Michele Riccio / mbed-dsp_F767ZI

mbed-os 6.10 versione

cmsis_dsp/TransformFunctions/arm_cfft_radix4_q31.c@3:7a284390b0ce, 2013-11-08 (annotated)

Committer:
mbed_official
Date:
Fri Nov 08 13:45:10 2013 +0000
Revision:
3:7a284390b0ce
Parent:
2:da51fb522205 
Synchronized with git revision e69956aba2f68a2a26ac26b051f8d349deaa1ce8

Who changed what in which revision?

UserRevisionLine numberNew contents of line
emilmont 1:fdd22bb7aa52 1 /* ----------------------------------------------------------------------
mbed_official 3:7a284390b0ce 2 * Copyright (C) 2010-2013 ARM Limited. All rights reserved.
emilmont 1:fdd22bb7aa52 3 *
mbed_official 3:7a284390b0ce 4 * $Date: 17. January 2013
mbed_official 3:7a284390b0ce 5 * $Revision: V1.4.1
emilmont 1:fdd22bb7aa52 6 *
emilmont 2:da51fb522205 7 * Project: CMSIS DSP Library
emilmont 2:da51fb522205 8 * Title: arm_cfft_radix4_q31.c
emilmont 1:fdd22bb7aa52 9 *
emilmont 2:da51fb522205 10 * Description: This file has function definition of Radix-4 FFT & IFFT function and
emilmont 2:da51fb522205 11 * In-place bit reversal using bit reversal table
emilmont 1:fdd22bb7aa52 12 *
emilmont 1:fdd22bb7aa52 13 * Target Processor: Cortex-M4/Cortex-M3/Cortex-M0
emilmont 1:fdd22bb7aa52 14 *
mbed_official 3:7a284390b0ce 15 * Redistribution and use in source and binary forms, with or without
mbed_official 3:7a284390b0ce 16 * modification, are permitted provided that the following conditions
mbed_official 3:7a284390b0ce 17 * are met:
mbed_official 3:7a284390b0ce 18 * - Redistributions of source code must retain the above copyright
mbed_official 3:7a284390b0ce 19 * notice, this list of conditions and the following disclaimer.
mbed_official 3:7a284390b0ce 20 * - Redistributions in binary form must reproduce the above copyright
mbed_official 3:7a284390b0ce 21 * notice, this list of conditions and the following disclaimer in
mbed_official 3:7a284390b0ce 22 * the documentation and/or other materials provided with the
mbed_official 3:7a284390b0ce 23 * distribution.
mbed_official 3:7a284390b0ce 24 * - Neither the name of ARM LIMITED nor the names of its contributors
mbed_official 3:7a284390b0ce 25 * may be used to endorse or promote products derived from this
mbed_official 3:7a284390b0ce 26 * software without specific prior written permission.
mbed_official 3:7a284390b0ce 27 *
mbed_official 3:7a284390b0ce 28 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
mbed_official 3:7a284390b0ce 29 * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
mbed_official 3:7a284390b0ce 30 * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
mbed_official 3:7a284390b0ce 31 * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE
mbed_official 3:7a284390b0ce 32 * COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
mbed_official 3:7a284390b0ce 33 * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING,
mbed_official 3:7a284390b0ce 34 * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
mbed_official 3:7a284390b0ce 35 * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
mbed_official 3:7a284390b0ce 36 * CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
mbed_official 3:7a284390b0ce 37 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN
mbed_official 3:7a284390b0ce 38 * ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
mbed_official 3:7a284390b0ce 39 * POSSIBILITY OF SUCH DAMAGE.
emilmont 1:fdd22bb7aa52 40 * -------------------------------------------------------------------- */
mbed_official 3:7a284390b0ce 41
emilmont 1:fdd22bb7aa52 42 #include "arm_math.h"
emilmont 1:fdd22bb7aa52 43
mbed_official 3:7a284390b0ce 44 void arm_radix4_butterfly_inverse_q31(
mbed_official 3:7a284390b0ce 45 q31_t * pSrc,
mbed_official 3:7a284390b0ce 46 uint32_t fftLen,
mbed_official 3:7a284390b0ce 47 q31_t * pCoef,
mbed_official 3:7a284390b0ce 48 uint32_t twidCoefModifier);
mbed_official 3:7a284390b0ce 49
mbed_official 3:7a284390b0ce 50 void arm_radix4_butterfly_q31(
mbed_official 3:7a284390b0ce 51 q31_t * pSrc,
mbed_official 3:7a284390b0ce 52 uint32_t fftLen,
mbed_official 3:7a284390b0ce 53 q31_t * pCoef,
mbed_official 3:7a284390b0ce 54 uint32_t twidCoefModifier);
mbed_official 3:7a284390b0ce 55
mbed_official 3:7a284390b0ce 56 void arm_bitreversal_q31(
mbed_official 3:7a284390b0ce 57 q31_t * pSrc,
mbed_official 3:7a284390b0ce 58 uint32_t fftLen,
mbed_official 3:7a284390b0ce 59 uint16_t bitRevFactor,
mbed_official 3:7a284390b0ce 60 uint16_t * pBitRevTab);
emilmont 1:fdd22bb7aa52 61
emilmont 1:fdd22bb7aa52 62 /**
emilmont 1:fdd22bb7aa52 63 * @ingroup groupTransforms
emilmont 1:fdd22bb7aa52 64 */
emilmont 1:fdd22bb7aa52 65
emilmont 1:fdd22bb7aa52 66 /**
mbed_official 3:7a284390b0ce 67 * @addtogroup ComplexFFT
emilmont 1:fdd22bb7aa52 68 * @{
emilmont 1:fdd22bb7aa52 69 */
emilmont 1:fdd22bb7aa52 70
emilmont 1:fdd22bb7aa52 71 /**
emilmont 1:fdd22bb7aa52 72 * @details
emilmont 1:fdd22bb7aa52 73 * @brief Processing function for the Q31 CFFT/CIFFT.
emilmont 1:fdd22bb7aa52 74 * @param[in] *S points to an instance of the Q31 CFFT/CIFFT structure.
emilmont 1:fdd22bb7aa52 75 * @param[in, out] *pSrc points to the complex data buffer of size <code>2*fftLen</code>. Processing occurs in-place.
emilmont 1:fdd22bb7aa52 76 * @return none.
emilmont 1:fdd22bb7aa52 77 *
emilmont 1:fdd22bb7aa52 78 * \par Input and output formats:
emilmont 1:fdd22bb7aa52 79 * \par
emilmont 1:fdd22bb7aa52 80 * Internally input is downscaled by 2 for every stage to avoid saturations inside CFFT/CIFFT process.
emilmont 1:fdd22bb7aa52 81 * Hence the output format is different for different FFT sizes.
emilmont 1:fdd22bb7aa52 82 * The input and output formats for different FFT sizes and number of bits to upscale are mentioned in the tables below for CFFT and CIFFT:
emilmont 1:fdd22bb7aa52 83 * \par
emilmont 1:fdd22bb7aa52 84 * \image html CFFTQ31.gif "Input and Output Formats for Q31 CFFT"
emilmont 1:fdd22bb7aa52 85 * \image html CIFFTQ31.gif "Input and Output Formats for Q31 CIFFT"
emilmont 1:fdd22bb7aa52 86 *
emilmont 1:fdd22bb7aa52 87 */
emilmont 1:fdd22bb7aa52 88
emilmont 1:fdd22bb7aa52 89 void arm_cfft_radix4_q31(
emilmont 1:fdd22bb7aa52 90 const arm_cfft_radix4_instance_q31 * S,
emilmont 1:fdd22bb7aa52 91 q31_t * pSrc)
emilmont 1:fdd22bb7aa52 92 {
emilmont 1:fdd22bb7aa52 93 if(S->ifftFlag == 1u)
emilmont 1:fdd22bb7aa52 94 {
emilmont 1:fdd22bb7aa52 95 /* Complex IFFT radix-4 */
emilmont 1:fdd22bb7aa52 96 arm_radix4_butterfly_inverse_q31(pSrc, S->fftLen, S->pTwiddle,
emilmont 1:fdd22bb7aa52 97 S->twidCoefModifier);
emilmont 1:fdd22bb7aa52 98 }
emilmont 1:fdd22bb7aa52 99 else
emilmont 1:fdd22bb7aa52 100 {
emilmont 1:fdd22bb7aa52 101 /* Complex FFT radix-4 */
emilmont 1:fdd22bb7aa52 102 arm_radix4_butterfly_q31(pSrc, S->fftLen, S->pTwiddle,
emilmont 1:fdd22bb7aa52 103 S->twidCoefModifier);
emilmont 1:fdd22bb7aa52 104 }
emilmont 1:fdd22bb7aa52 105
emilmont 1:fdd22bb7aa52 106
emilmont 1:fdd22bb7aa52 107 if(S->bitReverseFlag == 1u)
emilmont 1:fdd22bb7aa52 108 {
emilmont 1:fdd22bb7aa52 109 /* Bit Reversal */
emilmont 1:fdd22bb7aa52 110 arm_bitreversal_q31(pSrc, S->fftLen, S->bitRevFactor, S->pBitRevTable);
emilmont 1:fdd22bb7aa52 111 }
emilmont 1:fdd22bb7aa52 112
emilmont 1:fdd22bb7aa52 113 }
emilmont 1:fdd22bb7aa52 114
emilmont 1:fdd22bb7aa52 115 /**
mbed_official 3:7a284390b0ce 116 * @} end of ComplexFFT group
emilmont 1:fdd22bb7aa52 117 */
emilmont 1:fdd22bb7aa52 118
emilmont 1:fdd22bb7aa52 119 /*
emilmont 1:fdd22bb7aa52 120 * Radix-4 FFT algorithm used is :
emilmont 1:fdd22bb7aa52 121 *
emilmont 1:fdd22bb7aa52 122 * Input real and imaginary data:
emilmont 1:fdd22bb7aa52 123 * x(n) = xa + j * ya
emilmont 1:fdd22bb7aa52 124 * x(n+N/4 ) = xb + j * yb
emilmont 1:fdd22bb7aa52 125 * x(n+N/2 ) = xc + j * yc
emilmont 1:fdd22bb7aa52 126 * x(n+3N 4) = xd + j * yd
emilmont 1:fdd22bb7aa52 127 *
emilmont 1:fdd22bb7aa52 128 *
emilmont 1:fdd22bb7aa52 129 * Output real and imaginary data:
emilmont 1:fdd22bb7aa52 130 * x(4r) = xa'+ j * ya'
emilmont 1:fdd22bb7aa52 131 * x(4r+1) = xb'+ j * yb'
emilmont 1:fdd22bb7aa52 132 * x(4r+2) = xc'+ j * yc'
emilmont 1:fdd22bb7aa52 133 * x(4r+3) = xd'+ j * yd'
emilmont 1:fdd22bb7aa52 134 *
emilmont 1:fdd22bb7aa52 135 *
emilmont 1:fdd22bb7aa52 136 * Twiddle factors for radix-4 FFT:
emilmont 1:fdd22bb7aa52 137 * Wn = co1 + j * (- si1)
emilmont 1:fdd22bb7aa52 138 * W2n = co2 + j * (- si2)
emilmont 1:fdd22bb7aa52 139 * W3n = co3 + j * (- si3)
emilmont 1:fdd22bb7aa52 140 *
emilmont 1:fdd22bb7aa52 141 * Butterfly implementation:
emilmont 1:fdd22bb7aa52 142 * xa' = xa + xb + xc + xd
emilmont 1:fdd22bb7aa52 143 * ya' = ya + yb + yc + yd
emilmont 1:fdd22bb7aa52 144 * xb' = (xa+yb-xc-yd)* co1 + (ya-xb-yc+xd)* (si1)
emilmont 1:fdd22bb7aa52 145 * yb' = (ya-xb-yc+xd)* co1 - (xa+yb-xc-yd)* (si1)
emilmont 1:fdd22bb7aa52 146 * xc' = (xa-xb+xc-xd)* co2 + (ya-yb+yc-yd)* (si2)
emilmont 1:fdd22bb7aa52 147 * yc' = (ya-yb+yc-yd)* co2 - (xa-xb+xc-xd)* (si2)
emilmont 1:fdd22bb7aa52 148 * xd' = (xa-yb-xc+yd)* co3 + (ya+xb-yc-xd)* (si3)
emilmont 1:fdd22bb7aa52 149 * yd' = (ya+xb-yc-xd)* co3 - (xa-yb-xc+yd)* (si3)
emilmont 1:fdd22bb7aa52 150 *
emilmont 1:fdd22bb7aa52 151 */
emilmont 1:fdd22bb7aa52 152
emilmont 1:fdd22bb7aa52 153 /**
emilmont 1:fdd22bb7aa52 154 * @brief Core function for the Q31 CFFT butterfly process.
emilmont 1:fdd22bb7aa52 155 * @param[in, out] *pSrc points to the in-place buffer of Q31 data type.
emilmont 1:fdd22bb7aa52 156 * @param[in] fftLen length of the FFT.
emilmont 1:fdd22bb7aa52 157 * @param[in] *pCoef points to twiddle coefficient buffer.
emilmont 1:fdd22bb7aa52 158 * @param[in] twidCoefModifier twiddle coefficient modifier that supports different size FFTs with the same twiddle factor table.
emilmont 1:fdd22bb7aa52 159 * @return none.
emilmont 1:fdd22bb7aa52 160 */
emilmont 1:fdd22bb7aa52 161
emilmont 1:fdd22bb7aa52 162 void arm_radix4_butterfly_q31(
emilmont 1:fdd22bb7aa52 163 q31_t * pSrc,
emilmont 1:fdd22bb7aa52 164 uint32_t fftLen,
emilmont 1:fdd22bb7aa52 165 q31_t * pCoef,
emilmont 1:fdd22bb7aa52 166 uint32_t twidCoefModifier)
emilmont 1:fdd22bb7aa52 167 {
emilmont 1:fdd22bb7aa52 168 uint32_t n1, n2, ia1, ia2, ia3, i0, i1, i2, i3, j, k;
emilmont 1:fdd22bb7aa52 169 q31_t t1, t2, r1, r2, s1, s2, co1, co2, co3, si1, si2, si3;
emilmont 1:fdd22bb7aa52 170
emilmont 1:fdd22bb7aa52 171 q31_t xa, xb, xc, xd;
emilmont 1:fdd22bb7aa52 172 q31_t ya, yb, yc, yd;
emilmont 1:fdd22bb7aa52 173 q31_t xa_out, xb_out, xc_out, xd_out;
emilmont 1:fdd22bb7aa52 174 q31_t ya_out, yb_out, yc_out, yd_out;
emilmont 1:fdd22bb7aa52 175
emilmont 1:fdd22bb7aa52 176 q31_t *ptr1;
emilmont 1:fdd22bb7aa52 177 q63_t xaya, xbyb, xcyc, xdyd;
emilmont 1:fdd22bb7aa52 178 /* Total process is divided into three stages */
emilmont 1:fdd22bb7aa52 179
emilmont 1:fdd22bb7aa52 180 /* process first stage, middle stages, & last stage */
emilmont 1:fdd22bb7aa52 181
emilmont 1:fdd22bb7aa52 182
emilmont 1:fdd22bb7aa52 183 /* start of first stage process */
emilmont 1:fdd22bb7aa52 184
emilmont 1:fdd22bb7aa52 185 /* Initializations for the first stage */
emilmont 1:fdd22bb7aa52 186 n2 = fftLen;
emilmont 1:fdd22bb7aa52 187 n1 = n2;
emilmont 1:fdd22bb7aa52 188 /* n2 = fftLen/4 */
emilmont 1:fdd22bb7aa52 189 n2 >>= 2u;
emilmont 1:fdd22bb7aa52 190 i0 = 0u;
emilmont 1:fdd22bb7aa52 191 ia1 = 0u;
emilmont 1:fdd22bb7aa52 192
emilmont 1:fdd22bb7aa52 193 j = n2;
emilmont 1:fdd22bb7aa52 194
emilmont 1:fdd22bb7aa52 195 /* Calculation of first stage */
emilmont 1:fdd22bb7aa52 196 do
emilmont 1:fdd22bb7aa52 197 {
emilmont 1:fdd22bb7aa52 198 /* index calculation for the input as, */
emilmont 1:fdd22bb7aa52 199 /* pSrc[i0 + 0], pSrc[i0 + fftLen/4], pSrc[i0 + fftLen/2u], pSrc[i0 + 3fftLen/4] */
emilmont 1:fdd22bb7aa52 200 i1 = i0 + n2;
emilmont 1:fdd22bb7aa52 201 i2 = i1 + n2;
emilmont 1:fdd22bb7aa52 202 i3 = i2 + n2;
emilmont 1:fdd22bb7aa52 203
emilmont 1:fdd22bb7aa52 204 /* input is in 1.31(q31) format and provide 4 guard bits for the input */
emilmont 1:fdd22bb7aa52 205
emilmont 1:fdd22bb7aa52 206 /* Butterfly implementation */
emilmont 1:fdd22bb7aa52 207 /* xa + xc */
emilmont 1:fdd22bb7aa52 208 r1 = (pSrc[(2u * i0)] >> 4u) + (pSrc[(2u * i2)] >> 4u);
emilmont 1:fdd22bb7aa52 209 /* xa - xc */
emilmont 1:fdd22bb7aa52 210 r2 = (pSrc[2u * i0] >> 4u) - (pSrc[2u * i2] >> 4u);
emilmont 1:fdd22bb7aa52 211
emilmont 1:fdd22bb7aa52 212 /* xb + xd */
emilmont 1:fdd22bb7aa52 213 t1 = (pSrc[2u * i1] >> 4u) + (pSrc[2u * i3] >> 4u);
emilmont 1:fdd22bb7aa52 214
emilmont 1:fdd22bb7aa52 215 /* ya + yc */
emilmont 1:fdd22bb7aa52 216 s1 = (pSrc[(2u * i0) + 1u] >> 4u) + (pSrc[(2u * i2) + 1u] >> 4u);
emilmont 1:fdd22bb7aa52 217 /* ya - yc */
emilmont 1:fdd22bb7aa52 218 s2 = (pSrc[(2u * i0) + 1u] >> 4u) - (pSrc[(2u * i2) + 1u] >> 4u);
emilmont 1:fdd22bb7aa52 219
emilmont 1:fdd22bb7aa52 220 /* xa' = xa + xb + xc + xd */
emilmont 1:fdd22bb7aa52 221 pSrc[2u * i0] = (r1 + t1);
emilmont 1:fdd22bb7aa52 222 /* (xa + xc) - (xb + xd) */
emilmont 1:fdd22bb7aa52 223 r1 = r1 - t1;
emilmont 1:fdd22bb7aa52 224 /* yb + yd */
emilmont 1:fdd22bb7aa52 225 t2 = (pSrc[(2u * i1) + 1u] >> 4u) + (pSrc[(2u * i3) + 1u] >> 4u);
emilmont 1:fdd22bb7aa52 226
emilmont 1:fdd22bb7aa52 227 /* ya' = ya + yb + yc + yd */
emilmont 1:fdd22bb7aa52 228 pSrc[(2u * i0) + 1u] = (s1 + t2);
emilmont 1:fdd22bb7aa52 229
emilmont 1:fdd22bb7aa52 230 /* (ya + yc) - (yb + yd) */
emilmont 1:fdd22bb7aa52 231 s1 = s1 - t2;
emilmont 1:fdd22bb7aa52 232
emilmont 1:fdd22bb7aa52 233 /* yb - yd */
emilmont 1:fdd22bb7aa52 234 t1 = (pSrc[(2u * i1) + 1u] >> 4u) - (pSrc[(2u * i3) + 1u] >> 4u);
emilmont 1:fdd22bb7aa52 235 /* xb - xd */
emilmont 1:fdd22bb7aa52 236 t2 = (pSrc[2u * i1] >> 4u) - (pSrc[2u * i3] >> 4u);
emilmont 1:fdd22bb7aa52 237
emilmont 1:fdd22bb7aa52 238 /* index calculation for the coefficients */
emilmont 1:fdd22bb7aa52 239 ia2 = 2u * ia1;
emilmont 1:fdd22bb7aa52 240 co2 = pCoef[ia2 * 2u];
emilmont 1:fdd22bb7aa52 241 si2 = pCoef[(ia2 * 2u) + 1u];
emilmont 1:fdd22bb7aa52 242
emilmont 1:fdd22bb7aa52 243 /* xc' = (xa-xb+xc-xd)co2 + (ya-yb+yc-yd)(si2) */
emilmont 1:fdd22bb7aa52 244 pSrc[2u * i1] = (((int32_t) (((q63_t) r1 * co2) >> 32)) +
emilmont 1:fdd22bb7aa52 245 ((int32_t) (((q63_t) s1 * si2) >> 32))) << 1u;
emilmont 1:fdd22bb7aa52 246
emilmont 1:fdd22bb7aa52 247 /* yc' = (ya-yb+yc-yd)co2 - (xa-xb+xc-xd)(si2) */
emilmont 1:fdd22bb7aa52 248 pSrc[(2u * i1) + 1u] = (((int32_t) (((q63_t) s1 * co2) >> 32)) -
emilmont 1:fdd22bb7aa52 249 ((int32_t) (((q63_t) r1 * si2) >> 32))) << 1u;
emilmont 1:fdd22bb7aa52 250
emilmont 1:fdd22bb7aa52 251 /* (xa - xc) + (yb - yd) */
emilmont 1:fdd22bb7aa52 252 r1 = r2 + t1;
emilmont 1:fdd22bb7aa52 253 /* (xa - xc) - (yb - yd) */
emilmont 1:fdd22bb7aa52 254 r2 = r2 - t1;
emilmont 1:fdd22bb7aa52 255
emilmont 1:fdd22bb7aa52 256 /* (ya - yc) - (xb - xd) */
emilmont 1:fdd22bb7aa52 257 s1 = s2 - t2;
emilmont 1:fdd22bb7aa52 258 /* (ya - yc) + (xb - xd) */
emilmont 1:fdd22bb7aa52 259 s2 = s2 + t2;
emilmont 1:fdd22bb7aa52 260
emilmont 1:fdd22bb7aa52 261 co1 = pCoef[ia1 * 2u];
emilmont 1:fdd22bb7aa52 262 si1 = pCoef[(ia1 * 2u) + 1u];
emilmont 1:fdd22bb7aa52 263
emilmont 1:fdd22bb7aa52 264 /* xb' = (xa+yb-xc-yd)co1 + (ya-xb-yc+xd)(si1) */
emilmont 1:fdd22bb7aa52 265 pSrc[2u * i2] = (((int32_t) (((q63_t) r1 * co1) >> 32)) +
emilmont 1:fdd22bb7aa52 266 ((int32_t) (((q63_t) s1 * si1) >> 32))) << 1u;
emilmont 1:fdd22bb7aa52 267
emilmont 1:fdd22bb7aa52 268 /* yb' = (ya-xb-yc+xd)co1 - (xa+yb-xc-yd)(si1) */
emilmont 1:fdd22bb7aa52 269 pSrc[(2u * i2) + 1u] = (((int32_t) (((q63_t) s1 * co1) >> 32)) -
emilmont 1:fdd22bb7aa52 270 ((int32_t) (((q63_t) r1 * si1) >> 32))) << 1u;
emilmont 1:fdd22bb7aa52 271
emilmont 1:fdd22bb7aa52 272 /* index calculation for the coefficients */
emilmont 1:fdd22bb7aa52 273 ia3 = 3u * ia1;
emilmont 1:fdd22bb7aa52 274 co3 = pCoef[ia3 * 2u];
emilmont 1:fdd22bb7aa52 275 si3 = pCoef[(ia3 * 2u) + 1u];
emilmont 1:fdd22bb7aa52 276
emilmont 1:fdd22bb7aa52 277 /* xd' = (xa-yb-xc+yd)co3 + (ya+xb-yc-xd)(si3) */
emilmont 1:fdd22bb7aa52 278 pSrc[2u * i3] = (((int32_t) (((q63_t) r2 * co3) >> 32)) +
emilmont 1:fdd22bb7aa52 279 ((int32_t) (((q63_t) s2 * si3) >> 32))) << 1u;
emilmont 1:fdd22bb7aa52 280
emilmont 1:fdd22bb7aa52 281 /* yd' = (ya+xb-yc-xd)co3 - (xa-yb-xc+yd)(si3) */
emilmont 1:fdd22bb7aa52 282 pSrc[(2u * i3) + 1u] = (((int32_t) (((q63_t) s2 * co3) >> 32)) -
emilmont 1:fdd22bb7aa52 283 ((int32_t) (((q63_t) r2 * si3) >> 32))) << 1u;
emilmont 1:fdd22bb7aa52 284
emilmont 1:fdd22bb7aa52 285 /* Twiddle coefficients index modifier */
emilmont 1:fdd22bb7aa52 286 ia1 = ia1 + twidCoefModifier;
emilmont 1:fdd22bb7aa52 287
emilmont 1:fdd22bb7aa52 288 /* Updating input index */
emilmont 1:fdd22bb7aa52 289 i0 = i0 + 1u;
emilmont 1:fdd22bb7aa52 290
emilmont 1:fdd22bb7aa52 291 } while(--j);
emilmont 1:fdd22bb7aa52 292
emilmont 1:fdd22bb7aa52 293 /* end of first stage process */
emilmont 1:fdd22bb7aa52 294
emilmont 1:fdd22bb7aa52 295 /* data is in 5.27(q27) format */
emilmont 1:fdd22bb7aa52 296
emilmont 1:fdd22bb7aa52 297
emilmont 1:fdd22bb7aa52 298 /* start of Middle stages process */
emilmont 1:fdd22bb7aa52 299
emilmont 1:fdd22bb7aa52 300
emilmont 1:fdd22bb7aa52 301 /* each stage in middle stages provides two down scaling of the input */
emilmont 1:fdd22bb7aa52 302
emilmont 1:fdd22bb7aa52 303 twidCoefModifier <<= 2u;
emilmont 1:fdd22bb7aa52 304
emilmont 1:fdd22bb7aa52 305
emilmont 1:fdd22bb7aa52 306 for (k = fftLen / 4u; k > 4u; k >>= 2u)
emilmont 1:fdd22bb7aa52 307 {
emilmont 1:fdd22bb7aa52 308 /* Initializations for the first stage */
emilmont 1:fdd22bb7aa52 309 n1 = n2;
emilmont 1:fdd22bb7aa52 310 n2 >>= 2u;
emilmont 1:fdd22bb7aa52 311 ia1 = 0u;
emilmont 1:fdd22bb7aa52 312
emilmont 1:fdd22bb7aa52 313 /* Calculation of first stage */
emilmont 1:fdd22bb7aa52 314 for (j = 0u; j <= (n2 - 1u); j++)
emilmont 1:fdd22bb7aa52 315 {
emilmont 1:fdd22bb7aa52 316 /* index calculation for the coefficients */
emilmont 1:fdd22bb7aa52 317 ia2 = ia1 + ia1;
emilmont 1:fdd22bb7aa52 318 ia3 = ia2 + ia1;
emilmont 1:fdd22bb7aa52 319 co1 = pCoef[ia1 * 2u];
emilmont 1:fdd22bb7aa52 320 si1 = pCoef[(ia1 * 2u) + 1u];
emilmont 1:fdd22bb7aa52 321 co2 = pCoef[ia2 * 2u];
emilmont 1:fdd22bb7aa52 322 si2 = pCoef[(ia2 * 2u) + 1u];
emilmont 1:fdd22bb7aa52 323 co3 = pCoef[ia3 * 2u];
emilmont 1:fdd22bb7aa52 324 si3 = pCoef[(ia3 * 2u) + 1u];
emilmont 1:fdd22bb7aa52 325 /* Twiddle coefficients index modifier */
emilmont 1:fdd22bb7aa52 326 ia1 = ia1 + twidCoefModifier;
emilmont 1:fdd22bb7aa52 327
emilmont 1:fdd22bb7aa52 328 for (i0 = j; i0 < fftLen; i0 += n1)
emilmont 1:fdd22bb7aa52 329 {
emilmont 1:fdd22bb7aa52 330 /* index calculation for the input as, */
emilmont 1:fdd22bb7aa52 331 /* pSrc[i0 + 0], pSrc[i0 + fftLen/4], pSrc[i0 + fftLen/2u], pSrc[i0 + 3fftLen/4] */
emilmont 1:fdd22bb7aa52 332 i1 = i0 + n2;
emilmont 1:fdd22bb7aa52 333 i2 = i1 + n2;
emilmont 1:fdd22bb7aa52 334 i3 = i2 + n2;
emilmont 1:fdd22bb7aa52 335
emilmont 1:fdd22bb7aa52 336 /* Butterfly implementation */
emilmont 1:fdd22bb7aa52 337 /* xa + xc */
emilmont 1:fdd22bb7aa52 338 r1 = pSrc[2u * i0] + pSrc[2u * i2];
emilmont 1:fdd22bb7aa52 339 /* xa - xc */
emilmont 1:fdd22bb7aa52 340 r2 = pSrc[2u * i0] - pSrc[2u * i2];
emilmont 1:fdd22bb7aa52 341
emilmont 1:fdd22bb7aa52 342 /* ya + yc */
emilmont 1:fdd22bb7aa52 343 s1 = pSrc[(2u * i0) + 1u] + pSrc[(2u * i2) + 1u];
emilmont 1:fdd22bb7aa52 344 /* ya - yc */
emilmont 1:fdd22bb7aa52 345 s2 = pSrc[(2u * i0) + 1u] - pSrc[(2u * i2) + 1u];
emilmont 1:fdd22bb7aa52 346
emilmont 1:fdd22bb7aa52 347 /* xb + xd */
emilmont 1:fdd22bb7aa52 348 t1 = pSrc[2u * i1] + pSrc[2u * i3];
emilmont 1:fdd22bb7aa52 349
emilmont 1:fdd22bb7aa52 350 /* xa' = xa + xb + xc + xd */
emilmont 1:fdd22bb7aa52 351 pSrc[2u * i0] = (r1 + t1) >> 2u;
emilmont 1:fdd22bb7aa52 352 /* xa + xc -(xb + xd) */
emilmont 1:fdd22bb7aa52 353 r1 = r1 - t1;
emilmont 1:fdd22bb7aa52 354
emilmont 1:fdd22bb7aa52 355 /* yb + yd */
emilmont 1:fdd22bb7aa52 356 t2 = pSrc[(2u * i1) + 1u] + pSrc[(2u * i3) + 1u];
emilmont 1:fdd22bb7aa52 357 /* ya' = ya + yb + yc + yd */
emilmont 1:fdd22bb7aa52 358 pSrc[(2u * i0) + 1u] = (s1 + t2) >> 2u;
emilmont 1:fdd22bb7aa52 359
emilmont 1:fdd22bb7aa52 360 /* (ya + yc) - (yb + yd) */
emilmont 1:fdd22bb7aa52 361 s1 = s1 - t2;
emilmont 1:fdd22bb7aa52 362
emilmont 1:fdd22bb7aa52 363 /* (yb - yd) */
emilmont 1:fdd22bb7aa52 364 t1 = pSrc[(2u * i1) + 1u] - pSrc[(2u * i3) + 1u];
emilmont 1:fdd22bb7aa52 365 /* (xb - xd) */
emilmont 1:fdd22bb7aa52 366 t2 = pSrc[2u * i1] - pSrc[2u * i3];
emilmont 1:fdd22bb7aa52 367
emilmont 1:fdd22bb7aa52 368 /* xc' = (xa-xb+xc-xd)co2 + (ya-yb+yc-yd)(si2) */
emilmont 1:fdd22bb7aa52 369 pSrc[2u * i1] = (((int32_t) (((q63_t) r1 * co2) >> 32)) +
emilmont 1:fdd22bb7aa52 370 ((int32_t) (((q63_t) s1 * si2) >> 32))) >> 1u;
emilmont 1:fdd22bb7aa52 371
emilmont 1:fdd22bb7aa52 372 /* yc' = (ya-yb+yc-yd)co2 - (xa-xb+xc-xd)(si2) */
emilmont 1:fdd22bb7aa52 373 pSrc[(2u * i1) + 1u] = (((int32_t) (((q63_t) s1 * co2) >> 32)) -
emilmont 1:fdd22bb7aa52 374 ((int32_t) (((q63_t) r1 * si2) >> 32))) >> 1u;
emilmont 1:fdd22bb7aa52 375
emilmont 1:fdd22bb7aa52 376 /* (xa - xc) + (yb - yd) */
emilmont 1:fdd22bb7aa52 377 r1 = r2 + t1;
emilmont 1:fdd22bb7aa52 378 /* (xa - xc) - (yb - yd) */
emilmont 1:fdd22bb7aa52 379 r2 = r2 - t1;
emilmont 1:fdd22bb7aa52 380
emilmont 1:fdd22bb7aa52 381 /* (ya - yc) - (xb - xd) */
emilmont 1:fdd22bb7aa52 382 s1 = s2 - t2;
emilmont 1:fdd22bb7aa52 383 /* (ya - yc) + (xb - xd) */
emilmont 1:fdd22bb7aa52 384 s2 = s2 + t2;
emilmont 1:fdd22bb7aa52 385
emilmont 1:fdd22bb7aa52 386 /* xb' = (xa+yb-xc-yd)co1 + (ya-xb-yc+xd)(si1) */
emilmont 1:fdd22bb7aa52 387 pSrc[2u * i2] = (((int32_t) (((q63_t) r1 * co1) >> 32)) +
emilmont 1:fdd22bb7aa52 388 ((int32_t) (((q63_t) s1 * si1) >> 32))) >> 1u;
emilmont 1:fdd22bb7aa52 389
emilmont 1:fdd22bb7aa52 390 /* yb' = (ya-xb-yc+xd)co1 - (xa+yb-xc-yd)(si1) */
emilmont 1:fdd22bb7aa52 391 pSrc[(2u * i2) + 1u] = (((int32_t) (((q63_t) s1 * co1) >> 32)) -
emilmont 1:fdd22bb7aa52 392 ((int32_t) (((q63_t) r1 * si1) >> 32))) >> 1u;
emilmont 1:fdd22bb7aa52 393
emilmont 1:fdd22bb7aa52 394 /* xd' = (xa-yb-xc+yd)co3 + (ya+xb-yc-xd)(si3) */
emilmont 1:fdd22bb7aa52 395 pSrc[2u * i3] = (((int32_t) (((q63_t) r2 * co3) >> 32)) +
emilmont 1:fdd22bb7aa52 396 ((int32_t) (((q63_t) s2 * si3) >> 32))) >> 1u;
emilmont 1:fdd22bb7aa52 397
emilmont 1:fdd22bb7aa52 398 /* yd' = (ya+xb-yc-xd)co3 - (xa-yb-xc+yd)(si3) */
emilmont 1:fdd22bb7aa52 399 pSrc[(2u * i3) + 1u] = (((int32_t) (((q63_t) s2 * co3) >> 32)) -
emilmont 1:fdd22bb7aa52 400 ((int32_t) (((q63_t) r2 * si3) >> 32))) >> 1u;
emilmont 1:fdd22bb7aa52 401 }
emilmont 1:fdd22bb7aa52 402 }
emilmont 1:fdd22bb7aa52 403 twidCoefModifier <<= 2u;
emilmont 1:fdd22bb7aa52 404 }
emilmont 1:fdd22bb7aa52 405
emilmont 1:fdd22bb7aa52 406 /* End of Middle stages process */
emilmont 1:fdd22bb7aa52 407
emilmont 1:fdd22bb7aa52 408 /* data is in 11.21(q21) format for the 1024 point as there are 3 middle stages */
emilmont 1:fdd22bb7aa52 409 /* data is in 9.23(q23) format for the 256 point as there are 2 middle stages */
emilmont 1:fdd22bb7aa52 410 /* data is in 7.25(q25) format for the 64 point as there are 1 middle stage */
emilmont 1:fdd22bb7aa52 411 /* data is in 5.27(q27) format for the 16 point as there are no middle stages */
emilmont 1:fdd22bb7aa52 412
emilmont 1:fdd22bb7aa52 413
emilmont 1:fdd22bb7aa52 414 /* start of Last stage process */
emilmont 1:fdd22bb7aa52 415 /* Initializations for the last stage */
emilmont 1:fdd22bb7aa52 416 j = fftLen >> 2;
emilmont 1:fdd22bb7aa52 417 ptr1 = &pSrc[0];
emilmont 1:fdd22bb7aa52 418
emilmont 1:fdd22bb7aa52 419 /* Calculations of last stage */
emilmont 1:fdd22bb7aa52 420 do
emilmont 1:fdd22bb7aa52 421 {
emilmont 1:fdd22bb7aa52 422
emilmont 1:fdd22bb7aa52 423 #ifndef ARM_MATH_BIG_ENDIAN
emilmont 1:fdd22bb7aa52 424
emilmont 1:fdd22bb7aa52 425 /* Read xa (real), ya(imag) input */
emilmont 1:fdd22bb7aa52 426 xaya = *__SIMD64(ptr1)++;
emilmont 1:fdd22bb7aa52 427 xa = (q31_t) xaya;
emilmont 1:fdd22bb7aa52 428 ya = (q31_t) (xaya >> 32);
emilmont 1:fdd22bb7aa52 429
emilmont 1:fdd22bb7aa52 430 /* Read xb (real), yb(imag) input */
emilmont 1:fdd22bb7aa52 431 xbyb = *__SIMD64(ptr1)++;
emilmont 1:fdd22bb7aa52 432 xb = (q31_t) xbyb;
emilmont 1:fdd22bb7aa52 433 yb = (q31_t) (xbyb >> 32);
emilmont 1:fdd22bb7aa52 434
emilmont 1:fdd22bb7aa52 435 /* Read xc (real), yc(imag) input */
emilmont 1:fdd22bb7aa52 436 xcyc = *__SIMD64(ptr1)++;
emilmont 1:fdd22bb7aa52 437 xc = (q31_t) xcyc;
emilmont 1:fdd22bb7aa52 438 yc = (q31_t) (xcyc >> 32);
emilmont 1:fdd22bb7aa52 439
emilmont 1:fdd22bb7aa52 440 /* Read xc (real), yc(imag) input */
emilmont 1:fdd22bb7aa52 441 xdyd = *__SIMD64(ptr1)++;
emilmont 1:fdd22bb7aa52 442 xd = (q31_t) xdyd;
emilmont 1:fdd22bb7aa52 443 yd = (q31_t) (xdyd >> 32);
emilmont 1:fdd22bb7aa52 444
emilmont 1:fdd22bb7aa52 445 #else
emilmont 1:fdd22bb7aa52 446
emilmont 1:fdd22bb7aa52 447 /* Read xa (real), ya(imag) input */
emilmont 1:fdd22bb7aa52 448 xaya = *__SIMD64(ptr1)++;
emilmont 1:fdd22bb7aa52 449 ya = (q31_t) xaya;
emilmont 1:fdd22bb7aa52 450 xa = (q31_t) (xaya >> 32);
emilmont 1:fdd22bb7aa52 451
emilmont 1:fdd22bb7aa52 452 /* Read xb (real), yb(imag) input */
emilmont 1:fdd22bb7aa52 453 xbyb = *__SIMD64(ptr1)++;
emilmont 1:fdd22bb7aa52 454 yb = (q31_t) xbyb;
emilmont 1:fdd22bb7aa52 455 xb = (q31_t) (xbyb >> 32);
emilmont 1:fdd22bb7aa52 456
emilmont 1:fdd22bb7aa52 457 /* Read xc (real), yc(imag) input */
emilmont 1:fdd22bb7aa52 458 xcyc = *__SIMD64(ptr1)++;
emilmont 1:fdd22bb7aa52 459 yc = (q31_t) xcyc;
emilmont 1:fdd22bb7aa52 460 xc = (q31_t) (xcyc >> 32);
emilmont 1:fdd22bb7aa52 461
emilmont 1:fdd22bb7aa52 462 /* Read xc (real), yc(imag) input */
emilmont 1:fdd22bb7aa52 463 xdyd = *__SIMD64(ptr1)++;
emilmont 1:fdd22bb7aa52 464 yd = (q31_t) xdyd;
emilmont 1:fdd22bb7aa52 465 xd = (q31_t) (xdyd >> 32);
emilmont 1:fdd22bb7aa52 466
emilmont 1:fdd22bb7aa52 467
emilmont 1:fdd22bb7aa52 468 #endif
emilmont 1:fdd22bb7aa52 469
emilmont 1:fdd22bb7aa52 470 /* xa' = xa + xb + xc + xd */
emilmont 1:fdd22bb7aa52 471 xa_out = xa + xb + xc + xd;
emilmont 1:fdd22bb7aa52 472
emilmont 1:fdd22bb7aa52 473 /* ya' = ya + yb + yc + yd */
emilmont 1:fdd22bb7aa52 474 ya_out = ya + yb + yc + yd;
emilmont 1:fdd22bb7aa52 475
emilmont 1:fdd22bb7aa52 476 /* pointer updation for writing */
emilmont 1:fdd22bb7aa52 477 ptr1 = ptr1 - 8u;
emilmont 1:fdd22bb7aa52 478
emilmont 1:fdd22bb7aa52 479 /* writing xa' and ya' */
emilmont 1:fdd22bb7aa52 480 *ptr1++ = xa_out;
emilmont 1:fdd22bb7aa52 481 *ptr1++ = ya_out;
emilmont 1:fdd22bb7aa52 482
emilmont 1:fdd22bb7aa52 483 xc_out = (xa - xb + xc - xd);
emilmont 1:fdd22bb7aa52 484 yc_out = (ya - yb + yc - yd);
emilmont 1:fdd22bb7aa52 485
emilmont 1:fdd22bb7aa52 486 /* writing xc' and yc' */
emilmont 1:fdd22bb7aa52 487 *ptr1++ = xc_out;
emilmont 1:fdd22bb7aa52 488 *ptr1++ = yc_out;
emilmont 1:fdd22bb7aa52 489
emilmont 1:fdd22bb7aa52 490 xb_out = (xa + yb - xc - yd);
emilmont 1:fdd22bb7aa52 491 yb_out = (ya - xb - yc + xd);
emilmont 1:fdd22bb7aa52 492
emilmont 1:fdd22bb7aa52 493 /* writing xb' and yb' */
emilmont 1:fdd22bb7aa52 494 *ptr1++ = xb_out;
emilmont 1:fdd22bb7aa52 495 *ptr1++ = yb_out;
emilmont 1:fdd22bb7aa52 496
emilmont 1:fdd22bb7aa52 497 xd_out = (xa - yb - xc + yd);
emilmont 1:fdd22bb7aa52 498 yd_out = (ya + xb - yc - xd);
emilmont 1:fdd22bb7aa52 499
emilmont 1:fdd22bb7aa52 500 /* writing xd' and yd' */
emilmont 1:fdd22bb7aa52 501 *ptr1++ = xd_out;
emilmont 1:fdd22bb7aa52 502 *ptr1++ = yd_out;
emilmont 1:fdd22bb7aa52 503
emilmont 1:fdd22bb7aa52 504
emilmont 1:fdd22bb7aa52 505 } while(--j);
emilmont 1:fdd22bb7aa52 506
emilmont 1:fdd22bb7aa52 507 /* output is in 11.21(q21) format for the 1024 point */
emilmont 1:fdd22bb7aa52 508 /* output is in 9.23(q23) format for the 256 point */
emilmont 1:fdd22bb7aa52 509 /* output is in 7.25(q25) format for the 64 point */
emilmont 1:fdd22bb7aa52 510 /* output is in 5.27(q27) format for the 16 point */
emilmont 1:fdd22bb7aa52 511
emilmont 1:fdd22bb7aa52 512 /* End of last stage process */
emilmont 1:fdd22bb7aa52 513
emilmont 1:fdd22bb7aa52 514 }
emilmont 1:fdd22bb7aa52 515
emilmont 1:fdd22bb7aa52 516
emilmont 1:fdd22bb7aa52 517 /**
emilmont 1:fdd22bb7aa52 518 * @brief Core function for the Q31 CIFFT butterfly process.
emilmont 1:fdd22bb7aa52 519 * @param[in, out] *pSrc points to the in-place buffer of Q31 data type.
emilmont 1:fdd22bb7aa52 520 * @param[in] fftLen length of the FFT.
emilmont 1:fdd22bb7aa52 521 * @param[in] *pCoef points to twiddle coefficient buffer.
emilmont 1:fdd22bb7aa52 522 * @param[in] twidCoefModifier twiddle coefficient modifier that supports different size FFTs with the same twiddle factor table.
emilmont 1:fdd22bb7aa52 523 * @return none.
emilmont 1:fdd22bb7aa52 524 */
emilmont 1:fdd22bb7aa52 525
emilmont 1:fdd22bb7aa52 526
emilmont 1:fdd22bb7aa52 527 /*
emilmont 1:fdd22bb7aa52 528 * Radix-4 IFFT algorithm used is :
emilmont 1:fdd22bb7aa52 529 *
emilmont 1:fdd22bb7aa52 530 * CIFFT uses same twiddle coefficients as CFFT Function
emilmont 1:fdd22bb7aa52 531 * x[k] = x[n] + (j)k * x[n + fftLen/4] + (-1)k * x[n+fftLen/2] + (-j)k * x[n+3*fftLen/4]
emilmont 1:fdd22bb7aa52 532 *
emilmont 1:fdd22bb7aa52 533 *
emilmont 1:fdd22bb7aa52 534 * IFFT is implemented with following changes in equations from FFT
emilmont 1:fdd22bb7aa52 535 *
emilmont 1:fdd22bb7aa52 536 * Input real and imaginary data:
emilmont 1:fdd22bb7aa52 537 * x(n) = xa + j * ya
emilmont 1:fdd22bb7aa52 538 * x(n+N/4 ) = xb + j * yb
emilmont 1:fdd22bb7aa52 539 * x(n+N/2 ) = xc + j * yc
emilmont 1:fdd22bb7aa52 540 * x(n+3N 4) = xd + j * yd
emilmont 1:fdd22bb7aa52 541 *
emilmont 1:fdd22bb7aa52 542 *
emilmont 1:fdd22bb7aa52 543 * Output real and imaginary data:
emilmont 1:fdd22bb7aa52 544 * x(4r) = xa'+ j * ya'
emilmont 1:fdd22bb7aa52 545 * x(4r+1) = xb'+ j * yb'
emilmont 1:fdd22bb7aa52 546 * x(4r+2) = xc'+ j * yc'
emilmont 1:fdd22bb7aa52 547 * x(4r+3) = xd'+ j * yd'
emilmont 1:fdd22bb7aa52 548 *
emilmont 1:fdd22bb7aa52 549 *
emilmont 1:fdd22bb7aa52 550 * Twiddle factors for radix-4 IFFT:
emilmont 1:fdd22bb7aa52 551 * Wn = co1 + j * (si1)
emilmont 1:fdd22bb7aa52 552 * W2n = co2 + j * (si2)
emilmont 1:fdd22bb7aa52 553 * W3n = co3 + j * (si3)
emilmont 1:fdd22bb7aa52 554
emilmont 1:fdd22bb7aa52 555 * The real and imaginary output values for the radix-4 butterfly are
emilmont 1:fdd22bb7aa52 556 * xa' = xa + xb + xc + xd
emilmont 1:fdd22bb7aa52 557 * ya' = ya + yb + yc + yd
emilmont 1:fdd22bb7aa52 558 * xb' = (xa-yb-xc+yd)* co1 - (ya+xb-yc-xd)* (si1)
emilmont 1:fdd22bb7aa52 559 * yb' = (ya+xb-yc-xd)* co1 + (xa-yb-xc+yd)* (si1)
emilmont 1:fdd22bb7aa52 560 * xc' = (xa-xb+xc-xd)* co2 - (ya-yb+yc-yd)* (si2)
emilmont 1:fdd22bb7aa52 561 * yc' = (ya-yb+yc-yd)* co2 + (xa-xb+xc-xd)* (si2)
emilmont 1:fdd22bb7aa52 562 * xd' = (xa+yb-xc-yd)* co3 - (ya-xb-yc+xd)* (si3)
emilmont 1:fdd22bb7aa52 563 * yd' = (ya-xb-yc+xd)* co3 + (xa+yb-xc-yd)* (si3)
emilmont 1:fdd22bb7aa52 564 *
emilmont 1:fdd22bb7aa52 565 */
emilmont 1:fdd22bb7aa52 566
emilmont 1:fdd22bb7aa52 567 void arm_radix4_butterfly_inverse_q31(
emilmont 1:fdd22bb7aa52 568 q31_t * pSrc,
emilmont 1:fdd22bb7aa52 569 uint32_t fftLen,
emilmont 1:fdd22bb7aa52 570 q31_t * pCoef,
emilmont 1:fdd22bb7aa52 571 uint32_t twidCoefModifier)
emilmont 1:fdd22bb7aa52 572 {
emilmont 1:fdd22bb7aa52 573 uint32_t n1, n2, ia1, ia2, ia3, i0, i1, i2, i3, j, k;
emilmont 1:fdd22bb7aa52 574 q31_t t1, t2, r1, r2, s1, s2, co1, co2, co3, si1, si2, si3;
emilmont 1:fdd22bb7aa52 575 q31_t xa, xb, xc, xd;
emilmont 1:fdd22bb7aa52 576 q31_t ya, yb, yc, yd;
emilmont 1:fdd22bb7aa52 577 q31_t xa_out, xb_out, xc_out, xd_out;
emilmont 1:fdd22bb7aa52 578 q31_t ya_out, yb_out, yc_out, yd_out;
emilmont 1:fdd22bb7aa52 579
emilmont 1:fdd22bb7aa52 580 q31_t *ptr1;
emilmont 1:fdd22bb7aa52 581 q63_t xaya, xbyb, xcyc, xdyd;
emilmont 1:fdd22bb7aa52 582
emilmont 1:fdd22bb7aa52 583 /* input is be 1.31(q31) format for all FFT sizes */
emilmont 1:fdd22bb7aa52 584 /* Total process is divided into three stages */
emilmont 1:fdd22bb7aa52 585 /* process first stage, middle stages, & last stage */
emilmont 1:fdd22bb7aa52 586
emilmont 1:fdd22bb7aa52 587 /* Start of first stage process */
emilmont 1:fdd22bb7aa52 588
emilmont 1:fdd22bb7aa52 589 /* Initializations for the first stage */
emilmont 1:fdd22bb7aa52 590 n2 = fftLen;
emilmont 1:fdd22bb7aa52 591 n1 = n2;
emilmont 1:fdd22bb7aa52 592 /* n2 = fftLen/4 */
emilmont 1:fdd22bb7aa52 593 n2 >>= 2u;
emilmont 1:fdd22bb7aa52 594 i0 = 0u;
emilmont 1:fdd22bb7aa52 595 ia1 = 0u;
emilmont 1:fdd22bb7aa52 596
emilmont 1:fdd22bb7aa52 597 j = n2;
emilmont 1:fdd22bb7aa52 598
emilmont 1:fdd22bb7aa52 599 do
emilmont 1:fdd22bb7aa52 600 {
emilmont 1:fdd22bb7aa52 601
emilmont 1:fdd22bb7aa52 602 /* input is in 1.31(q31) format and provide 4 guard bits for the input */
emilmont 1:fdd22bb7aa52 603
emilmont 1:fdd22bb7aa52 604 /* index calculation for the input as, */
emilmont 1:fdd22bb7aa52 605 /* pSrc[i0 + 0], pSrc[i0 + fftLen/4], pSrc[i0 + fftLen/2u], pSrc[i0 + 3fftLen/4] */
emilmont 1:fdd22bb7aa52 606 i1 = i0 + n2;
emilmont 1:fdd22bb7aa52 607 i2 = i1 + n2;
emilmont 1:fdd22bb7aa52 608 i3 = i2 + n2;
emilmont 1:fdd22bb7aa52 609
emilmont 1:fdd22bb7aa52 610 /* Butterfly implementation */
emilmont 1:fdd22bb7aa52 611 /* xa + xc */
emilmont 1:fdd22bb7aa52 612 r1 = (pSrc[2u * i0] >> 4u) + (pSrc[2u * i2] >> 4u);
emilmont 1:fdd22bb7aa52 613 /* xa - xc */
emilmont 1:fdd22bb7aa52 614 r2 = (pSrc[2u * i0] >> 4u) - (pSrc[2u * i2] >> 4u);
emilmont 1:fdd22bb7aa52 615
emilmont 1:fdd22bb7aa52 616 /* xb + xd */
emilmont 1:fdd22bb7aa52 617 t1 = (pSrc[2u * i1] >> 4u) + (pSrc[2u * i3] >> 4u);
emilmont 1:fdd22bb7aa52 618
emilmont 1:fdd22bb7aa52 619 /* ya + yc */
emilmont 1:fdd22bb7aa52 620 s1 = (pSrc[(2u * i0) + 1u] >> 4u) + (pSrc[(2u * i2) + 1u] >> 4u);
emilmont 1:fdd22bb7aa52 621 /* ya - yc */
emilmont 1:fdd22bb7aa52 622 s2 = (pSrc[(2u * i0) + 1u] >> 4u) - (pSrc[(2u * i2) + 1u] >> 4u);
emilmont 1:fdd22bb7aa52 623
emilmont 1:fdd22bb7aa52 624 /* xa' = xa + xb + xc + xd */
emilmont 1:fdd22bb7aa52 625 pSrc[2u * i0] = (r1 + t1);
emilmont 1:fdd22bb7aa52 626 /* (xa + xc) - (xb + xd) */
emilmont 1:fdd22bb7aa52 627 r1 = r1 - t1;
emilmont 1:fdd22bb7aa52 628 /* yb + yd */
emilmont 1:fdd22bb7aa52 629 t2 = (pSrc[(2u * i1) + 1u] >> 4u) + (pSrc[(2u * i3) + 1u] >> 4u);
emilmont 1:fdd22bb7aa52 630 /* ya' = ya + yb + yc + yd */
emilmont 1:fdd22bb7aa52 631 pSrc[(2u * i0) + 1u] = (s1 + t2);
emilmont 1:fdd22bb7aa52 632
emilmont 1:fdd22bb7aa52 633 /* (ya + yc) - (yb + yd) */
emilmont 1:fdd22bb7aa52 634 s1 = s1 - t2;
emilmont 1:fdd22bb7aa52 635
emilmont 1:fdd22bb7aa52 636 /* yb - yd */
emilmont 1:fdd22bb7aa52 637 t1 = (pSrc[(2u * i1) + 1u] >> 4u) - (pSrc[(2u * i3) + 1u] >> 4u);
emilmont 1:fdd22bb7aa52 638 /* xb - xd */
emilmont 1:fdd22bb7aa52 639 t2 = (pSrc[2u * i1] >> 4u) - (pSrc[2u * i3] >> 4u);
emilmont 1:fdd22bb7aa52 640
emilmont 1:fdd22bb7aa52 641 /* index calculation for the coefficients */
emilmont 1:fdd22bb7aa52 642 ia2 = 2u * ia1;
emilmont 1:fdd22bb7aa52 643 co2 = pCoef[ia2 * 2u];
emilmont 1:fdd22bb7aa52 644 si2 = pCoef[(ia2 * 2u) + 1u];
emilmont 1:fdd22bb7aa52 645
emilmont 1:fdd22bb7aa52 646 /* xc' = (xa-xb+xc-xd)co2 - (ya-yb+yc-yd)(si2) */
emilmont 1:fdd22bb7aa52 647 pSrc[2u * i1] = (((int32_t) (((q63_t) r1 * co2) >> 32)) -
emilmont 1:fdd22bb7aa52 648 ((int32_t) (((q63_t) s1 * si2) >> 32))) << 1u;
emilmont 1:fdd22bb7aa52 649
emilmont 1:fdd22bb7aa52 650 /* yc' = (ya-yb+yc-yd)co2 + (xa-xb+xc-xd)(si2) */
emilmont 1:fdd22bb7aa52 651 pSrc[2u * i1 + 1u] = (((int32_t) (((q63_t) s1 * co2) >> 32)) +
emilmont 1:fdd22bb7aa52 652 ((int32_t) (((q63_t) r1 * si2) >> 32))) << 1u;
emilmont 1:fdd22bb7aa52 653
emilmont 1:fdd22bb7aa52 654 /* (xa - xc) - (yb - yd) */
emilmont 1:fdd22bb7aa52 655 r1 = r2 - t1;
emilmont 1:fdd22bb7aa52 656 /* (xa - xc) + (yb - yd) */
emilmont 1:fdd22bb7aa52 657 r2 = r2 + t1;
emilmont 1:fdd22bb7aa52 658
emilmont 1:fdd22bb7aa52 659 /* (ya - yc) + (xb - xd) */
emilmont 1:fdd22bb7aa52 660 s1 = s2 + t2;
emilmont 1:fdd22bb7aa52 661 /* (ya - yc) - (xb - xd) */
emilmont 1:fdd22bb7aa52 662 s2 = s2 - t2;
emilmont 1:fdd22bb7aa52 663
emilmont 1:fdd22bb7aa52 664 co1 = pCoef[ia1 * 2u];
emilmont 1:fdd22bb7aa52 665 si1 = pCoef[(ia1 * 2u) + 1u];
emilmont 1:fdd22bb7aa52 666
emilmont 1:fdd22bb7aa52 667 /* xb' = (xa+yb-xc-yd)co1 - (ya-xb-yc+xd)(si1) */
emilmont 1:fdd22bb7aa52 668 pSrc[2u * i2] = (((int32_t) (((q63_t) r1 * co1) >> 32)) -
emilmont 1:fdd22bb7aa52 669 ((int32_t) (((q63_t) s1 * si1) >> 32))) << 1u;
emilmont 1:fdd22bb7aa52 670
emilmont 1:fdd22bb7aa52 671 /* yb' = (ya-xb-yc+xd)co1 + (xa+yb-xc-yd)(si1) */
emilmont 1:fdd22bb7aa52 672 pSrc[(2u * i2) + 1u] = (((int32_t) (((q63_t) s1 * co1) >> 32)) +
emilmont 1:fdd22bb7aa52 673 ((int32_t) (((q63_t) r1 * si1) >> 32))) << 1u;
emilmont 1:fdd22bb7aa52 674
emilmont 1:fdd22bb7aa52 675 /* index calculation for the coefficients */
emilmont 1:fdd22bb7aa52 676 ia3 = 3u * ia1;
emilmont 1:fdd22bb7aa52 677 co3 = pCoef[ia3 * 2u];
emilmont 1:fdd22bb7aa52 678 si3 = pCoef[(ia3 * 2u) + 1u];
emilmont 1:fdd22bb7aa52 679
emilmont 1:fdd22bb7aa52 680 /* xd' = (xa-yb-xc+yd)co3 - (ya+xb-yc-xd)(si3) */
emilmont 1:fdd22bb7aa52 681 pSrc[2u * i3] = (((int32_t) (((q63_t) r2 * co3) >> 32)) -
emilmont 1:fdd22bb7aa52 682 ((int32_t) (((q63_t) s2 * si3) >> 32))) << 1u;
emilmont 1:fdd22bb7aa52 683
emilmont 1:fdd22bb7aa52 684 /* yd' = (ya+xb-yc-xd)co3 + (xa-yb-xc+yd)(si3) */
emilmont 1:fdd22bb7aa52 685 pSrc[(2u * i3) + 1u] = (((int32_t) (((q63_t) s2 * co3) >> 32)) +
emilmont 1:fdd22bb7aa52 686 ((int32_t) (((q63_t) r2 * si3) >> 32))) << 1u;
emilmont 1:fdd22bb7aa52 687
emilmont 1:fdd22bb7aa52 688 /* Twiddle coefficients index modifier */
emilmont 1:fdd22bb7aa52 689 ia1 = ia1 + twidCoefModifier;
emilmont 1:fdd22bb7aa52 690
emilmont 1:fdd22bb7aa52 691 /* Updating input index */
emilmont 1:fdd22bb7aa52 692 i0 = i0 + 1u;
emilmont 1:fdd22bb7aa52 693
emilmont 1:fdd22bb7aa52 694 } while(--j);
emilmont 1:fdd22bb7aa52 695
emilmont 1:fdd22bb7aa52 696 /* data is in 5.27(q27) format */
emilmont 1:fdd22bb7aa52 697 /* each stage provides two down scaling of the input */
emilmont 1:fdd22bb7aa52 698
emilmont 1:fdd22bb7aa52 699
emilmont 1:fdd22bb7aa52 700 /* Start of Middle stages process */
emilmont 1:fdd22bb7aa52 701
emilmont 1:fdd22bb7aa52 702 twidCoefModifier <<= 2u;
emilmont 1:fdd22bb7aa52 703
emilmont 1:fdd22bb7aa52 704 /* Calculation of second stage to excluding last stage */
emilmont 1:fdd22bb7aa52 705 for (k = fftLen / 4u; k > 4u; k >>= 2u)
emilmont 1:fdd22bb7aa52 706 {
emilmont 1:fdd22bb7aa52 707 /* Initializations for the first stage */
emilmont 1:fdd22bb7aa52 708 n1 = n2;
emilmont 1:fdd22bb7aa52 709 n2 >>= 2u;
emilmont 1:fdd22bb7aa52 710 ia1 = 0u;
emilmont 1:fdd22bb7aa52 711
emilmont 1:fdd22bb7aa52 712 for (j = 0; j <= (n2 - 1u); j++)
emilmont 1:fdd22bb7aa52 713 {
emilmont 1:fdd22bb7aa52 714 /* index calculation for the coefficients */
emilmont 1:fdd22bb7aa52 715 ia2 = ia1 + ia1;
emilmont 1:fdd22bb7aa52 716 ia3 = ia2 + ia1;
emilmont 1:fdd22bb7aa52 717 co1 = pCoef[ia1 * 2u];
emilmont 1:fdd22bb7aa52 718 si1 = pCoef[(ia1 * 2u) + 1u];
emilmont 1:fdd22bb7aa52 719 co2 = pCoef[ia2 * 2u];
emilmont 1:fdd22bb7aa52 720 si2 = pCoef[(ia2 * 2u) + 1u];
emilmont 1:fdd22bb7aa52 721 co3 = pCoef[ia3 * 2u];
emilmont 1:fdd22bb7aa52 722 si3 = pCoef[(ia3 * 2u) + 1u];
emilmont 1:fdd22bb7aa52 723 /* Twiddle coefficients index modifier */
emilmont 1:fdd22bb7aa52 724 ia1 = ia1 + twidCoefModifier;
emilmont 1:fdd22bb7aa52 725
emilmont 1:fdd22bb7aa52 726 for (i0 = j; i0 < fftLen; i0 += n1)
emilmont 1:fdd22bb7aa52 727 {
emilmont 1:fdd22bb7aa52 728 /* index calculation for the input as, */
emilmont 1:fdd22bb7aa52 729 /* pSrc[i0 + 0], pSrc[i0 + fftLen/4], pSrc[i0 + fftLen/2u], pSrc[i0 + 3fftLen/4] */
emilmont 1:fdd22bb7aa52 730 i1 = i0 + n2;
emilmont 1:fdd22bb7aa52 731 i2 = i1 + n2;
emilmont 1:fdd22bb7aa52 732 i3 = i2 + n2;
emilmont 1:fdd22bb7aa52 733
emilmont 1:fdd22bb7aa52 734 /* Butterfly implementation */
emilmont 1:fdd22bb7aa52 735 /* xa + xc */
emilmont 1:fdd22bb7aa52 736 r1 = pSrc[2u * i0] + pSrc[2u * i2];
emilmont 1:fdd22bb7aa52 737 /* xa - xc */
emilmont 1:fdd22bb7aa52 738 r2 = pSrc[2u * i0] - pSrc[2u * i2];
emilmont 1:fdd22bb7aa52 739
emilmont 1:fdd22bb7aa52 740 /* ya + yc */
emilmont 1:fdd22bb7aa52 741 s1 = pSrc[(2u * i0) + 1u] + pSrc[(2u * i2) + 1u];
emilmont 1:fdd22bb7aa52 742 /* ya - yc */
emilmont 1:fdd22bb7aa52 743 s2 = pSrc[(2u * i0) + 1u] - pSrc[(2u * i2) + 1u];
emilmont 1:fdd22bb7aa52 744
emilmont 1:fdd22bb7aa52 745 /* xb + xd */
emilmont 1:fdd22bb7aa52 746 t1 = pSrc[2u * i1] + pSrc[2u * i3];
emilmont 1:fdd22bb7aa52 747
emilmont 1:fdd22bb7aa52 748 /* xa' = xa + xb + xc + xd */
emilmont 1:fdd22bb7aa52 749 pSrc[2u * i0] = (r1 + t1) >> 2u;
emilmont 1:fdd22bb7aa52 750 /* xa + xc -(xb + xd) */
emilmont 1:fdd22bb7aa52 751 r1 = r1 - t1;
emilmont 1:fdd22bb7aa52 752 /* yb + yd */
emilmont 1:fdd22bb7aa52 753 t2 = pSrc[(2u * i1) + 1u] + pSrc[(2u * i3) + 1u];
emilmont 1:fdd22bb7aa52 754 /* ya' = ya + yb + yc + yd */
emilmont 1:fdd22bb7aa52 755 pSrc[(2u * i0) + 1u] = (s1 + t2) >> 2u;
emilmont 1:fdd22bb7aa52 756
emilmont 1:fdd22bb7aa52 757 /* (ya + yc) - (yb + yd) */
emilmont 1:fdd22bb7aa52 758 s1 = s1 - t2;
emilmont 1:fdd22bb7aa52 759
emilmont 1:fdd22bb7aa52 760 /* (yb - yd) */
emilmont 1:fdd22bb7aa52 761 t1 = pSrc[(2u * i1) + 1u] - pSrc[(2u * i3) + 1u];
emilmont 1:fdd22bb7aa52 762 /* (xb - xd) */
emilmont 1:fdd22bb7aa52 763 t2 = pSrc[2u * i1] - pSrc[2u * i3];
emilmont 1:fdd22bb7aa52 764
emilmont 1:fdd22bb7aa52 765 /* xc' = (xa-xb+xc-xd)co2 - (ya-yb+yc-yd)(si2) */
emilmont 1:fdd22bb7aa52 766 pSrc[2u * i1] = (((int32_t) (((q63_t) r1 * co2) >> 32u)) -
emilmont 1:fdd22bb7aa52 767 ((int32_t) (((q63_t) s1 * si2) >> 32u))) >> 1u;
emilmont 1:fdd22bb7aa52 768
emilmont 1:fdd22bb7aa52 769 /* yc' = (ya-yb+yc-yd)co2 + (xa-xb+xc-xd)(si2) */
emilmont 1:fdd22bb7aa52 770 pSrc[(2u * i1) + 1u] =
emilmont 1:fdd22bb7aa52 771 (((int32_t) (((q63_t) s1 * co2) >> 32u)) +
emilmont 1:fdd22bb7aa52 772 ((int32_t) (((q63_t) r1 * si2) >> 32u))) >> 1u;
emilmont 1:fdd22bb7aa52 773
emilmont 1:fdd22bb7aa52 774 /* (xa - xc) - (yb - yd) */
emilmont 1:fdd22bb7aa52 775 r1 = r2 - t1;
emilmont 1:fdd22bb7aa52 776 /* (xa - xc) + (yb - yd) */
emilmont 1:fdd22bb7aa52 777 r2 = r2 + t1;
emilmont 1:fdd22bb7aa52 778
emilmont 1:fdd22bb7aa52 779 /* (ya - yc) + (xb - xd) */
emilmont 1:fdd22bb7aa52 780 s1 = s2 + t2;
emilmont 1:fdd22bb7aa52 781 /* (ya - yc) - (xb - xd) */
emilmont 1:fdd22bb7aa52 782 s2 = s2 - t2;
emilmont 1:fdd22bb7aa52 783
emilmont 1:fdd22bb7aa52 784 /* xb' = (xa+yb-xc-yd)co1 - (ya-xb-yc+xd)(si1) */
emilmont 1:fdd22bb7aa52 785 pSrc[2u * i2] = (((int32_t) (((q63_t) r1 * co1) >> 32)) -
emilmont 1:fdd22bb7aa52 786 ((int32_t) (((q63_t) s1 * si1) >> 32))) >> 1u;
emilmont 1:fdd22bb7aa52 787
emilmont 1:fdd22bb7aa52 788 /* yb' = (ya-xb-yc+xd)co1 + (xa+yb-xc-yd)(si1) */
emilmont 1:fdd22bb7aa52 789 pSrc[(2u * i2) + 1u] = (((int32_t) (((q63_t) s1 * co1) >> 32)) +
emilmont 1:fdd22bb7aa52 790 ((int32_t) (((q63_t) r1 * si1) >> 32))) >> 1u;
emilmont 1:fdd22bb7aa52 791
emilmont 1:fdd22bb7aa52 792 /* xd' = (xa-yb-xc+yd)co3 - (ya+xb-yc-xd)(si3) */
emilmont 1:fdd22bb7aa52 793 pSrc[(2u * i3)] = (((int32_t) (((q63_t) r2 * co3) >> 32)) -
emilmont 1:fdd22bb7aa52 794 ((int32_t) (((q63_t) s2 * si3) >> 32))) >> 1u;
emilmont 1:fdd22bb7aa52 795
emilmont 1:fdd22bb7aa52 796 /* yd' = (ya+xb-yc-xd)co3 + (xa-yb-xc+yd)(si3) */
emilmont 1:fdd22bb7aa52 797 pSrc[(2u * i3) + 1u] = (((int32_t) (((q63_t) s2 * co3) >> 32)) +
emilmont 1:fdd22bb7aa52 798 ((int32_t) (((q63_t) r2 * si3) >> 32))) >> 1u;
emilmont 1:fdd22bb7aa52 799 }
emilmont 1:fdd22bb7aa52 800 }
emilmont 1:fdd22bb7aa52 801 twidCoefModifier <<= 2u;
emilmont 1:fdd22bb7aa52 802 }
emilmont 1:fdd22bb7aa52 803
emilmont 1:fdd22bb7aa52 804 /* End of Middle stages process */
emilmont 1:fdd22bb7aa52 805
emilmont 1:fdd22bb7aa52 806 /* data is in 11.21(q21) format for the 1024 point as there are 3 middle stages */
emilmont 1:fdd22bb7aa52 807 /* data is in 9.23(q23) format for the 256 point as there are 2 middle stages */
emilmont 1:fdd22bb7aa52 808 /* data is in 7.25(q25) format for the 64 point as there are 1 middle stage */
emilmont 1:fdd22bb7aa52 809 /* data is in 5.27(q27) format for the 16 point as there are no middle stages */
emilmont 1:fdd22bb7aa52 810
emilmont 1:fdd22bb7aa52 811
emilmont 1:fdd22bb7aa52 812 /* Start of last stage process */
emilmont 1:fdd22bb7aa52 813
emilmont 1:fdd22bb7aa52 814
emilmont 1:fdd22bb7aa52 815 /* Initializations for the last stage */
emilmont 1:fdd22bb7aa52 816 j = fftLen >> 2;
emilmont 1:fdd22bb7aa52 817 ptr1 = &pSrc[0];
emilmont 1:fdd22bb7aa52 818
emilmont 1:fdd22bb7aa52 819 /* Calculations of last stage */
emilmont 1:fdd22bb7aa52 820 do
emilmont 1:fdd22bb7aa52 821 {
emilmont 1:fdd22bb7aa52 822 #ifndef ARM_MATH_BIG_ENDIAN
emilmont 1:fdd22bb7aa52 823 /* Read xa (real), ya(imag) input */
emilmont 1:fdd22bb7aa52 824 xaya = *__SIMD64(ptr1)++;
emilmont 1:fdd22bb7aa52 825 xa = (q31_t) xaya;
emilmont 1:fdd22bb7aa52 826 ya = (q31_t) (xaya >> 32);
emilmont 1:fdd22bb7aa52 827
emilmont 1:fdd22bb7aa52 828 /* Read xb (real), yb(imag) input */
emilmont 1:fdd22bb7aa52 829 xbyb = *__SIMD64(ptr1)++;
emilmont 1:fdd22bb7aa52 830 xb = (q31_t) xbyb;
emilmont 1:fdd22bb7aa52 831 yb = (q31_t) (xbyb >> 32);
emilmont 1:fdd22bb7aa52 832
emilmont 1:fdd22bb7aa52 833 /* Read xc (real), yc(imag) input */
emilmont 1:fdd22bb7aa52 834 xcyc = *__SIMD64(ptr1)++;
emilmont 1:fdd22bb7aa52 835 xc = (q31_t) xcyc;
emilmont 1:fdd22bb7aa52 836 yc = (q31_t) (xcyc >> 32);
emilmont 1:fdd22bb7aa52 837
emilmont 1:fdd22bb7aa52 838 /* Read xc (real), yc(imag) input */
emilmont 1:fdd22bb7aa52 839 xdyd = *__SIMD64(ptr1)++;
emilmont 1:fdd22bb7aa52 840 xd = (q31_t) xdyd;
emilmont 1:fdd22bb7aa52 841 yd = (q31_t) (xdyd >> 32);
emilmont 1:fdd22bb7aa52 842
emilmont 1:fdd22bb7aa52 843 #else
emilmont 1:fdd22bb7aa52 844
emilmont 1:fdd22bb7aa52 845 /* Read xa (real), ya(imag) input */
emilmont 1:fdd22bb7aa52 846 xaya = *__SIMD64(ptr1)++;
emilmont 1:fdd22bb7aa52 847 ya = (q31_t) xaya;
emilmont 1:fdd22bb7aa52 848 xa = (q31_t) (xaya >> 32);
emilmont 1:fdd22bb7aa52 849
emilmont 1:fdd22bb7aa52 850 /* Read xb (real), yb(imag) input */
emilmont 1:fdd22bb7aa52 851 xbyb = *__SIMD64(ptr1)++;
emilmont 1:fdd22bb7aa52 852 yb = (q31_t) xbyb;
emilmont 1:fdd22bb7aa52 853 xb = (q31_t) (xbyb >> 32);
emilmont 1:fdd22bb7aa52 854
emilmont 1:fdd22bb7aa52 855 /* Read xc (real), yc(imag) input */
emilmont 1:fdd22bb7aa52 856 xcyc = *__SIMD64(ptr1)++;
emilmont 1:fdd22bb7aa52 857 yc = (q31_t) xcyc;
emilmont 1:fdd22bb7aa52 858 xc = (q31_t) (xcyc >> 32);
emilmont 1:fdd22bb7aa52 859
emilmont 1:fdd22bb7aa52 860 /* Read xc (real), yc(imag) input */
emilmont 1:fdd22bb7aa52 861 xdyd = *__SIMD64(ptr1)++;
emilmont 1:fdd22bb7aa52 862 yd = (q31_t) xdyd;
emilmont 1:fdd22bb7aa52 863 xd = (q31_t) (xdyd >> 32);
emilmont 1:fdd22bb7aa52 864
emilmont 1:fdd22bb7aa52 865
emilmont 1:fdd22bb7aa52 866 #endif
emilmont 1:fdd22bb7aa52 867
emilmont 1:fdd22bb7aa52 868 /* xa' = xa + xb + xc + xd */
emilmont 1:fdd22bb7aa52 869 xa_out = xa + xb + xc + xd;
emilmont 1:fdd22bb7aa52 870
emilmont 1:fdd22bb7aa52 871 /* ya' = ya + yb + yc + yd */
emilmont 1:fdd22bb7aa52 872 ya_out = ya + yb + yc + yd;
emilmont 1:fdd22bb7aa52 873
emilmont 1:fdd22bb7aa52 874 /* pointer updation for writing */
emilmont 1:fdd22bb7aa52 875 ptr1 = ptr1 - 8u;
emilmont 1:fdd22bb7aa52 876
emilmont 1:fdd22bb7aa52 877 /* writing xa' and ya' */
emilmont 1:fdd22bb7aa52 878 *ptr1++ = xa_out;
emilmont 1:fdd22bb7aa52 879 *ptr1++ = ya_out;
emilmont 1:fdd22bb7aa52 880
emilmont 1:fdd22bb7aa52 881 xc_out = (xa - xb + xc - xd);
emilmont 1:fdd22bb7aa52 882 yc_out = (ya - yb + yc - yd);
emilmont 1:fdd22bb7aa52 883
emilmont 1:fdd22bb7aa52 884 /* writing xc' and yc' */
emilmont 1:fdd22bb7aa52 885 *ptr1++ = xc_out;
emilmont 1:fdd22bb7aa52 886 *ptr1++ = yc_out;
emilmont 1:fdd22bb7aa52 887
emilmont 1:fdd22bb7aa52 888 xb_out = (xa - yb - xc + yd);
emilmont 1:fdd22bb7aa52 889 yb_out = (ya + xb - yc - xd);
emilmont 1:fdd22bb7aa52 890
emilmont 1:fdd22bb7aa52 891 /* writing xb' and yb' */
emilmont 1:fdd22bb7aa52 892 *ptr1++ = xb_out;
emilmont 1:fdd22bb7aa52 893 *ptr1++ = yb_out;
emilmont 1:fdd22bb7aa52 894
emilmont 1:fdd22bb7aa52 895 xd_out = (xa + yb - xc - yd);
emilmont 1:fdd22bb7aa52 896 yd_out = (ya - xb - yc + xd);
emilmont 1:fdd22bb7aa52 897
emilmont 1:fdd22bb7aa52 898 /* writing xd' and yd' */
emilmont 1:fdd22bb7aa52 899 *ptr1++ = xd_out;
emilmont 1:fdd22bb7aa52 900 *ptr1++ = yd_out;
emilmont 1:fdd22bb7aa52 901
emilmont 1:fdd22bb7aa52 902
emilmont 1:fdd22bb7aa52 903 } while(--j);
emilmont 1:fdd22bb7aa52 904
emilmont 1:fdd22bb7aa52 905 /* output is in 11.21(q21) format for the 1024 point */
emilmont 1:fdd22bb7aa52 906 /* output is in 9.23(q23) format for the 256 point */
emilmont 1:fdd22bb7aa52 907 /* output is in 7.25(q25) format for the 64 point */
emilmont 1:fdd22bb7aa52 908 /* output is in 5.27(q27) format for the 16 point */
emilmont 1:fdd22bb7aa52 909
emilmont 1:fdd22bb7aa52 910 /* End of last stage process */
emilmont 1:fdd22bb7aa52 911 }