Details | Last modification | View Log | RSS feed
| Rev | Author | Line No. | Line |
|---|---|---|---|
| 56 | mjames | 1 | /* ---------------------------------------------------------------------- |
| 2 | * Project: CMSIS DSP Library |
||
| 3 | * Title: arm_rfft_q31.c |
||
| 4 | * Description: FFT & RIFFT Q31 process function |
||
| 5 | * |
||
| 6 | * $Date: 27. January 2017 |
||
| 7 | * $Revision: V.1.5.1 |
||
| 8 | * |
||
| 9 | * Target Processor: Cortex-M cores |
||
| 10 | * -------------------------------------------------------------------- */ |
||
| 11 | /* |
||
| 12 | * Copyright (C) 2010-2017 ARM Limited or its affiliates. All rights reserved. |
||
| 13 | * |
||
| 14 | * SPDX-License-Identifier: Apache-2.0 |
||
| 15 | * |
||
| 16 | * Licensed under the Apache License, Version 2.0 (the License); you may |
||
| 17 | * not use this file except in compliance with the License. |
||
| 18 | * You may obtain a copy of the License at |
||
| 19 | * |
||
| 20 | * www.apache.org/licenses/LICENSE-2.0 |
||
| 21 | * |
||
| 22 | * Unless required by applicable law or agreed to in writing, software |
||
| 23 | * distributed under the License is distributed on an AS IS BASIS, WITHOUT |
||
| 24 | * WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. |
||
| 25 | * See the License for the specific language governing permissions and |
||
| 26 | * limitations under the License. |
||
| 27 | */ |
||
| 28 | |||
| 29 | #include "arm_math.h" |
||
| 30 | |||
| 31 | /* ---------------------------------------------------------------------- |
||
| 32 | * Internal functions prototypes |
||
| 33 | * -------------------------------------------------------------------- */ |
||
| 34 | |||
| 35 | void arm_split_rfft_q31( |
||
| 36 | q31_t * pSrc, |
||
| 37 | uint32_t fftLen, |
||
| 38 | q31_t * pATable, |
||
| 39 | q31_t * pBTable, |
||
| 40 | q31_t * pDst, |
||
| 41 | uint32_t modifier); |
||
| 42 | |||
| 43 | void arm_split_rifft_q31( |
||
| 44 | q31_t * pSrc, |
||
| 45 | uint32_t fftLen, |
||
| 46 | q31_t * pATable, |
||
| 47 | q31_t * pBTable, |
||
| 48 | q31_t * pDst, |
||
| 49 | uint32_t modifier); |
||
| 50 | |||
| 51 | /** |
||
| 52 | * @addtogroup RealFFT |
||
| 53 | * @{ |
||
| 54 | */ |
||
| 55 | |||
| 56 | /** |
||
| 57 | * @brief Processing function for the Q31 RFFT/RIFFT. |
||
| 58 | * @param[in] *S points to an instance of the Q31 RFFT/RIFFT structure. |
||
| 59 | * @param[in] *pSrc points to the input buffer. |
||
| 60 | * @param[out] *pDst points to the output buffer. |
||
| 61 | * @return none. |
||
| 62 | * |
||
| 63 | * \par Input an output formats: |
||
| 64 | * \par |
||
| 65 | * Internally input is downscaled by 2 for every stage to avoid saturations inside CFFT/CIFFT process. |
||
| 66 | * Hence the output format is different for different RFFT sizes. |
||
| 67 | * The input and output formats for different RFFT sizes and number of bits to upscale are mentioned in the tables below for RFFT and RIFFT: |
||
| 68 | * \par |
||
| 69 | * \image html RFFTQ31.gif "Input and Output Formats for Q31 RFFT" |
||
| 70 | * |
||
| 71 | * \par |
||
| 72 | * \image html RIFFTQ31.gif "Input and Output Formats for Q31 RIFFT" |
||
| 73 | */ |
||
| 74 | void arm_rfft_q31( |
||
| 75 | const arm_rfft_instance_q31 * S, |
||
| 76 | q31_t * pSrc, |
||
| 77 | q31_t * pDst) |
||
| 78 | { |
||
| 79 | const arm_cfft_instance_q31 *S_CFFT = S->pCfft; |
||
| 80 | uint32_t i; |
||
| 81 | uint32_t L2 = S->fftLenReal >> 1; |
||
| 82 | |||
| 83 | /* Calculation of RIFFT of input */ |
||
| 84 | if (S->ifftFlagR == 1U) |
||
| 85 | { |
||
| 86 | /* Real IFFT core process */ |
||
| 87 | arm_split_rifft_q31(pSrc, L2, S->pTwiddleAReal, |
||
| 88 | S->pTwiddleBReal, pDst, S->twidCoefRModifier); |
||
| 89 | |||
| 90 | /* Complex IFFT process */ |
||
| 91 | arm_cfft_q31(S_CFFT, pDst, S->ifftFlagR, S->bitReverseFlagR); |
||
| 92 | |||
| 93 | for(i=0;i<S->fftLenReal;i++) |
||
| 94 | { |
||
| 95 | pDst[i] = pDst[i] << 1; |
||
| 96 | } |
||
| 97 | } |
||
| 98 | else |
||
| 99 | { |
||
| 100 | /* Calculation of RFFT of input */ |
||
| 101 | |||
| 102 | /* Complex FFT process */ |
||
| 103 | arm_cfft_q31(S_CFFT, pSrc, S->ifftFlagR, S->bitReverseFlagR); |
||
| 104 | |||
| 105 | /* Real FFT core process */ |
||
| 106 | arm_split_rfft_q31(pSrc, L2, S->pTwiddleAReal, |
||
| 107 | S->pTwiddleBReal, pDst, S->twidCoefRModifier); |
||
| 108 | } |
||
| 109 | } |
||
| 110 | |||
| 111 | /** |
||
| 112 | * @} end of RealFFT group |
||
| 113 | */ |
||
| 114 | |||
| 115 | /** |
||
| 116 | * @brief Core Real FFT process |
||
| 117 | * @param[in] *pSrc points to the input buffer. |
||
| 118 | * @param[in] fftLen length of FFT. |
||
| 119 | * @param[in] *pATable points to the twiddle Coef A buffer. |
||
| 120 | * @param[in] *pBTable points to the twiddle Coef B buffer. |
||
| 121 | * @param[out] *pDst points to the output buffer. |
||
| 122 | * @param[in] modifier twiddle coefficient modifier that supports different size FFTs with the same twiddle factor table. |
||
| 123 | * @return none. |
||
| 124 | */ |
||
| 125 | void arm_split_rfft_q31( |
||
| 126 | q31_t * pSrc, |
||
| 127 | uint32_t fftLen, |
||
| 128 | q31_t * pATable, |
||
| 129 | q31_t * pBTable, |
||
| 130 | q31_t * pDst, |
||
| 131 | uint32_t modifier) |
||
| 132 | { |
||
| 133 | uint32_t i; /* Loop Counter */ |
||
| 134 | q31_t outR, outI; /* Temporary variables for output */ |
||
| 135 | q31_t *pCoefA, *pCoefB; /* Temporary pointers for twiddle factors */ |
||
| 136 | q31_t CoefA1, CoefA2, CoefB1; /* Temporary variables for twiddle coefficients */ |
||
| 137 | q31_t *pOut1 = &pDst[2], *pOut2 = &pDst[(4U * fftLen) - 1U]; |
||
| 138 | q31_t *pIn1 = &pSrc[2], *pIn2 = &pSrc[(2U * fftLen) - 1U]; |
||
| 139 | |||
| 140 | /* Init coefficient pointers */ |
||
| 141 | pCoefA = &pATable[modifier * 2U]; |
||
| 142 | pCoefB = &pBTable[modifier * 2U]; |
||
| 143 | |||
| 144 | i = fftLen - 1U; |
||
| 145 | |||
| 146 | while (i > 0U) |
||
| 147 | { |
||
| 148 | /* |
||
| 149 | outR = (pSrc[2 * i] * pATable[2 * i] - pSrc[2 * i + 1] * pATable[2 * i + 1] |
||
| 150 | + pSrc[2 * n - 2 * i] * pBTable[2 * i] + |
||
| 151 | pSrc[2 * n - 2 * i + 1] * pBTable[2 * i + 1]); |
||
| 152 | */ |
||
| 153 | |||
| 154 | /* outI = (pIn[2 * i + 1] * pATable[2 * i] + pIn[2 * i] * pATable[2 * i + 1] + |
||
| 155 | pIn[2 * n - 2 * i] * pBTable[2 * i + 1] - |
||
| 156 | pIn[2 * n - 2 * i + 1] * pBTable[2 * i]); */ |
||
| 157 | |||
| 158 | CoefA1 = *pCoefA++; |
||
| 159 | CoefA2 = *pCoefA; |
||
| 160 | |||
| 161 | /* outR = (pSrc[2 * i] * pATable[2 * i] */ |
||
| 162 | mult_32x32_keep32_R(outR, *pIn1, CoefA1); |
||
| 163 | |||
| 164 | /* outI = pIn[2 * i] * pATable[2 * i + 1] */ |
||
| 165 | mult_32x32_keep32_R(outI, *pIn1++, CoefA2); |
||
| 166 | |||
| 167 | /* - pSrc[2 * i + 1] * pATable[2 * i + 1] */ |
||
| 168 | multSub_32x32_keep32_R(outR, *pIn1, CoefA2); |
||
| 169 | |||
| 170 | /* (pIn[2 * i + 1] * pATable[2 * i] */ |
||
| 171 | multAcc_32x32_keep32_R(outI, *pIn1++, CoefA1); |
||
| 172 | |||
| 173 | /* pSrc[2 * n - 2 * i] * pBTable[2 * i] */ |
||
| 174 | multSub_32x32_keep32_R(outR, *pIn2, CoefA2); |
||
| 175 | CoefB1 = *pCoefB; |
||
| 176 | |||
| 177 | /* pIn[2 * n - 2 * i] * pBTable[2 * i + 1] */ |
||
| 178 | multSub_32x32_keep32_R(outI, *pIn2--, CoefB1); |
||
| 179 | |||
| 180 | /* pSrc[2 * n - 2 * i + 1] * pBTable[2 * i + 1] */ |
||
| 181 | multAcc_32x32_keep32_R(outR, *pIn2, CoefB1); |
||
| 182 | |||
| 183 | /* pIn[2 * n - 2 * i + 1] * pBTable[2 * i] */ |
||
| 184 | multSub_32x32_keep32_R(outI, *pIn2--, CoefA2); |
||
| 185 | |||
| 186 | /* write output */ |
||
| 187 | *pOut1++ = outR; |
||
| 188 | *pOut1++ = outI; |
||
| 189 | |||
| 190 | /* write complex conjugate output */ |
||
| 191 | *pOut2-- = -outI; |
||
| 192 | *pOut2-- = outR; |
||
| 193 | |||
| 194 | /* update coefficient pointer */ |
||
| 195 | pCoefB = pCoefB + (modifier * 2U); |
||
| 196 | pCoefA = pCoefA + ((modifier * 2U) - 1U); |
||
| 197 | |||
| 198 | i--; |
||
| 199 | } |
||
| 200 | pDst[2U * fftLen] = (pSrc[0] - pSrc[1]) >> 1; |
||
| 201 | pDst[(2U * fftLen) + 1U] = 0; |
||
| 202 | |||
| 203 | pDst[0] = (pSrc[0] + pSrc[1]) >> 1; |
||
| 204 | pDst[1] = 0; |
||
| 205 | } |
||
| 206 | |||
| 207 | /** |
||
| 208 | * @brief Core Real IFFT process |
||
| 209 | * @param[in] *pSrc points to the input buffer. |
||
| 210 | * @param[in] fftLen length of FFT. |
||
| 211 | * @param[in] *pATable points to the twiddle Coef A buffer. |
||
| 212 | * @param[in] *pBTable points to the twiddle Coef B buffer. |
||
| 213 | * @param[out] *pDst points to the output buffer. |
||
| 214 | * @param[in] modifier twiddle coefficient modifier that supports different size FFTs with the same twiddle factor table. |
||
| 215 | * @return none. |
||
| 216 | */ |
||
| 217 | void arm_split_rifft_q31( |
||
| 218 | q31_t * pSrc, |
||
| 219 | uint32_t fftLen, |
||
| 220 | q31_t * pATable, |
||
| 221 | q31_t * pBTable, |
||
| 222 | q31_t * pDst, |
||
| 223 | uint32_t modifier) |
||
| 224 | { |
||
| 225 | q31_t outR, outI; /* Temporary variables for output */ |
||
| 226 | q31_t *pCoefA, *pCoefB; /* Temporary pointers for twiddle factors */ |
||
| 227 | q31_t CoefA1, CoefA2, CoefB1; /* Temporary variables for twiddle coefficients */ |
||
| 228 | q31_t *pIn1 = &pSrc[0], *pIn2 = &pSrc[(2U * fftLen) + 1U]; |
||
| 229 | |||
| 230 | pCoefA = &pATable[0]; |
||
| 231 | pCoefB = &pBTable[0]; |
||
| 232 | |||
| 233 | while (fftLen > 0U) |
||
| 234 | { |
||
| 235 | /* |
||
| 236 | outR = (pIn[2 * i] * pATable[2 * i] + pIn[2 * i + 1] * pATable[2 * i + 1] + |
||
| 237 | pIn[2 * n - 2 * i] * pBTable[2 * i] - |
||
| 238 | pIn[2 * n - 2 * i + 1] * pBTable[2 * i + 1]); |
||
| 239 | |||
| 240 | outI = (pIn[2 * i + 1] * pATable[2 * i] - pIn[2 * i] * pATable[2 * i + 1] - |
||
| 241 | pIn[2 * n - 2 * i] * pBTable[2 * i + 1] - |
||
| 242 | pIn[2 * n - 2 * i + 1] * pBTable[2 * i]); |
||
| 243 | */ |
||
| 244 | CoefA1 = *pCoefA++; |
||
| 245 | CoefA2 = *pCoefA; |
||
| 246 | |||
| 247 | /* outR = (pIn[2 * i] * pATable[2 * i] */ |
||
| 248 | mult_32x32_keep32_R(outR, *pIn1, CoefA1); |
||
| 249 | |||
| 250 | /* - pIn[2 * i] * pATable[2 * i + 1] */ |
||
| 251 | mult_32x32_keep32_R(outI, *pIn1++, -CoefA2); |
||
| 252 | |||
| 253 | /* pIn[2 * i + 1] * pATable[2 * i + 1] */ |
||
| 254 | multAcc_32x32_keep32_R(outR, *pIn1, CoefA2); |
||
| 255 | |||
| 256 | /* pIn[2 * i + 1] * pATable[2 * i] */ |
||
| 257 | multAcc_32x32_keep32_R(outI, *pIn1++, CoefA1); |
||
| 258 | |||
| 259 | /* pIn[2 * n - 2 * i] * pBTable[2 * i] */ |
||
| 260 | multAcc_32x32_keep32_R(outR, *pIn2, CoefA2); |
||
| 261 | CoefB1 = *pCoefB; |
||
| 262 | |||
| 263 | /* pIn[2 * n - 2 * i] * pBTable[2 * i + 1] */ |
||
| 264 | multSub_32x32_keep32_R(outI, *pIn2--, CoefB1); |
||
| 265 | |||
| 266 | /* pIn[2 * n - 2 * i + 1] * pBTable[2 * i + 1] */ |
||
| 267 | multAcc_32x32_keep32_R(outR, *pIn2, CoefB1); |
||
| 268 | |||
| 269 | /* pIn[2 * n - 2 * i + 1] * pBTable[2 * i] */ |
||
| 270 | multAcc_32x32_keep32_R(outI, *pIn2--, CoefA2); |
||
| 271 | |||
| 272 | /* write output */ |
||
| 273 | *pDst++ = outR; |
||
| 274 | *pDst++ = outI; |
||
| 275 | |||
| 276 | /* update coefficient pointer */ |
||
| 277 | pCoefB = pCoefB + (modifier * 2U); |
||
| 278 | pCoefA = pCoefA + ((modifier * 2U) - 1U); |
||
| 279 | |||
| 280 | /* Decrement loop count */ |
||
| 281 | fftLen--; |
||
| 282 | } |
||
| 283 | } |