1.1 --- /dev/null Thu Jan 01 00:00:00 1970 +0000
1.2 +++ b/media/libvpx/vp8/common/idctllm.c Wed Dec 31 06:09:35 2014 +0100
1.3 @@ -0,0 +1,204 @@
1.4 +/*
1.5 + * Copyright (c) 2010 The WebM project authors. All Rights Reserved.
1.6 + *
1.7 + * Use of this source code is governed by a BSD-style license
1.8 + * that can be found in the LICENSE file in the root of the source
1.9 + * tree. An additional intellectual property rights grant can be found
1.10 + * in the file PATENTS. All contributing project authors may
1.11 + * be found in the AUTHORS file in the root of the source tree.
1.12 + */
1.13 +
1.14 +
1.15 +/****************************************************************************
1.16 + * Notes:
1.17 + *
1.18 + * This implementation makes use of 16 bit fixed point verio of two multiply
1.19 + * constants:
1.20 + * 1. sqrt(2) * cos (pi/8)
1.21 + * 2. sqrt(2) * sin (pi/8)
1.22 + * Becuase the first constant is bigger than 1, to maintain the same 16 bit
1.23 + * fixed point precision as the second one, we use a trick of
1.24 + * x * a = x + x*(a-1)
1.25 + * so
1.26 + * x * sqrt(2) * cos (pi/8) = x + x * (sqrt(2) *cos(pi/8)-1).
1.27 + **************************************************************************/
1.28 +static const int cospi8sqrt2minus1 = 20091;
1.29 +static const int sinpi8sqrt2 = 35468;
1.30 +
1.31 +void vp8_short_idct4x4llm_c(short *input, unsigned char *pred_ptr,
1.32 + int pred_stride, unsigned char *dst_ptr,
1.33 + int dst_stride)
1.34 +{
1.35 + int i;
1.36 + int r, c;
1.37 + int a1, b1, c1, d1;
1.38 + short output[16];
1.39 + short *ip = input;
1.40 + short *op = output;
1.41 + int temp1, temp2;
1.42 + int shortpitch = 4;
1.43 +
1.44 + for (i = 0; i < 4; i++)
1.45 + {
1.46 + a1 = ip[0] + ip[8];
1.47 + b1 = ip[0] - ip[8];
1.48 +
1.49 + temp1 = (ip[4] * sinpi8sqrt2) >> 16;
1.50 + temp2 = ip[12] + ((ip[12] * cospi8sqrt2minus1) >> 16);
1.51 + c1 = temp1 - temp2;
1.52 +
1.53 + temp1 = ip[4] + ((ip[4] * cospi8sqrt2minus1) >> 16);
1.54 + temp2 = (ip[12] * sinpi8sqrt2) >> 16;
1.55 + d1 = temp1 + temp2;
1.56 +
1.57 + op[shortpitch*0] = a1 + d1;
1.58 + op[shortpitch*3] = a1 - d1;
1.59 +
1.60 + op[shortpitch*1] = b1 + c1;
1.61 + op[shortpitch*2] = b1 - c1;
1.62 +
1.63 + ip++;
1.64 + op++;
1.65 + }
1.66 +
1.67 + ip = output;
1.68 + op = output;
1.69 +
1.70 + for (i = 0; i < 4; i++)
1.71 + {
1.72 + a1 = ip[0] + ip[2];
1.73 + b1 = ip[0] - ip[2];
1.74 +
1.75 + temp1 = (ip[1] * sinpi8sqrt2) >> 16;
1.76 + temp2 = ip[3] + ((ip[3] * cospi8sqrt2minus1) >> 16);
1.77 + c1 = temp1 - temp2;
1.78 +
1.79 + temp1 = ip[1] + ((ip[1] * cospi8sqrt2minus1) >> 16);
1.80 + temp2 = (ip[3] * sinpi8sqrt2) >> 16;
1.81 + d1 = temp1 + temp2;
1.82 +
1.83 +
1.84 + op[0] = (a1 + d1 + 4) >> 3;
1.85 + op[3] = (a1 - d1 + 4) >> 3;
1.86 +
1.87 + op[1] = (b1 + c1 + 4) >> 3;
1.88 + op[2] = (b1 - c1 + 4) >> 3;
1.89 +
1.90 + ip += shortpitch;
1.91 + op += shortpitch;
1.92 + }
1.93 +
1.94 + ip = output;
1.95 + for (r = 0; r < 4; r++)
1.96 + {
1.97 + for (c = 0; c < 4; c++)
1.98 + {
1.99 + int a = ip[c] + pred_ptr[c] ;
1.100 +
1.101 + if (a < 0)
1.102 + a = 0;
1.103 +
1.104 + if (a > 255)
1.105 + a = 255;
1.106 +
1.107 + dst_ptr[c] = (unsigned char) a ;
1.108 + }
1.109 + ip += 4;
1.110 + dst_ptr += dst_stride;
1.111 + pred_ptr += pred_stride;
1.112 + }
1.113 +}
1.114 +
1.115 +void vp8_dc_only_idct_add_c(short input_dc, unsigned char *pred_ptr,
1.116 + int pred_stride, unsigned char *dst_ptr,
1.117 + int dst_stride)
1.118 +{
1.119 + int a1 = ((input_dc + 4) >> 3);
1.120 + int r, c;
1.121 +
1.122 + for (r = 0; r < 4; r++)
1.123 + {
1.124 + for (c = 0; c < 4; c++)
1.125 + {
1.126 + int a = a1 + pred_ptr[c] ;
1.127 +
1.128 + if (a < 0)
1.129 + a = 0;
1.130 +
1.131 + if (a > 255)
1.132 + a = 255;
1.133 +
1.134 + dst_ptr[c] = (unsigned char) a ;
1.135 + }
1.136 +
1.137 + dst_ptr += dst_stride;
1.138 + pred_ptr += pred_stride;
1.139 + }
1.140 +
1.141 +}
1.142 +
1.143 +void vp8_short_inv_walsh4x4_c(short *input, short *mb_dqcoeff)
1.144 +{
1.145 + short output[16];
1.146 + int i;
1.147 + int a1, b1, c1, d1;
1.148 + int a2, b2, c2, d2;
1.149 + short *ip = input;
1.150 + short *op = output;
1.151 +
1.152 + for (i = 0; i < 4; i++)
1.153 + {
1.154 + a1 = ip[0] + ip[12];
1.155 + b1 = ip[4] + ip[8];
1.156 + c1 = ip[4] - ip[8];
1.157 + d1 = ip[0] - ip[12];
1.158 +
1.159 + op[0] = a1 + b1;
1.160 + op[4] = c1 + d1;
1.161 + op[8] = a1 - b1;
1.162 + op[12] = d1 - c1;
1.163 + ip++;
1.164 + op++;
1.165 + }
1.166 +
1.167 + ip = output;
1.168 + op = output;
1.169 +
1.170 + for (i = 0; i < 4; i++)
1.171 + {
1.172 + a1 = ip[0] + ip[3];
1.173 + b1 = ip[1] + ip[2];
1.174 + c1 = ip[1] - ip[2];
1.175 + d1 = ip[0] - ip[3];
1.176 +
1.177 + a2 = a1 + b1;
1.178 + b2 = c1 + d1;
1.179 + c2 = a1 - b1;
1.180 + d2 = d1 - c1;
1.181 +
1.182 + op[0] = (a2 + 3) >> 3;
1.183 + op[1] = (b2 + 3) >> 3;
1.184 + op[2] = (c2 + 3) >> 3;
1.185 + op[3] = (d2 + 3) >> 3;
1.186 +
1.187 + ip += 4;
1.188 + op += 4;
1.189 + }
1.190 +
1.191 + for(i = 0; i < 16; i++)
1.192 + {
1.193 + mb_dqcoeff[i * 16] = output[i];
1.194 + }
1.195 +}
1.196 +
1.197 +void vp8_short_inv_walsh4x4_1_c(short *input, short *mb_dqcoeff)
1.198 +{
1.199 + int i;
1.200 + int a1;
1.201 +
1.202 + a1 = ((input[0] + 3) >> 3);
1.203 + for(i = 0; i < 16; i++)
1.204 + {
1.205 + mb_dqcoeff[i * 16] = a1;
1.206 + }
1.207 +}
