frame.c 34.1 KB
Newer Older
Laurent Aimar's avatar
Laurent Aimar committed
1 2 3
/*****************************************************************************
 * frame.c: h264 encoder library
 *****************************************************************************
4
 * Copyright (C) 2003-2008 x264 project
Laurent Aimar's avatar
Laurent Aimar committed
5 6
 *
 * Authors: Laurent Aimar <fenrir@via.ecp.fr>
7
 *          Loren Merritt <lorenm@u.washington.edu>
Laurent Aimar's avatar
Laurent Aimar committed
8 9 10 11 12 13 14 15 16 17 18 19 20
 *
 * 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.
 *
 * 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.
 *
 * You should have received a copy of the GNU General Public License
 * along with this program; if not, write to the Free Software
21
 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA  02111, USA.
Laurent Aimar's avatar
Laurent Aimar committed
22 23 24 25
 *****************************************************************************/

#include "common.h"

26 27
#define ALIGN(x,a) (((x)+((a)-1))&~((a)-1))

Laurent Aimar's avatar
Laurent Aimar committed
28 29
x264_frame_t *x264_frame_new( x264_t *h )
{
30
    x264_frame_t *frame = x264_malloc( sizeof(x264_frame_t) );
31
    int i, j;
Laurent Aimar's avatar
Laurent Aimar committed
32

33
    int i_mb_count = h->mb.i_mb_count;
34
    int i_stride, i_width, i_lines;
35
    int i_padv = PADV << h->param.b_interlaced;
36
    int luma_plane_size;
37
    int align = h->param.cpu&X264_CPU_CACHELINE_64 ? 64 : h->param.cpu&X264_CPU_CACHELINE_32 ? 32 : 16;
Laurent Aimar's avatar
Laurent Aimar committed
38

39 40
    if( !frame ) return NULL;

41 42
    memset( frame, 0, sizeof(x264_frame_t) );

Laurent Aimar's avatar
Laurent Aimar committed
43
    /* allocate frame data (+64 for extra data for me) */
44 45 46
    i_width  = ALIGN( h->param.i_width, 16 );
    i_stride = ALIGN( i_width + 2*PADH, align );
    i_lines  = ALIGN( h->param.i_height, 16<<h->param.b_interlaced );
47

Laurent Aimar's avatar
Laurent Aimar committed
48 49 50
    frame->i_plane = 3;
    for( i = 0; i < 3; i++ )
    {
51 52 53
        frame->i_stride[i] = i_stride >> !!i;
        frame->i_width[i] = i_width >> !!i;
        frame->i_lines[i] = i_lines >> !!i;
Laurent Aimar's avatar
Laurent Aimar committed
54 55
    }

56 57
    luma_plane_size = (frame->i_stride[0] * ( frame->i_lines[0] + 2*i_padv ));
    for( i = 1; i < 3; i++ )
58
    {
59
        CHECKED_MALLOC( frame->buffer[i], luma_plane_size/4 );
60
        frame->plane[i] = frame->buffer[i] + (frame->i_stride[i] * i_padv + PADH)/2;
61
    }
62 63 64 65
    /* all 4 luma planes allocated together, since the cacheline split code
     * requires them to be in-phase wrt cacheline alignment. */
    CHECKED_MALLOC( frame->buffer[0], 4*luma_plane_size);
    for( i = 0; i < 4; i++ )
66
        frame->filtered[i] = frame->buffer[0] + i*luma_plane_size + frame->i_stride[0] * i_padv + PADH;
67
    frame->plane[0] = frame->filtered[0];
68

69
    if( h->frames.b_have_lowres )
70
    {
71
        frame->i_width_lowres = frame->i_width[0]/2;
72
        frame->i_stride_lowres = ALIGN( frame->i_width_lowres + 2*PADH, align );
73
        frame->i_lines_lowres = frame->i_lines[0]/2;
74 75 76 77

        luma_plane_size = frame->i_stride_lowres * ( frame->i_lines[0]/2 + 2*i_padv );

        CHECKED_MALLOC( frame->buffer_lowres[0], 4 * luma_plane_size );
78
        for( i = 0; i < 4; i++ )
79
            frame->lowres[i] = frame->buffer_lowres[0] + (frame->i_stride_lowres * i_padv + PADH) + i * luma_plane_size;
80
    }
81

82
    if( h->param.analyse.i_me_method >= X264_ME_ESA )
83
    {
84
        CHECKED_MALLOC( frame->buffer[3],
85
                        2 * frame->i_stride[0] * (frame->i_lines[0] + 2*i_padv) * sizeof(uint16_t) );
86
        frame->integral = (uint16_t*)frame->buffer[3] + frame->i_stride[0] * i_padv + PADH;
87 88
    }

Laurent Aimar's avatar
Laurent Aimar committed
89 90 91
    frame->i_poc = -1;
    frame->i_type = X264_TYPE_AUTO;
    frame->i_qpplus1 = 0;
92 93
    frame->i_pts = -1;
    frame->i_frame = -1;
94
    frame->i_frame_num = -1;
Loren Merritt's avatar
Loren Merritt committed
95
    frame->i_lines_completed = -1;
Laurent Aimar's avatar
Laurent Aimar committed
96

97 98 99
    CHECKED_MALLOC( frame->mb_type, i_mb_count * sizeof(int8_t));
    CHECKED_MALLOC( frame->mv[0], 2*16 * i_mb_count * sizeof(int16_t) );
    CHECKED_MALLOC( frame->ref[0], 4 * i_mb_count * sizeof(int8_t) );
100 101
    if( h->param.i_bframe )
    {
102 103
        CHECKED_MALLOC( frame->mv[1], 2*16 * i_mb_count * sizeof(int16_t) );
        CHECKED_MALLOC( frame->ref[1], 4 * i_mb_count * sizeof(int8_t) );
104 105 106 107 108 109 110
    }
    else
    {
        frame->mv[1]  = NULL;
        frame->ref[1] = NULL;
    }

111 112
    CHECKED_MALLOC( frame->i_row_bits, i_lines/16 * sizeof(int) );
    CHECKED_MALLOC( frame->i_row_qp, i_lines/16 * sizeof(int) );
113 114
    for( i = 0; i < h->param.i_bframe + 2; i++ )
        for( j = 0; j < h->param.i_bframe + 2; j++ )
115
            CHECKED_MALLOC( frame->i_row_satds[i][j], i_lines/16 * sizeof(int) );
116

117 118
    x264_pthread_mutex_init( &frame->mutex, NULL );
    x264_pthread_cond_init( &frame->cv, NULL );
119

Laurent Aimar's avatar
Laurent Aimar committed
120
    return frame;
121 122 123 124

fail:
    x264_frame_delete( frame );
    return NULL;
Laurent Aimar's avatar
Laurent Aimar committed
125 126 127 128
}

void x264_frame_delete( x264_frame_t *frame )
{
129
    int i, j;
130
    for( i = 0; i < 4; i++ )
131
        x264_free( frame->buffer[i] );
132 133
    for( i = 0; i < 4; i++ )
        x264_free( frame->buffer_lowres[i] );
134 135 136 137 138
    for( i = 0; i < X264_BFRAME_MAX+2; i++ )
        for( j = 0; j < X264_BFRAME_MAX+2; j++ )
            x264_free( frame->i_row_satds[i][j] );
    x264_free( frame->i_row_bits );
    x264_free( frame->i_row_qp );
139
    x264_free( frame->mb_type );
140 141 142 143
    x264_free( frame->mv[0] );
    x264_free( frame->mv[1] );
    x264_free( frame->ref[0] );
    x264_free( frame->ref[1] );
144 145
    x264_pthread_mutex_destroy( &frame->mutex );
    x264_pthread_cond_destroy( &frame->cv );
Laurent Aimar's avatar
Laurent Aimar committed
146 147 148
    x264_free( frame );
}

149
int x264_frame_copy_picture( x264_t *h, x264_frame_t *dst, x264_picture_t *src )
Laurent Aimar's avatar
Laurent Aimar committed
150
{
151
    int i_csp = src->img.i_csp & X264_CSP_MASK;
152 153 154 155 156 157 158
    int i;
    if( i_csp != X264_CSP_I420 && i_csp != X264_CSP_YV12 )
    {
        x264_log( h, X264_LOG_ERROR, "Arg invalid CSP\n" );
        return -1;
    }

Laurent Aimar's avatar
Laurent Aimar committed
159 160 161 162
    dst->i_type     = src->i_type;
    dst->i_qpplus1  = src->i_qpplus1;
    dst->i_pts      = src->i_pts;

163 164 165 166 167 168 169 170 171 172 173 174 175 176 177
    for( i=0; i<3; i++ )
    {
        int s = (i_csp == X264_CSP_YV12 && i) ? i^3 : i;
        uint8_t *plane = src->img.plane[s];
        int stride = src->img.i_stride[s];
        int width = h->param.i_width >> !!i;
        int height = h->param.i_height >> !!i;
        if( src->img.i_csp & X264_CSP_VFLIP )
        {
            plane += (height-1)*stride;
            stride = -stride;
        }
        h->mc.plane_copy( dst->plane[i], dst->i_stride[i], plane, stride, width, height );
    }
    return 0;
Laurent Aimar's avatar
Laurent Aimar committed
178 179 180 181
}



Loren Merritt's avatar
Loren Merritt committed
182
static void plane_expand_border( uint8_t *pix, int i_stride, int i_width, int i_height, int i_padh, int i_padv, int b_pad_top, int b_pad_bottom )
183 184 185 186 187 188
{
#define PPIXEL(x, y) ( pix + (x) + (y)*i_stride )
    int y;
    for( y = 0; y < i_height; y++ )
    {
        /* left band */
189
        memset( PPIXEL(-i_padh, y), PPIXEL(0, y)[0], i_padh );
190
        /* right band */
191
        memset( PPIXEL(i_width, y), PPIXEL(i_width-1, y)[0], i_padh );
192 193
    }
    /* upper band */
Loren Merritt's avatar
Loren Merritt committed
194
    if( b_pad_top )
195 196
    for( y = 0; y < i_padv; y++ )
        memcpy( PPIXEL(-i_padh, -y-1), PPIXEL(-i_padh, 0), i_width+2*i_padh );
197
    /* lower band */
Loren Merritt's avatar
Loren Merritt committed
198
    if( b_pad_bottom )
199 200
    for( y = 0; y < i_padv; y++ )
        memcpy( PPIXEL(-i_padh, i_height+y), PPIXEL(-i_padh, i_height-1), i_width+2*i_padh );
201 202 203
#undef PPIXEL
}

Loren Merritt's avatar
Loren Merritt committed
204
void x264_frame_expand_border( x264_t *h, x264_frame_t *frame, int mb_y, int b_end )
Laurent Aimar's avatar
Laurent Aimar committed
205
{
206
    int i;
Loren Merritt's avatar
Loren Merritt committed
207 208 209
    int b_start = !mb_y;
    if( mb_y & h->sh.b_mbaff )
        return;
Laurent Aimar's avatar
Laurent Aimar committed
210 211
    for( i = 0; i < frame->i_plane; i++ )
    {
212 213
        int stride = frame->i_stride[i];
        int width = 16*h->sps->i_mb_width >> !!i;
Loren Merritt's avatar
Loren Merritt committed
214
        int height = (b_end ? 16*(h->sps->i_mb_height - mb_y) >> h->sh.b_mbaff : 16) >> !!i;
215 216
        int padh = PADH >> !!i;
        int padv = PADV >> !!i;
Loren Merritt's avatar
Loren Merritt committed
217 218 219 220 221
        // buffer: 2 chroma, 3 luma (rounded to 4) because deblocking goes beyond the top of the mb
        uint8_t *pix = frame->plane[i] + X264_MAX(0, (16*mb_y-4)*stride >> !!i);
        if( b_end && !b_start )
            height += 4 >> (!!i + h->sh.b_mbaff);
        if( h->sh.b_mbaff )
222
        {
Loren Merritt's avatar
Loren Merritt committed
223 224
            plane_expand_border( pix, stride*2, width, height, padh, padv, b_start, b_end );
            plane_expand_border( pix+stride, stride*2, width, height, padh, padv, b_start, b_end );
225 226 227
        }
        else
        {
Loren Merritt's avatar
Loren Merritt committed
228
            plane_expand_border( pix, stride, width, height, padh, padv, b_start, b_end );
229
        }
Laurent Aimar's avatar
Laurent Aimar committed
230 231 232
    }
}

Loren Merritt's avatar
Loren Merritt committed
233
void x264_frame_expand_border_filtered( x264_t *h, x264_frame_t *frame, int mb_y, int b_end )
234
{
Loren Merritt's avatar
Loren Merritt committed
235 236
    /* during filtering, 8 extra pixels were filtered on each edge,
     * but up to 3 of the horizontal ones may be wrong. 
237
       we want to expand border from the last filtered pixel */
Loren Merritt's avatar
Loren Merritt committed
238
    int b_start = !mb_y;
239
    int stride = frame->i_stride[0];
Loren Merritt's avatar
Loren Merritt committed
240
    int width = 16*h->sps->i_mb_width + 8;
Loren Merritt's avatar
Loren Merritt committed
241
    int height = b_end ? (16*(h->sps->i_mb_height - mb_y) >> h->sh.b_mbaff) + 16 : 16;
Loren Merritt's avatar
Loren Merritt committed
242
    int padh = PADH - 4;
243
    int padv = PADV - 8;
244
    int i;
245
    for( i = 1; i < 4; i++ )
246
    {
Loren Merritt's avatar
Loren Merritt committed
247
        // buffer: 8 luma, to match the hpel filter
Loren Merritt's avatar
Loren Merritt committed
248
        uint8_t *pix = frame->filtered[i] + (16*mb_y - (8 << h->sh.b_mbaff)) * stride - 4;
Loren Merritt's avatar
Loren Merritt committed
249
        if( h->sh.b_mbaff )
250
        {
Loren Merritt's avatar
Loren Merritt committed
251 252
            plane_expand_border( pix, stride*2, width, height, padh, padv, b_start, b_end );
            plane_expand_border( pix+stride, stride*2, width, height, padh, padv, b_start, b_end );
253 254 255
        }
        else
        {
Loren Merritt's avatar
Loren Merritt committed
256
            plane_expand_border( pix, stride, width, height, padh, padv, b_start, b_end );
257 258
        }
    }
259 260
}

261
void x264_frame_expand_border_lowres( x264_frame_t *frame )
262
{
263
    int i;
264
    for( i = 0; i < 4; i++ )
Loren Merritt's avatar
Loren Merritt committed
265
        plane_expand_border( frame->lowres[i], frame->i_stride_lowres, frame->i_stride_lowres - 2*PADH, frame->i_lines_lowres, PADH, PADV, 1, 1 );
266 267
}

Loren Merritt's avatar
Loren Merritt committed
268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287
void x264_frame_expand_border_mod16( x264_t *h, x264_frame_t *frame )
{
    int i, y;
    for( i = 0; i < frame->i_plane; i++ )
    {
        int i_subsample = i ? 1 : 0;
        int i_width = h->param.i_width >> i_subsample;
        int i_height = h->param.i_height >> i_subsample;
        int i_padx = ( h->sps->i_mb_width * 16 - h->param.i_width ) >> i_subsample;
        int i_pady = ( h->sps->i_mb_height * 16 - h->param.i_height ) >> i_subsample;

        if( i_padx )
        {
            for( y = 0; y < i_height; y++ )
                memset( &frame->plane[i][y*frame->i_stride[i] + i_width],
                         frame->plane[i][y*frame->i_stride[i] + i_width - 1],
                         i_padx );
        }
        if( i_pady )
        {
288
            //FIXME interlace? or just let it pad using the wrong field
Loren Merritt's avatar
Loren Merritt committed
289
            for( y = i_height; y < i_height + i_pady; y++ )
Loren Merritt's avatar
Loren Merritt committed
290 291 292 293 294 295 296
                memcpy( &frame->plane[i][y*frame->i_stride[i]],
                        &frame->plane[i][(i_height-1)*frame->i_stride[i]],
                        i_width + i_padx );
        }
    }
}

297

298 299 300 301 302 303
/* cavlc + 8x8 transform stores nnz per 16 coeffs for the purpose of
 * entropy coding, but per 64 coeffs for the purpose of deblocking */
void munge_cavlc_nnz_row( x264_t *h, int mb_y, uint8_t (*buf)[16] )
{
    uint32_t (*src)[6] = (uint32_t(*)[6])h->mb.non_zero_count + mb_y * h->sps->i_mb_width;
    int8_t *transform = h->mb.mb_transform_size + mb_y * h->sps->i_mb_width;
304
    int x, nnz;
305 306 307 308 309
    for( x=0; x<h->sps->i_mb_width; x++ )
    {
        memcpy( buf+x, src+x, 16 );
        if( transform[x] )
        {
310 311 312 313
            nnz = src[x][0] | src[x][1];
            src[x][0] = src[x][1] = ((uint16_t)nnz ? 0x0101 : 0) + (nnz>>16 ? 0x01010000 : 0);
            nnz = src[x][2] | src[x][3];
            src[x][2] = src[x][3] = ((uint16_t)nnz ? 0x0101 : 0) + (nnz>>16 ? 0x01010000 : 0);
314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339
        }
    }
}

static void restore_cavlc_nnz_row( x264_t *h, int mb_y, uint8_t (*buf)[16] )
{
    uint8_t (*dst)[24] = h->mb.non_zero_count + mb_y * h->sps->i_mb_width;
    int x;
    for( x=0; x<h->sps->i_mb_width; x++ )
        memcpy( dst+x, buf+x, 16 );
}

static void munge_cavlc_nnz( x264_t *h, int mb_y, uint8_t (*buf)[16], void (*func)(x264_t*, int, uint8_t (*)[16]) )
{
    func( h, mb_y, buf );
    if( mb_y > 0 )
        func( h, mb_y-1, buf + h->sps->i_mb_width );
    if( h->sh.b_mbaff )
    {
        func( h, mb_y+1, buf + h->sps->i_mb_width * 2 );
        if( mb_y > 0 )
            func( h, mb_y-2, buf + h->sps->i_mb_width * 3 );
    }
}


340
/* Deblocking filter */
Laurent Aimar's avatar
Laurent Aimar committed
341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359

static const int i_alpha_table[52] =
{
     0,  0,  0,  0,  0,  0,  0,  0,  0,  0,
     0,  0,  0,  0,  0,  0,  4,  4,  5,  6,
     7,  8,  9, 10, 12, 13, 15, 17, 20, 22,
    25, 28, 32, 36, 40, 45, 50, 56, 63, 71,
    80, 90,101,113,127,144,162,182,203,226,
    255, 255
};
static const int i_beta_table[52] =
{
     0,  0,  0,  0,  0,  0,  0,  0,  0,  0,
     0,  0,  0,  0,  0,  0,  2,  2,  2,  3,
     3,  3,  3,  4,  4,  4,  6,  6,  7,  7,
     8,  8,  9,  9, 10, 10, 11, 11, 12, 12,
    13, 13, 14, 14, 15, 15, 16, 16, 17, 17,
    18, 18
};
Fiona Glaser's avatar
Fiona Glaser committed
360 361

static const int8_t i_tc0_table[52][4] =
Laurent Aimar's avatar
Laurent Aimar committed
362
{
Fiona Glaser's avatar
Fiona Glaser committed
363 364 365 366 367 368 369 370 371
    {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 },
    {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 },
    {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 1 },
    {-1, 0, 0, 1 }, {-1, 0, 0, 1 }, {-1, 0, 0, 1 }, {-1, 0, 1, 1 }, {-1, 0, 1, 1 }, {-1, 1, 1, 1 },
    {-1, 1, 1, 1 }, {-1, 1, 1, 1 }, {-1, 1, 1, 1 }, {-1, 1, 1, 2 }, {-1, 1, 1, 2 }, {-1, 1, 1, 2 },
    {-1, 1, 1, 2 }, {-1, 1, 2, 3 }, {-1, 1, 2, 3 }, {-1, 2, 2, 3 }, {-1, 2, 2, 4 }, {-1, 2, 3, 4 },
    {-1, 2, 3, 4 }, {-1, 3, 3, 5 }, {-1, 3, 4, 6 }, {-1, 3, 4, 6 }, {-1, 4, 5, 7 }, {-1, 4, 5, 8 },
    {-1, 4, 6, 9 }, {-1, 5, 7,10 }, {-1, 6, 8,11 }, {-1, 6, 8,13 }, {-1, 7,10,14 }, {-1, 8,11,16 },
    {-1, 9,12,18 }, {-1,10,13,20 }, {-1,11,15,23 }, {-1,13,17,25 }
Laurent Aimar's avatar
Laurent Aimar committed
372 373 374 375 376 377 378 379 380 381 382
};

/* From ffmpeg */
static inline int clip_uint8( int a )
{
    if (a&(~255))
        return (-a)>>31;
    else
        return a;
}

Loren Merritt's avatar
Loren Merritt committed
383
static inline void deblock_luma_c( uint8_t *pix, int xstride, int ystride, int alpha, int beta, int8_t *tc0 )
Laurent Aimar's avatar
Laurent Aimar committed
384 385
{
    int i, d;
Loren Merritt's avatar
Loren Merritt committed
386 387 388
    for( i = 0; i < 4; i++ ) {
        if( tc0[i] < 0 ) {
            pix += 4*ystride;
Laurent Aimar's avatar
Laurent Aimar committed
389 390
            continue;
        }
Loren Merritt's avatar
Loren Merritt committed
391 392 393 394 395 396 397 398
        for( d = 0; d < 4; 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];
   
Loren Merritt's avatar
Loren Merritt committed
399 400 401
            if( abs( p0 - q0 ) < alpha &&
                abs( p1 - p0 ) < beta &&
                abs( q1 - q0 ) < beta ) {
Loren Merritt's avatar
Loren Merritt committed
402 403 404 405
   
                int tc = tc0[i];
                int delta;
   
Loren Merritt's avatar
Loren Merritt committed
406
                if( abs( p2 - p0 ) < beta ) {
Loren Merritt's avatar
Loren Merritt committed
407
                    pix[-2*xstride] = p1 + x264_clip3( (( p2 + ((p0 + q0 + 1) >> 1)) >> 1) - p1, -tc0[i], tc0[i] );
Fiona Glaser's avatar
Fiona Glaser committed
408
                    tc++;
Laurent Aimar's avatar
Laurent Aimar committed
409
                }
Loren Merritt's avatar
Loren Merritt committed
410
                if( abs( q2 - q0 ) < beta ) {
Loren Merritt's avatar
Loren Merritt committed
411 412
                    pix[ 1*xstride] = q1 + x264_clip3( (( q2 + ((p0 + q0 + 1) >> 1)) >> 1) - q1, -tc0[i], tc0[i] );
                    tc++;
Laurent Aimar's avatar
Laurent Aimar committed
413
                }
Loren Merritt's avatar
Loren Merritt committed
414 415 416 417
    
                delta = x264_clip3( (((q0 - p0 ) << 2) + (p1 - q1) + 4) >> 3, -tc, tc );
                pix[-1*xstride] = clip_uint8( p0 + delta );    /* p0' */
                pix[ 0*xstride] = clip_uint8( q0 - delta );    /* q0' */
Laurent Aimar's avatar
Laurent Aimar committed
418
            }
Loren Merritt's avatar
Loren Merritt committed
419
            pix += ystride;
Laurent Aimar's avatar
Laurent Aimar committed
420 421 422
        }
    }
}
Loren Merritt's avatar
Loren Merritt committed
423 424 425 426 427 428 429 430
static void deblock_v_luma_c( uint8_t *pix, int stride, int alpha, int beta, int8_t *tc0 )
{
    deblock_luma_c( pix, stride, 1, alpha, beta, tc0 ); 
}
static void deblock_h_luma_c( uint8_t *pix, int stride, int alpha, int beta, int8_t *tc0 )
{
    deblock_luma_c( pix, 1, stride, alpha, beta, tc0 );
}
Laurent Aimar's avatar
Laurent Aimar committed
431

Loren Merritt's avatar
Loren Merritt committed
432
static inline void deblock_chroma_c( uint8_t *pix, int xstride, int ystride, int alpha, int beta, int8_t *tc0 )
Laurent Aimar's avatar
Laurent Aimar committed
433 434
{
    int i, d;
Loren Merritt's avatar
Loren Merritt committed
435 436 437 438
    for( i = 0; i < 4; i++ ) {
        const int tc = tc0[i];
        if( tc <= 0 ) {
            pix += 2*ystride;
Laurent Aimar's avatar
Laurent Aimar committed
439 440
            continue;
        }
Loren Merritt's avatar
Loren Merritt committed
441 442 443 444 445 446
        for( d = 0; d < 2; d++ ) {
            const int p1 = pix[-2*xstride];
            const int p0 = pix[-1*xstride];
            const int q0 = pix[ 0*xstride];
            const int q1 = pix[ 1*xstride];

Loren Merritt's avatar
Loren Merritt committed
447 448 449
            if( abs( p0 - q0 ) < alpha &&
                abs( p1 - p0 ) < beta &&
                abs( q1 - q0 ) < beta ) {
Loren Merritt's avatar
Loren Merritt committed
450 451 452 453

                int delta = x264_clip3( (((q0 - p0 ) << 2) + (p1 - q1) + 4) >> 3, -tc, tc );
                pix[-1*xstride] = clip_uint8( p0 + delta );    /* p0' */
                pix[ 0*xstride] = clip_uint8( q0 - delta );    /* q0' */
Laurent Aimar's avatar
Laurent Aimar committed
454
            }
Loren Merritt's avatar
Loren Merritt committed
455
            pix += ystride;
Laurent Aimar's avatar
Laurent Aimar committed
456 457 458
        }
    }
}
Loren Merritt's avatar
Loren Merritt committed
459 460 461 462 463 464 465 466
static void deblock_v_chroma_c( uint8_t *pix, int stride, int alpha, int beta, int8_t *tc0 )
{   
    deblock_chroma_c( pix, stride, 1, alpha, beta, tc0 );
}
static void deblock_h_chroma_c( uint8_t *pix, int stride, int alpha, int beta, int8_t *tc0 )
{   
    deblock_chroma_c( pix, 1, stride, alpha, beta, tc0 );
}
Laurent Aimar's avatar
Laurent Aimar committed
467

Loren Merritt's avatar
Loren Merritt committed
468
static inline void deblock_luma_intra_c( uint8_t *pix, int xstride, int ystride, int alpha, int beta )
Laurent Aimar's avatar
Laurent Aimar committed
469
{
Loren Merritt's avatar
Loren Merritt committed
470 471 472 473 474 475 476 477 478
    int d;
    for( d = 0; d < 16; 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];

Loren Merritt's avatar
Loren Merritt committed
479 480 481
        if( abs( p0 - q0 ) < alpha &&
            abs( p1 - p0 ) < beta &&
            abs( q1 - q0 ) < beta ) {
Loren Merritt's avatar
Loren Merritt committed
482

Loren Merritt's avatar
Loren Merritt committed
483 484
            if(abs( p0 - q0 ) < ((alpha >> 2) + 2) ){
                if( abs( p2 - p0 ) < beta)
Laurent Aimar's avatar
Laurent Aimar committed
485
                {
Loren Merritt's avatar
Loren Merritt committed
486 487 488 489 490 491 492 493
                    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;
Laurent Aimar's avatar
Laurent Aimar committed
494
                }
Loren Merritt's avatar
Loren Merritt committed
495
                if( abs( q2 - q0 ) < beta)
Laurent Aimar's avatar
Laurent Aimar committed
496
                {
Loren Merritt's avatar
Loren Merritt committed
497 498 499 500 501 502 503 504
                    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;
Laurent Aimar's avatar
Laurent Aimar committed
505
                }
Loren Merritt's avatar
Loren Merritt committed
506 507 508 509
            }else{
                /* p0', q0' */
                pix[-1*xstride] = ( 2*p1 + p0 + q1 + 2 ) >> 2;
                pix[ 0*xstride] = ( 2*q1 + q0 + p1 + 2 ) >> 2;
Laurent Aimar's avatar
Laurent Aimar committed
510 511
            }
        }
Loren Merritt's avatar
Loren Merritt committed
512
        pix += ystride;
Laurent Aimar's avatar
Laurent Aimar committed
513 514
    }
}
Loren Merritt's avatar
Loren Merritt committed
515 516 517 518 519 520 521 522
static void deblock_v_luma_intra_c( uint8_t *pix, int stride, int alpha, int beta )
{   
    deblock_luma_intra_c( pix, stride, 1, alpha, beta );
}
static void deblock_h_luma_intra_c( uint8_t *pix, int stride, int alpha, int beta )
{   
    deblock_luma_intra_c( pix, 1, stride, alpha, beta );
}
Laurent Aimar's avatar
Laurent Aimar committed
523

Loren Merritt's avatar
Loren Merritt committed
524 525 526 527 528 529 530 531 532
static inline void deblock_chroma_intra_c( uint8_t *pix, int xstride, int ystride, int alpha, int beta )
{   
    int d; 
    for( d = 0; d < 8; d++ ) {
        const int p1 = pix[-2*xstride];
        const int p0 = pix[-1*xstride];
        const int q0 = pix[ 0*xstride];
        const int q1 = pix[ 1*xstride];

Loren Merritt's avatar
Loren Merritt committed
533 534 535
        if( abs( p0 - q0 ) < alpha &&
            abs( p1 - p0 ) < beta &&
            abs( q1 - q0 ) < beta ) {
Loren Merritt's avatar
Loren Merritt committed
536 537 538

            pix[-1*xstride] = (2*p1 + p0 + q1 + 2) >> 2;   /* p0' */
            pix[ 0*xstride] = (2*q1 + q0 + p1 + 2) >> 2;   /* q0' */
Laurent Aimar's avatar
Laurent Aimar committed
539 540
        }

Loren Merritt's avatar
Loren Merritt committed
541 542 543 544 545 546 547 548 549 550 551 552
        pix += ystride;
    }
}
static void deblock_v_chroma_intra_c( uint8_t *pix, int stride, int alpha, int beta )
{   
    deblock_chroma_intra_c( pix, stride, 1, alpha, beta );
}
static void deblock_h_chroma_intra_c( uint8_t *pix, int stride, int alpha, int beta )
{   
    deblock_chroma_intra_c( pix, 1, stride, alpha, beta );
}

Fiona Glaser's avatar
Fiona Glaser committed
553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577
static inline void deblock_edge( x264_t *h, uint8_t *pix1, uint8_t *pix2, int i_stride, uint8_t bS[4], int i_qp, int b_chroma, x264_deblock_inter_t pf_inter )
{
    const int index_a = x264_clip3( i_qp + h->sh.i_alpha_c0_offset, 0, 51 );
    const int alpha = i_alpha_table[index_a];
    const int beta  = i_beta_table[x264_clip3( i_qp + h->sh.i_beta_offset, 0, 51 )];
    int8_t tc[4];

    if( !alpha || !beta )
        return;

    tc[0] = i_tc0_table[index_a][bS[0]] + b_chroma;
    tc[1] = i_tc0_table[index_a][bS[1]] + b_chroma;
    tc[2] = i_tc0_table[index_a][bS[2]] + b_chroma;
    tc[3] = i_tc0_table[index_a][bS[3]] + b_chroma;

    if(b_chroma)
    {
        pf_inter( pix1, i_stride, alpha, beta, tc );
        pf_inter( pix2, i_stride, alpha, beta, tc );
    }
    else
        pf_inter( pix1, i_stride, alpha, beta, tc );
}

static inline void deblock_edge_intra( x264_t *h, uint8_t *pix1, uint8_t *pix2, int i_stride, uint8_t bS[4], int i_qp, int b_chroma, x264_deblock_intra_t pf_intra )
Loren Merritt's avatar
Loren Merritt committed
578 579 580 581 582
{
    const int index_a = x264_clip3( i_qp + h->sh.i_alpha_c0_offset, 0, 51 );
    const int alpha = i_alpha_table[index_a];
    const int beta  = i_beta_table[x264_clip3( i_qp + h->sh.i_beta_offset, 0, 51 )];

Fiona Glaser's avatar
Fiona Glaser committed
583 584 585 586 587 588 589
    if( !alpha || !beta )
        return;

    if(b_chroma)
    {
        pf_intra( pix1, i_stride, alpha, beta );
        pf_intra( pix2, i_stride, alpha, beta );
Laurent Aimar's avatar
Laurent Aimar committed
590
    }
Fiona Glaser's avatar
Fiona Glaser committed
591 592
    else
        pf_intra( pix1, i_stride, alpha, beta );
Laurent Aimar's avatar
Laurent Aimar committed
593 594
}

Loren Merritt's avatar
Loren Merritt committed
595
void x264_frame_deblock_row( x264_t *h, int mb_y )
Laurent Aimar's avatar
Laurent Aimar committed
596 597 598
{
    const int s8x8 = 2 * h->mb.i_mb_stride;
    const int s4x4 = 4 * h->mb.i_mb_stride;
Loren Merritt's avatar
Loren Merritt committed
599
    const int b_interlaced = h->sh.b_mbaff;
600
    const int mvy_limit = 4 >> b_interlaced;
Loren Merritt's avatar
Loren Merritt committed
601
    int mb_x;
Laurent Aimar's avatar
Laurent Aimar committed
602

603 604 605 606
    int i_stride2[3] = { h->fdec->i_stride[0] << b_interlaced,
                         h->fdec->i_stride[1] << b_interlaced,
                         h->fdec->i_stride[2] << b_interlaced };

607 608 609
    if( !h->pps->b_cabac && h->pps->b_transform_8x8_mode )
        munge_cavlc_nnz( h, mb_y, h->mb.nnz_backup, munge_cavlc_nnz_row );

Loren Merritt's avatar
Loren Merritt committed
610
    for( mb_x = 0; mb_x < h->sps->i_mb_width; )
Laurent Aimar's avatar
Laurent Aimar committed
611 612 613 614
    {
        const int mb_xy  = mb_y * h->mb.i_mb_stride + mb_x;
        const int mb_8x8 = 2 * s8x8 * mb_y + 2 * mb_x;
        const int mb_4x4 = 4 * s4x4 * mb_y + 4 * mb_x;
615
        const int b_8x8_transform = h->mb.mb_transform_size[mb_xy];
Loren Merritt's avatar
Loren Merritt committed
616
        const int i_edge_end = (h->mb.type[mb_xy] == P_SKIP) ? 1 : 4;
617

618 619 620 621 622 623 624 625 626 627
        int i_pix_y[3] = { 16*mb_y*h->fdec->i_stride[0] + 16*mb_x,
                            8*mb_y*h->fdec->i_stride[1] +  8*mb_x,
                            8*mb_y*h->fdec->i_stride[2] +  8*mb_x };
        if( b_interlaced && (mb_y&1) )
        {
            i_pix_y[0] -= 15*h->fdec->i_stride[0];
            i_pix_y[1] -=  7*h->fdec->i_stride[1];
            i_pix_y[2] -=  7*h->fdec->i_stride[2];
        }

628 629
        x264_prefetch_fenc( h, h->fdec, mb_x, mb_y );

Fiona Glaser's avatar
Fiona Glaser committed
630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670
        #define FILTER_DIR(intra, i_dir)\
        {\
            /* Y plane */\
            i_qp = h->mb.qp[mb_xy];\
            i_qpn= h->mb.qp[mbn_xy];\
            if( i_dir == 0 )\
            {\
                /* vertical edge */\
                deblock_edge##intra( h, &h->fdec->plane[0][i_pix_y[0] + 4*i_edge], NULL,\
                              i_stride2[0], bS, (i_qp+i_qpn+1) >> 1, 0,\
                              h->loopf.deblock_h_luma##intra );\
                if( !(i_edge & 1) )\
                {\
                    /* U/V planes */\
                    int i_qpc = ( i_chroma_qp_table[x264_clip3( i_qp + h->pps->i_chroma_qp_index_offset, 0, 51 )] +\
                                  i_chroma_qp_table[x264_clip3( i_qpn + h->pps->i_chroma_qp_index_offset, 0, 51 )] + 1 ) >> 1;\
                    deblock_edge##intra( h, &h->fdec->plane[1][i_pix_y[1] + 2*i_edge],\
                                  &h->fdec->plane[2][i_pix_y[2] + 2*i_edge],\
                                  i_stride2[1], bS, i_qpc, 1,\
                                  h->loopf.deblock_h_chroma##intra );\
                }\
            }\
            else\
            {\
                /* horizontal edge */\
                deblock_edge##intra( h, &h->fdec->plane[0][i_pix_y[0] + 4*i_edge*i_stride2[0]], NULL,\
                              i_stride2[0], bS, (i_qp+i_qpn+1) >> 1, 0,\
                              h->loopf.deblock_v_luma##intra );\
                /* U/V planes */\
                if( !(i_edge & 1) )\
                {\
                    int i_qpc = ( i_chroma_qp_table[x264_clip3( i_qp + h->pps->i_chroma_qp_index_offset, 0, 51 )] +\
                                  i_chroma_qp_table[x264_clip3( i_qpn + h->pps->i_chroma_qp_index_offset, 0, 51 )] + 1 ) >> 1;\
                    deblock_edge##intra( h, &h->fdec->plane[1][i_pix_y[1] + 2*i_edge*i_stride2[1]],\
                                  &h->fdec->plane[2][i_pix_y[2] + 2*i_edge*i_stride2[2]],\
                                  i_stride2[1], bS, i_qpc, 1,\
                                  h->loopf.deblock_v_chroma##intra );\
                }\
            }\
        }

Laurent Aimar's avatar
Laurent Aimar committed
671 672
        /* i_dir == 0 -> vertical edge
         * i_dir == 1 -> horizontal edge */
Fiona Glaser's avatar
Fiona Glaser committed
673 674
        #define deblock_dir(i_dir)\
        {\
Fiona Glaser's avatar
Fiona Glaser committed
675 676 677 678
            int i_edge = (i_dir ? (mb_y <= b_interlaced) : (mb_x == 0));\
            int i_qp, i_qpn, i, l, mbn_xy, mbn_8x8, mbn_4x4;\
            DECLARE_ALIGNED_4( uint8_t bS[4] );  /* filtering strength */\
            if( i_edge )\
Fiona Glaser's avatar
Fiona Glaser committed
679
            {\
Fiona Glaser's avatar
Fiona Glaser committed
680 681 682 683 684 685 686 687 688 689 690
                i_edge+= b_8x8_transform;\
                mbn_xy  = mb_xy;\
                mbn_8x8 = mb_8x8;\
                mbn_4x4 = mb_4x4;\
            }\
            else\
            {\
                mbn_xy  = i_dir == 0 ? mb_xy  - 1 : mb_xy - h->mb.i_mb_stride;\
                mbn_8x8 = i_dir == 0 ? mb_8x8 - 2 : mb_8x8 - 2 * s8x8;\
                mbn_4x4 = i_dir == 0 ? mb_4x4 - 4 : mb_4x4 - 4 * s4x4;\
                if( b_interlaced && i_dir == 1 )\
Fiona Glaser's avatar
Fiona Glaser committed
691 692 693 694 695
                {\
                    mbn_xy -= h->mb.i_mb_stride;\
                    mbn_8x8 -= 2 * s8x8;\
                    mbn_4x4 -= 4 * s4x4;\
                }\
Fiona Glaser's avatar
Fiona Glaser committed
696 697 698 699 700 701 702 703 704 705 706
                else if( IS_INTRA( h->mb.type[mb_xy] ) || IS_INTRA( h->mb.type[mbn_xy]) )\
                {\
                    FILTER_DIR( _intra, i_dir );\
                    i_edge += b_8x8_transform+1;\
                    mbn_xy  = mb_xy;\
                    mbn_8x8 = mb_8x8;\
                    mbn_4x4 = mb_4x4;\
                }\
            }\
            for( ; i_edge < i_edge_end; i_edge+=b_8x8_transform+1 )\
            {\
Fiona Glaser's avatar
Fiona Glaser committed
707 708
                /* *** Get bS for each 4px for the current edge *** */\
                if( IS_INTRA( h->mb.type[mb_xy] ) || IS_INTRA( h->mb.type[mbn_xy] ) )\
Fiona Glaser's avatar
Fiona Glaser committed
709
                    *(uint32_t*)bS = 0x03030303;\
Fiona Glaser's avatar
Fiona Glaser committed
710 711
                else\
                {\
Fiona Glaser's avatar
Fiona Glaser committed
712
                    *(uint32_t*)bS = 0x00000000;\
Fiona Glaser's avatar
Fiona Glaser committed
713 714 715 716 717 718
                    for( i = 0; i < 4; i++ )\
                    {\
                        int x  = i_dir == 0 ? i_edge : i;\
                        int y  = i_dir == 0 ? i      : i_edge;\
                        int xn = (x - (i_dir == 0 ? 1 : 0 ))&0x03;\
                        int yn = (y - (i_dir == 0 ? 0 : 1 ))&0x03;\
719 720
                        if( h->mb.non_zero_count[mb_xy][x+y*4] != 0 ||\
                            h->mb.non_zero_count[mbn_xy][xn+yn*4] != 0 )\
Fiona Glaser's avatar
Fiona Glaser committed
721 722 723 724 725 726 727
                            bS[i] = 2;\
                        else\
                        {\
                            /* FIXME: A given frame may occupy more than one position in\
                             * the reference list. So we should compare the frame numbers,\
                             * not the indices in the ref list.\
                             * No harm yet, as we don't generate that case.*/\
Fiona Glaser's avatar
Fiona Glaser committed
728 729
                            int i8p= mb_8x8+(x>>1)+(y>>1)*s8x8;\
                            int i8q= mbn_8x8+(xn>>1)+(yn>>1)*s8x8;\
Fiona Glaser's avatar
Fiona Glaser committed
730 731 732 733 734 735 736 737 738 739 740 741 742
                            int i4p= mb_4x4+x+y*s4x4;\
                            int i4q= mbn_4x4+xn+yn*s4x4;\
                            for( l = 0; l < 1 + (h->sh.i_type == SLICE_TYPE_B); l++ )\
                                if( h->mb.ref[l][i8p] != h->mb.ref[l][i8q] ||\
                                    abs( h->mb.mv[l][i4p][0] - h->mb.mv[l][i4q][0] ) >= 4 ||\
                                    abs( h->mb.mv[l][i4p][1] - h->mb.mv[l][i4q][1] ) >= mvy_limit )\
                                {\
                                    bS[i] = 1;\
                                    break;\
                                }\
                        }\
                    }\
                }\
Fiona Glaser's avatar
Fiona Glaser committed
743 744 745 746 747
                if( *(uint32_t*)bS )\
                    FILTER_DIR( , i_dir);\
                mbn_xy  = mb_xy;\
                mbn_8x8 = mb_8x8;\
                mbn_4x4 = mb_4x4;\
Fiona Glaser's avatar
Fiona Glaser committed
748
            }\
Laurent Aimar's avatar
Laurent Aimar committed
749 750
        }

Fiona Glaser's avatar
Fiona Glaser committed
751 752 753
        deblock_dir(0);
        deblock_dir(1);

754 755 756
        /* next mb */
        if( !b_interlaced || (mb_y&1) )
            mb_x++;
Loren Merritt's avatar
Loren Merritt committed
757
        mb_y ^= b_interlaced;
Laurent Aimar's avatar
Laurent Aimar committed
758
    }
759 760 761

    if( !h->pps->b_cabac && h->pps->b_transform_8x8_mode )
        munge_cavlc_nnz( h, mb_y, h->mb.nnz_backup, restore_cavlc_nnz_row );
Laurent Aimar's avatar
Laurent Aimar committed
762 763
}

Loren Merritt's avatar
Loren Merritt committed
764 765 766 767 768 769 770
void x264_frame_deblock( x264_t *h )
{
    int mb_y;
    for( mb_y = 0; mb_y < h->sps->i_mb_height; mb_y += 1 + h->sh.b_mbaff )
        x264_frame_deblock_row( h, mb_y );
}

771
#ifdef HAVE_MMX
Loren Merritt's avatar
Loren Merritt committed
772 773 774 775 776 777 778
void x264_deblock_v_chroma_mmxext( uint8_t *pix, int stride, int alpha, int beta, int8_t *tc0 );
void x264_deblock_h_chroma_mmxext( uint8_t *pix, int stride, int alpha, int beta, int8_t *tc0 );
void x264_deblock_v_chroma_intra_mmxext( uint8_t *pix, int stride, int alpha, int beta );
void x264_deblock_h_chroma_intra_mmxext( uint8_t *pix, int stride, int alpha, int beta );

void x264_deblock_v_luma_sse2( uint8_t *pix, int stride, int alpha, int beta, int8_t *tc0 );
void x264_deblock_h_luma_sse2( uint8_t *pix, int stride, int alpha, int beta, int8_t *tc0 );
Loren Merritt's avatar
Loren Merritt committed
779 780
void x264_deblock_v_luma_intra_sse2( uint8_t *pix, int stride, int alpha, int beta );
void x264_deblock_h_luma_intra_sse2( uint8_t *pix, int stride, int alpha, int beta );
781
#ifdef ARCH_X86
Loren Merritt's avatar
Loren Merritt committed
782 783
void x264_deblock_h_luma_mmxext( uint8_t *pix, int stride, int alpha, int beta, int8_t *tc0 );
void x264_deblock_v8_luma_mmxext( uint8_t *pix, int stride, int alpha, int beta, int8_t *tc0 );
Loren Merritt's avatar
Loren Merritt committed
784 785
void x264_deblock_h_luma_intra_mmxext( uint8_t *pix, int stride, int alpha, int beta );
void x264_deblock_v8_luma_intra_mmxext( uint8_t *pix, int stride, int alpha, int beta );
Loren Merritt's avatar
Loren Merritt committed
786 787 788 789 790 791

void x264_deblock_v_luma_mmxext( uint8_t *pix, int stride, int alpha, int beta, int8_t *tc0 )
{
    x264_deblock_v8_luma_mmxext( pix,   stride, alpha, beta, tc0   );
    x264_deblock_v8_luma_mmxext( pix+8, stride, alpha, beta, tc0+2 );
}
Loren Merritt's avatar
Loren Merritt committed
792 793 794 795 796
void x264_deblock_v_luma_intra_mmxext( uint8_t *pix, int stride, int alpha, int beta )
{
    x264_deblock_v8_luma_intra_mmxext( pix,   stride, alpha, beta );
    x264_deblock_v8_luma_intra_mmxext( pix+8, stride, alpha, beta );
}
Loren Merritt's avatar
Loren Merritt committed
797
#endif
Loren Merritt's avatar
Loren Merritt committed
798
#endif
Laurent Aimar's avatar
Laurent Aimar committed
799

800 801 802 803 804
#ifdef ARCH_PPC
void x264_deblock_v_luma_altivec( uint8_t *pix, int stride, int alpha, int beta, int8_t *tc0 );
void x264_deblock_h_luma_altivec( uint8_t *pix, int stride, int alpha, int beta, int8_t *tc0 );
#endif // ARCH_PPC

Loren Merritt's avatar
Loren Merritt committed
805 806 807 808 809 810 811 812 813 814 815
void x264_deblock_init( int cpu, x264_deblock_function_t *pf )
{
    pf->deblock_v_luma = deblock_v_luma_c;
    pf->deblock_h_luma = deblock_h_luma_c;
    pf->deblock_v_chroma = deblock_v_chroma_c;
    pf->deblock_h_chroma = deblock_h_chroma_c;
    pf->deblock_v_luma_intra = deblock_v_luma_intra_c;
    pf->deblock_h_luma_intra = deblock_h_luma_intra_c;
    pf->deblock_v_chroma_intra = deblock_v_chroma_intra_c;
    pf->deblock_h_chroma_intra = deblock_h_chroma_intra_c;

816
#ifdef HAVE_MMX
Loren Merritt's avatar
Loren Merritt committed
817 818 819 820 821 822
    if( cpu&X264_CPU_MMXEXT )
    {
        pf->deblock_v_chroma = x264_deblock_v_chroma_mmxext;
        pf->deblock_h_chroma = x264_deblock_h_chroma_mmxext;
        pf->deblock_v_chroma_intra = x264_deblock_v_chroma_intra_mmxext;
        pf->deblock_h_chroma_intra = x264_deblock_h_chroma_intra_mmxext;
823 824 825
#ifdef ARCH_X86
        pf->deblock_v_luma = x264_deblock_v_luma_mmxext;
        pf->deblock_h_luma = x264_deblock_h_luma_mmxext;
Loren Merritt's avatar
Loren Merritt committed
826 827
        pf->deblock_v_luma_intra = x264_deblock_v_luma_intra_mmxext;
        pf->deblock_h_luma_intra = x264_deblock_h_luma_intra_mmxext;
828
#endif
829
        if( (cpu&X264_CPU_SSE2) && !(cpu&X264_CPU_STACK_MOD4) )
Loren Merritt's avatar
Loren Merritt committed
830 831 832
        {
            pf->deblock_v_luma = x264_deblock_v_luma_sse2;
            pf->deblock_h_luma = x264_deblock_h_luma_sse2;
Loren Merritt's avatar
Loren Merritt committed
833 834
            pf->deblock_v_luma_intra = x264_deblock_v_luma_intra_sse2;
            pf->deblock_h_luma_intra = x264_deblock_h_luma_intra_sse2;
Loren Merritt's avatar
Loren Merritt committed
835 836
        }
    }
837
#endif
838 839 840 841 842 843 844 845

#ifdef ARCH_PPC
    if( cpu&X264_CPU_ALTIVEC )
    {
        pf->deblock_v_luma = x264_deblock_v_luma_altivec;
        pf->deblock_h_luma = x264_deblock_h_luma_altivec;
   }
#endif // ARCH_PPC
Loren Merritt's avatar
Loren Merritt committed
846
}
Laurent Aimar's avatar
Laurent Aimar committed
847

848 849

/* threading */
850
void x264_frame_cond_broadcast( x264_frame_t *frame, int i_lines_completed )
851
{
852
    x264_pthread_mutex_lock( &frame->mutex );
853
    frame->i_lines_completed = i_lines_completed;
854 855
    x264_pthread_cond_broadcast( &frame->cv );
    x264_pthread_mutex_unlock( &frame->mutex );
856 857 858 859
}

void x264_frame_cond_wait( x264_frame_t *frame, int i_lines_completed )
{
860
    x264_pthread_mutex_lock( &frame->mutex );
861
    while( frame->i_lines_completed < i_lines_completed )
862 863
        x264_pthread_cond_wait( &frame->cv, &frame->mutex );
    x264_pthread_mutex_unlock( &frame->mutex );
864 865
}

866 867 868 869 870 871 872 873 874 875 876 877 878 879 880 881 882 883 884 885 886 887 888 889 890 891 892 893 894 895 896 897 898 899 900 901 902 903 904 905 906 907 908 909 910 911 912 913 914 915 916 917 918 919 920 921 922 923 924 925 926 927 928 929 930 931 932 933 934 935 936 937 938 939 940 941 942 943 944
/* list operators */

void x264_frame_push( x264_frame_t **list, x264_frame_t *frame )
{
    int i = 0;
    while( list[i] ) i++;
    list[i] = frame;
}

x264_frame_t *x264_frame_pop( x264_frame_t **list )
{
    x264_frame_t *frame;
    int i = 0;
    assert( list[0] );
    while( list[i+1] ) i++;
    frame = list[i];
    list[i] = NULL;
    return frame;
}

void x264_frame_unshift( x264_frame_t **list, x264_frame_t *frame )
{
    int i = 0;
    while( list[i] ) i++;
    while( i-- )
        list[i+1] = list[i];
    list[0] = frame;
}

x264_frame_t *x264_frame_shift( x264_frame_t **list )
{
    x264_frame_t *frame = list[0];
    int i;
    for( i = 0; list[i]; i++ )
        list[i] = list[i+1];
    assert(frame);
    return frame;
}

void x264_frame_push_unused( x264_t *h, x264_frame_t *frame )
{
    assert( frame->i_reference_count > 0 );
    frame->i_reference_count--;
    if( frame->i_reference_count == 0 )
        x264_frame_push( h->frames.unused, frame );
    assert( h->frames.unused[ sizeof(h->frames.unused) / sizeof(*h->frames.unused) - 1 ] == NULL );
}

x264_frame_t *x264_frame_pop_unused( x264_t *h )
{
    x264_frame_t *frame;
    if( h->frames.unused[0] )
        frame = x264_frame_pop( h->frames.unused );
    else
        frame = x264_frame_new( h );
    assert( frame->i_reference_count == 0 );
    frame->i_reference_count = 1;
    return frame;
}

void x264_frame_sort( x264_frame_t **list, int b_dts )
{
    int i, b_ok;
    do {
        b_ok = 1;
        for( i = 0; list[i+1]; i++ )
        {
            int dtype = list[i]->i_type - list[i+1]->i_type;
            int dtime = list[i]->i_frame - list[i+1]->i_frame;
            int swap = b_dts ? dtype > 0 || ( dtype == 0 && dtime > 0 )
                             : dtime > 0;
            if( swap )
            {
                XCHG( x264_frame_t*, list[i], list[i+1] );
                b_ok = 0;
            }
        }
    } while( !b_ok );
}