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 "snow_dwt.h"
yading@10: 
yading@10: int ff_slice_buffer_init(slice_buffer *buf, int line_count,
yading@10:                          int max_allocated_lines, int line_width,
yading@10:                          IDWTELEM *base_buffer)
yading@10: {
yading@10:     int i;
yading@10: 
yading@10:     buf->base_buffer = base_buffer;
yading@10:     buf->line_count  = line_count;
yading@10:     buf->line_width  = line_width;
yading@10:     buf->data_count  = max_allocated_lines;
yading@10:     buf->line        = av_mallocz(sizeof(IDWTELEM *) * line_count);
yading@10:     if (!buf->line)
yading@10:         return AVERROR(ENOMEM);
yading@10:     buf->data_stack  = av_malloc(sizeof(IDWTELEM *) * max_allocated_lines);
yading@10:     if (!buf->data_stack) {
yading@10:         av_freep(&buf->line);
yading@10:         return AVERROR(ENOMEM);
yading@10:     }
yading@10: 
yading@10:     for (i = 0; i < max_allocated_lines; i++) {
yading@10:         buf->data_stack[i] = av_malloc(sizeof(IDWTELEM) * line_width);
yading@10:         if (!buf->data_stack[i]) {
yading@10:             for (i--; i >=0; i--)
yading@10:                 av_freep(&buf->data_stack[i]);
yading@10:             av_freep(&buf->data_stack);
yading@10:             av_freep(&buf->line);
yading@10:             return AVERROR(ENOMEM);
yading@10:         }
yading@10:     }
yading@10: 
yading@10:     buf->data_stack_top = max_allocated_lines - 1;
yading@10:     return 0;
yading@10: }
yading@10: 
yading@10: IDWTELEM *ff_slice_buffer_load_line(slice_buffer *buf, int line)
yading@10: {
yading@10:     IDWTELEM *buffer;
yading@10: 
yading@10:     av_assert0(buf->data_stack_top >= 0);
yading@10: //  av_assert1(!buf->line[line]);
yading@10:     if (buf->line[line])
yading@10:         return buf->line[line];
yading@10: 
yading@10:     buffer = buf->data_stack[buf->data_stack_top];
yading@10:     buf->data_stack_top--;
yading@10:     buf->line[line] = buffer;
yading@10: 
yading@10:     return buffer;
yading@10: }
yading@10: 
yading@10: void ff_slice_buffer_release(slice_buffer *buf, int line)
yading@10: {
yading@10:     IDWTELEM *buffer;
yading@10: 
yading@10:     av_assert1(line >= 0 && line < buf->line_count);
yading@10:     av_assert1(buf->line[line]);
yading@10: 
yading@10:     buffer = buf->line[line];
yading@10:     buf->data_stack_top++;
yading@10:     buf->data_stack[buf->data_stack_top] = buffer;
yading@10:     buf->line[line]                      = NULL;
yading@10: }
yading@10: 
yading@10: void ff_slice_buffer_flush(slice_buffer *buf)
yading@10: {
yading@10:     int i;
yading@10:     for (i = 0; i < buf->line_count; i++)
yading@10:         if (buf->line[i])
yading@10:             ff_slice_buffer_release(buf, i);
yading@10: }
yading@10: 
yading@10: void ff_slice_buffer_destroy(slice_buffer *buf)
yading@10: {
yading@10:     int i;
yading@10:     ff_slice_buffer_flush(buf);
yading@10: 
yading@10:     for (i = buf->data_count - 1; i >= 0; i--)
yading@10:         av_freep(&buf->data_stack[i]);
yading@10:     av_freep(&buf->data_stack);
yading@10:     av_freep(&buf->line);
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 av_always_inline void lift(DWTELEM *dst, DWTELEM *src, DWTELEM *ref,
yading@10:                                   int dst_step, int src_step, int ref_step,
yading@10:                                   int width, int mul, int add, int shift,
yading@10:                                   int highpass, int inverse)
yading@10: {
yading@10:     const int mirror_left  = !highpass;
yading@10:     const int mirror_right = (width & 1) ^ highpass;
yading@10:     const int w            = (width >> 1) - 1 + (highpass & width);
yading@10:     int i;
yading@10: 
yading@10: #define LIFT(src, ref, inv) ((src) + ((inv) ? -(ref) : +(ref)))
yading@10:     if (mirror_left) {
yading@10:         dst[0] = LIFT(src[0], ((mul * 2 * ref[0] + add) >> shift), inverse);
yading@10:         dst   += dst_step;
yading@10:         src   += src_step;
yading@10:     }
yading@10: 
yading@10:     for (i = 0; i < w; i++)
yading@10:         dst[i * dst_step] = LIFT(src[i * src_step],
yading@10:                                  ((mul * (ref[i * ref_step] +
yading@10:                                           ref[(i + 1) * ref_step]) +
yading@10:                                    add) >> shift),
yading@10:                                  inverse);
yading@10: 
yading@10:     if (mirror_right)
yading@10:         dst[w * dst_step] = LIFT(src[w * src_step],
yading@10:                                  ((mul * 2 * ref[w * ref_step] + add) >> shift),
yading@10:                                  inverse);
yading@10: }
yading@10: 
yading@10: static av_always_inline void liftS(DWTELEM *dst, DWTELEM *src, DWTELEM *ref,
yading@10:                                    int dst_step, int src_step, int ref_step,
yading@10:                                    int width, int mul, int add, int shift,
yading@10:                                    int highpass, int inverse)
yading@10: {
yading@10:     const int mirror_left  = !highpass;
yading@10:     const int mirror_right = (width & 1) ^ highpass;
yading@10:     const int w            = (width >> 1) - 1 + (highpass & width);
yading@10:     int i;
yading@10: 
yading@10:     av_assert1(shift == 4);
yading@10: #define LIFTS(src, ref, inv)                                            \
yading@10:     ((inv) ? (src) + (((ref) + 4 * (src)) >> shift)                     \
yading@10:            : -((-16 * (src) + (ref) + add /                             \
yading@10:                 4 + 1 + (5 << 25)) / (5 * 4) - (1 << 23)))
yading@10:     if (mirror_left) {
yading@10:         dst[0] = LIFTS(src[0], mul * 2 * ref[0] + add, inverse);
yading@10:         dst   += dst_step;
yading@10:         src   += src_step;
yading@10:     }
yading@10: 
yading@10:     for (i = 0; i < w; i++)
yading@10:         dst[i * dst_step] = LIFTS(src[i * src_step],
yading@10:                                   mul * (ref[i * ref_step] +
yading@10:                                          ref[(i + 1) * ref_step]) + add,
yading@10:                                   inverse);
yading@10: 
yading@10:     if (mirror_right)
yading@10:         dst[w * dst_step] = LIFTS(src[w * src_step],
yading@10:                                   mul * 2 * ref[w * ref_step] + add,
yading@10:                                   inverse);
yading@10: }
yading@10: 
yading@10: static void horizontal_decompose53i(DWTELEM *b, DWTELEM *temp, int width)
yading@10: {
yading@10:     const int width2 = width >> 1;
yading@10:     int x;
yading@10:     const int w2 = (width + 1) >> 1;
yading@10: 
yading@10:     for (x = 0; x < width2; x++) {
yading@10:         temp[x]      = b[2 * x];
yading@10:         temp[x + w2] = b[2 * x + 1];
yading@10:     }
yading@10:     if (width & 1)
yading@10:         temp[x] = b[2 * x];
yading@10:     lift(b + w2, temp + w2, temp,   1, 1, 1, width, -1, 0, 1, 1, 0);
yading@10:     lift(b,      temp,      b + w2, 1, 1, 1, width,  1, 2, 2, 0, 0);
yading@10: }
yading@10: 
yading@10: static void vertical_decompose53iH0(DWTELEM *b0, DWTELEM *b1, DWTELEM *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]) >> 1;
yading@10: }
yading@10: 
yading@10: static void vertical_decompose53iL0(DWTELEM *b0, DWTELEM *b1, DWTELEM *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: static void spatial_decompose53i(DWTELEM *buffer, DWTELEM *temp,
yading@10:                                  int width, int height, int stride)
yading@10: {
yading@10:     int y;
yading@10:     DWTELEM *b0 = buffer + mirror(-2 - 1, height - 1) * stride;
yading@10:     DWTELEM *b1 = buffer + mirror(-2,     height - 1) * stride;
yading@10: 
yading@10:     for (y = -2; y < height; y += 2) {
yading@10:         DWTELEM *b2 = buffer + mirror(y + 1, height - 1) * stride;
yading@10:         DWTELEM *b3 = buffer + mirror(y + 2, height - 1) * stride;
yading@10: 
yading@10:         if (y + 1 < (unsigned)height)
yading@10:             horizontal_decompose53i(b2, temp, width);
yading@10:         if (y + 2 < (unsigned)height)
yading@10:             horizontal_decompose53i(b3, temp, width);
yading@10: 
yading@10:         if (y + 1 < (unsigned)height)
yading@10:             vertical_decompose53iH0(b1, b2, b3, width);
yading@10:         if (y + 0 < (unsigned)height)
yading@10:             vertical_decompose53iL0(b0, b1, b2, width);
yading@10: 
yading@10:         b0 = b2;
yading@10:         b1 = b3;
yading@10:     }
yading@10: }
yading@10: 
yading@10: static void horizontal_decompose97i(DWTELEM *b, DWTELEM *temp, int width)
yading@10: {
yading@10:     const int w2 = (width + 1) >> 1;
yading@10: 
yading@10:     lift(temp + w2, b + 1, b,         1, 2, 2, width, W_AM, W_AO, W_AS, 1, 1);
yading@10:     liftS(temp,     b,     temp + w2, 1, 2, 1, width, W_BM, W_BO, W_BS, 0, 0);
yading@10:     lift(b + w2, temp + w2, temp,     1, 1, 1, width, W_CM, W_CO, W_CS, 1, 0);
yading@10:     lift(b,      temp,      b + w2,   1, 1, 1, width, W_DM, W_DO, W_DS, 0, 0);
yading@10: }
yading@10: 
yading@10: static void vertical_decompose97iH0(DWTELEM *b0, DWTELEM *b1, DWTELEM *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] -= (W_AM * (b0[i] + b2[i]) + W_AO) >> W_AS;
yading@10: }
yading@10: 
yading@10: static void vertical_decompose97iH1(DWTELEM *b0, DWTELEM *b1, DWTELEM *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] += (W_CM * (b0[i] + b2[i]) + W_CO) >> W_CS;
yading@10: }
yading@10: 
yading@10: static void vertical_decompose97iL0(DWTELEM *b0, DWTELEM *b1, DWTELEM *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] = (16 * 4 * b1[i] - 4 * (b0[i] + b2[i]) + W_BO * 5 + (5 << 27)) /
yading@10:                 (5 * 16) - (1 << 23);
yading@10: }
yading@10: 
yading@10: static void vertical_decompose97iL1(DWTELEM *b0, DWTELEM *b1, DWTELEM *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] += (W_DM * (b0[i] + b2[i]) + W_DO) >> W_DS;
yading@10: }
yading@10: 
yading@10: static void spatial_decompose97i(DWTELEM *buffer, DWTELEM *temp,
yading@10:                                  int width, int height, int stride)
yading@10: {
yading@10:     int y;
yading@10:     DWTELEM *b0 = buffer + mirror(-4 - 1, height - 1) * stride;
yading@10:     DWTELEM *b1 = buffer + mirror(-4,     height - 1) * stride;
yading@10:     DWTELEM *b2 = buffer + mirror(-4 + 1, height - 1) * stride;
yading@10:     DWTELEM *b3 = buffer + mirror(-4 + 2, height - 1) * stride;
yading@10: 
yading@10:     for (y = -4; y < height; y += 2) {
yading@10:         DWTELEM *b4 = buffer + mirror(y + 3, height - 1) * stride;
yading@10:         DWTELEM *b5 = buffer + mirror(y + 4, height - 1) * stride;
yading@10: 
yading@10:         if (y + 3 < (unsigned)height)
yading@10:             horizontal_decompose97i(b4, temp, width);
yading@10:         if (y + 4 < (unsigned)height)
yading@10:             horizontal_decompose97i(b5, temp, width);
yading@10: 
yading@10:         if (y + 3 < (unsigned)height)
yading@10:             vertical_decompose97iH0(b3, b4, b5, width);
yading@10:         if (y + 2 < (unsigned)height)
yading@10:             vertical_decompose97iL0(b2, b3, b4, width);
yading@10:         if (y + 1 < (unsigned)height)
yading@10:             vertical_decompose97iH1(b1, b2, b3, width);
yading@10:         if (y + 0 < (unsigned)height)
yading@10:             vertical_decompose97iL1(b0, b1, b2, width);
yading@10: 
yading@10:         b0 = b2;
yading@10:         b1 = b3;
yading@10:         b2 = b4;
yading@10:         b3 = b5;
yading@10:     }
yading@10: }
yading@10: 
yading@10: void ff_spatial_dwt(DWTELEM *buffer, DWTELEM *temp, int width, int height,
yading@10:                     int stride, int type, int decomposition_count)
yading@10: {
yading@10:     int level;
yading@10: 
yading@10:     for (level = 0; level < decomposition_count; level++) {
yading@10:         switch (type) {
yading@10:         case DWT_97:
yading@10:             spatial_decompose97i(buffer, temp,
yading@10:                                  width >> level, height >> level,
yading@10:                                  stride << level);
yading@10:             break;
yading@10:         case DWT_53:
yading@10:             spatial_decompose53i(buffer, temp,
yading@10:                                  width >> level, height >> level,
yading@10:                                  stride << level);
yading@10:             break;
yading@10:         }
yading@10:     }
yading@10: }
yading@10: 
yading@10: static void horizontal_compose53i(IDWTELEM *b, IDWTELEM *temp, int width)
yading@10: {
yading@10:     const int width2 = width >> 1;
yading@10:     const int w2     = (width + 1) >> 1;
yading@10:     int x;
yading@10: 
yading@10:     for (x = 0; x < width2; x++) {
yading@10:         temp[2 * x]     = b[x];
yading@10:         temp[2 * x + 1] = b[x + w2];
yading@10:     }
yading@10:     if (width & 1)
yading@10:         temp[2 * x] = b[x];
yading@10: 
yading@10:     b[0] = temp[0] - ((temp[1] + 1) >> 1);
yading@10:     for (x = 2; x < width - 1; x += 2) {
yading@10:         b[x]     = temp[x]     - ((temp[x - 1] + temp[x + 1] + 2) >> 2);
yading@10:         b[x - 1] = temp[x - 1] + ((b[x - 2]    + b[x]        + 1) >> 1);
yading@10:     }
yading@10:     if (width & 1) {
yading@10:         b[x]     = temp[x]     - ((temp[x - 1]     + 1) >> 1);
yading@10:         b[x - 1] = temp[x - 1] + ((b[x - 2] + b[x] + 1) >> 1);
yading@10:     } else
yading@10:         b[x - 1] = temp[x - 1] + b[x - 2];
yading@10: }
yading@10: 
yading@10: static void vertical_compose53iH0(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]) >> 1;
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: static void spatial_compose53i_buffered_init(DWTCompose *cs, slice_buffer *sb,
yading@10:                                              int height, int stride_line)
yading@10: {
yading@10:     cs->b0 = slice_buffer_get_line(sb,
yading@10:                                    mirror(-1 - 1, height - 1) * stride_line);
yading@10:     cs->b1 = slice_buffer_get_line(sb, mirror(-1, height - 1) * stride_line);
yading@10:     cs->y  = -1;
yading@10: }
yading@10: 
yading@10: static void spatial_compose53i_init(DWTCompose *cs, IDWTELEM *buffer,
yading@10:                                     int height, int stride)
yading@10: {
yading@10:     cs->b0 = buffer + mirror(-1 - 1, height - 1) * stride;
yading@10:     cs->b1 = buffer + mirror(-1,     height - 1) * stride;
yading@10:     cs->y  = -1;
yading@10: }
yading@10: 
yading@10: static void spatial_compose53i_dy_buffered(DWTCompose *cs, slice_buffer *sb,
yading@10:                                            IDWTELEM *temp,
yading@10:                                            int width, int height,
yading@10:                                            int stride_line)
yading@10: {
yading@10:     int y = cs->y;
yading@10: 
yading@10:     IDWTELEM *b0 = cs->b0;
yading@10:     IDWTELEM *b1 = cs->b1;
yading@10:     IDWTELEM *b2 = slice_buffer_get_line(sb,
yading@10:                                          mirror(y + 1, height - 1) *
yading@10:                                          stride_line);
yading@10:     IDWTELEM *b3 = slice_buffer_get_line(sb,
yading@10:                                          mirror(y + 2, height - 1) *
yading@10:                                          stride_line);
yading@10: 
yading@10:     if (y + 1 < (unsigned)height && y < (unsigned)height) {
yading@10:         int x;
yading@10: 
yading@10:         for (x = 0; x < width; x++) {
yading@10:             b2[x] -= (b1[x] + b3[x] + 2) >> 2;
yading@10:             b1[x] += (b0[x] + b2[x])     >> 1;
yading@10:         }
yading@10:     } else {
yading@10:         if (y + 1 < (unsigned)height)
yading@10:             vertical_compose53iL0(b1, b2, b3, width);
yading@10:         if (y + 0 < (unsigned)height)
yading@10:             vertical_compose53iH0(b0, b1, b2, width);
yading@10:     }
yading@10: 
yading@10:     if (y - 1 < (unsigned)height)
yading@10:         horizontal_compose53i(b0, temp, width);
yading@10:     if (y + 0 < (unsigned)height)
yading@10:         horizontal_compose53i(b1, temp, width);
yading@10: 
yading@10:     cs->b0  = b2;
yading@10:     cs->b1  = b3;
yading@10:     cs->y  += 2;
yading@10: }
yading@10: 
yading@10: static void spatial_compose53i_dy(DWTCompose *cs, IDWTELEM *buffer,
yading@10:                                   IDWTELEM *temp, int width, int height,
yading@10:                                   int stride)
yading@10: {
yading@10:     int y        = cs->y;
yading@10:     IDWTELEM *b0 = cs->b0;
yading@10:     IDWTELEM *b1 = cs->b1;
yading@10:     IDWTELEM *b2 = buffer + mirror(y + 1, height - 1) * stride;
yading@10:     IDWTELEM *b3 = buffer + mirror(y + 2, height - 1) * stride;
yading@10: 
yading@10:     if (y + 1 < (unsigned)height)
yading@10:         vertical_compose53iL0(b1, b2, b3, width);
yading@10:     if (y + 0 < (unsigned)height)
yading@10:         vertical_compose53iH0(b0, b1, b2, width);
yading@10: 
yading@10:     if (y - 1 < (unsigned)height)
yading@10:         horizontal_compose53i(b0, temp, width);
yading@10:     if (y + 0 < (unsigned)height)
yading@10:         horizontal_compose53i(b1, temp, width);
yading@10: 
yading@10:     cs->b0  = b2;
yading@10:     cs->b1  = b3;
yading@10:     cs->y  += 2;
yading@10: }
yading@10: 
yading@10: void ff_snow_horizontal_compose97i(IDWTELEM *b, IDWTELEM *temp, int width)
yading@10: {
yading@10:     const int w2 = (width + 1) >> 1;
yading@10:     int x;
yading@10: 
yading@10:     temp[0] = b[0] - ((3 * b[w2] + 2) >> 2);
yading@10:     for (x = 1; x < (width >> 1); x++) {
yading@10:         temp[2 * x]     = b[x] - ((3 * (b[x + w2 - 1] + b[x + w2]) + 4) >> 3);
yading@10:         temp[2 * x - 1] = b[x + w2 - 1] - temp[2 * x - 2] - temp[2 * x];
yading@10:     }
yading@10:     if (width & 1) {
yading@10:         temp[2 * x]     = b[x] - ((3 * b[x + w2 - 1] + 2) >> 2);
yading@10:         temp[2 * x - 1] = b[x + w2 - 1] - temp[2 * x - 2] - temp[2 * x];
yading@10:     } else
yading@10:         temp[2 * x - 1] = b[x + w2 - 1] - 2 * temp[2 * x - 2];
yading@10: 
yading@10:     b[0] = temp[0] + ((2 * temp[0] + temp[1] + 4) >> 3);
yading@10:     for (x = 2; x < width - 1; x += 2) {
yading@10:         b[x]     = temp[x] + ((4 * temp[x] + temp[x - 1] + temp[x + 1] + 8) >> 4);
yading@10:         b[x - 1] = temp[x - 1] + ((3 * (b[x - 2] + b[x])) >> 1);
yading@10:     }
yading@10:     if (width & 1) {
yading@10:         b[x]     = temp[x] + ((2 * temp[x] + temp[x - 1] + 4) >> 3);
yading@10:         b[x - 1] = temp[x - 1] + ((3 * (b[x - 2] + b[x])) >> 1);
yading@10:     } else
yading@10:         b[x - 1] = temp[x - 1] + 3 * b[x - 2];
yading@10: }
yading@10: 
yading@10: static void vertical_compose97iH0(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] += (W_AM * (b0[i] + b2[i]) + W_AO) >> W_AS;
yading@10: }
yading@10: 
yading@10: static void vertical_compose97iH1(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] -= (W_CM * (b0[i] + b2[i]) + W_CO) >> W_CS;
yading@10: }
yading@10: 
yading@10: static void vertical_compose97iL0(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] += (W_BM * (b0[i] + b2[i]) + 4 * b1[i] + W_BO) >> W_BS;
yading@10: }
yading@10: 
yading@10: static void vertical_compose97iL1(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] -= (W_DM * (b0[i] + b2[i]) + W_DO) >> W_DS;
yading@10: }
yading@10: 
yading@10: void ff_snow_vertical_compose97i(IDWTELEM *b0, IDWTELEM *b1, IDWTELEM *b2,
yading@10:                                  IDWTELEM *b3, IDWTELEM *b4, IDWTELEM *b5,
yading@10:                                  int width)
yading@10: {
yading@10:     int i;
yading@10: 
yading@10:     for (i = 0; i < width; i++) {
yading@10:         b4[i] -= (W_DM * (b3[i] + b5[i]) + W_DO) >> W_DS;
yading@10:         b3[i] -= (W_CM * (b2[i] + b4[i]) + W_CO) >> W_CS;
yading@10:         b2[i] += (W_BM * (b1[i] + b3[i]) + 4 * b2[i] + W_BO) >> W_BS;
yading@10:         b1[i] += (W_AM * (b0[i] + b2[i]) + W_AO) >> W_AS;
yading@10:     }
yading@10: }
yading@10: 
yading@10: static void spatial_compose97i_buffered_init(DWTCompose *cs, slice_buffer *sb,
yading@10:                                              int height, int stride_line)
yading@10: {
yading@10:     cs->b0 = slice_buffer_get_line(sb, mirror(-3 - 1, height - 1) * stride_line);
yading@10:     cs->b1 = slice_buffer_get_line(sb, mirror(-3,     height - 1) * stride_line);
yading@10:     cs->b2 = slice_buffer_get_line(sb, mirror(-3 + 1, height - 1) * stride_line);
yading@10:     cs->b3 = slice_buffer_get_line(sb, mirror(-3 + 2, height - 1) * stride_line);
yading@10:     cs->y  = -3;
yading@10: }
yading@10: 
yading@10: static void spatial_compose97i_init(DWTCompose *cs, IDWTELEM *buffer, int height,
yading@10:                                     int stride)
yading@10: {
yading@10:     cs->b0 = buffer + mirror(-3 - 1, height - 1) * stride;
yading@10:     cs->b1 = buffer + mirror(-3,     height - 1) * stride;
yading@10:     cs->b2 = buffer + mirror(-3 + 1, height - 1) * stride;
yading@10:     cs->b3 = buffer + mirror(-3 + 2, height - 1) * stride;
yading@10:     cs->y  = -3;
yading@10: }
yading@10: 
yading@10: static void spatial_compose97i_dy_buffered(SnowDWTContext *dsp, DWTCompose *cs,
yading@10:                                            slice_buffer * sb, IDWTELEM *temp,
yading@10:                                            int width, int height,
yading@10:                                            int stride_line)
yading@10: {
yading@10:     int y = cs->y;
yading@10: 
yading@10:     IDWTELEM *b0 = cs->b0;
yading@10:     IDWTELEM *b1 = cs->b1;
yading@10:     IDWTELEM *b2 = cs->b2;
yading@10:     IDWTELEM *b3 = cs->b3;
yading@10:     IDWTELEM *b4 = slice_buffer_get_line(sb,
yading@10:                                          mirror(y + 3, height - 1) *
yading@10:                                          stride_line);
yading@10:     IDWTELEM *b5 = slice_buffer_get_line(sb,
yading@10:                                          mirror(y + 4, height - 1) *
yading@10:                                          stride_line);
yading@10: 
yading@10:     if (y > 0 && y + 4 < height) {
yading@10:         dsp->vertical_compose97i(b0, b1, b2, b3, b4, b5, width);
yading@10:     } else {
yading@10:         if (y + 3 < (unsigned)height)
yading@10:             vertical_compose97iL1(b3, b4, b5, width);
yading@10:         if (y + 2 < (unsigned)height)
yading@10:             vertical_compose97iH1(b2, b3, b4, width);
yading@10:         if (y + 1 < (unsigned)height)
yading@10:             vertical_compose97iL0(b1, b2, b3, width);
yading@10:         if (y + 0 < (unsigned)height)
yading@10:             vertical_compose97iH0(b0, b1, b2, width);
yading@10:     }
yading@10: 
yading@10:     if (y - 1 < (unsigned)height)
yading@10:         dsp->horizontal_compose97i(b0, temp, width);
yading@10:     if (y + 0 < (unsigned)height)
yading@10:         dsp->horizontal_compose97i(b1, temp, width);
yading@10: 
yading@10:     cs->b0  = b2;
yading@10:     cs->b1  = b3;
yading@10:     cs->b2  = b4;
yading@10:     cs->b3  = b5;
yading@10:     cs->y  += 2;
yading@10: }
yading@10: 
yading@10: static void spatial_compose97i_dy(DWTCompose *cs, IDWTELEM *buffer,
yading@10:                                   IDWTELEM *temp, int width, int height,
yading@10:                                   int stride)
yading@10: {
yading@10:     int y        = cs->y;
yading@10:     IDWTELEM *b0 = cs->b0;
yading@10:     IDWTELEM *b1 = cs->b1;
yading@10:     IDWTELEM *b2 = cs->b2;
yading@10:     IDWTELEM *b3 = cs->b3;
yading@10:     IDWTELEM *b4 = buffer + mirror(y + 3, height - 1) * stride;
yading@10:     IDWTELEM *b5 = buffer + mirror(y + 4, height - 1) * stride;
yading@10: 
yading@10:     if (y + 3 < (unsigned)height)
yading@10:         vertical_compose97iL1(b3, b4, b5, width);
yading@10:     if (y + 2 < (unsigned)height)
yading@10:         vertical_compose97iH1(b2, b3, b4, width);
yading@10:     if (y + 1 < (unsigned)height)
yading@10:         vertical_compose97iL0(b1, b2, b3, width);
yading@10:     if (y + 0 < (unsigned)height)
yading@10:         vertical_compose97iH0(b0, b1, b2, width);
yading@10: 
yading@10:     if (y - 1 < (unsigned)height)
yading@10:         ff_snow_horizontal_compose97i(b0, temp, width);
yading@10:     if (y + 0 < (unsigned)height)
yading@10:         ff_snow_horizontal_compose97i(b1, temp, width);
yading@10: 
yading@10:     cs->b0  = b2;
yading@10:     cs->b1  = b3;
yading@10:     cs->b2  = b4;
yading@10:     cs->b3  = b5;
yading@10:     cs->y  += 2;
yading@10: }
yading@10: 
yading@10: void ff_spatial_idwt_buffered_init(DWTCompose *cs, slice_buffer *sb, int width,
yading@10:                                    int height, int stride_line, int type,
yading@10:                                    int decomposition_count)
yading@10: {
yading@10:     int level;
yading@10:     for (level = decomposition_count - 1; level >= 0; level--) {
yading@10:         switch (type) {
yading@10:         case DWT_97:
yading@10:             spatial_compose97i_buffered_init(cs + level, sb, height >> level,
yading@10:                                              stride_line << level);
yading@10:             break;
yading@10:         case DWT_53:
yading@10:             spatial_compose53i_buffered_init(cs + level, sb, height >> level,
yading@10:                                              stride_line << level);
yading@10:             break;
yading@10:         }
yading@10:     }
yading@10: }
yading@10: 
yading@10: void ff_spatial_idwt_buffered_slice(SnowDWTContext *dsp, DWTCompose *cs,
yading@10:                                     slice_buffer *slice_buf, IDWTELEM *temp,
yading@10:                                     int width, int height, int stride_line,
yading@10:                                     int type, int decomposition_count, int y)
yading@10: {
yading@10:     const int support = type == 1 ? 3 : 5;
yading@10:     int level;
yading@10:     if (type == 2)
yading@10:         return;
yading@10: 
yading@10:     for (level = decomposition_count - 1; level >= 0; level--)
yading@10:         while (cs[level].y <= FFMIN((y >> level) + support, height >> level)) {
yading@10:             switch (type) {
yading@10:             case DWT_97:
yading@10:                 spatial_compose97i_dy_buffered(dsp, cs + level, slice_buf, temp,
yading@10:                                                width >> level,
yading@10:                                                height >> level,
yading@10:                                                stride_line << level);
yading@10:                 break;
yading@10:             case DWT_53:
yading@10:                 spatial_compose53i_dy_buffered(cs + level, slice_buf, temp,
yading@10:                                                width >> level,
yading@10:                                                height >> level,
yading@10:                                                stride_line << level);
yading@10:                 break;
yading@10:             }
yading@10:         }
yading@10: }
yading@10: 
yading@10: static void ff_spatial_idwt_init(DWTCompose *cs, IDWTELEM *buffer, int width,
yading@10:                                  int height, int stride, int type,
yading@10:                                  int decomposition_count)
yading@10: {
yading@10:     int level;
yading@10:     for (level = decomposition_count - 1; level >= 0; level--) {
yading@10:         switch (type) {
yading@10:         case DWT_97:
yading@10:             spatial_compose97i_init(cs + level, buffer, height >> level,
yading@10:                                     stride << level);
yading@10:             break;
yading@10:         case DWT_53:
yading@10:             spatial_compose53i_init(cs + level, buffer, height >> level,
yading@10:                                     stride << level);
yading@10:             break;
yading@10:         }
yading@10:     }
yading@10: }
yading@10: 
yading@10: static void ff_spatial_idwt_slice(DWTCompose *cs, IDWTELEM *buffer,
yading@10:                                   IDWTELEM *temp, int width, int height,
yading@10:                                   int stride, int type,
yading@10:                                   int decomposition_count, int y)
yading@10: {
yading@10:     const int support = type == 1 ? 3 : 5;
yading@10:     int level;
yading@10:     if (type == 2)
yading@10:         return;
yading@10: 
yading@10:     for (level = decomposition_count - 1; level >= 0; level--)
yading@10:         while (cs[level].y <= FFMIN((y >> level) + support, height >> level)) {
yading@10:             switch (type) {
yading@10:             case DWT_97:
yading@10:                 spatial_compose97i_dy(cs + level, buffer, temp, width >> level,
yading@10:                                       height >> level, stride << level);
yading@10:                 break;
yading@10:             case DWT_53:
yading@10:                 spatial_compose53i_dy(cs + level, buffer, temp, width >> level,
yading@10:                                       height >> level, stride << level);
yading@10:                 break;
yading@10:             }
yading@10:         }
yading@10: }
yading@10: 
yading@10: void ff_spatial_idwt(IDWTELEM *buffer, IDWTELEM *temp, int width, int height,
yading@10:                      int stride, int type, int decomposition_count)
yading@10: {
yading@10:     DWTCompose cs[MAX_DECOMPOSITIONS];
yading@10:     int y;
yading@10:     ff_spatial_idwt_init(cs, buffer, width, height, stride, type,
yading@10:                          decomposition_count);
yading@10:     for (y = 0; y < height; y += 4)
yading@10:         ff_spatial_idwt_slice(cs, buffer, temp, width, height, stride, type,
yading@10:                               decomposition_count, y);
yading@10: }
yading@10: 
yading@10: static inline int w_c(void *v, uint8_t *pix1, uint8_t *pix2, int line_size,
yading@10:                       int w, int h, int type)
yading@10: {
yading@10:     int s, i, j;
yading@10:     const int dec_count = w == 8 ? 3 : 4;
yading@10:     int tmp[32 * 32], tmp2[32];
yading@10:     int level, ori;
yading@10:     static const int scale[2][2][4][4] = {
yading@10:         {
yading@10:             { // 9/7 8x8 dec=3
yading@10:                 { 268, 239, 239, 213 },
yading@10:                 { 0,   224, 224, 152 },
yading@10:                 { 0,   135, 135, 110 },
yading@10:             },
yading@10:             { // 9/7 16x16 or 32x32 dec=4
yading@10:                 { 344, 310, 310, 280 },
yading@10:                 { 0,   320, 320, 228 },
yading@10:                 { 0,   175, 175, 136 },
yading@10:                 { 0,   129, 129, 102 },
yading@10:             }
yading@10:         },
yading@10:         {
yading@10:             { // 5/3 8x8 dec=3
yading@10:                 { 275, 245, 245, 218 },
yading@10:                 { 0,   230, 230, 156 },
yading@10:                 { 0,   138, 138, 113 },
yading@10:             },
yading@10:             { // 5/3 16x16 or 32x32 dec=4
yading@10:                 { 352, 317, 317, 286 },
yading@10:                 { 0,   328, 328, 233 },
yading@10:                 { 0,   180, 180, 140 },
yading@10:                 { 0,   132, 132, 105 },
yading@10:             }
yading@10:         }
yading@10:     };
yading@10: 
yading@10:     for (i = 0; i < h; i++) {
yading@10:         for (j = 0; j < w; j += 4) {
yading@10:             tmp[32 * i + j + 0] = (pix1[j + 0] - pix2[j + 0]) << 4;
yading@10:             tmp[32 * i + j + 1] = (pix1[j + 1] - pix2[j + 1]) << 4;
yading@10:             tmp[32 * i + j + 2] = (pix1[j + 2] - pix2[j + 2]) << 4;
yading@10:             tmp[32 * i + j + 3] = (pix1[j + 3] - pix2[j + 3]) << 4;
yading@10:         }
yading@10:         pix1 += line_size;
yading@10:         pix2 += line_size;
yading@10:     }
yading@10: 
yading@10:     ff_spatial_dwt(tmp, tmp2, w, h, 32, type, dec_count);
yading@10: 
yading@10:     s = 0;
yading@10:     av_assert1(w == h);
yading@10:     for (level = 0; level < dec_count; level++)
yading@10:         for (ori = level ? 1 : 0; ori < 4; ori++) {
yading@10:             int size   = w >> (dec_count - level);
yading@10:             int sx     = (ori & 1) ? size : 0;
yading@10:             int stride = 32 << (dec_count - level);
yading@10:             int sy     = (ori & 2) ? stride >> 1 : 0;
yading@10: 
yading@10:             for (i = 0; i < size; i++)
yading@10:                 for (j = 0; j < size; j++) {
yading@10:                     int v = tmp[sx + sy + i * stride + j] *
yading@10:                             scale[type][dec_count - 3][level][ori];
yading@10:                     s += FFABS(v);
yading@10:                 }
yading@10:         }
yading@10:     av_assert1(s >= 0);
yading@10:     return s >> 9;
yading@10: }
yading@10: 
yading@10: static int w53_8_c(void *v, uint8_t *pix1, uint8_t *pix2, int line_size, int h)
yading@10: {
yading@10:     return w_c(v, pix1, pix2, line_size, 8, h, 1);
yading@10: }
yading@10: 
yading@10: static int w97_8_c(void *v, uint8_t *pix1, uint8_t *pix2, int line_size, int h)
yading@10: {
yading@10:     return w_c(v, pix1, pix2, line_size, 8, h, 0);
yading@10: }
yading@10: 
yading@10: static int w53_16_c(void *v, uint8_t *pix1, uint8_t *pix2, int line_size, int h)
yading@10: {
yading@10:     return w_c(v, pix1, pix2, line_size, 16, h, 1);
yading@10: }
yading@10: 
yading@10: static int w97_16_c(void *v, uint8_t *pix1, uint8_t *pix2, int line_size, int h)
yading@10: {
yading@10:     return w_c(v, pix1, pix2, line_size, 16, h, 0);
yading@10: }
yading@10: 
yading@10: int ff_w53_32_c(void *v, uint8_t *pix1, uint8_t *pix2, int line_size, int h)
yading@10: {
yading@10:     return w_c(v, pix1, pix2, line_size, 32, h, 1);
yading@10: }
yading@10: 
yading@10: int ff_w97_32_c(void *v, uint8_t *pix1, uint8_t *pix2, int line_size, int h)
yading@10: {
yading@10:     return w_c(v, pix1, pix2, line_size, 32, h, 0);
yading@10: }
yading@10: 
yading@10: void ff_dsputil_init_dwt(DSPContext *c)
yading@10: {
yading@10:     c->w53[0] = w53_16_c;
yading@10:     c->w53[1] = w53_8_c;
yading@10:     c->w97[0] = w97_16_c;
yading@10:     c->w97[1] = w97_8_c;
yading@10: }
yading@10: 
yading@10: void ff_dwt_init(SnowDWTContext *c)
yading@10: {
yading@10:     c->vertical_compose97i   = ff_snow_vertical_compose97i;
yading@10:     c->horizontal_compose97i = ff_snow_horizontal_compose97i;
yading@10:     c->inner_add_yblock      = ff_snow_inner_add_yblock;
yading@10: 
yading@10:     if (HAVE_MMX)
yading@10:         ff_dwt_init_x86(c);
yading@10: }
yading@10: 
yading@10: