 2*q1 + q2 + 4 ) >> 3;
190 pix[1*xstride] = ( p0 + q0 + q1 + q2 + 2 ) >> 2;
191 pix[2*xstride] = ( 2*q3 + 3*q2 + q1 + q0 + p0 + 4 ) >> 3;
194 pix[0*xstride] = ( 2*q1 + q0 + p1 + 2 ) >> 2;
198 pix[-1*xstride] = ( 2*p1 + p0 + q1 + 2 ) >> 2;
199 pix[ 0*xstride] = ( 2*q1 + q0 + p1 + 2 ) >> 2;
205 static void h264_v_loop_filter_luma_intra_c(uint8_t *pix, int stride, int alpha, int beta)
207 h264_loop_filter_luma_intra_c(pix, stride, 1, 4, alpha, beta);
209 static void h264_h_loop_filter_luma_intra_c(uint8_t *pix, int stride, int alpha, int beta)
211 h264_loop_filter_luma_intra_c(pix, 1, stride, 4, alpha, beta);
213 static void h264_h_loop_filter_luma_mbaff_intra_c(uint8_t *pix, int stride, int alpha, int beta)
215 h264_loop_filter_luma_intra_c(pix, 1, stride, 2, alpha, beta);
218 static av_always_inline av_flatten void h264_loop_filter_chroma_c(uint8_t *pix, int xstride, int ystride, int inner_iters, int alpha, int beta, int8_t *tc0)
221 for( i = 0; i < 4; i++ ) {
222 const int tc = tc0[i];
224 pix += inner_iters*ystride;
227 for( d = 0; d < inner_iters; d++ ) {
228 const int p0 = pix[-1*xstride];
229 const int p1 = pix[-2*xstride];
230 const int q0 = pix[0];
231 const int q1 = pix[1*xstride];
233 if( FFABS( p0 - q0 ) < alpha &&
234 FFABS( p1 - p0 ) < beta &&
235 FFABS( q1 - q0 ) < beta ) {
237 int delta = av_clip( (((q0 - p0 ) << 2) + (p1 - q1) + 4) >> 3, -tc, tc );
239 pix[-xstride] = av_clip_uint8( p0 + delta ); /* p0' */
240 pix[0] = av_clip_uint8( q0 - delta ); /* q0' */
246 static void h264_v_loop_filter_chroma_c(uint8_t *pix, int stride, int alpha, int beta, int8_t *tc0)
248 h264_loop_filter_chroma_c(pix, stride, 1, 2, alpha, beta, tc0);
250 static void h264_h_loop_filter_chroma_c(uint8_t *pix, int stride, int alpha, int beta, int8_t *tc0)
252 h264_loop_filter_chroma_c(pix, 1, stride, 2, alpha, beta, tc0);
254 static void h264_h_loop_filter_chroma_mbaff_c(uint8_t *pix, int stride, int alpha, int beta, int8_t *tc0)
256 h264_loop_filter_chroma_c(pix, 1, stride, 1, alpha, beta, tc0);
259 static av_always_inline av_flatten void h264_loop_filter_chroma_intra_c(uint8_t *pix, int xstride, int ystride, int inner_iters, int alpha, int beta)
262 for( d = 0; d < 4 * inner_iters; d++ ) {
263 const int p0 = pix[-1*xstride];
264 const int p1 = pix[-2*xstride];
265 const int q0 = pix[0];
266 const int q1 = pix[1*xstride];
268 if( FFABS( p0 - q0 ) < alpha &&
269 FFABS( p1 - p0 ) < beta &&
270 FFABS( q1 - q0 ) < beta ) {
272 pix[-xstride] = ( 2*p1 + p0 + q1 + 2 ) >> 2; /* p0' */
273 pix[0] = ( 2*q1 + q0 + p1 + 2 ) >> 2; /* q0' */
278 static void h264_v_loop_filter_chroma_intra_c(uint8_t *pix, int stride, int alpha, int beta)
280 h264_loop_filter_chroma_intra_c(pix, stride, 1, 2, alpha, beta);
282 static void h264_h_loop_filter_chroma_intra_c(uint8_t *pix, int stride, int alpha, int beta)
284 h264_loop_filter_chroma_intra_c(pix, 1, stride, 2, alpha, beta);
286 static void h264_h_loop_filter_chroma_mbaff_intra_c(uint8_t *pix, int stride, int alpha, int beta)
288 h264_loop_filter_chroma_intra_c(pix, 1, stride, 1, alpha, beta);
291 void ff_h264dsp_init(H264DSPContext *c)
293 c->h264_idct_add= ff_h264_idct_add_c;
294 c->h264_idct8_add= ff_h264_idct8_add_c;
295 c->h264_idct_dc_add= ff_h264_idct_dc_add_c;
296 c->h264_idct8_dc_add= ff_h264_idct8_dc_add_c;
297 c->h264_idct_add16 = ff_h264_idct_add16_c;
298 c->h264_idct8_add4 = ff_h264_idct8_add4_c;
299 c->h264_idct_add8 = ff_h264_idct_add8_c;
300 c->h264_idct_add16intra= ff_h264_idct_add16intra_c;
301 c->h264_luma_dc_dequant_idct= ff_h264_luma_dc_dequant_idct_c;
303 c->weight_h264_pixels_tab[0]= weight_h264_pixels16x16_c;
304 c->weight_h264_pixels_tab[1]= weight_h264_pixels16x8_c;
305 c->weight_h264_pixels_tab[2]= weight_h264_pixels8x16_c;
306 c->weight_h264_pixels_tab[3]= weight_h264_pixels8x8_c;
307 c->weight_h264_pixels_tab[4]= weight_h264_pixels8x4_c;
308 c->weight_h264_pixels_tab[5]= weight_h264_pixels4x8_c;
309 c->weight_h264_pixels_tab[6]= weight_h264_pixels4x4_c;
310 c->weight_h264_pixels_tab[7]= weight_h264_pixels4x2_c;
311 c->weight_h264_pixels_tab[8]= weight_h264_pixels2x4_c;
312 c->weight_h264_pixels_tab[9]= weight_h264_pixels2x2_c;
313 c->biweight_h264_pixels_tab[0]= biweight_h264_pixels16x16_c;
314 c->biweight_h264_pixels_tab[1]= biweight_h264_pixels16x8_c;
315 c->biweight_h264_pixels_tab[2]= biweight_h264_pixels8x16_c;
316 c->biweight_h264_pixels_tab[3]= biweight_h264_pixels8x8_c;
317 c->biweight_h264_pixels_tab[4]= biweight_h264_pixels8x4_c;
318 c->biweight_h264_pixels_tab[5]= biweight_h264_pixels4x8_c;
319 c->biweight_h264_pixels_tab[6]= biweight_h264_pixels4x4_c;
320 c->biweight_h264_pixels_tab[7]= biweight_h264_pixels4x2_c;
321 c->biweight_h264_pixels_tab[8]= biweight_h264_pixels2x4_c;
322 c->biweight_h264_pixels_tab[9]= biweight_h264_pixels2x2_c;
324 c->h264_v_loop_filter_luma= h264_v_loop_filter_luma_c;
325 c->h264_h_loop_filter_luma= h264_h_loop_filter_luma_c;
326 c->h264_h_loop_filter_luma_mbaff= h264_h_loop_filter_luma_mbaff_c;
327 c->h264_v_loop_filter_luma_intra= h264_v_loop_filter_luma_intra_c;
328 c->h264_h_loop_filter_luma_intra= h264_h_loop_filter_luma_intra_c;
329 c->h264_h_loop_filter_luma_mbaff_intra= h264_h_loop_filter_luma_mbaff_intra_c;
330 c->h264_v_loop_filter_chroma= h264_v_loop_filter_chroma_c;
331 c->h264_h_loop_filter_chroma= h264_h_loop_filter_chroma_c;
332 c->h264_h_loop_filter_chroma_mbaff= h264_h_loop_filter_chroma_mbaff_c;
333 c->h264_v_loop_filter_chroma_intra= h264_v_loop_filter_chroma_intra_c;
334 c->h264_h_loop_filter_chroma_intra= h264_h_loop_filter_chroma_intra_c;
335 c->h264_h_loop_filter_chroma_mbaff_intra= h264_h_loop_filter_chroma_mbaff_intra_c;
336 c->h264_loop_filter_strength= NULL;
338 if (ARCH_ARM) ff_h264dsp_init_arm(c);
339 if (HAVE_ALTIVEC) ff_h264dsp_init_ppc(c);
340 if (HAVE_MMX) ff_h264dsp_init_x86(c);