Thu, 22 Jan 2015 13:21:57 +0100
Incorporate requested changes from Mozilla in review:
https://bugzilla.mozilla.org/show_bug.cgi?id=1123480#c6
michael@0 | 1 | /* |
michael@0 | 2 | * Copyright (c) 2010 The WebM project authors. All Rights Reserved. |
michael@0 | 3 | * |
michael@0 | 4 | * Use of this source code is governed by a BSD-style license |
michael@0 | 5 | * that can be found in the LICENSE file in the root of the source |
michael@0 | 6 | * tree. An additional intellectual property rights grant can be found |
michael@0 | 7 | * in the file PATENTS. All contributing project authors may |
michael@0 | 8 | * be found in the AUTHORS file in the root of the source tree. |
michael@0 | 9 | */ |
michael@0 | 10 | |
michael@0 | 11 | |
michael@0 | 12 | /**************************************************************************** |
michael@0 | 13 | * Notes: |
michael@0 | 14 | * |
michael@0 | 15 | * This implementation makes use of 16 bit fixed point verio of two multiply |
michael@0 | 16 | * constants: |
michael@0 | 17 | * 1. sqrt(2) * cos (pi/8) |
michael@0 | 18 | * 2. sqrt(2) * sin (pi/8) |
michael@0 | 19 | * Becuase the first constant is bigger than 1, to maintain the same 16 bit |
michael@0 | 20 | * fixed point precision as the second one, we use a trick of |
michael@0 | 21 | * x * a = x + x*(a-1) |
michael@0 | 22 | * so |
michael@0 | 23 | * x * sqrt(2) * cos (pi/8) = x + x * (sqrt(2) *cos(pi/8)-1). |
michael@0 | 24 | **************************************************************************/ |
michael@0 | 25 | static const int cospi8sqrt2minus1 = 20091; |
michael@0 | 26 | static const int sinpi8sqrt2 = 35468; |
michael@0 | 27 | |
michael@0 | 28 | void vp8_short_idct4x4llm_c(short *input, unsigned char *pred_ptr, |
michael@0 | 29 | int pred_stride, unsigned char *dst_ptr, |
michael@0 | 30 | int dst_stride) |
michael@0 | 31 | { |
michael@0 | 32 | int i; |
michael@0 | 33 | int r, c; |
michael@0 | 34 | int a1, b1, c1, d1; |
michael@0 | 35 | short output[16]; |
michael@0 | 36 | short *ip = input; |
michael@0 | 37 | short *op = output; |
michael@0 | 38 | int temp1, temp2; |
michael@0 | 39 | int shortpitch = 4; |
michael@0 | 40 | |
michael@0 | 41 | for (i = 0; i < 4; i++) |
michael@0 | 42 | { |
michael@0 | 43 | a1 = ip[0] + ip[8]; |
michael@0 | 44 | b1 = ip[0] - ip[8]; |
michael@0 | 45 | |
michael@0 | 46 | temp1 = (ip[4] * sinpi8sqrt2) >> 16; |
michael@0 | 47 | temp2 = ip[12] + ((ip[12] * cospi8sqrt2minus1) >> 16); |
michael@0 | 48 | c1 = temp1 - temp2; |
michael@0 | 49 | |
michael@0 | 50 | temp1 = ip[4] + ((ip[4] * cospi8sqrt2minus1) >> 16); |
michael@0 | 51 | temp2 = (ip[12] * sinpi8sqrt2) >> 16; |
michael@0 | 52 | d1 = temp1 + temp2; |
michael@0 | 53 | |
michael@0 | 54 | op[shortpitch*0] = a1 + d1; |
michael@0 | 55 | op[shortpitch*3] = a1 - d1; |
michael@0 | 56 | |
michael@0 | 57 | op[shortpitch*1] = b1 + c1; |
michael@0 | 58 | op[shortpitch*2] = b1 - c1; |
michael@0 | 59 | |
michael@0 | 60 | ip++; |
michael@0 | 61 | op++; |
michael@0 | 62 | } |
michael@0 | 63 | |
michael@0 | 64 | ip = output; |
michael@0 | 65 | op = output; |
michael@0 | 66 | |
michael@0 | 67 | for (i = 0; i < 4; i++) |
michael@0 | 68 | { |
michael@0 | 69 | a1 = ip[0] + ip[2]; |
michael@0 | 70 | b1 = ip[0] - ip[2]; |
michael@0 | 71 | |
michael@0 | 72 | temp1 = (ip[1] * sinpi8sqrt2) >> 16; |
michael@0 | 73 | temp2 = ip[3] + ((ip[3] * cospi8sqrt2minus1) >> 16); |
michael@0 | 74 | c1 = temp1 - temp2; |
michael@0 | 75 | |
michael@0 | 76 | temp1 = ip[1] + ((ip[1] * cospi8sqrt2minus1) >> 16); |
michael@0 | 77 | temp2 = (ip[3] * sinpi8sqrt2) >> 16; |
michael@0 | 78 | d1 = temp1 + temp2; |
michael@0 | 79 | |
michael@0 | 80 | |
michael@0 | 81 | op[0] = (a1 + d1 + 4) >> 3; |
michael@0 | 82 | op[3] = (a1 - d1 + 4) >> 3; |
michael@0 | 83 | |
michael@0 | 84 | op[1] = (b1 + c1 + 4) >> 3; |
michael@0 | 85 | op[2] = (b1 - c1 + 4) >> 3; |
michael@0 | 86 | |
michael@0 | 87 | ip += shortpitch; |
michael@0 | 88 | op += shortpitch; |
michael@0 | 89 | } |
michael@0 | 90 | |
michael@0 | 91 | ip = output; |
michael@0 | 92 | for (r = 0; r < 4; r++) |
michael@0 | 93 | { |
michael@0 | 94 | for (c = 0; c < 4; c++) |
michael@0 | 95 | { |
michael@0 | 96 | int a = ip[c] + pred_ptr[c] ; |
michael@0 | 97 | |
michael@0 | 98 | if (a < 0) |
michael@0 | 99 | a = 0; |
michael@0 | 100 | |
michael@0 | 101 | if (a > 255) |
michael@0 | 102 | a = 255; |
michael@0 | 103 | |
michael@0 | 104 | dst_ptr[c] = (unsigned char) a ; |
michael@0 | 105 | } |
michael@0 | 106 | ip += 4; |
michael@0 | 107 | dst_ptr += dst_stride; |
michael@0 | 108 | pred_ptr += pred_stride; |
michael@0 | 109 | } |
michael@0 | 110 | } |
michael@0 | 111 | |
michael@0 | 112 | void vp8_dc_only_idct_add_c(short input_dc, unsigned char *pred_ptr, |
michael@0 | 113 | int pred_stride, unsigned char *dst_ptr, |
michael@0 | 114 | int dst_stride) |
michael@0 | 115 | { |
michael@0 | 116 | int a1 = ((input_dc + 4) >> 3); |
michael@0 | 117 | int r, c; |
michael@0 | 118 | |
michael@0 | 119 | for (r = 0; r < 4; r++) |
michael@0 | 120 | { |
michael@0 | 121 | for (c = 0; c < 4; c++) |
michael@0 | 122 | { |
michael@0 | 123 | int a = a1 + pred_ptr[c] ; |
michael@0 | 124 | |
michael@0 | 125 | if (a < 0) |
michael@0 | 126 | a = 0; |
michael@0 | 127 | |
michael@0 | 128 | if (a > 255) |
michael@0 | 129 | a = 255; |
michael@0 | 130 | |
michael@0 | 131 | dst_ptr[c] = (unsigned char) a ; |
michael@0 | 132 | } |
michael@0 | 133 | |
michael@0 | 134 | dst_ptr += dst_stride; |
michael@0 | 135 | pred_ptr += pred_stride; |
michael@0 | 136 | } |
michael@0 | 137 | |
michael@0 | 138 | } |
michael@0 | 139 | |
michael@0 | 140 | void vp8_short_inv_walsh4x4_c(short *input, short *mb_dqcoeff) |
michael@0 | 141 | { |
michael@0 | 142 | short output[16]; |
michael@0 | 143 | int i; |
michael@0 | 144 | int a1, b1, c1, d1; |
michael@0 | 145 | int a2, b2, c2, d2; |
michael@0 | 146 | short *ip = input; |
michael@0 | 147 | short *op = output; |
michael@0 | 148 | |
michael@0 | 149 | for (i = 0; i < 4; i++) |
michael@0 | 150 | { |
michael@0 | 151 | a1 = ip[0] + ip[12]; |
michael@0 | 152 | b1 = ip[4] + ip[8]; |
michael@0 | 153 | c1 = ip[4] - ip[8]; |
michael@0 | 154 | d1 = ip[0] - ip[12]; |
michael@0 | 155 | |
michael@0 | 156 | op[0] = a1 + b1; |
michael@0 | 157 | op[4] = c1 + d1; |
michael@0 | 158 | op[8] = a1 - b1; |
michael@0 | 159 | op[12] = d1 - c1; |
michael@0 | 160 | ip++; |
michael@0 | 161 | op++; |
michael@0 | 162 | } |
michael@0 | 163 | |
michael@0 | 164 | ip = output; |
michael@0 | 165 | op = output; |
michael@0 | 166 | |
michael@0 | 167 | for (i = 0; i < 4; i++) |
michael@0 | 168 | { |
michael@0 | 169 | a1 = ip[0] + ip[3]; |
michael@0 | 170 | b1 = ip[1] + ip[2]; |
michael@0 | 171 | c1 = ip[1] - ip[2]; |
michael@0 | 172 | d1 = ip[0] - ip[3]; |
michael@0 | 173 | |
michael@0 | 174 | a2 = a1 + b1; |
michael@0 | 175 | b2 = c1 + d1; |
michael@0 | 176 | c2 = a1 - b1; |
michael@0 | 177 | d2 = d1 - c1; |
michael@0 | 178 | |
michael@0 | 179 | op[0] = (a2 + 3) >> 3; |
michael@0 | 180 | op[1] = (b2 + 3) >> 3; |
michael@0 | 181 | op[2] = (c2 + 3) >> 3; |
michael@0 | 182 | op[3] = (d2 + 3) >> 3; |
michael@0 | 183 | |
michael@0 | 184 | ip += 4; |
michael@0 | 185 | op += 4; |
michael@0 | 186 | } |
michael@0 | 187 | |
michael@0 | 188 | for(i = 0; i < 16; i++) |
michael@0 | 189 | { |
michael@0 | 190 | mb_dqcoeff[i * 16] = output[i]; |
michael@0 | 191 | } |
michael@0 | 192 | } |
michael@0 | 193 | |
michael@0 | 194 | void vp8_short_inv_walsh4x4_1_c(short *input, short *mb_dqcoeff) |
michael@0 | 195 | { |
michael@0 | 196 | int i; |
michael@0 | 197 | int a1; |
michael@0 | 198 | |
michael@0 | 199 | a1 = ((input[0] + 3) >> 3); |
michael@0 | 200 | for(i = 0; i < 16; i++) |
michael@0 | 201 | { |
michael@0 | 202 | mb_dqcoeff[i * 16] = a1; |
michael@0 | 203 | } |
michael@0 | 204 | } |