1/****************************************************************************** 2 * 3 * Copyright 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_dct.h" 26#include "sbc_enc_func_declare.h" 27#include "sbc_encoder.h" 28 29/******************************************************************************* 30 * 31 * Function SBC_FastIDCT8 32 * 33 * Description implementation of fast DCT algorithm by Feig and Winograd 34 * 35 * 36 * Returns y = dct(pInVect) 37 * 38 * 39 ******************************************************************************/ 40 41#if (SBC_IS_64_MULT_IN_IDCT == FALSE) 42#define SBC_COS_PI_SUR_4 \ 43 (0x00005a82) /* ((0x8000) * 0.7071) = cos(pi/4) \ 44 */ 45#define SBC_COS_PI_SUR_8 \ 46 (0x00007641) /* ((0x8000) * 0.9239) = (cos(pi/8)) */ 47#define SBC_COS_3PI_SUR_8 \ 48 (0x000030fb) /* ((0x8000) * 0.3827) = (cos(3*pi/8)) */ 49#define SBC_COS_PI_SUR_16 \ 50 (0x00007d8a) /* ((0x8000) * 0.9808)) = (cos(pi/16)) */ 51#define SBC_COS_3PI_SUR_16 \ 52 (0x00006a6d) /* ((0x8000) * 0.8315)) = (cos(3*pi/16)) */ 53#define SBC_COS_5PI_SUR_16 \ 54 (0x0000471c) /* ((0x8000) * 0.5556)) = (cos(5*pi/16)) */ 55#define SBC_COS_7PI_SUR_16 \ 56 (0x000018f8) /* ((0x8000) * 0.1951)) = (cos(7*pi/16)) */ 57#define SBC_IDCT_MULT(a, b, c) SBC_MULT_32_16_SIMPLIFIED(a, b, c) 58#else 59#define SBC_COS_PI_SUR_4 \ 60 (0x5A827999) /* ((0x80000000) * 0.707106781) = (cos(pi/4) ) */ 61#define SBC_COS_PI_SUR_8 \ 62 (0x7641AF3C) /* ((0x80000000) * 0.923879533) = (cos(pi/8) ) */ 63#define SBC_COS_3PI_SUR_8 \ 64 (0x30FBC54D) /* ((0x80000000) * 0.382683432) = (cos(3*pi/8) ) */ 65#define SBC_COS_PI_SUR_16 \ 66 (0x7D8A5F3F) /* ((0x80000000) * 0.98078528 )) = (cos(pi/16) ) */ 67#define SBC_COS_3PI_SUR_16 \ 68 (0x6A6D98A4) /* ((0x80000000) * 0.831469612)) = (cos(3*pi/16)) */ 69#define SBC_COS_5PI_SUR_16 \ 70 (0x471CECE6) /* ((0x80000000) * 0.555570233)) = (cos(5*pi/16)) */ 71#define SBC_COS_7PI_SUR_16 \ 72 (0x18F8B83C) /* ((0x80000000) * 0.195090322)) = (cos(7*pi/16)) */ 73#define SBC_IDCT_MULT(a, b, c) SBC_MULT_32_32(a, b, c) 74#endif /* SBC_IS_64_MULT_IN_IDCT */ 75 76#if (SBC_FAST_DCT == FALSE) 77extern const int16_t gas16AnalDCTcoeff8[]; 78extern const int16_t gas16AnalDCTcoeff4[]; 79#endif 80 81void SBC_FastIDCT8(int32_t* pInVect, int32_t* pOutVect) { 82#if (SBC_FAST_DCT == TRUE) 83#if (SBC_ARM_ASM_OPT == TRUE) 84#else 85#if (SBC_IPAQ_OPT == TRUE) 86#if (SBC_IS_64_MULT_IN_IDCT == TRUE) 87 int64_t s64Temp; 88#endif 89#else 90#if (SBC_IS_64_MULT_IN_IDCT == TRUE) 91 int32_t s32HiTemp; 92#else 93 int32_t s32In2Temp; 94 register int32_t s32In1Temp; 95#endif 96#endif 97#endif 98 99 register int32_t x0, x1, x2, x3, x4, x5, x6, x7, temp; 100 int32_t res_even[4], res_odd[4]; 101 /*x0= (pInVect[4])/2 ;*/ 102 SBC_IDCT_MULT(SBC_COS_PI_SUR_4, pInVect[4], x0); 103 /*printf("x0 0x%x = %d = %d * %d\n", x0, x0, SBC_COS_PI_SUR_4, pInVect[4]);*/ 104 105 x1 = (pInVect[3] + pInVect[5]) >> 1; 106 x2 = (pInVect[2] + pInVect[6]) >> 1; 107 x3 = (pInVect[1] + pInVect[7]) >> 1; 108 x4 = (pInVect[0] + pInVect[8]) >> 1; 109 x5 = (pInVect[9] - pInVect[15]) >> 1; 110 x6 = (pInVect[10] - pInVect[14]) >> 1; 111 x7 = (pInVect[11] - pInVect[13]) >> 1; 112 113 /* 2-point IDCT of x0 and x4 as in (11) */ 114 temp = x0; 115 SBC_IDCT_MULT(SBC_COS_PI_SUR_4, (x0 + x4), 116 x0); /*x0 = ( x0 + x4 ) * cos(1*pi/4) ; */ 117 SBC_IDCT_MULT(SBC_COS_PI_SUR_4, (temp - x4), 118 x4); /*x4 = ( temp - x4 ) * cos(1*pi/4) ; */ 119 120 /* rearrangement of x2 and x6 as in (15) */ 121 x2 -= x6; 122 x6 <<= 1; 123 124 /* 2-point IDCT of x2 and x6 and post-multiplication as in (15) */ 125 SBC_IDCT_MULT(SBC_COS_PI_SUR_4, x6, x6); /*x6 = x6 * cos(1*pi/4) ; */ 126 temp = x2; 127 SBC_IDCT_MULT(SBC_COS_PI_SUR_8, (x2 + x6), 128 x2); /*x2 = ( x2 + x6 ) * cos(1*pi/8) ; */ 129 SBC_IDCT_MULT(SBC_COS_3PI_SUR_8, (temp - x6), 130 x6); /*x6 = ( temp - x6 ) * cos(3*pi/8) ;*/ 131 132 /* 4-point IDCT of x0,x2,x4 and x6 as in (11) */ 133 res_even[0] = x0 + x2; 134 res_even[1] = x4 + x6; 135 res_even[2] = x4 - x6; 136 res_even[3] = x0 - x2; 137 138 /* rearrangement of x1,x3,x5,x7 as in (15) */ 139 x7 <<= 1; 140 x5 = (x5 << 1) - x7; 141 x3 = (x3 << 1) - x5; 142 x1 -= x3 >> 1; 143 144 /* two-dimensional IDCT of x1 and x5 */ 145 SBC_IDCT_MULT(SBC_COS_PI_SUR_4, x5, x5); /*x5 = x5 * cos(1*pi/4) ; */ 146 temp = x1; 147 x1 = x1 + x5; 148 x5 = temp - x5; 149 150 /* rearrangement of x3 and x7 as in (15) */ 151 x3 -= x7; 152 x7 <<= 1; 153 SBC_IDCT_MULT(SBC_COS_PI_SUR_4, x7, x7); /*x7 = x7 * cos(1*pi/4) ; */ 154 155 /* 2-point IDCT of x3 and x7 and post-multiplication as in (15) */ 156 temp = x3; 157 SBC_IDCT_MULT(SBC_COS_PI_SUR_8, (x3 + x7), 158 x3); /*x3 = ( x3 + x7 ) * cos(1*pi/8) ; */ 159 SBC_IDCT_MULT(SBC_COS_3PI_SUR_8, (temp - x7), 160 x7); /*x7 = ( temp - x7 ) * cos(3*pi/8) ;*/ 161 162 /* 4-point IDCT of x1,x3,x5 and x7 and post multiplication by diagonal matrix 163 * as in (14) */ 164 SBC_IDCT_MULT((SBC_COS_PI_SUR_16), (x1 + x3), 165 res_odd[0]); /*res_odd[ 0 ] = ( x1 + x3 ) * cos(1*pi/16) ; */ 166 SBC_IDCT_MULT((SBC_COS_3PI_SUR_16), (x5 + x7), 167 res_odd[1]); /*res_odd[ 1 ] = ( x5 + x7 ) * cos(3*pi/16) ; */ 168 SBC_IDCT_MULT((SBC_COS_5PI_SUR_16), (x5 - x7), 169 res_odd[2]); /*res_odd[ 2 ] = ( x5 - x7 ) * cos(5*pi/16) ; */ 170 SBC_IDCT_MULT((SBC_COS_7PI_SUR_16), (x1 - x3), 171 res_odd[3]); /*res_odd[ 3 ] = ( x1 - x3 ) * cos(7*pi/16) ; */ 172 173 /* additions and subtractions as in (9) */ 174 pOutVect[0] = (res_even[0] + res_odd[0]); 175 pOutVect[1] = (res_even[1] + res_odd[1]); 176 pOutVect[2] = (res_even[2] + res_odd[2]); 177 pOutVect[3] = (res_even[3] + res_odd[3]); 178 pOutVect[7] = (res_even[0] - res_odd[0]); 179 pOutVect[6] = (res_even[1] - res_odd[1]); 180 pOutVect[5] = (res_even[2] - res_odd[2]); 181 pOutVect[4] = (res_even[3] - res_odd[3]); 182#else 183 uint8_t Index, k; 184 int32_t temp; 185 /*Calculate 4 subband samples by matrixing*/ 186 for (Index = 0; Index < 8; Index++) { 187 temp = 0; 188 for (k = 0; k < 16; k++) { 189 /*temp += (int32_t)(((int64_t)M[(Index*strEncParams->numOfSubBands*2)+k] * 190 * Y[k]) >> 16 );*/ 191 temp += (gas16AnalDCTcoeff8[(Index * 8 * 2) + k] * (pInVect[k] >> 16)); 192 temp += 193 ((gas16AnalDCTcoeff8[(Index * 8 * 2) + k] * (pInVect[k] & 0xFFFF)) >> 194 16); 195 } 196 pOutVect[Index] = temp; 197 } 198#endif 199 /* printf("pOutVect: 0x%x;0x%x;0x%x;0x%x;0x%x;0x%x;0x%x;0x%x\n",\ 200 pOutVect[0],pOutVect[1],pOutVect[2],pOutVect[3],pOutVect[4],pOutVect[5],pOutVect[6],pOutVect[7]);*/ 201} 202 203/******************************************************************************* 204 * 205 * Function SBC_FastIDCT4 206 * 207 * Description implementation of fast DCT algorithm by Feig and Winograd 208 * 209 * 210 * Returns y = dct(x0) 211 * 212 * 213 ******************************************************************************/ 214void SBC_FastIDCT4(int32_t* pInVect, int32_t* pOutVect) { 215#if (SBC_FAST_DCT == TRUE) 216#if (SBC_ARM_ASM_OPT == TRUE) 217#else 218#if (SBC_IPAQ_OPT == TRUE) 219#if (SBC_IS_64_MULT_IN_IDCT == TRUE) 220 int64_t s64Temp; 221#endif 222#else 223#if (SBC_IS_64_MULT_IN_IDCT == TRUE) 224 int32_t s32HiTemp; 225#else 226 uint16_t s32In2Temp; 227 int32_t s32In1Temp; 228#endif 229#endif 230#endif 231 int32_t temp, x2; 232 int32_t tmp[8]; 233 234 x2 = pInVect[2] >> 1; 235 temp = (pInVect[0] + pInVect[4]); 236 SBC_IDCT_MULT((SBC_COS_PI_SUR_4 >> 1), temp, tmp[0]); 237 tmp[1] = x2 - tmp[0]; 238 tmp[0] += x2; 239 temp = (pInVect[1] + pInVect[3]); 240 SBC_IDCT_MULT((SBC_COS_3PI_SUR_8 >> 1), temp, tmp[3]); 241 SBC_IDCT_MULT((SBC_COS_PI_SUR_8 >> 1), temp, tmp[2]); 242 temp = (pInVect[5] - pInVect[7]); 243 SBC_IDCT_MULT((SBC_COS_3PI_SUR_8 >> 1), temp, tmp[5]); 244 SBC_IDCT_MULT((SBC_COS_PI_SUR_8 >> 1), temp, tmp[4]); 245 tmp[6] = tmp[2] + tmp[5]; 246 tmp[7] = tmp[3] - tmp[4]; 247 pOutVect[0] = (tmp[0] + tmp[6]); 248 pOutVect[1] = (tmp[1] + tmp[7]); 249 pOutVect[2] = (tmp[1] - tmp[7]); 250 pOutVect[3] = (tmp[0] - tmp[6]); 251#else 252 uint8_t Index, k; 253 int32_t temp; 254 /*Calculate 4 subband samples by matrixing*/ 255 for (Index = 0; Index < 4; Index++) { 256 temp = 0; 257 for (k = 0; k < 8; k++) { 258 /*temp += (int32_t)(((int64_t)M[(Index*strEncParams->numOfSubBands*2)+k] * 259 * Y[k]) >> 16 ); */ 260 temp += (gas16AnalDCTcoeff4[(Index * 4 * 2) + k] * (pInVect[k] >> 16)); 261 temp += 262 ((gas16AnalDCTcoeff4[(Index * 4 * 2) + k] * (pInVect[k] & 0xFFFF)) >> 263 16); 264 } 265 pOutVect[Index] = temp; 266 } 267#endif 268} 269