h264dsp_template.c 11.9 KB
Newer Older
1 2
/*
 * H.26L/H.264/AVC/JVT/14496-10/... encoder/decoder
3
 * Copyright (c) 2003-2011 Michael Niedermayer <michaelni@gmx.at>
4
 *
5
 * This file is part of FFmpeg.
6
 *
7
 * FFmpeg is free software; you can redistribute it and/or
8 9 10 11
 * modify it under the terms of the GNU Lesser General Public
 * License as published by the Free Software Foundation; either
 * version 2.1 of the License, or (at your option) any later version.
 *
12
 * FFmpeg is distributed in the hope that it will be useful,
13 14 15 16 17
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
 * Lesser General Public License for more details.
 *
 * You should have received a copy of the GNU Lesser General Public
18
 * License along with FFmpeg; if not, write to the Free Software
19 20 21 22 23 24 25 26 27
 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
 */

/**
 * @file
 * H.264 / AVC / MPEG4 part10 DSP functions.
 * @author Michael Niedermayer <michaelni@gmx.at>
 */

28
#include "bit_depth_template.c"
29 30 31

#define op_scale1(x)  block[x] = av_clip_pixel( (block[x]*weight + offset) >> log2_denom )
#define op_scale2(x)  dst[x] = av_clip_pixel( (src[x]*weights + dst[x]*weightd + offset) >> (log2_denom+1))
32 33 34 35
#define H264_WEIGHT(W) \
static void FUNCC(weight_h264_pixels ## W)(uint8_t *_block, int stride, int height, \
                                           int log2_denom, int weight, int offset) \
{ \
36
    int y; \
37
    pixel *block = (pixel*)_block; \
38
    stride >>= sizeof(pixel)-1; \
39
    offset <<= (log2_denom + (BIT_DEPTH-8)); \
40
    if(log2_denom) offset += 1<<(log2_denom-1); \
41
    for (y = 0; y < height; y++, block += stride) { \
42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62
        op_scale1(0); \
        op_scale1(1); \
        if(W==2) continue; \
        op_scale1(2); \
        op_scale1(3); \
        if(W==4) continue; \
        op_scale1(4); \
        op_scale1(5); \
        op_scale1(6); \
        op_scale1(7); \
        if(W==8) continue; \
        op_scale1(8); \
        op_scale1(9); \
        op_scale1(10); \
        op_scale1(11); \
        op_scale1(12); \
        op_scale1(13); \
        op_scale1(14); \
        op_scale1(15); \
    } \
} \
63 64 65
static void FUNCC(biweight_h264_pixels ## W)(uint8_t *_dst, uint8_t *_src, int stride, int height, \
                                             int log2_denom, int weightd, int weights, int offset) \
{ \
66
    int y; \
67 68
    pixel *dst = (pixel*)_dst; \
    pixel *src = (pixel*)_src; \
69
    stride >>= sizeof(pixel)-1; \
70
    offset <<= (BIT_DEPTH-8); \
71
    offset = ((offset + 1) | 1) << log2_denom; \
72
    for (y = 0; y < height; y++, dst += stride, src += stride) { \
73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94
        op_scale2(0); \
        op_scale2(1); \
        if(W==2) continue; \
        op_scale2(2); \
        op_scale2(3); \
        if(W==4) continue; \
        op_scale2(4); \
        op_scale2(5); \
        op_scale2(6); \
        op_scale2(7); \
        if(W==8) continue; \
        op_scale2(8); \
        op_scale2(9); \
        op_scale2(10); \
        op_scale2(11); \
        op_scale2(12); \
        op_scale2(13); \
        op_scale2(14); \
        op_scale2(15); \
    } \
}

95 96 97 98
H264_WEIGHT(16)
H264_WEIGHT(8)
H264_WEIGHT(4)
H264_WEIGHT(2)
99 100 101 102 103

#undef op_scale1
#undef op_scale2
#undef H264_WEIGHT

104
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)
105
{
106
    pixel *pix = (pixel*)p_pix;
107
    int i, d;
108 109
    xstride >>= sizeof(pixel)-1;
    ystride >>= sizeof(pixel)-1;
110 111
    alpha <<= BIT_DEPTH - 8;
    beta  <<= BIT_DEPTH - 8;
112
    for( i = 0; i < 4; i++ ) {
113 114
        const int tc_orig = tc0[i] << (BIT_DEPTH - 8);
        if( tc_orig < 0 ) {
115 116 117 118 119 120 121 122 123 124 125 126 127 128 129
            pix += inner_iters*ystride;
            continue;
        }
        for( d = 0; d < inner_iters; d++ ) {
            const int p0 = pix[-1*xstride];
            const int p1 = pix[-2*xstride];
            const int p2 = pix[-3*xstride];
            const int q0 = pix[0];
            const int q1 = pix[1*xstride];
            const int q2 = pix[2*xstride];

            if( FFABS( p0 - q0 ) < alpha &&
                FFABS( p1 - p0 ) < beta &&
                FFABS( q1 - q0 ) < beta ) {

130
                int tc = tc_orig;
131 132 133
                int i_delta;

                if( FFABS( p2 - p0 ) < beta ) {
134 135
                    if(tc_orig)
                    pix[-2*xstride] = p1 + av_clip( (( p2 + ( ( p0 + q0 + 1 ) >> 1 ) ) >> 1) - p1, -tc_orig, tc_orig );
136 137 138
                    tc++;
                }
                if( FFABS( q2 - q0 ) < beta ) {
139 140
                    if(tc_orig)
                    pix[   xstride] = q1 + av_clip( (( q2 + ( ( p0 + q0 + 1 ) >> 1 ) ) >> 1) - q1, -tc_orig, tc_orig );
141 142 143 144
                    tc++;
                }

                i_delta = av_clip( (((q0 - p0 ) << 2) + (p1 - q1) + 4) >> 3, -tc, tc );
145 146
                pix[-xstride] = av_clip_pixel( p0 + i_delta );    /* p0' */
                pix[0]        = av_clip_pixel( q0 - i_delta );    /* q0' */
147 148 149 150 151
            }
            pix += ystride;
        }
    }
}
152
static void FUNCC(h264_v_loop_filter_luma)(uint8_t *pix, int stride, int alpha, int beta, int8_t *tc0)
153
{
154
    FUNCC(h264_loop_filter_luma)(pix, stride, sizeof(pixel), 4, alpha, beta, tc0);
155
}
156
static void FUNCC(h264_h_loop_filter_luma)(uint8_t *pix, int stride, int alpha, int beta, int8_t *tc0)
157
{
158
    FUNCC(h264_loop_filter_luma)(pix, sizeof(pixel), stride, 4, alpha, beta, tc0);
159
}
160
static void FUNCC(h264_h_loop_filter_luma_mbaff)(uint8_t *pix, int stride, int alpha, int beta, int8_t *tc0)
161
{
162
    FUNCC(h264_loop_filter_luma)(pix, sizeof(pixel), stride, 2, alpha, beta, tc0);
163 164
}

165
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)
166
{
167
    pixel *pix = (pixel*)p_pix;
168
    int d;
169 170
    xstride >>= sizeof(pixel)-1;
    ystride >>= sizeof(pixel)-1;
171 172
    alpha <<= BIT_DEPTH - 8;
    beta  <<= BIT_DEPTH - 8;
173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217
    for( d = 0; d < 4 * inner_iters; d++ ) {
        const int p2 = pix[-3*xstride];
        const int p1 = pix[-2*xstride];
        const int p0 = pix[-1*xstride];

        const int q0 = pix[ 0*xstride];
        const int q1 = pix[ 1*xstride];
        const int q2 = pix[ 2*xstride];

        if( FFABS( p0 - q0 ) < alpha &&
            FFABS( p1 - p0 ) < beta &&
            FFABS( q1 - q0 ) < beta ) {

            if(FFABS( p0 - q0 ) < (( alpha >> 2 ) + 2 )){
                if( FFABS( p2 - p0 ) < beta)
                {
                    const int p3 = pix[-4*xstride];
                    /* p0', p1', p2' */
                    pix[-1*xstride] = ( p2 + 2*p1 + 2*p0 + 2*q0 + q1 + 4 ) >> 3;
                    pix[-2*xstride] = ( p2 + p1 + p0 + q0 + 2 ) >> 2;
                    pix[-3*xstride] = ( 2*p3 + 3*p2 + p1 + p0 + q0 + 4 ) >> 3;
                } else {
                    /* p0' */
                    pix[-1*xstride] = ( 2*p1 + p0 + q1 + 2 ) >> 2;
                }
                if( FFABS( q2 - q0 ) < beta)
                {
                    const int q3 = pix[3*xstride];
                    /* q0', q1', q2' */
                    pix[0*xstride] = ( p1 + 2*p0 + 2*q0 + 2*q1 + q2 + 4 ) >> 3;
                    pix[1*xstride] = ( p0 + q0 + q1 + q2 + 2 ) >> 2;
                    pix[2*xstride] = ( 2*q3 + 3*q2 + q1 + q0 + p0 + 4 ) >> 3;
                } else {
                    /* q0' */
                    pix[0*xstride] = ( 2*q1 + q0 + p1 + 2 ) >> 2;
                }
            }else{
                /* p0', q0' */
                pix[-1*xstride] = ( 2*p1 + p0 + q1 + 2 ) >> 2;
                pix[ 0*xstride] = ( 2*q1 + q0 + p1 + 2 ) >> 2;
            }
        }
        pix += ystride;
    }
}
218
static void FUNCC(h264_v_loop_filter_luma_intra)(uint8_t *pix, int stride, int alpha, int beta)
219
{
220
    FUNCC(h264_loop_filter_luma_intra)(pix, stride, sizeof(pixel), 4, alpha, beta);
221
}
222
static void FUNCC(h264_h_loop_filter_luma_intra)(uint8_t *pix, int stride, int alpha, int beta)
223
{
224
    FUNCC(h264_loop_filter_luma_intra)(pix, sizeof(pixel), stride, 4, alpha, beta);
225
}
226
static void FUNCC(h264_h_loop_filter_luma_mbaff_intra)(uint8_t *pix, int stride, int alpha, int beta)
227
{
228
    FUNCC(h264_loop_filter_luma_intra)(pix, sizeof(pixel), stride, 2, alpha, beta);
229 230
}

231
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)
232
{
233
    pixel *pix = (pixel*)p_pix;
234
    int i, d;
235 236
    alpha <<= BIT_DEPTH - 8;
    beta  <<= BIT_DEPTH - 8;
237 238
    xstride >>= sizeof(pixel)-1;
    ystride >>= sizeof(pixel)-1;
239
    for( i = 0; i < 4; i++ ) {
240
        const int tc = ((tc0[i] - 1) << (BIT_DEPTH - 8)) + 1;
241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256
        if( tc <= 0 ) {
            pix += inner_iters*ystride;
            continue;
        }
        for( d = 0; d < inner_iters; d++ ) {
            const int p0 = pix[-1*xstride];
            const int p1 = pix[-2*xstride];
            const int q0 = pix[0];
            const int q1 = pix[1*xstride];

            if( FFABS( p0 - q0 ) < alpha &&
                FFABS( p1 - p0 ) < beta &&
                FFABS( q1 - q0 ) < beta ) {

                int delta = av_clip( (((q0 - p0 ) << 2) + (p1 - q1) + 4) >> 3, -tc, tc );

257 258
                pix[-xstride] = av_clip_pixel( p0 + delta );    /* p0' */
                pix[0]        = av_clip_pixel( q0 - delta );    /* q0' */
259 260 261 262 263
            }
            pix += ystride;
        }
    }
}
264
static void FUNCC(h264_v_loop_filter_chroma)(uint8_t *pix, int stride, int alpha, int beta, int8_t *tc0)
265
{
266
    FUNCC(h264_loop_filter_chroma)(pix, stride, sizeof(pixel), 2, alpha, beta, tc0);
267
}
268
static void FUNCC(h264_h_loop_filter_chroma)(uint8_t *pix, int stride, int alpha, int beta, int8_t *tc0)
269
{
270
    FUNCC(h264_loop_filter_chroma)(pix, sizeof(pixel), stride, 2, alpha, beta, tc0);
271
}
272
static void FUNCC(h264_h_loop_filter_chroma_mbaff)(uint8_t *pix, int stride, int alpha, int beta, int8_t *tc0)
273
{
274
    FUNCC(h264_loop_filter_chroma)(pix, sizeof(pixel), stride, 1, alpha, beta, tc0);
275
}
276 277 278 279 280 281 282 283
static void FUNCC(h264_h_loop_filter_chroma422)(uint8_t *pix, int stride, int alpha, int beta, int8_t *tc0)
{
    FUNCC(h264_loop_filter_chroma)(pix, sizeof(pixel), stride, 4, alpha, beta, tc0);
}
static void FUNCC(h264_h_loop_filter_chroma422_mbaff)(uint8_t *pix, int stride, int alpha, int beta, int8_t *tc0)
{
    FUNCC(h264_loop_filter_chroma)(pix, sizeof(pixel), stride, 2, alpha, beta, tc0);
}
284

285
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)
286
{
287
    pixel *pix = (pixel*)p_pix;
288
    int d;
289 290
    xstride >>= sizeof(pixel)-1;
    ystride >>= sizeof(pixel)-1;
291 292
    alpha <<= BIT_DEPTH - 8;
    beta  <<= BIT_DEPTH - 8;
293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308
    for( d = 0; d < 4 * inner_iters; d++ ) {
        const int p0 = pix[-1*xstride];
        const int p1 = pix[-2*xstride];
        const int q0 = pix[0];
        const int q1 = pix[1*xstride];

        if( FFABS( p0 - q0 ) < alpha &&
            FFABS( p1 - p0 ) < beta &&
            FFABS( q1 - q0 ) < beta ) {

            pix[-xstride] = ( 2*p1 + p0 + q1 + 2 ) >> 2;   /* p0' */
            pix[0]        = ( 2*q1 + q0 + p1 + 2 ) >> 2;   /* q0' */
        }
        pix += ystride;
    }
}
309
static void FUNCC(h264_v_loop_filter_chroma_intra)(uint8_t *pix, int stride, int alpha, int beta)
310
{
311
    FUNCC(h264_loop_filter_chroma_intra)(pix, stride, sizeof(pixel), 2, alpha, beta);
312
}
313
static void FUNCC(h264_h_loop_filter_chroma_intra)(uint8_t *pix, int stride, int alpha, int beta)
314
{
315
    FUNCC(h264_loop_filter_chroma_intra)(pix, sizeof(pixel), stride, 2, alpha, beta);
316
}
317
static void FUNCC(h264_h_loop_filter_chroma_mbaff_intra)(uint8_t *pix, int stride, int alpha, int beta)
318
{
319
    FUNCC(h264_loop_filter_chroma_intra)(pix, sizeof(pixel), stride, 1, alpha, beta);
320
}
321 322 323 324 325 326 327 328
static void FUNCC(h264_h_loop_filter_chroma422_intra)(uint8_t *pix, int stride, int alpha, int beta)
{
    FUNCC(h264_loop_filter_chroma_intra)(pix, sizeof(pixel), stride, 4, alpha, beta);
}
static void FUNCC(h264_h_loop_filter_chroma422_mbaff_intra)(uint8_t *pix, int stride, int alpha, int beta)
{
    FUNCC(h264_loop_filter_chroma_intra)(pix, sizeof(pixel), stride, 2, alpha, beta);
}