1 /******************************************************************************
2 *
3 * Copyright (C) 1999-2012 Broadcom Corporation
4 *
5 * Licensed under the Apache License, Version 2.0 (the "License");
6 * you may not use this file except in compliance with the License.
7 * You may obtain a copy of the License at:
8 *
9 * http://www.apache.org/licenses/LICENSE-2.0
10 *
11 * Unless required by applicable law or agreed to in writing, software
12 * distributed under the License is distributed on an "AS IS" BASIS,
13 * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
14 * See the License for the specific language governing permissions and
15 * limitations under the License.
16 *
17 ******************************************************************************/
18
19 /******************************************************************************
20 *
21 * source file for fast dct operations
22 *
23 ******************************************************************************/
24 #include "common/bt_target.h"
25 #include "sbc_encoder.h"
26 #include "sbc_enc_func_declare.h"
27 #include "sbc_dct.h"
28
29
30 #if (defined(SBC_ENC_INCLUDED) && SBC_ENC_INCLUDED == TRUE)
31
32 /*******************************************************************************
33 **
34 ** Function SBC_FastIDCT8
35 **
36 ** Description implementation of fast DCT algorithm by Feig and Winograd
37 **
38 **
39 ** Returns y = dct(pInVect)
40 **
41 **
42 *******************************************************************************/
43
44 #if (SBC_IS_64_MULT_IN_IDCT == FALSE)
45 #define SBC_COS_PI_SUR_4 (0x00005a82) /* ((0x8000) * 0.7071) = cos(pi/4) */
46 #define SBC_COS_PI_SUR_8 (0x00007641) /* ((0x8000) * 0.9239) = (cos(pi/8)) */
47 #define SBC_COS_3PI_SUR_8 (0x000030fb) /* ((0x8000) * 0.3827) = (cos(3*pi/8)) */
48 #define SBC_COS_PI_SUR_16 (0x00007d8a) /* ((0x8000) * 0.9808)) = (cos(pi/16)) */
49 #define SBC_COS_3PI_SUR_16 (0x00006a6d) /* ((0x8000) * 0.8315)) = (cos(3*pi/16)) */
50 #define SBC_COS_5PI_SUR_16 (0x0000471c) /* ((0x8000) * 0.5556)) = (cos(5*pi/16)) */
51 #define SBC_COS_7PI_SUR_16 (0x000018f8) /* ((0x8000) * 0.1951)) = (cos(7*pi/16)) */
52 #define SBC_IDCT_MULT(a,b,c) SBC_MULT_32_16_SIMPLIFIED(a,b,c)
53 #else
54 #define SBC_COS_PI_SUR_4 (0x5A827999) /* ((0x80000000) * 0.707106781) = (cos(pi/4) ) */
55 #define SBC_COS_PI_SUR_8 (0x7641AF3C) /* ((0x80000000) * 0.923879533) = (cos(pi/8) ) */
56 #define SBC_COS_3PI_SUR_8 (0x30FBC54D) /* ((0x80000000) * 0.382683432) = (cos(3*pi/8) ) */
57 #define SBC_COS_PI_SUR_16 (0x7D8A5F3F) /* ((0x80000000) * 0.98078528 )) = (cos(pi/16) ) */
58 #define SBC_COS_3PI_SUR_16 (0x6A6D98A4) /* ((0x80000000) * 0.831469612)) = (cos(3*pi/16)) */
59 #define SBC_COS_5PI_SUR_16 (0x471CECE6) /* ((0x80000000) * 0.555570233)) = (cos(5*pi/16)) */
60 #define SBC_COS_7PI_SUR_16 (0x18F8B83C) /* ((0x80000000) * 0.195090322)) = (cos(7*pi/16)) */
61 #define SBC_IDCT_MULT(a,b,c) SBC_MULT_32_32(a,b,c)
62 #endif /* SBC_IS_64_MULT_IN_IDCT */
63
64 #if (SBC_FAST_DCT == FALSE)
65 extern const SINT16 gas16AnalDCTcoeff8[];
66 extern const SINT16 gas16AnalDCTcoeff4[];
67 #endif
68
SBC_FastIDCT8(SINT32 * pInVect,SINT32 * pOutVect)69 void SBC_FastIDCT8(SINT32 *pInVect, SINT32 *pOutVect)
70 {
71 #if (SBC_FAST_DCT == TRUE)
72 #if (SBC_ARM_ASM_OPT==TRUE)
73 #else
74 #if (SBC_IPAQ_OPT==TRUE)
75 #if (SBC_IS_64_MULT_IN_IDCT == TRUE)
76 SINT64 s64Temp;
77 #endif
78 #else
79 #if (SBC_IS_64_MULT_IN_IDCT == TRUE)
80 SINT32 s32HiTemp;
81 #else
82 SINT32 s32In2Temp;
83 register SINT32 s32In1Temp;
84 #endif
85 #endif
86 #endif
87
88 register SINT32 x0, x1, x2, x3, x4, x5, x6, x7, temp;
89 SINT32 res_even[4], res_odd[4];
90 /*x0= (pInVect[4])/2 ;*/
91 SBC_IDCT_MULT(SBC_COS_PI_SUR_4, pInVect[4], x0);
92 /*printf("x0 0x%x = %d = %d * %d\n", x0, x0, SBC_COS_PI_SUR_4, pInVect[4]);*/
93
94 x1 = (pInVect[3] + pInVect[5]) >> 1;
95 x2 = (pInVect[2] + pInVect[6]) >> 1;
96 x3 = (pInVect[1] + pInVect[7]) >> 1;
97 x4 = (pInVect[0] + pInVect[8]) >> 1;
98 x5 = (pInVect[9] - pInVect[15]) >> 1;
99 x6 = (pInVect[10] - pInVect[14]) >> 1;
100 x7 = (pInVect[11] - pInVect[13]) >> 1;
101
102 /* 2-point IDCT of x0 and x4 as in (11) */
103 temp = x0 ;
104 SBC_IDCT_MULT(SBC_COS_PI_SUR_4, ( x0 + x4 ), x0); /*x0 = ( x0 + x4 ) * cos(1*pi/4) ; */
105 SBC_IDCT_MULT(SBC_COS_PI_SUR_4, ( temp - x4 ), x4); /*x4 = ( temp - x4 ) * cos(1*pi/4) ; */
106
107 /* rearrangement of x2 and x6 as in (15) */
108 x2 -= x6;
109 x6 <<= 1 ;
110
111 /* 2-point IDCT of x2 and x6 and post-multiplication as in (15) */
112 SBC_IDCT_MULT(SBC_COS_PI_SUR_4, x6, x6); /*x6 = x6 * cos(1*pi/4) ; */
113 temp = x2 ;
114 SBC_IDCT_MULT(SBC_COS_PI_SUR_8, ( x2 + x6 ), x2); /*x2 = ( x2 + x6 ) * cos(1*pi/8) ; */
115 SBC_IDCT_MULT(SBC_COS_3PI_SUR_8, ( temp - x6 ), x6); /*x6 = ( temp - x6 ) * cos(3*pi/8) ;*/
116
117 /* 4-point IDCT of x0,x2,x4 and x6 as in (11) */
118 res_even[ 0 ] = x0 + x2 ;
119 res_even[ 1 ] = x4 + x6 ;
120 res_even[ 2 ] = x4 - x6 ;
121 res_even[ 3 ] = x0 - x2 ;
122
123
124 /* rearrangement of x1,x3,x5,x7 as in (15) */
125 x7 <<= 1 ;
126 x5 = ( x5 << 1 ) - x7 ;
127 x3 = ( x3 << 1 ) - x5 ;
128 x1 -= x3 >> 1 ;
129
130 /* two-dimensional IDCT of x1 and x5 */
131 SBC_IDCT_MULT(SBC_COS_PI_SUR_4, x5, x5); /*x5 = x5 * cos(1*pi/4) ; */
132 temp = x1 ;
133 x1 = x1 + x5 ;
134 x5 = temp - x5 ;
135
136 /* rearrangement of x3 and x7 as in (15) */
137 x3 -= x7;
138 x7 <<= 1 ;
139 SBC_IDCT_MULT(SBC_COS_PI_SUR_4, x7, x7); /*x7 = x7 * cos(1*pi/4) ; */
140
141 /* 2-point IDCT of x3 and x7 and post-multiplication as in (15) */
142 temp = x3 ;
143 SBC_IDCT_MULT( SBC_COS_PI_SUR_8, ( x3 + x7 ), x3); /*x3 = ( x3 + x7 ) * cos(1*pi/8) ; */
144 SBC_IDCT_MULT( SBC_COS_3PI_SUR_8, ( temp - x7 ), x7); /*x7 = ( temp - x7 ) * cos(3*pi/8) ;*/
145
146 /* 4-point IDCT of x1,x3,x5 and x7 and post multiplication by diagonal matrix as in (14) */
147 SBC_IDCT_MULT((SBC_COS_PI_SUR_16), ( x1 + x3 ) , res_odd[0]); /*res_odd[ 0 ] = ( x1 + x3 ) * cos(1*pi/16) ; */
148 SBC_IDCT_MULT((SBC_COS_3PI_SUR_16), ( x5 + x7 ) , res_odd[1]); /*res_odd[ 1 ] = ( x5 + x7 ) * cos(3*pi/16) ; */
149 SBC_IDCT_MULT((SBC_COS_5PI_SUR_16), ( x5 - x7 ) , res_odd[2]); /*res_odd[ 2 ] = ( x5 - x7 ) * cos(5*pi/16) ; */
150 SBC_IDCT_MULT((SBC_COS_7PI_SUR_16), ( x1 - x3 ) , res_odd[3]); /*res_odd[ 3 ] = ( x1 - x3 ) * cos(7*pi/16) ; */
151
152 /* additions and subtractions as in (9) */
153 pOutVect[0] = (res_even[ 0 ] + res_odd[ 0 ]) ;
154 pOutVect[1] = (res_even[ 1 ] + res_odd[ 1 ]) ;
155 pOutVect[2] = (res_even[ 2 ] + res_odd[ 2 ]) ;
156 pOutVect[3] = (res_even[ 3 ] + res_odd[ 3 ]) ;
157 pOutVect[7] = (res_even[ 0 ] - res_odd[ 0 ]) ;
158 pOutVect[6] = (res_even[ 1 ] - res_odd[ 1 ]) ;
159 pOutVect[5] = (res_even[ 2 ] - res_odd[ 2 ]) ;
160 pOutVect[4] = (res_even[ 3 ] - res_odd[ 3 ]) ;
161 #else
162 UINT8 Index, k;
163 SINT32 temp;
164 /*Calculate 4 subband samples by matrixing*/
165 for (Index = 0; Index < 8; Index++) {
166 temp = 0;
167 for (k = 0; k < 16; k++) {
168 /*temp += (SINT32)(((SINT64)M[(Index*strEncParams->numOfSubBands*2)+k] * Y[k]) >> 16 );*/
169 temp += (gas16AnalDCTcoeff8[(Index * 8 * 2) + k] * (pInVect[k] >> 16));
170 temp += ((gas16AnalDCTcoeff8[(Index * 8 * 2) + k] * (pInVect[k] & 0xFFFF)) >> 16);
171 }
172 pOutVect[Index] = temp;
173 }
174 #endif
175 /* printf("pOutVect: 0x%x;0x%x;0x%x;0x%x;0x%x;0x%x;0x%x;0x%x\n",\
176 pOutVect[0],pOutVect[1],pOutVect[2],pOutVect[3],pOutVect[4],pOutVect[5],pOutVect[6],pOutVect[7]);*/
177 }
178
179 /*******************************************************************************
180 **
181 ** Function SBC_FastIDCT4
182 **
183 ** Description implementation of fast DCT algorithm by Feig and Winograd
184 **
185 **
186 ** Returns y = dct(x0)
187 **
188 **
189 *******************************************************************************/
SBC_FastIDCT4(SINT32 * pInVect,SINT32 * pOutVect)190 void SBC_FastIDCT4(SINT32 *pInVect, SINT32 *pOutVect)
191 {
192 #if (SBC_FAST_DCT == TRUE)
193 #if (SBC_ARM_ASM_OPT==TRUE)
194 #else
195 #if (SBC_IPAQ_OPT==TRUE)
196 #if (SBC_IS_64_MULT_IN_IDCT == TRUE)
197 SINT64 s64Temp;
198 #endif
199 #else
200 #if (SBC_IS_64_MULT_IN_IDCT == TRUE)
201 SINT32 s32HiTemp;
202 #else
203 UINT16 s32In2Temp;
204 SINT32 s32In1Temp;
205 #endif
206 #endif
207 #endif
208 SINT32 temp, x2;
209 SINT32 tmp[8];
210
211 x2 = pInVect[2] >> 1;
212 temp = (pInVect[0] + pInVect[4]);
213 SBC_IDCT_MULT((SBC_COS_PI_SUR_4 >> 1), temp , tmp[0]);
214 tmp[1] = x2 - tmp[0];
215 tmp[0] += x2;
216 temp = (pInVect[1] + pInVect[3]);
217 SBC_IDCT_MULT((SBC_COS_3PI_SUR_8 >> 1), temp , tmp[3]);
218 SBC_IDCT_MULT((SBC_COS_PI_SUR_8 >> 1), temp , tmp[2]);
219 temp = (pInVect[5] - pInVect[7]);
220 SBC_IDCT_MULT((SBC_COS_3PI_SUR_8 >> 1), temp , tmp[5]);
221 SBC_IDCT_MULT((SBC_COS_PI_SUR_8 >> 1), temp , tmp[4]);
222 tmp[6] = tmp[2] + tmp[5];
223 tmp[7] = tmp[3] - tmp[4];
224 pOutVect[0] = (tmp[0] + tmp[6]);
225 pOutVect[1] = (tmp[1] + tmp[7]);
226 pOutVect[2] = (tmp[1] - tmp[7]);
227 pOutVect[3] = (tmp[0] - tmp[6]);
228 #else
229 UINT8 Index, k;
230 SINT32 temp;
231 /*Calculate 4 subband samples by matrixing*/
232 for (Index = 0; Index < 4; Index++) {
233 temp = 0;
234 for (k = 0; k < 8; k++) {
235 /*temp += (SINT32)(((SINT64)M[(Index*strEncParams->numOfSubBands*2)+k] * Y[k]) >> 16 ); */
236 temp += (gas16AnalDCTcoeff4[(Index * 4 * 2) + k] * (pInVect[k] >> 16));
237 temp += ((gas16AnalDCTcoeff4[(Index * 4 * 2) + k] * (pInVect[k] & 0xFFFF)) >> 16);
238 }
239 pOutVect[Index] = temp;
240 }
241 #endif
242 }
243
244 #endif /* #if (defined(SBC_ENC_INCLUDED) && SBC_ENC_INCLUDED == TRUE) */
245