Line data Source code
1 : /*
2 : * Copyright (c) 2010 The WebM project authors. All Rights Reserved.
3 : *
4 : * Use of this source code is governed by a BSD-style license
5 : * that can be found in the LICENSE file in the root of the source
6 : * tree. An additional intellectual property rights grant can be found
7 : * in the file PATENTS. All contributing project authors may
8 : * be found in the AUTHORS file in the root of the source tree.
9 : */
10 :
11 : #include "./vp8_rtcd.h"
12 :
13 : /****************************************************************************
14 : * Notes:
15 : *
16 : * This implementation makes use of 16 bit fixed point verio of two multiply
17 : * constants:
18 : * 1. sqrt(2) * cos (pi/8)
19 : * 2. sqrt(2) * sin (pi/8)
20 : * Becuase the first constant is bigger than 1, to maintain the same 16 bit
21 : * fixed point precision as the second one, we use a trick of
22 : * x * a = x + x*(a-1)
23 : * so
24 : * x * sqrt(2) * cos (pi/8) = x + x * (sqrt(2) *cos(pi/8)-1).
25 : **************************************************************************/
26 : static const int cospi8sqrt2minus1 = 20091;
27 : static const int sinpi8sqrt2 = 35468;
28 :
29 0 : void vp8_short_idct4x4llm_c(short *input, unsigned char *pred_ptr,
30 : int pred_stride, unsigned char *dst_ptr,
31 : int dst_stride) {
32 : int i;
33 : int r, c;
34 : int a1, b1, c1, d1;
35 : short output[16];
36 0 : short *ip = input;
37 0 : short *op = output;
38 : int temp1, temp2;
39 0 : int shortpitch = 4;
40 :
41 0 : for (i = 0; i < 4; ++i) {
42 0 : a1 = ip[0] + ip[8];
43 0 : b1 = ip[0] - ip[8];
44 :
45 0 : temp1 = (ip[4] * sinpi8sqrt2) >> 16;
46 0 : temp2 = ip[12] + ((ip[12] * cospi8sqrt2minus1) >> 16);
47 0 : c1 = temp1 - temp2;
48 :
49 0 : temp1 = ip[4] + ((ip[4] * cospi8sqrt2minus1) >> 16);
50 0 : temp2 = (ip[12] * sinpi8sqrt2) >> 16;
51 0 : d1 = temp1 + temp2;
52 :
53 0 : op[shortpitch * 0] = a1 + d1;
54 0 : op[shortpitch * 3] = a1 - d1;
55 :
56 0 : op[shortpitch * 1] = b1 + c1;
57 0 : op[shortpitch * 2] = b1 - c1;
58 :
59 0 : ip++;
60 0 : op++;
61 : }
62 :
63 0 : ip = output;
64 0 : op = output;
65 :
66 0 : for (i = 0; i < 4; ++i) {
67 0 : a1 = ip[0] + ip[2];
68 0 : b1 = ip[0] - ip[2];
69 :
70 0 : temp1 = (ip[1] * sinpi8sqrt2) >> 16;
71 0 : temp2 = ip[3] + ((ip[3] * cospi8sqrt2minus1) >> 16);
72 0 : c1 = temp1 - temp2;
73 :
74 0 : temp1 = ip[1] + ((ip[1] * cospi8sqrt2minus1) >> 16);
75 0 : temp2 = (ip[3] * sinpi8sqrt2) >> 16;
76 0 : d1 = temp1 + temp2;
77 :
78 0 : op[0] = (a1 + d1 + 4) >> 3;
79 0 : op[3] = (a1 - d1 + 4) >> 3;
80 :
81 0 : op[1] = (b1 + c1 + 4) >> 3;
82 0 : op[2] = (b1 - c1 + 4) >> 3;
83 :
84 0 : ip += shortpitch;
85 0 : op += shortpitch;
86 : }
87 :
88 0 : ip = output;
89 0 : for (r = 0; r < 4; ++r) {
90 0 : for (c = 0; c < 4; ++c) {
91 0 : int a = ip[c] + pred_ptr[c];
92 :
93 0 : if (a < 0) a = 0;
94 :
95 0 : if (a > 255) a = 255;
96 :
97 0 : dst_ptr[c] = (unsigned char)a;
98 : }
99 0 : ip += 4;
100 0 : dst_ptr += dst_stride;
101 0 : pred_ptr += pred_stride;
102 : }
103 0 : }
104 :
105 0 : void vp8_dc_only_idct_add_c(short input_dc, unsigned char *pred_ptr,
106 : int pred_stride, unsigned char *dst_ptr,
107 : int dst_stride) {
108 0 : int a1 = ((input_dc + 4) >> 3);
109 : int r, c;
110 :
111 0 : for (r = 0; r < 4; ++r) {
112 0 : for (c = 0; c < 4; ++c) {
113 0 : int a = a1 + pred_ptr[c];
114 :
115 0 : if (a < 0) a = 0;
116 :
117 0 : if (a > 255) a = 255;
118 :
119 0 : dst_ptr[c] = (unsigned char)a;
120 : }
121 :
122 0 : dst_ptr += dst_stride;
123 0 : pred_ptr += pred_stride;
124 : }
125 0 : }
126 :
127 0 : void vp8_short_inv_walsh4x4_c(short *input, short *mb_dqcoeff) {
128 : short output[16];
129 : int i;
130 : int a1, b1, c1, d1;
131 : int a2, b2, c2, d2;
132 0 : short *ip = input;
133 0 : short *op = output;
134 :
135 0 : for (i = 0; i < 4; ++i) {
136 0 : a1 = ip[0] + ip[12];
137 0 : b1 = ip[4] + ip[8];
138 0 : c1 = ip[4] - ip[8];
139 0 : d1 = ip[0] - ip[12];
140 :
141 0 : op[0] = a1 + b1;
142 0 : op[4] = c1 + d1;
143 0 : op[8] = a1 - b1;
144 0 : op[12] = d1 - c1;
145 0 : ip++;
146 0 : op++;
147 : }
148 :
149 0 : ip = output;
150 0 : op = output;
151 :
152 0 : for (i = 0; i < 4; ++i) {
153 0 : a1 = ip[0] + ip[3];
154 0 : b1 = ip[1] + ip[2];
155 0 : c1 = ip[1] - ip[2];
156 0 : d1 = ip[0] - ip[3];
157 :
158 0 : a2 = a1 + b1;
159 0 : b2 = c1 + d1;
160 0 : c2 = a1 - b1;
161 0 : d2 = d1 - c1;
162 :
163 0 : op[0] = (a2 + 3) >> 3;
164 0 : op[1] = (b2 + 3) >> 3;
165 0 : op[2] = (c2 + 3) >> 3;
166 0 : op[3] = (d2 + 3) >> 3;
167 :
168 0 : ip += 4;
169 0 : op += 4;
170 : }
171 :
172 0 : for (i = 0; i < 16; ++i) {
173 0 : mb_dqcoeff[i * 16] = output[i];
174 : }
175 0 : }
176 :
177 0 : void vp8_short_inv_walsh4x4_1_c(short *input, short *mb_dqcoeff) {
178 : int i;
179 : int a1;
180 :
181 0 : a1 = ((input[0] + 3) >> 3);
182 0 : for (i = 0; i < 16; ++i) {
183 0 : mb_dqcoeff[i * 16] = a1;
184 : }
185 0 : }
|