yading@10: /*
yading@10:  * Copyright (C) 2004-2010 Michael Niedermayer <michaelni@gmx.at>
yading@10:  * Copyright (C) 2008 David Conrad
yading@10:  *
yading@10:  * This file is part of FFmpeg.
yading@10:  *
yading@10:  * FFmpeg is free software; you can redistribute it and/or
yading@10:  * modify it under the terms of the GNU Lesser General Public
yading@10:  * License as published by the Free Software Foundation; either
yading@10:  * version 2.1 of the License, or (at your option) any later version.
yading@10:  *
yading@10:  * FFmpeg is distributed in the hope that it will be useful,
yading@10:  * but WITHOUT ANY WARRANTY; without even the implied warranty of
yading@10:  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
yading@10:  * Lesser General Public License for more details.
yading@10:  *
yading@10:  * You should have received a copy of the GNU Lesser General Public
yading@10:  * License along with FFmpeg; if not, write to the Free Software
yading@10:  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
yading@10:  */
yading@10: 
yading@10: #include "libavutil/attributes.h"
yading@10: #include "libavutil/avassert.h"
yading@10: #include "libavutil/common.h"
yading@10: #include "dsputil.h"
yading@10: #include "dirac_dwt.h"
yading@10: #include "libavcodec/x86/dirac_dwt.h"
yading@10: 
yading@10: 
yading@10: static inline int mirror(int v, int m)
yading@10: {
yading@10:     while ((unsigned)v > (unsigned)m) {
yading@10:         v = -v;
yading@10:         if (v < 0)
yading@10:             v += 2 * m;
yading@10:     }
yading@10:     return v;
yading@10: }
yading@10: 
yading@10: static void vertical_compose53iL0(IDWTELEM *b0, IDWTELEM *b1, IDWTELEM *b2,
yading@10:                                   int width)
yading@10: {
yading@10:     int i;
yading@10: 
yading@10:     for (i = 0; i < width; i++)
yading@10:         b1[i] -= (b0[i] + b2[i] + 2) >> 2;
yading@10: }
yading@10: 
yading@10: 
yading@10: static av_always_inline
yading@10: void interleave(IDWTELEM *dst, IDWTELEM *src0, IDWTELEM *src1, int w2, int add, int shift)
yading@10: {
yading@10:     int i;
yading@10:     for (i = 0; i < w2; i++) {
yading@10:         dst[2*i  ] = (src0[i] + add) >> shift;
yading@10:         dst[2*i+1] = (src1[i] + add) >> shift;
yading@10:     }
yading@10: }
yading@10: 
yading@10: static void horizontal_compose_dirac53i(IDWTELEM *b, IDWTELEM *temp, int w)
yading@10: {
yading@10:     const int w2 = w >> 1;
yading@10:     int x;
yading@10: 
yading@10:     temp[0] = COMPOSE_53iL0(b[w2], b[0], b[w2]);
yading@10:     for (x = 1; x < w2; x++) {
yading@10:         temp[x     ] = COMPOSE_53iL0     (b[x+w2-1], b[x     ], b[x+w2]);
yading@10:         temp[x+w2-1] = COMPOSE_DIRAC53iH0(temp[x-1], b[x+w2-1], temp[x]);
yading@10:     }
yading@10:     temp[w-1] = COMPOSE_DIRAC53iH0(temp[w2-1], b[w-1], temp[w2-1]);
yading@10: 
yading@10:     interleave(b, temp, temp+w2, w2, 1, 1);
yading@10: }
yading@10: 
yading@10: static void horizontal_compose_dd97i(IDWTELEM *b, IDWTELEM *tmp, int w)
yading@10: {
yading@10:     const int w2 = w >> 1;
yading@10:     int x;
yading@10: 
yading@10:     tmp[0] = COMPOSE_53iL0(b[w2], b[0], b[w2]);
yading@10:     for (x = 1; x < w2; x++)
yading@10:         tmp[x] = COMPOSE_53iL0(b[x+w2-1], b[x], b[x+w2]);
yading@10: 
yading@10:     // extend the edges
yading@10:     tmp[-1]   = tmp[0];
yading@10:     tmp[w2+1] = tmp[w2] = tmp[w2-1];
yading@10: 
yading@10:     for (x = 0; x < w2; x++) {
yading@10:         b[2*x  ] = (tmp[x] + 1)>>1;
yading@10:         b[2*x+1] = (COMPOSE_DD97iH0(tmp[x-1], tmp[x], b[x+w2], tmp[x+1], tmp[x+2]) + 1)>>1;
yading@10:     }
yading@10: }
yading@10: 
yading@10: static void horizontal_compose_dd137i(IDWTELEM *b, IDWTELEM *tmp, int w)
yading@10: {
yading@10:     const int w2 = w >> 1;
yading@10:     int x;
yading@10: 
yading@10:     tmp[0] = COMPOSE_DD137iL0(b[w2], b[w2], b[0], b[w2  ], b[w2+1]);
yading@10:     tmp[1] = COMPOSE_DD137iL0(b[w2], b[w2], b[1], b[w2+1], b[w2+2]);
yading@10:     for (x = 2; x < w2-1; x++)
yading@10:         tmp[x] = COMPOSE_DD137iL0(b[x+w2-2], b[x+w2-1], b[x], b[x+w2], b[x+w2+1]);
yading@10:     tmp[w2-1] = COMPOSE_DD137iL0(b[w-3], b[w-2], b[w2-1], b[w-1], b[w-1]);
yading@10: 
yading@10:     // extend the edges
yading@10:     tmp[-1]   = tmp[0];
yading@10:     tmp[w2+1] = tmp[w2] = tmp[w2-1];
yading@10: 
yading@10:     for (x = 0; x < w2; x++) {
yading@10:         b[2*x  ] = (tmp[x] + 1)>>1;
yading@10:         b[2*x+1] = (COMPOSE_DD97iH0(tmp[x-1], tmp[x], b[x+w2], tmp[x+1], tmp[x+2]) + 1)>>1;
yading@10:     }
yading@10: }
yading@10: 
yading@10: static av_always_inline
yading@10: void horizontal_compose_haari(IDWTELEM *b, IDWTELEM *temp, int w, int shift)
yading@10: {
yading@10:     const int w2 = w >> 1;
yading@10:     int x;
yading@10: 
yading@10:     for (x = 0; x < w2; x++) {
yading@10:         temp[x   ] = COMPOSE_HAARiL0(b[x   ], b[x+w2]);
yading@10:         temp[x+w2] = COMPOSE_HAARiH0(b[x+w2], temp[x]);
yading@10:     }
yading@10: 
yading@10:     interleave(b, temp, temp+w2, w2, shift, shift);
yading@10: }
yading@10: 
yading@10: static void horizontal_compose_haar0i(IDWTELEM *b, IDWTELEM *temp, int w)
yading@10: {
yading@10:     horizontal_compose_haari(b, temp, w, 0);
yading@10: }
yading@10: 
yading@10: static void horizontal_compose_haar1i(IDWTELEM *b, IDWTELEM *temp, int w)
yading@10: {
yading@10:     horizontal_compose_haari(b, temp, w, 1);
yading@10: }
yading@10: 
yading@10: static void horizontal_compose_fidelityi(IDWTELEM *b, IDWTELEM *tmp, int w)
yading@10: {
yading@10:     const int w2 = w >> 1;
yading@10:     int i, x;
yading@10:     IDWTELEM v[8];
yading@10: 
yading@10:     for (x = 0; x < w2; x++) {
yading@10:         for (i = 0; i < 8; i++)
yading@10:             v[i] = b[av_clip(x-3+i, 0, w2-1)];
yading@10:         tmp[x] = COMPOSE_FIDELITYiH0(v[0], v[1], v[2], v[3], b[x+w2], v[4], v[5], v[6], v[7]);
yading@10:     }
yading@10: 
yading@10:     for (x = 0; x < w2; x++) {
yading@10:         for (i = 0; i < 8; i++)
yading@10:             v[i] = tmp[av_clip(x-4+i, 0, w2-1)];
yading@10:         tmp[x+w2] = COMPOSE_FIDELITYiL0(v[0], v[1], v[2], v[3], b[x], v[4], v[5], v[6], v[7]);
yading@10:     }
yading@10: 
yading@10:     interleave(b, tmp+w2, tmp, w2, 0, 0);
yading@10: }
yading@10: 
yading@10: static void horizontal_compose_daub97i(IDWTELEM *b, IDWTELEM *temp, int w)
yading@10: {
yading@10:     const int w2 = w >> 1;
yading@10:     int x, b0, b1, b2;
yading@10: 
yading@10:     temp[0] = COMPOSE_DAUB97iL1(b[w2], b[0], b[w2]);
yading@10:     for (x = 1; x < w2; x++) {
yading@10:         temp[x     ] = COMPOSE_DAUB97iL1(b[x+w2-1], b[x     ], b[x+w2]);
yading@10:         temp[x+w2-1] = COMPOSE_DAUB97iH1(temp[x-1], b[x+w2-1], temp[x]);
yading@10:     }
yading@10:     temp[w-1] = COMPOSE_DAUB97iH1(temp[w2-1], b[w-1], temp[w2-1]);
yading@10: 
yading@10:     // second stage combined with interleave and shift
yading@10:     b0 = b2 = COMPOSE_DAUB97iL0(temp[w2], temp[0], temp[w2]);
yading@10:     b[0] = (b0 + 1) >> 1;
yading@10:     for (x = 1; x < w2; x++) {
yading@10:         b2 = COMPOSE_DAUB97iL0(temp[x+w2-1], temp[x     ], temp[x+w2]);
yading@10:         b1 = COMPOSE_DAUB97iH0(          b0, temp[x+w2-1], b2        );
yading@10:         b[2*x-1] = (b1 + 1) >> 1;
yading@10:         b[2*x  ] = (b2 + 1) >> 1;
yading@10:         b0 = b2;
yading@10:     }
yading@10:     b[w-1] = (COMPOSE_DAUB97iH0(b2, temp[w-1], b2) + 1) >> 1;
yading@10: }
yading@10: 
yading@10: static void vertical_compose_dirac53iH0(IDWTELEM *b0, IDWTELEM *b1, IDWTELEM *b2, int width)
yading@10: {
yading@10:     int i;
yading@10: 
yading@10:     for(i=0; i<width; i++){
yading@10:         b1[i] = COMPOSE_DIRAC53iH0(b0[i], b1[i], b2[i]);
yading@10:     }
yading@10: }
yading@10: 
yading@10: static void vertical_compose_dd97iH0(IDWTELEM *b0, IDWTELEM *b1, IDWTELEM *b2,
yading@10:                                   IDWTELEM *b3, IDWTELEM *b4, int width)
yading@10: {
yading@10:     int i;
yading@10: 
yading@10:     for(i=0; i<width; i++){
yading@10:         b2[i] = COMPOSE_DD97iH0(b0[i], b1[i], b2[i], b3[i], b4[i]);
yading@10:     }
yading@10: }
yading@10: 
yading@10: static void vertical_compose_dd137iL0(IDWTELEM *b0, IDWTELEM *b1, IDWTELEM *b2,
yading@10:                                       IDWTELEM *b3, IDWTELEM *b4, int width)
yading@10: {
yading@10:     int i;
yading@10: 
yading@10:     for(i=0; i<width; i++){
yading@10:         b2[i] = COMPOSE_DD137iL0(b0[i], b1[i], b2[i], b3[i], b4[i]);
yading@10:     }
yading@10: }
yading@10: 
yading@10: static void vertical_compose_haar(IDWTELEM *b0, IDWTELEM *b1, int width)
yading@10: {
yading@10:     int i;
yading@10: 
yading@10:     for (i = 0; i < width; i++) {
yading@10:         b0[i] = COMPOSE_HAARiL0(b0[i], b1[i]);
yading@10:         b1[i] = COMPOSE_HAARiH0(b1[i], b0[i]);
yading@10:     }
yading@10: }
yading@10: 
yading@10: static void vertical_compose_fidelityiH0(IDWTELEM *dst, IDWTELEM *b[8], int width)
yading@10: {
yading@10:     int i;
yading@10: 
yading@10:     for(i=0; i<width; i++){
yading@10:         dst[i] = COMPOSE_FIDELITYiH0(b[0][i], b[1][i], b[2][i], b[3][i], dst[i], b[4][i], b[5][i], b[6][i], b[7][i]);
yading@10:     }
yading@10: }
yading@10: 
yading@10: static void vertical_compose_fidelityiL0(IDWTELEM *dst, IDWTELEM *b[8], int width)
yading@10: {
yading@10:     int i;
yading@10: 
yading@10:     for(i=0; i<width; i++){
yading@10:         dst[i] = COMPOSE_FIDELITYiL0(b[0][i], b[1][i], b[2][i], b[3][i], dst[i], b[4][i], b[5][i], b[6][i], b[7][i]);
yading@10:     }
yading@10: }
yading@10: 
yading@10: static void vertical_compose_daub97iH0(IDWTELEM *b0, IDWTELEM *b1, IDWTELEM *b2, int width)
yading@10: {
yading@10:     int i;
yading@10: 
yading@10:     for(i=0; i<width; i++){
yading@10:         b1[i] = COMPOSE_DAUB97iH0(b0[i], b1[i], b2[i]);
yading@10:     }
yading@10: }
yading@10: 
yading@10: static void vertical_compose_daub97iH1(IDWTELEM *b0, IDWTELEM *b1, IDWTELEM *b2, int width)
yading@10: {
yading@10:     int i;
yading@10: 
yading@10:     for(i=0; i<width; i++){
yading@10:         b1[i] = COMPOSE_DAUB97iH1(b0[i], b1[i], b2[i]);
yading@10:     }
yading@10: }
yading@10: 
yading@10: static void vertical_compose_daub97iL0(IDWTELEM *b0, IDWTELEM *b1, IDWTELEM *b2, int width)
yading@10: {
yading@10:     int i;
yading@10: 
yading@10:     for(i=0; i<width; i++){
yading@10:         b1[i] = COMPOSE_DAUB97iL0(b0[i], b1[i], b2[i]);
yading@10:     }
yading@10: }
yading@10: 
yading@10: static void vertical_compose_daub97iL1(IDWTELEM *b0, IDWTELEM *b1, IDWTELEM *b2, int width)
yading@10: {
yading@10:     int i;
yading@10: 
yading@10:     for(i=0; i<width; i++){
yading@10:         b1[i] = COMPOSE_DAUB97iL1(b0[i], b1[i], b2[i]);
yading@10:     }
yading@10: }
yading@10: 
yading@10: 
yading@10: static void spatial_compose_dd97i_dy(DWTContext *d, int level, int width, int height, int stride)
yading@10: {
yading@10:     vertical_compose_3tap vertical_compose_l0 = (void*)d->vertical_compose_l0;
yading@10:     vertical_compose_5tap vertical_compose_h0 = (void*)d->vertical_compose_h0;
yading@10:     DWTCompose *cs = d->cs + level;
yading@10: 
yading@10:     int i, y = cs->y;
yading@10:     IDWTELEM *b[8];
yading@10:     for (i = 0; i < 6; i++)
yading@10:         b[i] = cs->b[i];
yading@10:     b[6] = d->buffer + av_clip(y+5, 0, height-2)*stride;
yading@10:     b[7] = d->buffer + av_clip(y+6, 1, height-1)*stride;
yading@10: 
yading@10:         if(y+5<(unsigned)height) vertical_compose_l0(      b[5], b[6], b[7],       width);
yading@10:         if(y+1<(unsigned)height) vertical_compose_h0(b[0], b[2], b[3], b[4], b[6], width);
yading@10: 
yading@10:         if(y-1<(unsigned)height) d->horizontal_compose(b[0], d->temp, width);
yading@10:         if(y+0<(unsigned)height) d->horizontal_compose(b[1], d->temp, width);
yading@10: 
yading@10:     for (i = 0; i < 6; i++)
yading@10:         cs->b[i] = b[i+2];
yading@10:     cs->y += 2;
yading@10: }
yading@10: 
yading@10: static void spatial_compose_dirac53i_dy(DWTContext *d, int level, int width, int height, int stride)
yading@10: {
yading@10:     vertical_compose_3tap vertical_compose_l0 = (void*)d->vertical_compose_l0;
yading@10:     vertical_compose_3tap vertical_compose_h0 = (void*)d->vertical_compose_h0;
yading@10:     DWTCompose *cs = d->cs + level;
yading@10: 
yading@10:     int y= cs->y;
yading@10:     IDWTELEM *b[4] = { cs->b[0], cs->b[1] };
yading@10:     b[2] = d->buffer + mirror(y+1, height-1)*stride;
yading@10:     b[3] = d->buffer + mirror(y+2, height-1)*stride;
yading@10: 
yading@10:         if(y+1<(unsigned)height) vertical_compose_l0(b[1], b[2], b[3], width);
yading@10:         if(y+0<(unsigned)height) vertical_compose_h0(b[0], b[1], b[2], width);
yading@10: 
yading@10:         if(y-1<(unsigned)height) d->horizontal_compose(b[0], d->temp, width);
yading@10:         if(y+0<(unsigned)height) d->horizontal_compose(b[1], d->temp, width);
yading@10: 
yading@10:     cs->b[0] = b[2];
yading@10:     cs->b[1] = b[3];
yading@10:     cs->y += 2;
yading@10: }
yading@10: 
yading@10: 
yading@10: static void spatial_compose_dd137i_dy(DWTContext *d, int level, int width, int height, int stride)
yading@10: {
yading@10:     vertical_compose_5tap vertical_compose_l0 = (void*)d->vertical_compose_l0;
yading@10:     vertical_compose_5tap vertical_compose_h0 = (void*)d->vertical_compose_h0;
yading@10:     DWTCompose *cs = d->cs + level;
yading@10: 
yading@10:     int i, y = cs->y;
yading@10:     IDWTELEM *b[10];
yading@10:     for (i = 0; i < 8; i++)
yading@10:         b[i] = cs->b[i];
yading@10:     b[8] = d->buffer + av_clip(y+7, 0, height-2)*stride;
yading@10:     b[9] = d->buffer + av_clip(y+8, 1, height-1)*stride;
yading@10: 
yading@10:         if(y+5<(unsigned)height) vertical_compose_l0(b[3], b[5], b[6], b[7], b[9], width);
yading@10:         if(y+1<(unsigned)height) vertical_compose_h0(b[0], b[2], b[3], b[4], b[6], width);
yading@10: 
yading@10:         if(y-1<(unsigned)height) d->horizontal_compose(b[0], d->temp, width);
yading@10:         if(y+0<(unsigned)height) d->horizontal_compose(b[1], d->temp, width);
yading@10: 
yading@10:     for (i = 0; i < 8; i++)
yading@10:         cs->b[i] = b[i+2];
yading@10:     cs->y += 2;
yading@10: }
yading@10: 
yading@10: // haar makes the assumption that height is even (always true for dirac)
yading@10: static void spatial_compose_haari_dy(DWTContext *d, int level, int width, int height, int stride)
yading@10: {
yading@10:     vertical_compose_2tap vertical_compose = (void*)d->vertical_compose;
yading@10:     int y = d->cs[level].y;
yading@10:     IDWTELEM *b0 = d->buffer + (y-1)*stride;
yading@10:     IDWTELEM *b1 = d->buffer + (y  )*stride;
yading@10: 
yading@10:     vertical_compose(b0, b1, width);
yading@10:     d->horizontal_compose(b0, d->temp, width);
yading@10:     d->horizontal_compose(b1, d->temp, width);
yading@10: 
yading@10:     d->cs[level].y += 2;
yading@10: }
yading@10: 
yading@10: // Don't do sliced idwt for fidelity; the 9 tap filter makes it a bit annoying
yading@10: // Fortunately, this filter isn't used in practice.
yading@10: static void spatial_compose_fidelity(DWTContext *d, int level, int width, int height, int stride)
yading@10: {
yading@10:     vertical_compose_9tap vertical_compose_l0 = (void*)d->vertical_compose_l0;
yading@10:     vertical_compose_9tap vertical_compose_h0 = (void*)d->vertical_compose_h0;
yading@10:     int i, y;
yading@10:     IDWTELEM *b[8];
yading@10: 
yading@10:     for (y = 1; y < height; y += 2) {
yading@10:         for (i = 0; i < 8; i++)
yading@10:             b[i] = d->buffer + av_clip((y-7 + 2*i), 0, height-2)*stride;
yading@10:         vertical_compose_h0(d->buffer + y*stride, b, width);
yading@10:     }
yading@10: 
yading@10:     for (y = 0; y < height; y += 2) {
yading@10:         for (i = 0; i < 8; i++)
yading@10:             b[i] = d->buffer + av_clip((y-7 + 2*i), 1, height-1)*stride;
yading@10:         vertical_compose_l0(d->buffer + y*stride, b, width);
yading@10:     }
yading@10: 
yading@10:     for (y = 0; y < height; y++)
yading@10:         d->horizontal_compose(d->buffer + y*stride, d->temp, width);
yading@10: 
yading@10:     d->cs[level].y = height+1;
yading@10: }
yading@10: 
yading@10: static void spatial_compose_daub97i_dy(DWTContext *d, int level, int width, int height, int stride)
yading@10: {
yading@10:     vertical_compose_3tap vertical_compose_l0 = (void*)d->vertical_compose_l0;
yading@10:     vertical_compose_3tap vertical_compose_h0 = (void*)d->vertical_compose_h0;
yading@10:     vertical_compose_3tap vertical_compose_l1 = (void*)d->vertical_compose_l1;
yading@10:     vertical_compose_3tap vertical_compose_h1 = (void*)d->vertical_compose_h1;
yading@10:     DWTCompose *cs = d->cs + level;
yading@10: 
yading@10:     int i, y = cs->y;
yading@10:     IDWTELEM *b[6];
yading@10:     for (i = 0; i < 4; i++)
yading@10:         b[i] = cs->b[i];
yading@10:     b[4] = d->buffer + mirror(y+3, height-1)*stride;
yading@10:     b[5] = d->buffer + mirror(y+4, height-1)*stride;
yading@10: 
yading@10:         if(y+3<(unsigned)height) vertical_compose_l1(b[3], b[4], b[5], width);
yading@10:         if(y+2<(unsigned)height) vertical_compose_h1(b[2], b[3], b[4], width);
yading@10:         if(y+1<(unsigned)height) vertical_compose_l0(b[1], b[2], b[3], width);
yading@10:         if(y+0<(unsigned)height) vertical_compose_h0(b[0], b[1], b[2], width);
yading@10: 
yading@10:         if(y-1<(unsigned)height) d->horizontal_compose(b[0], d->temp, width);
yading@10:         if(y+0<(unsigned)height) d->horizontal_compose(b[1], d->temp, width);
yading@10: 
yading@10:     for (i = 0; i < 4; i++)
yading@10:         cs->b[i] = b[i+2];
yading@10:     cs->y += 2;
yading@10: }
yading@10: 
yading@10: 
yading@10: static void spatial_compose97i_init2(DWTCompose *cs, IDWTELEM *buffer, int height, int stride)
yading@10: {
yading@10:     cs->b[0] = buffer + mirror(-3-1, height-1)*stride;
yading@10:     cs->b[1] = buffer + mirror(-3  , height-1)*stride;
yading@10:     cs->b[2] = buffer + mirror(-3+1, height-1)*stride;
yading@10:     cs->b[3] = buffer + mirror(-3+2, height-1)*stride;
yading@10:     cs->y = -3;
yading@10: }
yading@10: 
yading@10: static void spatial_compose53i_init2(DWTCompose *cs, IDWTELEM *buffer, int height, int stride)
yading@10: {
yading@10:     cs->b[0] = buffer + mirror(-1-1, height-1)*stride;
yading@10:     cs->b[1] = buffer + mirror(-1  , height-1)*stride;
yading@10:     cs->y = -1;
yading@10: }
yading@10: 
yading@10: static void spatial_compose_dd97i_init(DWTCompose *cs, IDWTELEM *buffer, int height, int stride)
yading@10: {
yading@10:     cs->b[0] = buffer + av_clip(-5-1, 0, height-2)*stride;
yading@10:     cs->b[1] = buffer + av_clip(-5  , 1, height-1)*stride;
yading@10:     cs->b[2] = buffer + av_clip(-5+1, 0, height-2)*stride;
yading@10:     cs->b[3] = buffer + av_clip(-5+2, 1, height-1)*stride;
yading@10:     cs->b[4] = buffer + av_clip(-5+3, 0, height-2)*stride;
yading@10:     cs->b[5] = buffer + av_clip(-5+4, 1, height-1)*stride;
yading@10:     cs->y = -5;
yading@10: }
yading@10: 
yading@10: static void spatial_compose_dd137i_init(DWTCompose *cs, IDWTELEM *buffer, int height, int stride)
yading@10: {
yading@10:     cs->b[0] = buffer + av_clip(-5-1, 0, height-2)*stride;
yading@10:     cs->b[1] = buffer + av_clip(-5  , 1, height-1)*stride;
yading@10:     cs->b[2] = buffer + av_clip(-5+1, 0, height-2)*stride;
yading@10:     cs->b[3] = buffer + av_clip(-5+2, 1, height-1)*stride;
yading@10:     cs->b[4] = buffer + av_clip(-5+3, 0, height-2)*stride;
yading@10:     cs->b[5] = buffer + av_clip(-5+4, 1, height-1)*stride;
yading@10:     cs->b[6] = buffer + av_clip(-5+5, 0, height-2)*stride;
yading@10:     cs->b[7] = buffer + av_clip(-5+6, 1, height-1)*stride;
yading@10:     cs->y = -5;
yading@10: }
yading@10: 
yading@10: int ff_spatial_idwt_init2(DWTContext *d, IDWTELEM *buffer, int width, int height,
yading@10:                           int stride, enum dwt_type type, int decomposition_count,
yading@10:                           IDWTELEM *temp)
yading@10: {
yading@10:     int level;
yading@10: 
yading@10:     d->buffer = buffer;
yading@10:     d->width = width;
yading@10:     d->height = height;
yading@10:     d->stride = stride;
yading@10:     d->decomposition_count = decomposition_count;
yading@10:     d->temp = temp + 8;
yading@10: 
yading@10:     for(level=decomposition_count-1; level>=0; level--){
yading@10:         int hl = height >> level;
yading@10:         int stride_l = stride << level;
yading@10: 
yading@10:         switch(type){
yading@10:         case DWT_DIRAC_DD9_7:
yading@10:             spatial_compose_dd97i_init(d->cs+level, buffer, hl, stride_l);
yading@10:             break;
yading@10:         case DWT_DIRAC_LEGALL5_3:
yading@10:             spatial_compose53i_init2(d->cs+level, buffer, hl, stride_l);
yading@10:             break;
yading@10:         case DWT_DIRAC_DD13_7:
yading@10:             spatial_compose_dd137i_init(d->cs+level, buffer, hl, stride_l);
yading@10:             break;
yading@10:         case DWT_DIRAC_HAAR0:
yading@10:         case DWT_DIRAC_HAAR1:
yading@10:             d->cs[level].y = 1;
yading@10:             break;
yading@10:         case DWT_DIRAC_DAUB9_7:
yading@10:             spatial_compose97i_init2(d->cs+level, buffer, hl, stride_l);
yading@10:             break;
yading@10:         default:
yading@10:             d->cs[level].y = 0;
yading@10:             break;
yading@10:         }
yading@10:     }
yading@10: 
yading@10:     switch (type) {
yading@10:     case DWT_DIRAC_DD9_7:
yading@10:         d->spatial_compose = spatial_compose_dd97i_dy;
yading@10:         d->vertical_compose_l0 = (void*)vertical_compose53iL0;
yading@10:         d->vertical_compose_h0 = (void*)vertical_compose_dd97iH0;
yading@10:         d->horizontal_compose = horizontal_compose_dd97i;
yading@10:         d->support = 7;
yading@10:         break;
yading@10:     case DWT_DIRAC_LEGALL5_3:
yading@10:         d->spatial_compose = spatial_compose_dirac53i_dy;
yading@10:         d->vertical_compose_l0 = (void*)vertical_compose53iL0;
yading@10:         d->vertical_compose_h0 = (void*)vertical_compose_dirac53iH0;
yading@10:         d->horizontal_compose = horizontal_compose_dirac53i;
yading@10:         d->support = 3;
yading@10:         break;
yading@10:     case DWT_DIRAC_DD13_7:
yading@10:         d->spatial_compose = spatial_compose_dd137i_dy;
yading@10:         d->vertical_compose_l0 = (void*)vertical_compose_dd137iL0;
yading@10:         d->vertical_compose_h0 = (void*)vertical_compose_dd97iH0;
yading@10:         d->horizontal_compose = horizontal_compose_dd137i;
yading@10:         d->support = 7;
yading@10:         break;
yading@10:     case DWT_DIRAC_HAAR0:
yading@10:     case DWT_DIRAC_HAAR1:
yading@10:         d->spatial_compose = spatial_compose_haari_dy;
yading@10:         d->vertical_compose = (void*)vertical_compose_haar;
yading@10:         if (type == DWT_DIRAC_HAAR0)
yading@10:             d->horizontal_compose = horizontal_compose_haar0i;
yading@10:         else
yading@10:             d->horizontal_compose = horizontal_compose_haar1i;
yading@10:         d->support = 1;
yading@10:         break;
yading@10:     case DWT_DIRAC_FIDELITY:
yading@10:         d->spatial_compose = spatial_compose_fidelity;
yading@10:         d->vertical_compose_l0 = (void*)vertical_compose_fidelityiL0;
yading@10:         d->vertical_compose_h0 = (void*)vertical_compose_fidelityiH0;
yading@10:         d->horizontal_compose = horizontal_compose_fidelityi;
yading@10:         break;
yading@10:     case DWT_DIRAC_DAUB9_7:
yading@10:         d->spatial_compose = spatial_compose_daub97i_dy;
yading@10:         d->vertical_compose_l0 = (void*)vertical_compose_daub97iL0;
yading@10:         d->vertical_compose_h0 = (void*)vertical_compose_daub97iH0;
yading@10:         d->vertical_compose_l1 = (void*)vertical_compose_daub97iL1;
yading@10:         d->vertical_compose_h1 = (void*)vertical_compose_daub97iH1;
yading@10:         d->horizontal_compose = horizontal_compose_daub97i;
yading@10:         d->support = 5;
yading@10:         break;
yading@10:     default:
yading@10:         av_log(NULL, AV_LOG_ERROR, "Unknown wavelet type %d\n", type);
yading@10:         return -1;
yading@10:     }
yading@10: 
yading@10:     if (HAVE_MMX) ff_spatial_idwt_init_mmx(d, type);
yading@10: 
yading@10:     return 0;
yading@10: }
yading@10: 
yading@10: void ff_spatial_idwt_slice2(DWTContext *d, int y)
yading@10: {
yading@10:     int level, support = d->support;
yading@10: 
yading@10:     for (level = d->decomposition_count-1; level >= 0; level--) {
yading@10:         int wl = d->width  >> level;
yading@10:         int hl = d->height >> level;
yading@10:         int stride_l = d->stride << level;
yading@10: 
yading@10:         while (d->cs[level].y <= FFMIN((y>>level)+support, hl))
yading@10:             d->spatial_compose(d, level, wl, hl, stride_l);
yading@10:     }
yading@10: }
yading@10: 
yading@10: int ff_spatial_idwt2(IDWTELEM *buffer, int width, int height, int stride,
yading@10:                      enum dwt_type type, int decomposition_count, IDWTELEM *temp)
yading@10: {
yading@10:     DWTContext d;
yading@10:     int y;
yading@10: 
yading@10:     if (ff_spatial_idwt_init2(&d, buffer, width, height, stride, type, decomposition_count, temp))
yading@10:         return -1;
yading@10: 
yading@10:     for (y = 0; y < d.height; y += 4)
yading@10:         ff_spatial_idwt_slice2(&d, y);
yading@10: 
yading@10:     return 0;
yading@10: }