frame.c 33.5 KB
Newer Older
Laurent Aimar's avatar
Laurent Aimar committed
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27
/*****************************************************************************
 * frame.c: h264 encoder library
 *****************************************************************************
 * Copyright (C) 2003 Laurent Aimar
 * $Id: frame.c,v 1.1 2004/06/03 19:27:06 fenrir Exp $
 *
 * Authors: Laurent Aimar <fenrir@via.ecp.fr>
 *
 * 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
 * Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA  02111, USA.
 *****************************************************************************/

#include "common.h"

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

31
    int i_mb_count = h->mb.i_mb_count;
32
    int i_stride, i_width, i_lines;
33
    int i_padv = PADV << h->param.b_interlaced;
34
    int luma_plane_size;
Laurent Aimar's avatar
Laurent Aimar committed
35

36 37
    if( !frame ) return NULL;

38 39
    memset( frame, 0, sizeof(x264_frame_t) );

Laurent Aimar's avatar
Laurent Aimar committed
40
    /* allocate frame data (+64 for extra data for me) */
41 42
    i_width  = ( ( h->param.i_width  + 15 ) & -16 );
    i_stride = i_width + 2*PADH;
43 44 45
    i_lines  = ( ( h->param.i_height + 15 ) & -16 );
    if( h->param.b_interlaced )
        i_lines = ( i_lines + 31 ) & -32;
Laurent Aimar's avatar
Laurent Aimar committed
46

47 48 49 50
    if( h->param.cpu&X264_CPU_CACHELINE_64 )
        i_stride = (i_stride + 63) & ~63;
    else if( h->param.cpu&X264_CPU_CACHELINE_32 )
        i_stride = (i_stride + 31) & ~31;
51

Laurent Aimar's avatar
Laurent Aimar committed
52 53 54
    frame->i_plane = 3;
    for( i = 0; i < 3; i++ )
    {
55 56 57
        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
58 59
    }

60 61
    luma_plane_size = (frame->i_stride[0] * ( frame->i_lines[0] + 2*i_padv ));
    for( i = 1; i < 3; i++ )
62
    {
63
        CHECKED_MALLOC( frame->buffer[i], luma_plane_size/4 );
64
        frame->plane[i] = (uint8_t*)frame->buffer[i] + (frame->i_stride[i] * i_padv + PADH)/2;
65
    }
66 67 68 69
    /* 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++ )
70
        frame->filtered[i] = (uint8_t*)frame->buffer[0] + i*luma_plane_size + frame->i_stride[0] * i_padv + PADH;
71
    frame->plane[0] = frame->filtered[0];
72

73
    if( h->frames.b_have_lowres )
74
    {
75 76
        frame->i_width_lowres = frame->i_width[0]/2;
        frame->i_stride_lowres = frame->i_width_lowres + 2*PADH;
77 78 79
        frame->i_lines_lowres = frame->i_lines[0]/2;
        for( i = 0; i < 4; i++ )
        {
80 81 82 83
            CHECKED_MALLOC( frame->buffer_lowres[i],
                            frame->i_stride_lowres * ( frame->i_lines[0]/2 + 2*i_padv ) );
            frame->lowres[i] = ((uint8_t*)frame->buffer_lowres[i]) +
                                frame->i_stride_lowres * i_padv + PADH;
84
        }
85
    }
86

87
    if( h->param.analyse.i_me_method >= X264_ME_ESA )
88
    {
89
        CHECKED_MALLOC( frame->buffer[3],
90
                        2 * frame->i_stride[0] * (frame->i_lines[0] + 2*i_padv) * sizeof(uint16_t) );
91
        frame->integral = (uint16_t*)frame->buffer[3] + frame->i_stride[0] * i_padv + PADH;
92 93
    }

Laurent Aimar's avatar
Laurent Aimar committed
94 95 96
    frame->i_poc = -1;
    frame->i_type = X264_TYPE_AUTO;
    frame->i_qpplus1 = 0;
97 98
    frame->i_pts = -1;
    frame->i_frame = -1;
99
    frame->i_frame_num = -1;
Loren Merritt's avatar
Loren Merritt committed
100
    frame->i_lines_completed = -1;
Laurent Aimar's avatar
Laurent Aimar committed
101

102 103 104
    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) );
105 106
    if( h->param.i_bframe )
    {
107 108
        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) );
109 110 111 112 113 114 115
    }
    else
    {
        frame->mv[1]  = NULL;
        frame->ref[1] = NULL;
    }

116 117
    CHECKED_MALLOC( frame->i_row_bits, i_lines/16 * sizeof(int) );
    CHECKED_MALLOC( frame->i_row_qp, i_lines/16 * sizeof(int) );
118 119
    for( i = 0; i < h->param.i_bframe + 2; i++ )
        for( j = 0; j < h->param.i_bframe + 2; j++ )
120
            CHECKED_MALLOC( frame->i_row_satds[i][j], i_lines/16 * sizeof(int) );
121

122 123
    x264_pthread_mutex_init( &frame->mutex, NULL );
    x264_pthread_cond_init( &frame->cv, NULL );
124

Laurent Aimar's avatar
Laurent Aimar committed
125
    return frame;
126 127 128 129

fail:
    x264_frame_delete( frame );
    return NULL;
Laurent Aimar's avatar
Laurent Aimar committed
130 131 132 133
}

void x264_frame_delete( x264_frame_t *frame )
{
134
    int i, j;
135
    for( i = 0; i < 4; i++ )
136
        x264_free( frame->buffer[i] );
137 138
    for( i = 0; i < 4; i++ )
        x264_free( frame->buffer_lowres[i] );
139 140 141 142 143
    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 );
144
    x264_free( frame->mb_type );
145 146 147 148
    x264_free( frame->mv[0] );
    x264_free( frame->mv[1] );
    x264_free( frame->ref[0] );
    x264_free( frame->ref[1] );
149 150
    x264_pthread_mutex_destroy( &frame->mutex );
    x264_pthread_cond_destroy( &frame->cv );
Laurent Aimar's avatar
Laurent Aimar committed
151 152 153
    x264_free( frame );
}

154
int x264_frame_copy_picture( x264_t *h, x264_frame_t *dst, x264_picture_t *src )
Laurent Aimar's avatar
Laurent Aimar committed
155
{
156
    int i_csp = src->img.i_csp & X264_CSP_MASK;
157 158 159 160 161 162 163
    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
164 165 166 167
    dst->i_type     = src->i_type;
    dst->i_qpplus1  = src->i_qpplus1;
    dst->i_pts      = src->i_pts;

168 169 170 171 172 173 174 175 176 177 178 179 180 181 182
    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
183 184 185 186
}



Loren Merritt's avatar
Loren Merritt committed
187
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 )
188 189 190 191 192 193
{
#define PPIXEL(x, y) ( pix + (x) + (y)*i_stride )
    int y;
    for( y = 0; y < i_height; y++ )
    {
        /* left band */
194
        memset( PPIXEL(-i_padh, y), PPIXEL(0, y)[0], i_padh );
195
        /* right band */
196
        memset( PPIXEL(i_width, y), PPIXEL(i_width-1, y)[0], i_padh );
197 198
    }
    /* upper band */
Loren Merritt's avatar
Loren Merritt committed
199
    if( b_pad_top )
200 201
    for( y = 0; y < i_padv; y++ )
        memcpy( PPIXEL(-i_padh, -y-1), PPIXEL(-i_padh, 0), i_width+2*i_padh );
202
    /* lower band */
Loren Merritt's avatar
Loren Merritt committed
203
    if( b_pad_bottom )
204 205
    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 );
206 207 208
#undef PPIXEL
}

Loren Merritt's avatar
Loren Merritt committed
209
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
210
{
211
    int i;
Loren Merritt's avatar
Loren Merritt committed
212 213 214
    int b_start = !mb_y;
    if( mb_y & h->sh.b_mbaff )
        return;
Laurent Aimar's avatar
Laurent Aimar committed
215 216
    for( i = 0; i < frame->i_plane; i++ )
    {
217 218
        int stride = frame->i_stride[i];
        int width = 16*h->sps->i_mb_width >> !!i;
Loren Merritt's avatar
Loren Merritt committed
219
        int height = (b_end ? 16*(h->sps->i_mb_height - mb_y) >> h->sh.b_mbaff : 16) >> !!i;
220 221
        int padh = PADH >> !!i;
        int padv = PADV >> !!i;
Loren Merritt's avatar
Loren Merritt committed
222 223 224 225 226
        // 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 )
227
        {
Loren Merritt's avatar
Loren Merritt committed
228 229
            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 );
230 231 232
        }
        else
        {
Loren Merritt's avatar
Loren Merritt committed
233
            plane_expand_border( pix, stride, width, height, padh, padv, b_start, b_end );
234
        }
Laurent Aimar's avatar
Laurent Aimar committed
235 236 237
    }
}

Loren Merritt's avatar
Loren Merritt committed
238
void x264_frame_expand_border_filtered( x264_t *h, x264_frame_t *frame, int mb_y, int b_end )
239
{
Loren Merritt's avatar
Loren Merritt committed
240 241
    /* during filtering, 8 extra pixels were filtered on each edge,
     * but up to 3 of the horizontal ones may be wrong. 
242
       we want to expand border from the last filtered pixel */
Loren Merritt's avatar
Loren Merritt committed
243
    int b_start = !mb_y;
244
    int stride = frame->i_stride[0];
Loren Merritt's avatar
Loren Merritt committed
245
    int width = 16*h->sps->i_mb_width + 8;
Loren Merritt's avatar
Loren Merritt committed
246
    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
247
    int padh = PADH - 4;
248
    int padv = PADV - 8;
249
    int i;
250
    for( i = 1; i < 4; i++ )
251
    {
Loren Merritt's avatar
Loren Merritt committed
252
        // buffer: 8 luma, to match the hpel filter
Loren Merritt's avatar
Loren Merritt committed
253
        uint8_t *pix = frame->filtered[i] + (16*mb_y - (8 << h->sh.b_mbaff)) * stride - 4;
Loren Merritt's avatar
Loren Merritt committed
254
        if( h->sh.b_mbaff )
255
        {
Loren Merritt's avatar
Loren Merritt committed
256 257
            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 );
258 259 260
        }
        else
        {
Loren Merritt's avatar
Loren Merritt committed
261
            plane_expand_border( pix, stride, width, height, padh, padv, b_start, b_end );
262 263
        }
    }
264 265
}

266
void x264_frame_expand_border_lowres( x264_frame_t *frame )
267
{
268
    int i;
269
    for( i = 0; i < 4; i++ )
Loren Merritt's avatar
Loren Merritt committed
270
        plane_expand_border( frame->lowres[i], frame->i_stride_lowres, frame->i_stride_lowres - 2*PADH, frame->i_lines_lowres, PADH, PADV, 1, 1 );
271 272
}

Loren Merritt's avatar
Loren Merritt committed
273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292
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 )
        {
293
            //FIXME interlace? or just let it pad using the wrong field
Loren Merritt's avatar
Loren Merritt committed
294
            for( y = i_height; y < i_height + i_pady; y++ )
Loren Merritt's avatar
Loren Merritt committed
295 296 297 298 299 300 301
                memcpy( &frame->plane[i][y*frame->i_stride[i]],
                        &frame->plane[i][(i_height-1)*frame->i_stride[i]],
                        i_width + i_padx );
        }
    }
}

302

303 304 305 306 307 308
/* 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;
309
    int x, nnz;
310 311 312 313 314
    for( x=0; x<h->sps->i_mb_width; x++ )
    {
        memcpy( buf+x, src+x, 16 );
        if( transform[x] )
        {
315 316 317 318
            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);
319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344
        }
    }
}

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 );
    }
}


345
/* Deblocking filter */
Laurent Aimar's avatar
Laurent Aimar committed
346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386

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
};
static const int i_tc0_table[52][3] =
{
    { 0, 0, 0 }, { 0, 0, 0 }, { 0, 0, 0 }, { 0, 0, 0 }, { 0, 0, 0 }, { 0, 0, 0 },
    { 0, 0, 0 }, { 0, 0, 0 }, { 0, 0, 0 }, { 0, 0, 0 }, { 0, 0, 0 }, { 0, 0, 0 },
    { 0, 0, 0 }, { 0, 0, 0 }, { 0, 0, 0 }, { 0, 0, 0 }, { 0, 0, 0 }, { 0, 0, 1 },
    { 0, 0, 1 }, { 0, 0, 1 }, { 0, 0, 1 }, { 0, 1, 1 }, { 0, 1, 1 }, { 1, 1, 1 },
    { 1, 1, 1 }, { 1, 1, 1 }, { 1, 1, 1 }, { 1, 1, 2 }, { 1, 1, 2 }, { 1, 1, 2 },
    { 1, 1, 2 }, { 1, 2, 3 }, { 1, 2, 3 }, { 2, 2, 3 }, { 2, 2, 4 }, { 2, 3, 4 },
    { 2, 3, 4 }, { 3, 3, 5 }, { 3, 4, 6 }, { 3, 4, 6 }, { 4, 5, 7 }, { 4, 5, 8 },
    { 4, 6, 9 }, { 5, 7,10 }, { 6, 8,11 }, { 6, 8,13 }, { 7,10,14 }, { 8,11,16 },
    { 9,12,18 }, {10,13,20 }, {11,15,23 }, {13,17,25 }
};

/* 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
387
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
388 389
{
    int i, d;
Loren Merritt's avatar
Loren Merritt committed
390 391 392
    for( i = 0; i < 4; i++ ) {
        if( tc0[i] < 0 ) {
            pix += 4*ystride;
Laurent Aimar's avatar
Laurent Aimar committed
393 394
            continue;
        }
Loren Merritt's avatar
Loren Merritt committed
395 396 397 398 399 400 401 402
        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
403 404 405
            if( abs( p0 - q0 ) < alpha &&
                abs( p1 - p0 ) < beta &&
                abs( q1 - q0 ) < beta ) {
Loren Merritt's avatar
Loren Merritt committed
406 407 408 409
   
                int tc = tc0[i];
                int delta;
   
Loren Merritt's avatar
Loren Merritt committed
410
                if( abs( p2 - p0 ) < beta ) {
Loren Merritt's avatar
Loren Merritt committed
411 412
                    pix[-2*xstride] = p1 + x264_clip3( (( p2 + ((p0 + q0 + 1) >> 1)) >> 1) - p1, -tc0[i], tc0[i] );
                    tc++; 
Laurent Aimar's avatar
Laurent Aimar committed
413
                }
Loren Merritt's avatar
Loren Merritt committed
414
                if( abs( q2 - q0 ) < beta ) {
Loren Merritt's avatar
Loren Merritt committed
415 416
                    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
417
                }
Loren Merritt's avatar
Loren Merritt committed
418 419 420 421
    
                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
422
            }
Loren Merritt's avatar
Loren Merritt committed
423
            pix += ystride;
Laurent Aimar's avatar
Laurent Aimar committed
424 425 426
        }
    }
}
Loren Merritt's avatar
Loren Merritt committed
427 428 429 430 431 432 433 434
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
435

Loren Merritt's avatar
Loren Merritt committed
436
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
437 438
{
    int i, d;
Loren Merritt's avatar
Loren Merritt committed
439 440 441 442
    for( i = 0; i < 4; i++ ) {
        const int tc = tc0[i];
        if( tc <= 0 ) {
            pix += 2*ystride;
Laurent Aimar's avatar
Laurent Aimar committed
443 444
            continue;
        }
Loren Merritt's avatar
Loren Merritt committed
445 446 447 448 449 450
        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
451 452 453
            if( abs( p0 - q0 ) < alpha &&
                abs( p1 - p0 ) < beta &&
                abs( q1 - q0 ) < beta ) {
Loren Merritt's avatar
Loren Merritt committed
454 455 456 457

                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
458
            }
Loren Merritt's avatar
Loren Merritt committed
459
            pix += ystride;
Laurent Aimar's avatar
Laurent Aimar committed
460 461 462
        }
    }
}
Loren Merritt's avatar
Loren Merritt committed
463 464 465 466 467 468 469 470
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
471

Loren Merritt's avatar
Loren Merritt committed
472
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
473
{
Loren Merritt's avatar
Loren Merritt committed
474 475 476 477 478 479 480 481 482
    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
483 484 485
        if( abs( p0 - q0 ) < alpha &&
            abs( p1 - p0 ) < beta &&
            abs( q1 - q0 ) < beta ) {
Loren Merritt's avatar
Loren Merritt committed
486

Loren Merritt's avatar
Loren Merritt committed
487 488
            if(abs( p0 - q0 ) < ((alpha >> 2) + 2) ){
                if( abs( p2 - p0 ) < beta)
Laurent Aimar's avatar
Laurent Aimar committed
489
                {
Loren Merritt's avatar
Loren Merritt committed
490 491 492 493 494 495 496 497
                    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
498
                }
Loren Merritt's avatar
Loren Merritt committed
499
                if( abs( q2 - q0 ) < beta)
Laurent Aimar's avatar
Laurent Aimar committed
500
                {
Loren Merritt's avatar
Loren Merritt committed
501 502 503 504 505 506 507 508
                    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
509
                }
Loren Merritt's avatar
Loren Merritt committed
510 511 512 513
            }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
514 515
            }
        }
Loren Merritt's avatar
Loren Merritt committed
516
        pix += ystride;
Laurent Aimar's avatar
Laurent Aimar committed
517 518
    }
}
Loren Merritt's avatar
Loren Merritt committed
519 520 521 522 523 524 525 526
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
527

Loren Merritt's avatar
Loren Merritt committed
528 529 530 531 532 533 534 535 536
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
537 538 539
        if( abs( p0 - q0 ) < alpha &&
            abs( p1 - p0 ) < beta &&
            abs( q1 - q0 ) < beta ) {
Loren Merritt's avatar
Loren Merritt committed
540 541 542

            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
543 544
        }

Loren Merritt's avatar
Loren Merritt committed
545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571
        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 );
}

static inline void deblock_edge( x264_t *h, uint8_t *pix, int i_stride, int bS[4], int i_qp, int b_chroma,
                                 x264_deblock_inter_t pf_inter, x264_deblock_intra_t pf_intra )
{
    int i;
    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 )];

    if( bS[0] < 4 ) {
        int8_t tc[4]; 
        for(i=0; i<4; i++)
            tc[i] = (bS[i] ? i_tc0_table[index_a][bS[i] - 1] : -1) + b_chroma;
        pf_inter( pix, i_stride, alpha, beta, tc );
    } else {
        pf_intra( pix, i_stride, alpha, beta );
Laurent Aimar's avatar
Laurent Aimar committed
572 573 574
    }
}

Loren Merritt's avatar
Loren Merritt committed
575
void x264_frame_deblock_row( x264_t *h, int mb_y )
Laurent Aimar's avatar
Laurent Aimar committed
576 577 578
{
    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
579
    const int b_interlaced = h->sh.b_mbaff;
580
    const int mvy_limit = 4 >> b_interlaced;
Loren Merritt's avatar
Loren Merritt committed
581
    int mb_x;
Laurent Aimar's avatar
Laurent Aimar committed
582

583 584 585 586
    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 };

587 588 589
    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
590
    for( mb_x = 0; mb_x < h->sps->i_mb_width; )
Laurent Aimar's avatar
Laurent Aimar committed
591 592 593 594
    {
        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;
595
        const int b_8x8_transform = h->mb.mb_transform_size[mb_xy];
Loren Merritt's avatar
Loren Merritt committed
596
        const int i_edge_end = (h->mb.type[mb_xy] == P_SKIP) ? 1 : 4;
Fiona Glaser's avatar
Fiona Glaser committed
597
        int i_edge;
598

599 600 601 602 603 604 605 606 607 608
        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];
        }

609 610
        x264_prefetch_fenc( h, h->fdec, mb_x, mb_y );

Laurent Aimar's avatar
Laurent Aimar committed
611 612
        /* i_dir == 0 -> vertical edge
         * i_dir == 1 -> horizontal edge */
613

Fiona Glaser's avatar
Fiona Glaser committed
614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644
        #define deblock_dir(i_dir)\
        {\
            int i_start = (i_dir ? (mb_y <= b_interlaced) : (mb_x == 0));\
            int i_qp, i_qpn;\
            for( i_edge = i_start; i_edge < i_edge_end; i_edge++ )\
            {\
                int mbn_xy, mbn_8x8, mbn_4x4;\
                int bS[4];  /* filtering strength */\
                if( b_8x8_transform && (i_edge&1) )\
                    continue;\
                mbn_xy  = i_edge > 0 ? mb_xy  : ( i_dir == 0 ? mb_xy  - 1 : mb_xy - h->mb.i_mb_stride );\
                mbn_8x8 = i_edge > 0 ? mb_8x8 : ( i_dir == 0 ? mb_8x8 - 2 : mb_8x8 - 2 * s8x8 );\
                mbn_4x4 = i_edge > 0 ? mb_4x4 : ( i_dir == 0 ? mb_4x4 - 4 : mb_4x4 - 4 * s4x4 );\
                if( b_interlaced && i_edge == 0 && i_dir == 1 )\
                {\
                    mbn_xy -= h->mb.i_mb_stride;\
                    mbn_8x8 -= 2 * s8x8;\
                    mbn_4x4 -= 4 * s4x4;\
                }\
                /* *** Get bS for each 4px for the current edge *** */\
                if( IS_INTRA( h->mb.type[mb_xy] ) || IS_INTRA( h->mb.type[mbn_xy] ) )\
                    bS[0] = bS[1] = bS[2] = bS[3] = ( i_edge == 0 && !(b_interlaced && i_dir) ? 4 : 3 );\
                else\
                {\
                    int i;\
                    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;\
645 646
                        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
647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717
                        {\
                            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.*/\
                            int i8p= mb_8x8+(x/2)+(y/2)*s8x8;\
                            int i8q= mbn_8x8+(xn/2)+(yn/2)*s8x8;\
                            int i4p= mb_4x4+x+y*s4x4;\
                            int i4q= mbn_4x4+xn+yn*s4x4;\
                            int l;\
                            bS[i] = 0;\
                            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;\
                                }\
                            }\
                        }\
                    }\
                }\
                /* *** filter *** */\
                /* Y plane */\
                i_qp = h->mb.qp[mb_xy];\
                i_qpn= h->mb.qp[mbn_xy];\
                if( i_dir == 0 )\
                {\
                    /* vertical edge */\
                    deblock_edge( h, &h->fdec->plane[0][i_pix_y[0] + 4*i_edge],\
                                  i_stride2[0], bS, (i_qp+i_qpn+1) >> 1, 0,\
                                  h->loopf.deblock_h_luma, 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( h, &h->fdec->plane[1][i_pix_y[1] + 2*i_edge],\
                                      i_stride2[1], bS, i_qpc, 1,\
                                      h->loopf.deblock_h_chroma, h->loopf.deblock_h_chroma_intra );\
                        deblock_edge( h, &h->fdec->plane[2][i_pix_y[2] + 2*i_edge],\
                                      i_stride2[2], bS, i_qpc, 1,\
                                      h->loopf.deblock_h_chroma, h->loopf.deblock_h_chroma_intra );\
                    }\
                }\
                else\
                {\
                    /* horizontal edge */\
                    deblock_edge( h, &h->fdec->plane[0][i_pix_y[0] + 4*i_edge*i_stride2[0]],\
                                  i_stride2[0], bS, (i_qp+i_qpn+1) >> 1, 0,\
                                  h->loopf.deblock_v_luma, 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( h, &h->fdec->plane[1][i_pix_y[1] + 2*i_edge*i_stride2[1]],\
                                      i_stride2[1], bS, i_qpc, 1,\
                                      h->loopf.deblock_v_chroma, h->loopf.deblock_v_chroma_intra );\
                        deblock_edge( h, &h->fdec->plane[2][i_pix_y[2] + 2*i_edge*i_stride2[2]],\
                                      i_stride2[2], bS, i_qpc, 1,\
                                      h->loopf.deblock_v_chroma, h->loopf.deblock_v_chroma_intra );\
                    }\
                }\
            }\
Laurent Aimar's avatar
Laurent Aimar committed
718 719
        }

Fiona Glaser's avatar
Fiona Glaser committed
720 721 722
        deblock_dir(0);
        deblock_dir(1);

723 724 725
        /* next mb */
        if( !b_interlaced || (mb_y&1) )
            mb_x++;
Loren Merritt's avatar
Loren Merritt committed
726
        mb_y ^= b_interlaced;
Laurent Aimar's avatar
Laurent Aimar committed
727
    }
728 729 730

    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
731 732
}

Loren Merritt's avatar
Loren Merritt committed
733 734 735 736 737 738 739
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 );
}

740
#ifdef HAVE_MMX
Loren Merritt's avatar
Loren Merritt committed
741 742 743 744 745 746 747
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
748 749
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 );
750
#ifdef ARCH_X86
Loren Merritt's avatar
Loren Merritt committed
751 752
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
753 754
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
755 756 757 758 759 760

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
761 762 763 764 765
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
766
#endif
Loren Merritt's avatar
Loren Merritt committed
767
#endif
Laurent Aimar's avatar
Laurent Aimar committed
768

769 770 771 772 773
#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
774 775 776 777 778 779 780 781 782 783 784
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;

785
#ifdef HAVE_MMX
Loren Merritt's avatar
Loren Merritt committed
786 787 788 789 790 791
    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;
792 793 794
#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
795 796
        pf->deblock_v_luma_intra = x264_deblock_v_luma_intra_mmxext;
        pf->deblock_h_luma_intra = x264_deblock_h_luma_intra_mmxext;
797
#endif
798
        if( (cpu&X264_CPU_SSE2) && !(cpu&X264_CPU_STACK_MOD4) )
Loren Merritt's avatar
Loren Merritt committed
799 800 801
        {
            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
802 803
            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
804 805
        }
    }
806
#endif
807 808 809 810 811 812 813 814

#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
815
}
Laurent Aimar's avatar
Laurent Aimar committed
816

817 818

/* threading */
819
void x264_frame_cond_broadcast( x264_frame_t *frame, int i_lines_completed )
820
{
821
    x264_pthread_mutex_lock( &frame->mutex );
822
    frame->i_lines_completed = i_lines_completed;
823 824
    x264_pthread_cond_broadcast( &frame->cv );
    x264_pthread_mutex_unlock( &frame->mutex );
825 826 827 828
}

void x264_frame_cond_wait( x264_frame_t *frame, int i_lines_completed )
{
829
    x264_pthread_mutex_lock( &frame->mutex );
830
    while( frame->i_lines_completed < i_lines_completed )
831 832
        x264_pthread_cond_wait( &frame->cv, &frame->mutex );
    x264_pthread_mutex_unlock( &frame->mutex );
833 834
}

835 836 837 838 839 840 841 842 843 844 845 846 847 848 849 850 851 852 853 854 855 856 857 858 859 860 861 862 863 864 865 866 867 868 869 870 871 872 873 874 875 876 877 878 879 880 881 882 883 884 885 886 887 888 889 890 891 892 893 894 895 896 897 898 899 900 901 902 903 904 905 906 907 908 909 910 911 912 913
/* 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 );
}