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
25 #include "sbc_encoder.h"
26 #include "sbc_enc_func_declare.h"
27 #include "sbc_dct.h"
28
29
30
31 /*******************************************************************************
32 **
33 ** Function SBC_FastIDCT8
34 **
35 ** Description implementation of fast DCT algorithm by Feig and Winograd
36 **
37 **
38 ** Returns y = dct(pInVect)
39 **
40 **
41 *******************************************************************************/
42
43 #if (SBC_IS_64_MULT_IN_IDCT == FALSE)
44 #define SBC_COS_PI_SUR_4 (0x00005a82) /* ((0x8000) * 0.7071) = cos(pi/4) */
45 #define SBC_COS_PI_SUR_8 (0x00007641) /* ((0x8000) * 0.9239) = (cos(pi/8)) */
46 #define SBC_COS_3PI_SUR_8 (0x000030fb) /* ((0x8000) * 0.3827) = (cos(3*pi/8)) */
47 #define SBC_COS_PI_SUR_16 (0x00007d8a) /* ((0x8000) * 0.9808)) = (cos(pi/16)) */
48 #define SBC_COS_3PI_SUR_16 (0x00006a6d) /* ((0x8000) * 0.8315)) = (cos(3*pi/16)) */
49 #define SBC_COS_5PI_SUR_16 (0x0000471c) /* ((0x8000) * 0.5556)) = (cos(5*pi/16)) */
50 #define SBC_COS_7PI_SUR_16 (0x000018f8) /* ((0x8000) * 0.1951)) = (cos(7*pi/16)) */
51 #define SBC_IDCT_MULT(a,b,c) SBC_MULT_32_16_SIMPLIFIED(a,b,c)
52 #else
53 #define SBC_COS_PI_SUR_4 (0x5A827999) /* ((0x80000000) * 0.707106781) = (cos(pi/4) ) */
54 #define SBC_COS_PI_SUR_8 (0x7641AF3C) /* ((0x80000000) * 0.923879533) = (cos(pi/8) ) */
55 #define SBC_COS_3PI_SUR_8 (0x30FBC54D) /* ((0x80000000) * 0.382683432) = (cos(3*pi/8) ) */
56 #define SBC_COS_PI_SUR_16 (0x7D8A5F3F) /* ((0x80000000) * 0.98078528 )) = (cos(pi/16) ) */
57 #define SBC_COS_3PI_SUR_16 (0x6A6D98A4) /* ((0x80000000) * 0.831469612)) = (cos(3*pi/16)) */
58 #define SBC_COS_5PI_SUR_16 (0x471CECE6) /* ((0x80000000) * 0.555570233)) = (cos(5*pi/16)) */
59 #define SBC_COS_7PI_SUR_16 (0x18F8B83C) /* ((0x80000000) * 0.195090322)) = (cos(7*pi/16)) */
60 #define SBC_IDCT_MULT(a,b,c) SBC_MULT_32_32(a,b,c)
61 #endif /* SBC_IS_64_MULT_IN_IDCT */
62
63 #if (SBC_FAST_DCT == FALSE)
64 extern const SINT16 gas16AnalDCTcoeff8[];
65 extern const SINT16 gas16AnalDCTcoeff4[];
66 #endif
67
SBC_FastIDCT8(SINT32 * pInVect,SINT32 * pOutVect)68 void SBC_FastIDCT8(SINT32 *pInVect, SINT32 *pOutVect)
69 {
70 #if (SBC_FAST_DCT == TRUE)
71 #if (SBC_ARM_ASM_OPT==TRUE)
72 #else
73 #if (SBC_IPAQ_OPT==TRUE)
74 #if (SBC_IS_64_MULT_IN_IDCT == TRUE)
75 SINT64 s64Temp;
76 #endif
77 #else
78 #if (SBC_IS_64_MULT_IN_IDCT == TRUE)
79 SINT32 s32HiTemp;
80 #else
81 SINT32 s32In2Temp;
82 register SINT32 s32In1Temp;
83 #endif
84 #endif
85 #endif
86
87 register SINT32 x0, x1, x2, x3, x4, x5, x6, x7,temp;
88 SINT32 res_even[4], res_odd[4];
89 /*x0= (pInVect[4])/2 ;*/
90 SBC_IDCT_MULT(SBC_COS_PI_SUR_4,pInVect[4], x0);
91 /*printf("x0 0x%x = %d = %d * %d\n", x0, x0, SBC_COS_PI_SUR_4, pInVect[4]);*/
92
93 x1 = (pInVect[3] + pInVect[5]) >>1;
94 x2 = (pInVect[2] + pInVect[6]) >>1;
95 x3 = (pInVect[1] + pInVect[7]) >>1;
96 x4 = (pInVect[0] + pInVect[8]) >>1;
97 x5 = (pInVect[9] - pInVect[15]) >>1;
98 x6 = (pInVect[10] - pInVect[14])>>1;
99 x7 = (pInVect[11] - pInVect[13])>>1;
100
101 /* 2-point IDCT of x0 and x4 as in (11) */
102 temp = x0 ;
103 SBC_IDCT_MULT(SBC_COS_PI_SUR_4, ( x0 + x4 ), x0); /*x0 = ( x0 + x4 ) * cos(1*pi/4) ; */
104 SBC_IDCT_MULT(SBC_COS_PI_SUR_4, ( temp - x4 ), x4); /*x4 = ( temp - x4 ) * cos(1*pi/4) ; */
105
106 /* rearrangement of x2 and x6 as in (15) */
107 x2 -=x6;
108 x6 <<= 1 ;
109
110 /* 2-point IDCT of x2 and x6 and post-multiplication as in (15) */
111 SBC_IDCT_MULT(SBC_COS_PI_SUR_4,x6, x6); /*x6 = x6 * cos(1*pi/4) ; */
112 temp = x2 ;
113 SBC_IDCT_MULT(SBC_COS_PI_SUR_8,( x2 + x6 ), x2); /*x2 = ( x2 + x6 ) * cos(1*pi/8) ; */
114 SBC_IDCT_MULT(SBC_COS_3PI_SUR_8,( temp - x6 ), x6); /*x6 = ( temp - x6 ) * cos(3*pi/8) ;*/
115
116 /* 4-point IDCT of x0,x2,x4 and x6 as in (11) */
117 res_even[ 0 ] = x0 + x2 ;
118 res_even[ 1 ] = x4 + x6 ;
119 res_even[ 2 ] = x4 - x6 ;
120 res_even[ 3 ] = x0 - x2 ;
121
122
123 /* rearrangement of x1,x3,x5,x7 as in (15) */
124 x7 <<= 1 ;
125 x5 = ( x5 <<1 ) - x7 ;
126 x3 = ( x3 <<1 ) - x5 ;
127 x1 -= x3 >>1 ;
128
129 /* two-dimensional IDCT of x1 and x5 */
130 SBC_IDCT_MULT(SBC_COS_PI_SUR_4, x5, x5); /*x5 = x5 * cos(1*pi/4) ; */
131 temp = x1 ;
132 x1 = x1 + x5 ;
133 x5 = temp - x5 ;
134
135 /* rearrangement of x3 and x7 as in (15) */
136 x3 -= x7;
137 x7 <<= 1 ;
138 SBC_IDCT_MULT(SBC_COS_PI_SUR_4, x7, x7); /*x7 = x7 * cos(1*pi/4) ; */
139
140 /* 2-point IDCT of x3 and x7 and post-multiplication as in (15) */
141 temp = x3 ;
142 SBC_IDCT_MULT( SBC_COS_PI_SUR_8,( x3 + x7 ), x3); /*x3 = ( x3 + x7 ) * cos(1*pi/8) ; */
143 SBC_IDCT_MULT( SBC_COS_3PI_SUR_8,( temp - x7 ), x7); /*x7 = ( temp - x7 ) * cos(3*pi/8) ;*/
144
145 /* 4-point IDCT of x1,x3,x5 and x7 and post multiplication by diagonal matrix as in (14) */
146 SBC_IDCT_MULT((SBC_COS_PI_SUR_16), ( x1 + x3 ) , res_odd[0]); /*res_odd[ 0 ] = ( x1 + x3 ) * cos(1*pi/16) ; */
147 SBC_IDCT_MULT((SBC_COS_3PI_SUR_16), ( x5 + x7 ) , res_odd[1]); /*res_odd[ 1 ] = ( x5 + x7 ) * cos(3*pi/16) ; */
148 SBC_IDCT_MULT((SBC_COS_5PI_SUR_16), ( x5 - x7 ) , res_odd[2]); /*res_odd[ 2 ] = ( x5 - x7 ) * cos(5*pi/16) ; */
149 SBC_IDCT_MULT((SBC_COS_7PI_SUR_16), ( x1 - x3 ) , res_odd[3]); /*res_odd[ 3 ] = ( x1 - x3 ) * cos(7*pi/16) ; */
150
151 /* additions and subtractions as in (9) */
152 pOutVect[0] = (res_even[ 0 ] + res_odd[ 0 ]) ;
153 pOutVect[1] = (res_even[ 1 ] + res_odd[ 1 ]) ;
154 pOutVect[2] = (res_even[ 2 ] + res_odd[ 2 ]) ;
155 pOutVect[3] = (res_even[ 3 ] + res_odd[ 3 ]) ;
156 pOutVect[7] = (res_even[ 0 ] - res_odd[ 0 ]) ;
157 pOutVect[6] = (res_even[ 1 ] - res_odd[ 1 ]) ;
158 pOutVect[5] = (res_even[ 2 ] - res_odd[ 2 ]) ;
159 pOutVect[4] = (res_even[ 3 ] - res_odd[ 3 ]) ;
160 #else
161 UINT8 Index, k;
162 SINT32 temp;
163 /*Calculate 4 subband samples by matrixing*/
164 for(Index=0; Index<8; Index++)
165 {
166 temp = 0;
167 for(k=0; k<16; k++)
168 {
169 /*temp += (SINT32)(((SINT64)M[(Index*strEncParams->numOfSubBands*2)+k] * Y[k]) >> 16 );*/
170 temp += (gas16AnalDCTcoeff8[(Index*8*2)+k] * (pInVect[k] >> 16));
171 temp += ((gas16AnalDCTcoeff8[(Index*8*2)+k] * (pInVect[k] & 0xFFFF)) >> 16);
172 }
173 pOutVect[Index] = temp;
174 }
175 #endif
176 /* printf("pOutVect: 0x%x;0x%x;0x%x;0x%x;0x%x;0x%x;0x%x;0x%x\n",\
177 pOutVect[0],pOutVect[1],pOutVect[2],pOutVect[3],pOutVect[4],pOutVect[5],pOutVect[6],pOutVect[7]);*/
178 }
179
180 /*******************************************************************************
181 **
182 ** Function SBC_FastIDCT4
183 **
184 ** Description implementation of fast DCT algorithm by Feig and Winograd
185 **
186 **
187 ** Returns y = dct(x0)
188 **
189 **
190 *******************************************************************************/
SBC_FastIDCT4(SINT32 * pInVect,SINT32 * pOutVect)191 void SBC_FastIDCT4(SINT32 *pInVect, SINT32 *pOutVect)
192 {
193 #if (SBC_FAST_DCT == TRUE)
194 #if (SBC_ARM_ASM_OPT==TRUE)
195 #else
196 #if (SBC_IPAQ_OPT==TRUE)
197 #if (SBC_IS_64_MULT_IN_IDCT == TRUE)
198 SINT64 s64Temp;
199 #endif
200 #else
201 #if (SBC_IS_64_MULT_IN_IDCT == TRUE)
202 SINT32 s32HiTemp;
203 #else
204 UINT16 s32In2Temp;
205 SINT32 s32In1Temp;
206 #endif
207 #endif
208 #endif
209 SINT32 temp,x2;
210 SINT32 tmp[8];
211
212 x2=pInVect[2]>>1;
213 temp=(pInVect[0]+pInVect[4]);
214 SBC_IDCT_MULT((SBC_COS_PI_SUR_4>>1), temp , tmp[0]);
215 tmp[1]=x2-tmp[0];
216 tmp[0]+=x2;
217 temp=(pInVect[1]+pInVect[3]);
218 SBC_IDCT_MULT((SBC_COS_3PI_SUR_8>>1), temp , tmp[3]);
219 SBC_IDCT_MULT((SBC_COS_PI_SUR_8>>1), temp , tmp[2]);
220 temp=(pInVect[5]-pInVect[7]);
221 SBC_IDCT_MULT((SBC_COS_3PI_SUR_8>>1), temp , tmp[5]);
222 SBC_IDCT_MULT((SBC_COS_PI_SUR_8>>1), temp , tmp[4]);
223 tmp[6]=tmp[2]+tmp[5];
224 tmp[7]=tmp[3]-tmp[4];
225 pOutVect[0] = (tmp[0]+tmp[6]);
226 pOutVect[1] = (tmp[1]+tmp[7]);
227 pOutVect[2] = (tmp[1]-tmp[7]);
228 pOutVect[3] = (tmp[0]-tmp[6]);
229 #else
230 UINT8 Index, k;
231 SINT32 temp;
232 /*Calculate 4 subband samples by matrixing*/
233 for(Index=0; Index<4; Index++)
234 {
235 temp = 0;
236 for(k=0; k<8; k++)
237 {
238 /*temp += (SINT32)(((SINT64)M[(Index*strEncParams->numOfSubBands*2)+k] * Y[k]) >> 16 ); */
239 temp += (gas16AnalDCTcoeff4[(Index*4*2)+k] * (pInVect[k] >> 16));
240 temp += ((gas16AnalDCTcoeff4[(Index*4*2)+k] * (pInVect[k] & 0xFFFF)) >> 16);
241 }
242 pOutVect[Index] = temp;
243 }
244 #endif
245 }
246