media/video/av1/src/cdef_tmpl.c

   1 /*
   2  * Copyright © 2018, VideoLAN and dav1d authors
   3  * Copyright © 2018, Two Orioles, LLC
   4  * All rights reserved.
   5  *
   6  * Redistribution and use in source and binary forms, with or without
   7  * modification, are permitted provided that the following conditions are met:
   8  *
   9  * 1. Redistributions of source code must retain the above copyright notice, this
  10  *    list of conditions and the following disclaimer.
  11  *
  12  * 2. Redistributions in binary form must reproduce the above copyright notice,
  13  *    this list of conditions and the following disclaimer in the documentation
  14  *    and/or other materials provided with the distribution.
  15  *
  16  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND
  17  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
  18  * WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
  19  * DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE FOR
  20  * ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
  21  * (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
  22  * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND
  23  * ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
  24  * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
  25  * SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
  26  */
  27
  28 #include "config.h"
  29
  30 #include <stdlib.h>
  31 #include <stdio.h>
  32
  33 #include "common/intops.h"
  34
  35 #include "src/cdef.h"
  36 #include "src/tables.h"
  37
  38 static int cdef_find_dir_c(const pixel *img, const ptrdiff_t stride,
  39                            unsigned *const var HIGHBD_DECL_SUFFIX)
  40 {
  41     const int bitdepth_min_8 = bitdepth_from_max(bitdepth_max) - 8;
  42     printf("bitdepth_max = %d, bitdepth_min_8 = %d\n", bitdepth_max, bitdepth_min_8);
  43     int partial_sum_hv[2][8] = { { 0 } };
  44     int partial_sum_diag[2][15] = { { 0 } };
  45     int partial_sum_alt[4][11] = { { 0 } };
  46
  47     for (int y = 0; y < 8; y++) {
  48         printf("img: y: %d : %p -> \t", y, img);
  49         for (int x = 0; x < 8; x++) {
  50             const int px = (img[x] >> bitdepth_min_8) - 128;
  51             printf("%04x/%04x ", img[x], px);
  52
  53             partial_sum_diag[0][     y       +  x      ] += px;
  54             partial_sum_alt [0][     y       + (x >> 1)] += px;
  55             partial_sum_hv  [0][     y                 ] += px;
  56             partial_sum_alt [1][3 +  y       - (x >> 1)] += px;
  57             partial_sum_diag[1][7 +  y       -  x      ] += px;
  58             partial_sum_alt [2][3 - (y >> 1) +  x      ] += px;
  59             partial_sum_hv  [1][                x      ] += px;
  60             partial_sum_alt [3][    (y >> 1) +  x      ] += px;
  61         }
  62         printf("\n");
  63         img += PXSTRIDE(stride);
  64     }
  65
  66     printf("partial_sum_alt : \n");
  67     for (int y = 0; y < 4; y++) {
  68       for (int x = 0; x < 11; x++) {
  69         printf("%08x ", partial_sum_alt[y][x]);
  70       }
  71       printf("\n");
  72     }
  73     printf("\n");
  74
  75     unsigned cost[8] = { 0 };
  76     for (int n = 0; n < 8; n++) {
  77         cost[2] += partial_sum_hv[0][n] * partial_sum_hv[0][n];
  78         cost[6] += partial_sum_hv[1][n] * partial_sum_hv[1][n];
  79     }
  80     cost[2] *= 105;
  81     cost[6] *= 105;
  82
  83     static const uint16_t div_table[7] = { 840, 420, 280, 210, 168, 140, 120 };
  84     for (int n = 0; n < 7; n++) {
  85         const int d = div_table[n];
  86         printf("n: %d\n", n);
  87         int t = partial_sum_diag[0][14 - n];
  88         printf("partial_sum_diag[0][14 - %d] = %d/%08x, partial_sum_diag[0][14 - %d]^2 = %d/%08x\n", n, t, t , n, t * t, t * t);
  89         t = (partial_sum_diag[0][n]      * partial_sum_diag[0][n] +
  90                     partial_sum_diag[0][14 - n] * partial_sum_diag[0][14 - n]) * d;
  91         printf("t = %d/%08x, d = %d/%08x, t * d = %d/%08x\n", t, t, d, d, t * d, t * d);
  92         cost[0] += t * d;
  93         cost[4] += (partial_sum_diag[1][n]      * partial_sum_diag[1][n] +
  94                     partial_sum_diag[1][14 - n] * partial_sum_diag[1][14 - n]) * d;
  95     }
  96     cost[0] += partial_sum_diag[0][7] * partial_sum_diag[0][7] * 105;
  97     cost[4] += partial_sum_diag[1][7] * partial_sum_diag[1][7] * 105;
  98
  99     printf("cost: \n");
 100     for (int y = 0; y < 8; y++) {
 101       printf("%08x ", cost[y]);
 102     }
 103     printf("\n");
 104
 105     for (int n = 0; n < 4; n++) {
 106         unsigned *const cost_ptr = &cost[n * 2 + 1];
 107         for (int m = 0; m < 5; m++)
 108             *cost_ptr += partial_sum_alt[n][3 + m] * partial_sum_alt[n][3 + m];
 109         *cost_ptr *= 105;
 110         for (int m = 0; m < 3; m++) {
 111             const int d = div_table[2 * m + 1];
 112             *cost_ptr += (partial_sum_alt[n][m]      * partial_sum_alt[n][m] +
 113                           partial_sum_alt[n][10 - m] * partial_sum_alt[n][10 - m]) * d;
 114         }
 115     }
 116     printf("cost: \n");
 117     for (int y = 0; y < 8; y++) {
 118       printf("%08x ", cost[y]);
 119     }
 120     printf("\n");
 121
 122     int best_dir = 0;
 123     unsigned best_cost = cost[0];
 124     for (int n = 1; n < 8; n++) {
 125         if (cost[n] > best_cost) {
 126             best_cost = cost[n];
 127             best_dir = n;
 128         }
 129     }
 130
 131     *var = (best_cost - (cost[best_dir ^ 4])) >> 10;
 132     return best_dir;
 133 }
 134
 135 #if HAVE_ASM
 136 #if ARCH_AARCH64 || ARCH_ARM
 137 #include "src/arm/cdef.h"
 138 #elif ARCH_PPC64LE
 139 #include "src/ppc/cdef.h"
 140 #elif ARCH_X86
 141 #include "src/x86/cdef.h"
 142 #endif
 143 #endif
 144
 145 COLD void bitfn(dav1d_cdef_dsp_init)(Dav1dCdefDSPContext *const c) {
 146     c->dir = cdef_find_dir_c;
 147
 148 #if HAVE_ASM
 149 #if ARCH_AARCH64 || ARCH_ARM
 150     cdef_dsp_init_arm(c);
 151 #elif ARCH_PPC64LE
 152     cdef_dsp_init_ppc(c);
 153 #elif ARCH_X86
 154     cdef_dsp_init_x86(c);
 155 #endif
 156 #endif
 157 }