mbed official / mbed-dsp

CMSIS DSP library

Dependents:   performance_timer Surfboard_ gps2rtty Capstone ... more

Legacy Warning

This is an mbed 2 library. To learn more about mbed OS 5, visit the docs.

cmsis_dsp/TransformFunctions/arm_cfft_radix4_q15.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
Child:
5:3762170b6d4d
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_q15.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 * -------------------------------------------------------------------- */
emilmont 1:fdd22bb7aa52 41
emilmont 1:fdd22bb7aa52 42 #include "arm_math.h"
emilmont 1:fdd22bb7aa52 43
mbed_official 3:7a284390b0ce 44
mbed_official 3:7a284390b0ce 45 void arm_radix4_butterfly_q15(
mbed_official 3:7a284390b0ce 46 q15_t * pSrc16,
mbed_official 3:7a284390b0ce 47 uint32_t fftLen,
mbed_official 3:7a284390b0ce 48 q15_t * pCoef16,
mbed_official 3:7a284390b0ce 49 uint32_t twidCoefModifier);
mbed_official 3:7a284390b0ce 50
mbed_official 3:7a284390b0ce 51 void arm_radix4_butterfly_inverse_q15(
mbed_official 3:7a284390b0ce 52 q15_t * pSrc16,
mbed_official 3:7a284390b0ce 53 uint32_t fftLen,
mbed_official 3:7a284390b0ce 54 q15_t * pCoef16,
mbed_official 3:7a284390b0ce 55 uint32_t twidCoefModifier);
mbed_official 3:7a284390b0ce 56
mbed_official 3:7a284390b0ce 57 void arm_bitreversal_q15(
mbed_official 3:7a284390b0ce 58 q15_t * pSrc,
mbed_official 3:7a284390b0ce 59 uint32_t fftLen,
mbed_official 3:7a284390b0ce 60 uint16_t bitRevFactor,
mbed_official 3:7a284390b0ce 61 uint16_t * pBitRevTab);
mbed_official 3:7a284390b0ce 62
emilmont 1:fdd22bb7aa52 63 /**
emilmont 1:fdd22bb7aa52 64 * @ingroup groupTransforms
emilmont 1:fdd22bb7aa52 65 */
emilmont 1:fdd22bb7aa52 66
emilmont 1:fdd22bb7aa52 67 /**
mbed_official 3:7a284390b0ce 68 * @addtogroup ComplexFFT
emilmont 1:fdd22bb7aa52 69 * @{
emilmont 1:fdd22bb7aa52 70 */
emilmont 1:fdd22bb7aa52 71
emilmont 1:fdd22bb7aa52 72
emilmont 1:fdd22bb7aa52 73 /**
emilmont 1:fdd22bb7aa52 74 * @details
emilmont 1:fdd22bb7aa52 75 * @brief Processing function for the Q15 CFFT/CIFFT.
emilmont 1:fdd22bb7aa52 76 * @param[in] *S points to an instance of the Q15 CFFT/CIFFT structure.
emilmont 1:fdd22bb7aa52 77 * @param[in, out] *pSrc points to the complex data buffer. Processing occurs in-place.
emilmont 1:fdd22bb7aa52 78 * @return none.
emilmont 1:fdd22bb7aa52 79 *
emilmont 1:fdd22bb7aa52 80 * \par Input and output formats:
emilmont 1:fdd22bb7aa52 81 * \par
emilmont 1:fdd22bb7aa52 82 * Internally input is downscaled by 2 for every stage to avoid saturations inside CFFT/CIFFT process.
emilmont 1:fdd22bb7aa52 83 * Hence the output format is different for different FFT sizes.
emilmont 1:fdd22bb7aa52 84 * 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 85 * \par
emilmont 1:fdd22bb7aa52 86 * \image html CFFTQ15.gif "Input and Output Formats for Q15 CFFT"
emilmont 1:fdd22bb7aa52 87 * \image html CIFFTQ15.gif "Input and Output Formats for Q15 CIFFT"
emilmont 1:fdd22bb7aa52 88 */
emilmont 1:fdd22bb7aa52 89
emilmont 1:fdd22bb7aa52 90 void arm_cfft_radix4_q15(
emilmont 1:fdd22bb7aa52 91 const arm_cfft_radix4_instance_q15 * S,
emilmont 1:fdd22bb7aa52 92 q15_t * pSrc)
emilmont 1:fdd22bb7aa52 93 {
emilmont 1:fdd22bb7aa52 94 if(S->ifftFlag == 1u)
emilmont 1:fdd22bb7aa52 95 {
emilmont 1:fdd22bb7aa52 96 /* Complex IFFT radix-4 */
emilmont 1:fdd22bb7aa52 97 arm_radix4_butterfly_inverse_q15(pSrc, S->fftLen, S->pTwiddle,
emilmont 1:fdd22bb7aa52 98 S->twidCoefModifier);
emilmont 1:fdd22bb7aa52 99 }
emilmont 1:fdd22bb7aa52 100 else
emilmont 1:fdd22bb7aa52 101 {
emilmont 1:fdd22bb7aa52 102 /* Complex FFT radix-4 */
emilmont 1:fdd22bb7aa52 103 arm_radix4_butterfly_q15(pSrc, S->fftLen, S->pTwiddle,
emilmont 1:fdd22bb7aa52 104 S->twidCoefModifier);
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_q15(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 * The real and imaginary output values for the radix-4 butterfly are
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 Q15 CFFT butterfly process.
emilmont 1:fdd22bb7aa52 155 * @param[in, out] *pSrc16 points to the in-place buffer of Q15 data type.
emilmont 1:fdd22bb7aa52 156 * @param[in] fftLen length of the FFT.
emilmont 1:fdd22bb7aa52 157 * @param[in] *pCoef16 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_q15(
emilmont 1:fdd22bb7aa52 163 q15_t * pSrc16,
emilmont 1:fdd22bb7aa52 164 uint32_t fftLen,
emilmont 1:fdd22bb7aa52 165 q15_t * pCoef16,
emilmont 1:fdd22bb7aa52 166 uint32_t twidCoefModifier)
emilmont 1:fdd22bb7aa52 167 {
emilmont 1:fdd22bb7aa52 168
mbed_official 3:7a284390b0ce 169 #ifndef ARM_MATH_CM0_FAMILY
emilmont 1:fdd22bb7aa52 170
emilmont 1:fdd22bb7aa52 171 /* Run the below code for Cortex-M4 and Cortex-M3 */
emilmont 1:fdd22bb7aa52 172
emilmont 1:fdd22bb7aa52 173 q31_t R, S, T, U;
emilmont 1:fdd22bb7aa52 174 q31_t C1, C2, C3, out1, out2;
emilmont 1:fdd22bb7aa52 175 uint32_t n1, n2, ic, i0, i1, i2, i3, j, k;
emilmont 1:fdd22bb7aa52 176 q15_t in;
emilmont 1:fdd22bb7aa52 177
emilmont 1:fdd22bb7aa52 178 q15_t *ptr1;
emilmont 1:fdd22bb7aa52 179
emilmont 1:fdd22bb7aa52 180
emilmont 1:fdd22bb7aa52 181
emilmont 1:fdd22bb7aa52 182 q31_t xaya, xbyb, xcyc, xdyd;
emilmont 1:fdd22bb7aa52 183
emilmont 1:fdd22bb7aa52 184 /* Total process is divided into three stages */
emilmont 1:fdd22bb7aa52 185
emilmont 1:fdd22bb7aa52 186 /* process first stage, middle stages, & last stage */
emilmont 1:fdd22bb7aa52 187
emilmont 1:fdd22bb7aa52 188 /* Initializations for the first stage */
emilmont 1:fdd22bb7aa52 189 n2 = fftLen;
emilmont 1:fdd22bb7aa52 190 n1 = n2;
emilmont 1:fdd22bb7aa52 191
emilmont 1:fdd22bb7aa52 192 /* n2 = fftLen/4 */
emilmont 1:fdd22bb7aa52 193 n2 >>= 2u;
emilmont 1:fdd22bb7aa52 194
emilmont 1:fdd22bb7aa52 195 /* Index for twiddle coefficient */
emilmont 1:fdd22bb7aa52 196 ic = 0u;
emilmont 1:fdd22bb7aa52 197
emilmont 1:fdd22bb7aa52 198 /* Index for input read and output write */
emilmont 1:fdd22bb7aa52 199 i0 = 0u;
emilmont 1:fdd22bb7aa52 200 j = n2;
emilmont 1:fdd22bb7aa52 201
emilmont 1:fdd22bb7aa52 202 /* Input is in 1.15(q15) format */
emilmont 1:fdd22bb7aa52 203
emilmont 1:fdd22bb7aa52 204 /* start of first stage process */
emilmont 1:fdd22bb7aa52 205 do
emilmont 1:fdd22bb7aa52 206 {
emilmont 1:fdd22bb7aa52 207 /* Butterfly implementation */
emilmont 1:fdd22bb7aa52 208
emilmont 1:fdd22bb7aa52 209 /* index calculation for the input as, */
emilmont 1:fdd22bb7aa52 210 /* pSrc16[i0 + 0], pSrc16[i0 + fftLen/4], pSrc16[i0 + fftLen/2], pSrc16[i0 + 3fftLen/4] */
emilmont 1:fdd22bb7aa52 211 i1 = i0 + n2;
emilmont 1:fdd22bb7aa52 212 i2 = i1 + n2;
emilmont 1:fdd22bb7aa52 213 i3 = i2 + n2;
emilmont 1:fdd22bb7aa52 214
emilmont 1:fdd22bb7aa52 215 /* Reading i0, i0+fftLen/2 inputs */
emilmont 1:fdd22bb7aa52 216 /* Read ya (real), xa(imag) input */
emilmont 1:fdd22bb7aa52 217 T = _SIMD32_OFFSET(pSrc16 + (2u * i0));
emilmont 1:fdd22bb7aa52 218 in = ((int16_t) (T & 0xFFFF)) >> 2;
emilmont 1:fdd22bb7aa52 219 T = ((T >> 2) & 0xFFFF0000) | (in & 0xFFFF);
emilmont 1:fdd22bb7aa52 220
emilmont 1:fdd22bb7aa52 221 /* Read yc (real), xc(imag) input */
emilmont 1:fdd22bb7aa52 222 S = _SIMD32_OFFSET(pSrc16 + (2u * i2));
emilmont 1:fdd22bb7aa52 223 in = ((int16_t) (S & 0xFFFF)) >> 2;
emilmont 1:fdd22bb7aa52 224 S = ((S >> 2) & 0xFFFF0000) | (in & 0xFFFF);
emilmont 1:fdd22bb7aa52 225
emilmont 1:fdd22bb7aa52 226 /* R = packed((ya + yc), (xa + xc) ) */
emilmont 1:fdd22bb7aa52 227 R = __QADD16(T, S);
emilmont 1:fdd22bb7aa52 228
emilmont 1:fdd22bb7aa52 229 /* S = packed((ya - yc), (xa - xc) ) */
emilmont 1:fdd22bb7aa52 230 S = __QSUB16(T, S);
emilmont 1:fdd22bb7aa52 231
emilmont 1:fdd22bb7aa52 232 /* Reading i0+fftLen/4 , i0+3fftLen/4 inputs */
emilmont 1:fdd22bb7aa52 233 /* Read yb (real), xb(imag) input */
emilmont 1:fdd22bb7aa52 234 T = _SIMD32_OFFSET(pSrc16 + (2u * i1));
emilmont 1:fdd22bb7aa52 235 in = ((int16_t) (T & 0xFFFF)) >> 2;
emilmont 1:fdd22bb7aa52 236 T = ((T >> 2) & 0xFFFF0000) | (in & 0xFFFF);
emilmont 1:fdd22bb7aa52 237
emilmont 1:fdd22bb7aa52 238 /* Read yd (real), xd(imag) input */
emilmont 1:fdd22bb7aa52 239 U = _SIMD32_OFFSET(pSrc16 + (2u * i3));
emilmont 1:fdd22bb7aa52 240 in = ((int16_t) (U & 0xFFFF)) >> 2;
emilmont 1:fdd22bb7aa52 241 U = ((U >> 2) & 0xFFFF0000) | (in & 0xFFFF);
emilmont 1:fdd22bb7aa52 242
emilmont 1:fdd22bb7aa52 243 /* T = packed((yb + yd), (xb + xd) ) */
emilmont 1:fdd22bb7aa52 244 T = __QADD16(T, U);
emilmont 1:fdd22bb7aa52 245
emilmont 1:fdd22bb7aa52 246 /* writing the butterfly processed i0 sample */
emilmont 1:fdd22bb7aa52 247 /* xa' = xa + xb + xc + xd */
emilmont 1:fdd22bb7aa52 248 /* ya' = ya + yb + yc + yd */
emilmont 1:fdd22bb7aa52 249 _SIMD32_OFFSET(pSrc16 + (2u * i0)) = __SHADD16(R, T);
emilmont 1:fdd22bb7aa52 250
emilmont 1:fdd22bb7aa52 251 /* R = packed((ya + yc) - (yb + yd), (xa + xc)- (xb + xd)) */
emilmont 1:fdd22bb7aa52 252 R = __QSUB16(R, T);
emilmont 1:fdd22bb7aa52 253
emilmont 1:fdd22bb7aa52 254 /* co2 & si2 are read from SIMD Coefficient pointer */
emilmont 1:fdd22bb7aa52 255 C2 = _SIMD32_OFFSET(pCoef16 + (4u * ic));
emilmont 1:fdd22bb7aa52 256
emilmont 1:fdd22bb7aa52 257 #ifndef ARM_MATH_BIG_ENDIAN
emilmont 1:fdd22bb7aa52 258
emilmont 1:fdd22bb7aa52 259 /* xc' = (xa-xb+xc-xd)* co2 + (ya-yb+yc-yd)* (si2) */
emilmont 1:fdd22bb7aa52 260 out1 = __SMUAD(C2, R) >> 16u;
emilmont 1:fdd22bb7aa52 261 /* yc' = (ya-yb+yc-yd)* co2 - (xa-xb+xc-xd)* (si2) */
emilmont 1:fdd22bb7aa52 262 out2 = __SMUSDX(C2, R);
emilmont 1:fdd22bb7aa52 263
emilmont 1:fdd22bb7aa52 264 #else
emilmont 1:fdd22bb7aa52 265
emilmont 1:fdd22bb7aa52 266 /* xc' = (ya-yb+yc-yd)* co2 - (xa-xb+xc-xd)* (si2) */
emilmont 1:fdd22bb7aa52 267 out1 = __SMUSDX(R, C2) >> 16u;
emilmont 1:fdd22bb7aa52 268 /* yc' = (xa-xb+xc-xd)* co2 + (ya-yb+yc-yd)* (si2) */
emilmont 1:fdd22bb7aa52 269 out2 = __SMUAD(C2, R);
emilmont 1:fdd22bb7aa52 270
emilmont 1:fdd22bb7aa52 271 #endif /* #ifndef ARM_MATH_BIG_ENDIAN */
emilmont 1:fdd22bb7aa52 272
emilmont 1:fdd22bb7aa52 273 /* Reading i0+fftLen/4 */
emilmont 1:fdd22bb7aa52 274 /* T = packed(yb, xb) */
emilmont 1:fdd22bb7aa52 275 T = _SIMD32_OFFSET(pSrc16 + (2u * i1));
emilmont 1:fdd22bb7aa52 276 in = ((int16_t) (T & 0xFFFF)) >> 2;
emilmont 1:fdd22bb7aa52 277 T = ((T >> 2) & 0xFFFF0000) | (in & 0xFFFF);
emilmont 1:fdd22bb7aa52 278
emilmont 1:fdd22bb7aa52 279 /* writing the butterfly processed i0 + fftLen/4 sample */
emilmont 1:fdd22bb7aa52 280 /* writing output(xc', yc') in little endian format */
emilmont 1:fdd22bb7aa52 281 _SIMD32_OFFSET(pSrc16 + (2u * i1)) =
emilmont 1:fdd22bb7aa52 282 (q31_t) ((out2) & 0xFFFF0000) | (out1 & 0x0000FFFF);
emilmont 1:fdd22bb7aa52 283
emilmont 1:fdd22bb7aa52 284 /* Butterfly calculations */
emilmont 1:fdd22bb7aa52 285 /* U = packed(yd, xd) */
emilmont 1:fdd22bb7aa52 286 U = _SIMD32_OFFSET(pSrc16 + (2u * i3));
emilmont 1:fdd22bb7aa52 287 in = ((int16_t) (U & 0xFFFF)) >> 2;
emilmont 1:fdd22bb7aa52 288 U = ((U >> 2) & 0xFFFF0000) | (in & 0xFFFF);
emilmont 1:fdd22bb7aa52 289
emilmont 1:fdd22bb7aa52 290 /* T = packed(yb-yd, xb-xd) */
emilmont 1:fdd22bb7aa52 291 T = __QSUB16(T, U);
emilmont 1:fdd22bb7aa52 292
emilmont 1:fdd22bb7aa52 293 #ifndef ARM_MATH_BIG_ENDIAN
emilmont 1:fdd22bb7aa52 294
emilmont 1:fdd22bb7aa52 295 /* R = packed((ya-yc) + (xb- xd) , (xa-xc) - (yb-yd)) */
emilmont 1:fdd22bb7aa52 296 R = __QASX(S, T);
emilmont 1:fdd22bb7aa52 297 /* S = packed((ya-yc) - (xb- xd), (xa-xc) + (yb-yd)) */
emilmont 1:fdd22bb7aa52 298 S = __QSAX(S, T);
emilmont 1:fdd22bb7aa52 299
emilmont 1:fdd22bb7aa52 300 #else
emilmont 1:fdd22bb7aa52 301
emilmont 1:fdd22bb7aa52 302 /* R = packed((ya-yc) + (xb- xd) , (xa-xc) - (yb-yd)) */
emilmont 1:fdd22bb7aa52 303 R = __QSAX(S, T);
emilmont 1:fdd22bb7aa52 304 /* S = packed((ya-yc) - (xb- xd), (xa-xc) + (yb-yd)) */
emilmont 1:fdd22bb7aa52 305 S = __QASX(S, T);
emilmont 1:fdd22bb7aa52 306
emilmont 1:fdd22bb7aa52 307 #endif /* #ifndef ARM_MATH_BIG_ENDIAN */
emilmont 1:fdd22bb7aa52 308
emilmont 1:fdd22bb7aa52 309 /* co1 & si1 are read from SIMD Coefficient pointer */
emilmont 1:fdd22bb7aa52 310 C1 = _SIMD32_OFFSET(pCoef16 + (2u * ic));
emilmont 1:fdd22bb7aa52 311 /* Butterfly process for the i0+fftLen/2 sample */
emilmont 1:fdd22bb7aa52 312
emilmont 1:fdd22bb7aa52 313 #ifndef ARM_MATH_BIG_ENDIAN
emilmont 1:fdd22bb7aa52 314
emilmont 1:fdd22bb7aa52 315 /* xb' = (xa+yb-xc-yd)* co1 + (ya-xb-yc+xd)* (si1) */
emilmont 1:fdd22bb7aa52 316 out1 = __SMUAD(C1, S) >> 16u;
emilmont 1:fdd22bb7aa52 317 /* yb' = (ya-xb-yc+xd)* co1 - (xa+yb-xc-yd)* (si1) */
emilmont 1:fdd22bb7aa52 318 out2 = __SMUSDX(C1, S);
emilmont 1:fdd22bb7aa52 319
emilmont 1:fdd22bb7aa52 320 #else
emilmont 1:fdd22bb7aa52 321
emilmont 1:fdd22bb7aa52 322 /* xb' = (ya-xb-yc+xd)* co1 - (xa+yb-xc-yd)* (si1) */
emilmont 1:fdd22bb7aa52 323 out1 = __SMUSDX(S, C1) >> 16u;
emilmont 1:fdd22bb7aa52 324 /* yb' = (xa+yb-xc-yd)* co1 + (ya-xb-yc+xd)* (si1) */
emilmont 1:fdd22bb7aa52 325 out2 = __SMUAD(C1, S);
emilmont 1:fdd22bb7aa52 326
emilmont 1:fdd22bb7aa52 327 #endif /* #ifndef ARM_MATH_BIG_ENDIAN */
emilmont 1:fdd22bb7aa52 328
emilmont 1:fdd22bb7aa52 329 /* writing output(xb', yb') in little endian format */
emilmont 1:fdd22bb7aa52 330 _SIMD32_OFFSET(pSrc16 + (2u * i2)) =
emilmont 1:fdd22bb7aa52 331 ((out2) & 0xFFFF0000) | ((out1) & 0x0000FFFF);
emilmont 1:fdd22bb7aa52 332
emilmont 1:fdd22bb7aa52 333
emilmont 1:fdd22bb7aa52 334 /* co3 & si3 are read from SIMD Coefficient pointer */
emilmont 1:fdd22bb7aa52 335 C3 = _SIMD32_OFFSET(pCoef16 + (6u * ic));
emilmont 1:fdd22bb7aa52 336 /* Butterfly process for the i0+3fftLen/4 sample */
emilmont 1:fdd22bb7aa52 337
emilmont 1:fdd22bb7aa52 338 #ifndef ARM_MATH_BIG_ENDIAN
emilmont 1:fdd22bb7aa52 339
emilmont 1:fdd22bb7aa52 340 /* xd' = (xa-yb-xc+yd)* co3 + (ya+xb-yc-xd)* (si3) */
emilmont 1:fdd22bb7aa52 341 out1 = __SMUAD(C3, R) >> 16u;
emilmont 1:fdd22bb7aa52 342 /* yd' = (ya+xb-yc-xd)* co3 - (xa-yb-xc+yd)* (si3) */
emilmont 1:fdd22bb7aa52 343 out2 = __SMUSDX(C3, R);
emilmont 1:fdd22bb7aa52 344
emilmont 1:fdd22bb7aa52 345 #else
emilmont 1:fdd22bb7aa52 346
emilmont 1:fdd22bb7aa52 347 /* xd' = (ya+xb-yc-xd)* co3 - (xa-yb-xc+yd)* (si3) */
emilmont 1:fdd22bb7aa52 348 out1 = __SMUSDX(R, C3) >> 16u;
emilmont 1:fdd22bb7aa52 349 /* yd' = (xa-yb-xc+yd)* co3 + (ya+xb-yc-xd)* (si3) */
emilmont 1:fdd22bb7aa52 350 out2 = __SMUAD(C3, R);
emilmont 1:fdd22bb7aa52 351
emilmont 1:fdd22bb7aa52 352 #endif /* #ifndef ARM_MATH_BIG_ENDIAN */
emilmont 1:fdd22bb7aa52 353
emilmont 1:fdd22bb7aa52 354 /* writing output(xd', yd') in little endian format */
emilmont 1:fdd22bb7aa52 355 _SIMD32_OFFSET(pSrc16 + (2u * i3)) =
emilmont 1:fdd22bb7aa52 356 ((out2) & 0xFFFF0000) | (out1 & 0x0000FFFF);
emilmont 1:fdd22bb7aa52 357
emilmont 1:fdd22bb7aa52 358 /* Twiddle coefficients index modifier */
emilmont 1:fdd22bb7aa52 359 ic = ic + twidCoefModifier;
emilmont 1:fdd22bb7aa52 360
emilmont 1:fdd22bb7aa52 361 /* Updating input index */
emilmont 1:fdd22bb7aa52 362 i0 = i0 + 1u;
emilmont 1:fdd22bb7aa52 363
emilmont 1:fdd22bb7aa52 364 } while(--j);
emilmont 1:fdd22bb7aa52 365 /* data is in 4.11(q11) format */
emilmont 1:fdd22bb7aa52 366
emilmont 1:fdd22bb7aa52 367 /* end of first stage process */
emilmont 1:fdd22bb7aa52 368
emilmont 1:fdd22bb7aa52 369
emilmont 1:fdd22bb7aa52 370 /* start of middle stage process */
emilmont 1:fdd22bb7aa52 371
emilmont 1:fdd22bb7aa52 372 /* Twiddle coefficients index modifier */
emilmont 1:fdd22bb7aa52 373 twidCoefModifier <<= 2u;
emilmont 1:fdd22bb7aa52 374
emilmont 1:fdd22bb7aa52 375 /* Calculation of Middle stage */
emilmont 1:fdd22bb7aa52 376 for (k = fftLen / 4u; k > 4u; k >>= 2u)
emilmont 1:fdd22bb7aa52 377 {
emilmont 1:fdd22bb7aa52 378 /* Initializations for the middle stage */
emilmont 1:fdd22bb7aa52 379 n1 = n2;
emilmont 1:fdd22bb7aa52 380 n2 >>= 2u;
emilmont 1:fdd22bb7aa52 381 ic = 0u;
emilmont 1:fdd22bb7aa52 382
emilmont 1:fdd22bb7aa52 383 for (j = 0u; j <= (n2 - 1u); j++)
emilmont 1:fdd22bb7aa52 384 {
emilmont 1:fdd22bb7aa52 385 /* index calculation for the coefficients */
emilmont 1:fdd22bb7aa52 386 C1 = _SIMD32_OFFSET(pCoef16 + (2u * ic));
emilmont 1:fdd22bb7aa52 387 C2 = _SIMD32_OFFSET(pCoef16 + (4u * ic));
emilmont 1:fdd22bb7aa52 388 C3 = _SIMD32_OFFSET(pCoef16 + (6u * ic));
emilmont 1:fdd22bb7aa52 389
emilmont 1:fdd22bb7aa52 390 /* Twiddle coefficients index modifier */
emilmont 1:fdd22bb7aa52 391 ic = ic + twidCoefModifier;
emilmont 1:fdd22bb7aa52 392
emilmont 1:fdd22bb7aa52 393 /* Butterfly implementation */
emilmont 1:fdd22bb7aa52 394 for (i0 = j; i0 < fftLen; i0 += n1)
emilmont 1:fdd22bb7aa52 395 {
emilmont 1:fdd22bb7aa52 396 /* index calculation for the input as, */
emilmont 1:fdd22bb7aa52 397 /* pSrc16[i0 + 0], pSrc16[i0 + fftLen/4], pSrc16[i0 + fftLen/2], pSrc16[i0 + 3fftLen/4] */
emilmont 1:fdd22bb7aa52 398 i1 = i0 + n2;
emilmont 1:fdd22bb7aa52 399 i2 = i1 + n2;
emilmont 1:fdd22bb7aa52 400 i3 = i2 + n2;
emilmont 1:fdd22bb7aa52 401
emilmont 1:fdd22bb7aa52 402 /* Reading i0, i0+fftLen/2 inputs */
emilmont 1:fdd22bb7aa52 403 /* Read ya (real), xa(imag) input */
emilmont 1:fdd22bb7aa52 404 T = _SIMD32_OFFSET(pSrc16 + (2u * i0));
emilmont 1:fdd22bb7aa52 405
emilmont 1:fdd22bb7aa52 406 /* Read yc (real), xc(imag) input */
emilmont 1:fdd22bb7aa52 407 S = _SIMD32_OFFSET(pSrc16 + (2u * i2));
emilmont 1:fdd22bb7aa52 408
emilmont 1:fdd22bb7aa52 409 /* R = packed( (ya + yc), (xa + xc)) */
emilmont 1:fdd22bb7aa52 410 R = __QADD16(T, S);
emilmont 1:fdd22bb7aa52 411
emilmont 1:fdd22bb7aa52 412 /* S = packed((ya - yc), (xa - xc)) */
emilmont 1:fdd22bb7aa52 413 S = __QSUB16(T, S);
emilmont 1:fdd22bb7aa52 414
emilmont 1:fdd22bb7aa52 415 /* Reading i0+fftLen/4 , i0+3fftLen/4 inputs */
emilmont 1:fdd22bb7aa52 416 /* Read yb (real), xb(imag) input */
emilmont 1:fdd22bb7aa52 417 T = _SIMD32_OFFSET(pSrc16 + (2u * i1));
emilmont 1:fdd22bb7aa52 418
emilmont 1:fdd22bb7aa52 419 /* Read yd (real), xd(imag) input */
emilmont 1:fdd22bb7aa52 420 U = _SIMD32_OFFSET(pSrc16 + (2u * i3));
emilmont 1:fdd22bb7aa52 421
emilmont 1:fdd22bb7aa52 422 /* T = packed( (yb + yd), (xb + xd)) */
emilmont 1:fdd22bb7aa52 423 T = __QADD16(T, U);
emilmont 1:fdd22bb7aa52 424
emilmont 1:fdd22bb7aa52 425 /* writing the butterfly processed i0 sample */
emilmont 1:fdd22bb7aa52 426
emilmont 1:fdd22bb7aa52 427 /* xa' = xa + xb + xc + xd */
emilmont 1:fdd22bb7aa52 428 /* ya' = ya + yb + yc + yd */
emilmont 1:fdd22bb7aa52 429 out1 = __SHADD16(R, T);
emilmont 1:fdd22bb7aa52 430 in = ((int16_t) (out1 & 0xFFFF)) >> 1;
emilmont 1:fdd22bb7aa52 431 out1 = ((out1 >> 1) & 0xFFFF0000) | (in & 0xFFFF);
emilmont 1:fdd22bb7aa52 432 _SIMD32_OFFSET(pSrc16 + (2u * i0)) = out1;
emilmont 1:fdd22bb7aa52 433
emilmont 1:fdd22bb7aa52 434 /* R = packed( (ya + yc) - (yb + yd), (xa + xc) - (xb + xd)) */
emilmont 1:fdd22bb7aa52 435 R = __SHSUB16(R, T);
emilmont 1:fdd22bb7aa52 436
emilmont 1:fdd22bb7aa52 437 #ifndef ARM_MATH_BIG_ENDIAN
emilmont 1:fdd22bb7aa52 438
emilmont 1:fdd22bb7aa52 439 /* (ya-yb+yc-yd)* (si2) + (xa-xb+xc-xd)* co2 */
emilmont 1:fdd22bb7aa52 440 out1 = __SMUAD(C2, R) >> 16u;
emilmont 1:fdd22bb7aa52 441
emilmont 1:fdd22bb7aa52 442 /* (ya-yb+yc-yd)* co2 - (xa-xb+xc-xd)* (si2) */
emilmont 1:fdd22bb7aa52 443 out2 = __SMUSDX(C2, R);
emilmont 1:fdd22bb7aa52 444
emilmont 1:fdd22bb7aa52 445 #else
emilmont 1:fdd22bb7aa52 446
emilmont 1:fdd22bb7aa52 447 /* (ya-yb+yc-yd)* co2 - (xa-xb+xc-xd)* (si2) */
emilmont 1:fdd22bb7aa52 448 out1 = __SMUSDX(R, C2) >> 16u;
emilmont 1:fdd22bb7aa52 449
emilmont 1:fdd22bb7aa52 450 /* (ya-yb+yc-yd)* (si2) + (xa-xb+xc-xd)* co2 */
emilmont 1:fdd22bb7aa52 451 out2 = __SMUAD(C2, R);
emilmont 1:fdd22bb7aa52 452
emilmont 1:fdd22bb7aa52 453 #endif /* #ifndef ARM_MATH_BIG_ENDIAN */
emilmont 1:fdd22bb7aa52 454
emilmont 1:fdd22bb7aa52 455 /* Reading i0+3fftLen/4 */
emilmont 1:fdd22bb7aa52 456 /* Read yb (real), xb(imag) input */
emilmont 1:fdd22bb7aa52 457 T = _SIMD32_OFFSET(pSrc16 + (2u * i1));
emilmont 1:fdd22bb7aa52 458
emilmont 1:fdd22bb7aa52 459 /* writing the butterfly processed i0 + fftLen/4 sample */
emilmont 1:fdd22bb7aa52 460 /* xc' = (xa-xb+xc-xd)* co2 + (ya-yb+yc-yd)* (si2) */
emilmont 1:fdd22bb7aa52 461 /* yc' = (ya-yb+yc-yd)* co2 - (xa-xb+xc-xd)* (si2) */
emilmont 1:fdd22bb7aa52 462 _SIMD32_OFFSET(pSrc16 + (2u * i1)) =
emilmont 1:fdd22bb7aa52 463 ((out2) & 0xFFFF0000) | (out1 & 0x0000FFFF);
emilmont 1:fdd22bb7aa52 464
emilmont 1:fdd22bb7aa52 465 /* Butterfly calculations */
emilmont 1:fdd22bb7aa52 466
emilmont 1:fdd22bb7aa52 467 /* Read yd (real), xd(imag) input */
emilmont 1:fdd22bb7aa52 468 U = _SIMD32_OFFSET(pSrc16 + (2u * i3));
emilmont 1:fdd22bb7aa52 469
emilmont 1:fdd22bb7aa52 470 /* T = packed(yb-yd, xb-xd) */
emilmont 1:fdd22bb7aa52 471 T = __QSUB16(T, U);
emilmont 1:fdd22bb7aa52 472
emilmont 1:fdd22bb7aa52 473 #ifndef ARM_MATH_BIG_ENDIAN
emilmont 1:fdd22bb7aa52 474
emilmont 1:fdd22bb7aa52 475 /* R = packed((ya-yc) + (xb- xd) , (xa-xc) - (yb-yd)) */
emilmont 1:fdd22bb7aa52 476 R = __SHASX(S, T);
emilmont 1:fdd22bb7aa52 477
emilmont 1:fdd22bb7aa52 478 /* S = packed((ya-yc) - (xb- xd), (xa-xc) + (yb-yd)) */
emilmont 1:fdd22bb7aa52 479 S = __SHSAX(S, T);
emilmont 1:fdd22bb7aa52 480
emilmont 1:fdd22bb7aa52 481
emilmont 1:fdd22bb7aa52 482 /* Butterfly process for the i0+fftLen/2 sample */
emilmont 1:fdd22bb7aa52 483 out1 = __SMUAD(C1, S) >> 16u;
emilmont 1:fdd22bb7aa52 484 out2 = __SMUSDX(C1, S);
emilmont 1:fdd22bb7aa52 485
emilmont 1:fdd22bb7aa52 486 #else
emilmont 1:fdd22bb7aa52 487
emilmont 1:fdd22bb7aa52 488 /* R = packed((ya-yc) + (xb- xd) , (xa-xc) - (yb-yd)) */
emilmont 1:fdd22bb7aa52 489 R = __SHSAX(S, T);
emilmont 1:fdd22bb7aa52 490
emilmont 1:fdd22bb7aa52 491 /* S = packed((ya-yc) - (xb- xd), (xa-xc) + (yb-yd)) */
emilmont 1:fdd22bb7aa52 492 S = __SHASX(S, T);
emilmont 1:fdd22bb7aa52 493
emilmont 1:fdd22bb7aa52 494
emilmont 1:fdd22bb7aa52 495 /* Butterfly process for the i0+fftLen/2 sample */
emilmont 1:fdd22bb7aa52 496 out1 = __SMUSDX(S, C1) >> 16u;
emilmont 1:fdd22bb7aa52 497 out2 = __SMUAD(C1, S);
emilmont 1:fdd22bb7aa52 498
emilmont 1:fdd22bb7aa52 499 #endif /* #ifndef ARM_MATH_BIG_ENDIAN */
emilmont 1:fdd22bb7aa52 500
emilmont 1:fdd22bb7aa52 501 /* xb' = (xa+yb-xc-yd)* co1 + (ya-xb-yc+xd)* (si1) */
emilmont 1:fdd22bb7aa52 502 /* yb' = (ya-xb-yc+xd)* co1 - (xa+yb-xc-yd)* (si1) */
emilmont 1:fdd22bb7aa52 503 _SIMD32_OFFSET(pSrc16 + (2u * i2)) =
emilmont 1:fdd22bb7aa52 504 ((out2) & 0xFFFF0000) | (out1 & 0x0000FFFF);
emilmont 1:fdd22bb7aa52 505
emilmont 1:fdd22bb7aa52 506 /* Butterfly process for the i0+3fftLen/4 sample */
emilmont 1:fdd22bb7aa52 507
emilmont 1:fdd22bb7aa52 508 #ifndef ARM_MATH_BIG_ENDIAN
emilmont 1:fdd22bb7aa52 509
emilmont 1:fdd22bb7aa52 510 out1 = __SMUAD(C3, R) >> 16u;
emilmont 1:fdd22bb7aa52 511 out2 = __SMUSDX(C3, R);
emilmont 1:fdd22bb7aa52 512
emilmont 1:fdd22bb7aa52 513 #else
emilmont 1:fdd22bb7aa52 514
emilmont 1:fdd22bb7aa52 515 out1 = __SMUSDX(R, C3) >> 16u;
emilmont 1:fdd22bb7aa52 516 out2 = __SMUAD(C3, R);
emilmont 1:fdd22bb7aa52 517
emilmont 1:fdd22bb7aa52 518 #endif /* #ifndef ARM_MATH_BIG_ENDIAN */
emilmont 1:fdd22bb7aa52 519
emilmont 1:fdd22bb7aa52 520 /* xd' = (xa-yb-xc+yd)* co3 + (ya+xb-yc-xd)* (si3) */
emilmont 1:fdd22bb7aa52 521 /* yd' = (ya+xb-yc-xd)* co3 - (xa-yb-xc+yd)* (si3) */
emilmont 1:fdd22bb7aa52 522 _SIMD32_OFFSET(pSrc16 + (2u * i3)) =
emilmont 1:fdd22bb7aa52 523 ((out2) & 0xFFFF0000) | (out1 & 0x0000FFFF);
emilmont 1:fdd22bb7aa52 524 }
emilmont 1:fdd22bb7aa52 525 }
emilmont 1:fdd22bb7aa52 526 /* Twiddle coefficients index modifier */
emilmont 1:fdd22bb7aa52 527 twidCoefModifier <<= 2u;
emilmont 1:fdd22bb7aa52 528 }
emilmont 1:fdd22bb7aa52 529 /* end of middle stage process */
emilmont 1:fdd22bb7aa52 530
emilmont 1:fdd22bb7aa52 531
emilmont 1:fdd22bb7aa52 532 /* data is in 10.6(q6) format for the 1024 point */
emilmont 1:fdd22bb7aa52 533 /* data is in 8.8(q8) format for the 256 point */
emilmont 1:fdd22bb7aa52 534 /* data is in 6.10(q10) format for the 64 point */
emilmont 1:fdd22bb7aa52 535 /* data is in 4.12(q12) format for the 16 point */
emilmont 1:fdd22bb7aa52 536
emilmont 1:fdd22bb7aa52 537 /* Initializations for the last stage */
emilmont 1:fdd22bb7aa52 538 j = fftLen >> 2;
emilmont 1:fdd22bb7aa52 539
emilmont 1:fdd22bb7aa52 540 ptr1 = &pSrc16[0];
emilmont 1:fdd22bb7aa52 541
emilmont 1:fdd22bb7aa52 542 /* start of last stage process */
emilmont 1:fdd22bb7aa52 543
emilmont 1:fdd22bb7aa52 544 /* Butterfly implementation */
emilmont 1:fdd22bb7aa52 545 do
emilmont 1:fdd22bb7aa52 546 {
emilmont 1:fdd22bb7aa52 547 /* Read xa (real), ya(imag) input */
emilmont 1:fdd22bb7aa52 548 xaya = *__SIMD32(ptr1)++;
emilmont 1:fdd22bb7aa52 549
emilmont 1:fdd22bb7aa52 550 /* Read xb (real), yb(imag) input */
emilmont 1:fdd22bb7aa52 551 xbyb = *__SIMD32(ptr1)++;
emilmont 1:fdd22bb7aa52 552
emilmont 1:fdd22bb7aa52 553 /* Read xc (real), yc(imag) input */
emilmont 1:fdd22bb7aa52 554 xcyc = *__SIMD32(ptr1)++;
emilmont 1:fdd22bb7aa52 555
emilmont 1:fdd22bb7aa52 556 /* Read xd (real), yd(imag) input */
emilmont 1:fdd22bb7aa52 557 xdyd = *__SIMD32(ptr1)++;
emilmont 1:fdd22bb7aa52 558
emilmont 1:fdd22bb7aa52 559 /* R = packed((ya + yc), (xa + xc)) */
emilmont 1:fdd22bb7aa52 560 R = __QADD16(xaya, xcyc);
emilmont 1:fdd22bb7aa52 561
emilmont 1:fdd22bb7aa52 562 /* T = packed((yb + yd), (xb + xd)) */
emilmont 1:fdd22bb7aa52 563 T = __QADD16(xbyb, xdyd);
emilmont 1:fdd22bb7aa52 564
emilmont 1:fdd22bb7aa52 565 /* pointer updation for writing */
emilmont 1:fdd22bb7aa52 566 ptr1 = ptr1 - 8u;
emilmont 1:fdd22bb7aa52 567
emilmont 1:fdd22bb7aa52 568
emilmont 1:fdd22bb7aa52 569 /* xa' = xa + xb + xc + xd */
emilmont 1:fdd22bb7aa52 570 /* ya' = ya + yb + yc + yd */
emilmont 1:fdd22bb7aa52 571 *__SIMD32(ptr1)++ = __SHADD16(R, T);
emilmont 1:fdd22bb7aa52 572
emilmont 1:fdd22bb7aa52 573 /* T = packed((yb + yd), (xb + xd)) */
emilmont 1:fdd22bb7aa52 574 T = __QADD16(xbyb, xdyd);
emilmont 1:fdd22bb7aa52 575
emilmont 1:fdd22bb7aa52 576 /* xc' = (xa-xb+xc-xd) */
emilmont 1:fdd22bb7aa52 577 /* yc' = (ya-yb+yc-yd) */
emilmont 1:fdd22bb7aa52 578 *__SIMD32(ptr1)++ = __SHSUB16(R, T);
emilmont 1:fdd22bb7aa52 579
emilmont 1:fdd22bb7aa52 580 /* S = packed((ya - yc), (xa - xc)) */
emilmont 1:fdd22bb7aa52 581 S = __QSUB16(xaya, xcyc);
emilmont 1:fdd22bb7aa52 582
emilmont 1:fdd22bb7aa52 583 /* Read yd (real), xd(imag) input */
emilmont 1:fdd22bb7aa52 584 /* T = packed( (yb - yd), (xb - xd)) */
emilmont 1:fdd22bb7aa52 585 U = __QSUB16(xbyb, xdyd);
emilmont 1:fdd22bb7aa52 586
emilmont 1:fdd22bb7aa52 587 #ifndef ARM_MATH_BIG_ENDIAN
emilmont 1:fdd22bb7aa52 588
emilmont 1:fdd22bb7aa52 589 /* xb' = (xa+yb-xc-yd) */
emilmont 1:fdd22bb7aa52 590 /* yb' = (ya-xb-yc+xd) */
emilmont 1:fdd22bb7aa52 591 *__SIMD32(ptr1)++ = __SHSAX(S, U);
emilmont 1:fdd22bb7aa52 592
emilmont 1:fdd22bb7aa52 593
emilmont 1:fdd22bb7aa52 594 /* xd' = (xa-yb-xc+yd) */
emilmont 1:fdd22bb7aa52 595 /* yd' = (ya+xb-yc-xd) */
emilmont 1:fdd22bb7aa52 596 *__SIMD32(ptr1)++ = __SHASX(S, U);
emilmont 1:fdd22bb7aa52 597
emilmont 1:fdd22bb7aa52 598 #else
emilmont 1:fdd22bb7aa52 599
emilmont 1:fdd22bb7aa52 600 /* xb' = (xa+yb-xc-yd) */
emilmont 1:fdd22bb7aa52 601 /* yb' = (ya-xb-yc+xd) */
emilmont 1:fdd22bb7aa52 602 *__SIMD32(ptr1)++ = __SHASX(S, U);
emilmont 1:fdd22bb7aa52 603
emilmont 1:fdd22bb7aa52 604
emilmont 1:fdd22bb7aa52 605 /* xd' = (xa-yb-xc+yd) */
emilmont 1:fdd22bb7aa52 606 /* yd' = (ya+xb-yc-xd) */
emilmont 1:fdd22bb7aa52 607 *__SIMD32(ptr1)++ = __SHSAX(S, U);
emilmont 1:fdd22bb7aa52 608
emilmont 1:fdd22bb7aa52 609 #endif /* #ifndef ARM_MATH_BIG_ENDIAN */
emilmont 1:fdd22bb7aa52 610
emilmont 1:fdd22bb7aa52 611 } while(--j);
emilmont 1:fdd22bb7aa52 612
emilmont 1:fdd22bb7aa52 613 /* end of last stage process */
emilmont 1:fdd22bb7aa52 614
emilmont 1:fdd22bb7aa52 615 /* output is in 11.5(q5) format for the 1024 point */
emilmont 1:fdd22bb7aa52 616 /* output is in 9.7(q7) format for the 256 point */
emilmont 1:fdd22bb7aa52 617 /* output is in 7.9(q9) format for the 64 point */
emilmont 1:fdd22bb7aa52 618 /* output is in 5.11(q11) format for the 16 point */
emilmont 1:fdd22bb7aa52 619
emilmont 1:fdd22bb7aa52 620
emilmont 1:fdd22bb7aa52 621 #else
emilmont 1:fdd22bb7aa52 622
emilmont 1:fdd22bb7aa52 623 /* Run the below code for Cortex-M0 */
emilmont 1:fdd22bb7aa52 624
emilmont 1:fdd22bb7aa52 625 q15_t R0, R1, S0, S1, T0, T1, U0, U1;
emilmont 1:fdd22bb7aa52 626 q15_t Co1, Si1, Co2, Si2, Co3, Si3, out1, out2;
emilmont 1:fdd22bb7aa52 627 uint32_t n1, n2, ic, i0, i1, i2, i3, j, k;
emilmont 1:fdd22bb7aa52 628
emilmont 1:fdd22bb7aa52 629 /* Total process is divided into three stages */
emilmont 1:fdd22bb7aa52 630
emilmont 1:fdd22bb7aa52 631 /* process first stage, middle stages, & last stage */
emilmont 1:fdd22bb7aa52 632
emilmont 1:fdd22bb7aa52 633 /* Initializations for the first stage */
emilmont 1:fdd22bb7aa52 634 n2 = fftLen;
emilmont 1:fdd22bb7aa52 635 n1 = n2;
emilmont 1:fdd22bb7aa52 636
emilmont 1:fdd22bb7aa52 637 /* n2 = fftLen/4 */
emilmont 1:fdd22bb7aa52 638 n2 >>= 2u;
emilmont 1:fdd22bb7aa52 639
emilmont 1:fdd22bb7aa52 640 /* Index for twiddle coefficient */
emilmont 1:fdd22bb7aa52 641 ic = 0u;
emilmont 1:fdd22bb7aa52 642
emilmont 1:fdd22bb7aa52 643 /* Index for input read and output write */
emilmont 1:fdd22bb7aa52 644 i0 = 0u;
emilmont 1:fdd22bb7aa52 645 j = n2;
emilmont 1:fdd22bb7aa52 646
emilmont 1:fdd22bb7aa52 647 /* Input is in 1.15(q15) format */
emilmont 1:fdd22bb7aa52 648
emilmont 1:fdd22bb7aa52 649 /* start of first stage process */
emilmont 1:fdd22bb7aa52 650 do
emilmont 1:fdd22bb7aa52 651 {
emilmont 1:fdd22bb7aa52 652 /* Butterfly implementation */
emilmont 1:fdd22bb7aa52 653
emilmont 1:fdd22bb7aa52 654 /* index calculation for the input as, */
emilmont 1:fdd22bb7aa52 655 /* pSrc16[i0 + 0], pSrc16[i0 + fftLen/4], pSrc16[i0 + fftLen/2], pSrc16[i0 + 3fftLen/4] */
emilmont 1:fdd22bb7aa52 656 i1 = i0 + n2;
emilmont 1:fdd22bb7aa52 657 i2 = i1 + n2;
emilmont 1:fdd22bb7aa52 658 i3 = i2 + n2;
emilmont 1:fdd22bb7aa52 659
emilmont 1:fdd22bb7aa52 660 /* Reading i0, i0+fftLen/2 inputs */
emilmont 1:fdd22bb7aa52 661
emilmont 1:fdd22bb7aa52 662 /* input is down scale by 4 to avoid overflow */
emilmont 1:fdd22bb7aa52 663 /* Read ya (real), xa(imag) input */
emilmont 1:fdd22bb7aa52 664 T0 = pSrc16[i0 * 2u] >> 2u;
emilmont 1:fdd22bb7aa52 665 T1 = pSrc16[(i0 * 2u) + 1u] >> 2u;
emilmont 1:fdd22bb7aa52 666
emilmont 1:fdd22bb7aa52 667 /* input is down scale by 4 to avoid overflow */
emilmont 1:fdd22bb7aa52 668 /* Read yc (real), xc(imag) input */
emilmont 1:fdd22bb7aa52 669 S0 = pSrc16[i2 * 2u] >> 2u;
emilmont 1:fdd22bb7aa52 670 S1 = pSrc16[(i2 * 2u) + 1u] >> 2u;
emilmont 1:fdd22bb7aa52 671
emilmont 1:fdd22bb7aa52 672 /* R0 = (ya + yc) */
emilmont 1:fdd22bb7aa52 673 R0 = __SSAT(T0 + S0, 16u);
emilmont 1:fdd22bb7aa52 674 /* R1 = (xa + xc) */
emilmont 1:fdd22bb7aa52 675 R1 = __SSAT(T1 + S1, 16u);
emilmont 1:fdd22bb7aa52 676
emilmont 1:fdd22bb7aa52 677 /* S0 = (ya - yc) */
emilmont 1:fdd22bb7aa52 678 S0 = __SSAT(T0 - S0, 16);
emilmont 1:fdd22bb7aa52 679 /* S1 = (xa - xc) */
emilmont 1:fdd22bb7aa52 680 S1 = __SSAT(T1 - S1, 16);
emilmont 1:fdd22bb7aa52 681
emilmont 1:fdd22bb7aa52 682 /* Reading i0+fftLen/4 , i0+3fftLen/4 inputs */
emilmont 1:fdd22bb7aa52 683 /* input is down scale by 4 to avoid overflow */
emilmont 1:fdd22bb7aa52 684 /* Read yb (real), xb(imag) input */
emilmont 1:fdd22bb7aa52 685 T0 = pSrc16[i1 * 2u] >> 2u;
emilmont 1:fdd22bb7aa52 686 T1 = pSrc16[(i1 * 2u) + 1u] >> 2u;
emilmont 1:fdd22bb7aa52 687
emilmont 1:fdd22bb7aa52 688 /* input is down scale by 4 to avoid overflow */
emilmont 1:fdd22bb7aa52 689 /* Read yd (real), xd(imag) input */
emilmont 1:fdd22bb7aa52 690 U0 = pSrc16[i3 * 2u] >> 2u;
emilmont 1:fdd22bb7aa52 691 U1 = pSrc16[(i3 * 2u) + 1] >> 2u;
emilmont 1:fdd22bb7aa52 692
emilmont 1:fdd22bb7aa52 693 /* T0 = (yb + yd) */
emilmont 1:fdd22bb7aa52 694 T0 = __SSAT(T0 + U0, 16u);
emilmont 1:fdd22bb7aa52 695 /* T1 = (xb + xd) */
emilmont 1:fdd22bb7aa52 696 T1 = __SSAT(T1 + U1, 16u);
emilmont 1:fdd22bb7aa52 697
emilmont 1:fdd22bb7aa52 698 /* writing the butterfly processed i0 sample */
emilmont 1:fdd22bb7aa52 699 /* ya' = ya + yb + yc + yd */
emilmont 1:fdd22bb7aa52 700 /* xa' = xa + xb + xc + xd */
emilmont 1:fdd22bb7aa52 701 pSrc16[i0 * 2u] = (R0 >> 1u) + (T0 >> 1u);
emilmont 1:fdd22bb7aa52 702 pSrc16[(i0 * 2u) + 1u] = (R1 >> 1u) + (T1 >> 1u);
emilmont 1:fdd22bb7aa52 703
emilmont 1:fdd22bb7aa52 704 /* R0 = (ya + yc) - (yb + yd) */
emilmont 1:fdd22bb7aa52 705 /* R1 = (xa + xc) - (xb + xd) */
emilmont 1:fdd22bb7aa52 706 R0 = __SSAT(R0 - T0, 16u);
emilmont 1:fdd22bb7aa52 707 R1 = __SSAT(R1 - T1, 16u);
emilmont 1:fdd22bb7aa52 708
emilmont 1:fdd22bb7aa52 709 /* co2 & si2 are read from Coefficient pointer */
emilmont 1:fdd22bb7aa52 710 Co2 = pCoef16[2u * ic * 2u];
emilmont 1:fdd22bb7aa52 711 Si2 = pCoef16[(2u * ic * 2u) + 1];
emilmont 1:fdd22bb7aa52 712
emilmont 1:fdd22bb7aa52 713 /* xc' = (xa-xb+xc-xd)* co2 + (ya-yb+yc-yd)* (si2) */
emilmont 1:fdd22bb7aa52 714 out1 = (short) ((Co2 * R0 + Si2 * R1) >> 16u);
emilmont 1:fdd22bb7aa52 715 /* yc' = (ya-yb+yc-yd)* co2 - (xa-xb+xc-xd)* (si2) */
emilmont 1:fdd22bb7aa52 716 out2 = (short) ((-Si2 * R0 + Co2 * R1) >> 16u);
emilmont 1:fdd22bb7aa52 717
emilmont 1:fdd22bb7aa52 718 /* Reading i0+fftLen/4 */
emilmont 1:fdd22bb7aa52 719 /* input is down scale by 4 to avoid overflow */
emilmont 1:fdd22bb7aa52 720 /* T0 = yb, T1 = xb */
emilmont 1:fdd22bb7aa52 721 T0 = pSrc16[i1 * 2u] >> 2;
emilmont 1:fdd22bb7aa52 722 T1 = pSrc16[(i1 * 2u) + 1] >> 2;
emilmont 1:fdd22bb7aa52 723
emilmont 1:fdd22bb7aa52 724 /* writing the butterfly processed i0 + fftLen/4 sample */
emilmont 1:fdd22bb7aa52 725 /* writing output(xc', yc') in little endian format */
emilmont 1:fdd22bb7aa52 726 pSrc16[i1 * 2u] = out1;
emilmont 1:fdd22bb7aa52 727 pSrc16[(i1 * 2u) + 1] = out2;
emilmont 1:fdd22bb7aa52 728
emilmont 1:fdd22bb7aa52 729 /* Butterfly calculations */
emilmont 1:fdd22bb7aa52 730 /* input is down scale by 4 to avoid overflow */
emilmont 1:fdd22bb7aa52 731 /* U0 = yd, U1 = xd */
emilmont 1:fdd22bb7aa52 732 U0 = pSrc16[i3 * 2u] >> 2;
emilmont 1:fdd22bb7aa52 733 U1 = pSrc16[(i3 * 2u) + 1] >> 2;
emilmont 1:fdd22bb7aa52 734 /* T0 = yb-yd */
emilmont 1:fdd22bb7aa52 735 T0 = __SSAT(T0 - U0, 16);
emilmont 1:fdd22bb7aa52 736 /* T1 = xb-xd */
emilmont 1:fdd22bb7aa52 737 T1 = __SSAT(T1 - U1, 16);
emilmont 1:fdd22bb7aa52 738
emilmont 1:fdd22bb7aa52 739 /* R1 = (ya-yc) + (xb- xd), R0 = (xa-xc) - (yb-yd)) */
emilmont 1:fdd22bb7aa52 740 R0 = (short) __SSAT((q31_t) (S0 - T1), 16);
emilmont 1:fdd22bb7aa52 741 R1 = (short) __SSAT((q31_t) (S1 + T0), 16);
emilmont 1:fdd22bb7aa52 742
emilmont 1:fdd22bb7aa52 743 /* S1 = (ya-yc) - (xb- xd), S0 = (xa-xc) + (yb-yd)) */
emilmont 1:fdd22bb7aa52 744 S0 = (short) __SSAT(((q31_t) S0 + T1), 16u);
emilmont 1:fdd22bb7aa52 745 S1 = (short) __SSAT(((q31_t) S1 - T0), 16u);
emilmont 1:fdd22bb7aa52 746
emilmont 1:fdd22bb7aa52 747 /* co1 & si1 are read from Coefficient pointer */
emilmont 1:fdd22bb7aa52 748 Co1 = pCoef16[ic * 2u];
emilmont 1:fdd22bb7aa52 749 Si1 = pCoef16[(ic * 2u) + 1];
emilmont 1:fdd22bb7aa52 750 /* Butterfly process for the i0+fftLen/2 sample */
emilmont 1:fdd22bb7aa52 751 /* xb' = (xa+yb-xc-yd)* co1 + (ya-xb-yc+xd)* (si1) */
emilmont 1:fdd22bb7aa52 752 out1 = (short) ((Si1 * S1 + Co1 * S0) >> 16);
emilmont 1:fdd22bb7aa52 753 /* yb' = (ya-xb-yc+xd)* co1 - (xa+yb-xc-yd)* (si1) */
emilmont 1:fdd22bb7aa52 754 out2 = (short) ((-Si1 * S0 + Co1 * S1) >> 16);
emilmont 1:fdd22bb7aa52 755
emilmont 1:fdd22bb7aa52 756 /* writing output(xb', yb') in little endian format */
emilmont 1:fdd22bb7aa52 757 pSrc16[i2 * 2u] = out1;
emilmont 1:fdd22bb7aa52 758 pSrc16[(i2 * 2u) + 1] = out2;
emilmont 1:fdd22bb7aa52 759
emilmont 1:fdd22bb7aa52 760 /* Co3 & si3 are read from Coefficient pointer */
emilmont 1:fdd22bb7aa52 761 Co3 = pCoef16[3u * (ic * 2u)];
emilmont 1:fdd22bb7aa52 762 Si3 = pCoef16[(3u * (ic * 2u)) + 1];
emilmont 1:fdd22bb7aa52 763 /* Butterfly process for the i0+3fftLen/4 sample */
emilmont 1:fdd22bb7aa52 764 /* xd' = (xa-yb-xc+yd)* Co3 + (ya+xb-yc-xd)* (si3) */
emilmont 1:fdd22bb7aa52 765 out1 = (short) ((Si3 * R1 + Co3 * R0) >> 16u);
emilmont 1:fdd22bb7aa52 766 /* yd' = (ya+xb-yc-xd)* Co3 - (xa-yb-xc+yd)* (si3) */
emilmont 1:fdd22bb7aa52 767 out2 = (short) ((-Si3 * R0 + Co3 * R1) >> 16u);
emilmont 1:fdd22bb7aa52 768 /* writing output(xd', yd') in little endian format */
emilmont 1:fdd22bb7aa52 769 pSrc16[i3 * 2u] = out1;
emilmont 1:fdd22bb7aa52 770 pSrc16[(i3 * 2u) + 1] = out2;
emilmont 1:fdd22bb7aa52 771
emilmont 1:fdd22bb7aa52 772 /* Twiddle coefficients index modifier */
emilmont 1:fdd22bb7aa52 773 ic = ic + twidCoefModifier;
emilmont 1:fdd22bb7aa52 774
emilmont 1:fdd22bb7aa52 775 /* Updating input index */
emilmont 1:fdd22bb7aa52 776 i0 = i0 + 1u;
emilmont 1:fdd22bb7aa52 777
emilmont 1:fdd22bb7aa52 778 } while(--j);
emilmont 1:fdd22bb7aa52 779 /* data is in 4.11(q11) format */
emilmont 1:fdd22bb7aa52 780
emilmont 1:fdd22bb7aa52 781 /* end of first stage process */
emilmont 1:fdd22bb7aa52 782
emilmont 1:fdd22bb7aa52 783
emilmont 1:fdd22bb7aa52 784 /* start of middle stage process */
emilmont 1:fdd22bb7aa52 785
emilmont 1:fdd22bb7aa52 786 /* Twiddle coefficients index modifier */
emilmont 1:fdd22bb7aa52 787 twidCoefModifier <<= 2u;
emilmont 1:fdd22bb7aa52 788
emilmont 1:fdd22bb7aa52 789 /* Calculation of Middle stage */
emilmont 1:fdd22bb7aa52 790 for (k = fftLen / 4u; k > 4u; k >>= 2u)
emilmont 1:fdd22bb7aa52 791 {
emilmont 1:fdd22bb7aa52 792 /* Initializations for the middle stage */
emilmont 1:fdd22bb7aa52 793 n1 = n2;
emilmont 1:fdd22bb7aa52 794 n2 >>= 2u;
emilmont 1:fdd22bb7aa52 795 ic = 0u;
emilmont 1:fdd22bb7aa52 796
emilmont 1:fdd22bb7aa52 797 for (j = 0u; j <= (n2 - 1u); j++)
emilmont 1:fdd22bb7aa52 798 {
emilmont 1:fdd22bb7aa52 799 /* index calculation for the coefficients */
emilmont 1:fdd22bb7aa52 800 Co1 = pCoef16[ic * 2u];
emilmont 1:fdd22bb7aa52 801 Si1 = pCoef16[(ic * 2u) + 1u];
emilmont 1:fdd22bb7aa52 802 Co2 = pCoef16[2u * (ic * 2u)];
emilmont 1:fdd22bb7aa52 803 Si2 = pCoef16[(2u * (ic * 2u)) + 1u];
emilmont 1:fdd22bb7aa52 804 Co3 = pCoef16[3u * (ic * 2u)];
emilmont 1:fdd22bb7aa52 805 Si3 = pCoef16[(3u * (ic * 2u)) + 1u];
emilmont 1:fdd22bb7aa52 806
emilmont 1:fdd22bb7aa52 807 /* Twiddle coefficients index modifier */
emilmont 1:fdd22bb7aa52 808 ic = ic + twidCoefModifier;
emilmont 1:fdd22bb7aa52 809
emilmont 1:fdd22bb7aa52 810 /* Butterfly implementation */
emilmont 1:fdd22bb7aa52 811 for (i0 = j; i0 < fftLen; i0 += n1)
emilmont 1:fdd22bb7aa52 812 {
emilmont 1:fdd22bb7aa52 813 /* index calculation for the input as, */
emilmont 1:fdd22bb7aa52 814 /* pSrc16[i0 + 0], pSrc16[i0 + fftLen/4], pSrc16[i0 + fftLen/2], pSrc16[i0 + 3fftLen/4] */
emilmont 1:fdd22bb7aa52 815 i1 = i0 + n2;
emilmont 1:fdd22bb7aa52 816 i2 = i1 + n2;
emilmont 1:fdd22bb7aa52 817 i3 = i2 + n2;
emilmont 1:fdd22bb7aa52 818
emilmont 1:fdd22bb7aa52 819 /* Reading i0, i0+fftLen/2 inputs */
emilmont 1:fdd22bb7aa52 820 /* Read ya (real), xa(imag) input */
emilmont 1:fdd22bb7aa52 821 T0 = pSrc16[i0 * 2u];
emilmont 1:fdd22bb7aa52 822 T1 = pSrc16[(i0 * 2u) + 1u];
emilmont 1:fdd22bb7aa52 823
emilmont 1:fdd22bb7aa52 824 /* Read yc (real), xc(imag) input */
emilmont 1:fdd22bb7aa52 825 S0 = pSrc16[i2 * 2u];
emilmont 1:fdd22bb7aa52 826 S1 = pSrc16[(i2 * 2u) + 1u];
emilmont 1:fdd22bb7aa52 827
emilmont 1:fdd22bb7aa52 828 /* R0 = (ya + yc), R1 = (xa + xc) */
emilmont 1:fdd22bb7aa52 829 R0 = __SSAT(T0 + S0, 16);
emilmont 1:fdd22bb7aa52 830 R1 = __SSAT(T1 + S1, 16);
emilmont 1:fdd22bb7aa52 831
emilmont 1:fdd22bb7aa52 832 /* S0 = (ya - yc), S1 =(xa - xc) */
emilmont 1:fdd22bb7aa52 833 S0 = __SSAT(T0 - S0, 16);
emilmont 1:fdd22bb7aa52 834 S1 = __SSAT(T1 - S1, 16);
emilmont 1:fdd22bb7aa52 835
emilmont 1:fdd22bb7aa52 836 /* Reading i0+fftLen/4 , i0+3fftLen/4 inputs */
emilmont 1:fdd22bb7aa52 837 /* Read yb (real), xb(imag) input */
emilmont 1:fdd22bb7aa52 838 T0 = pSrc16[i1 * 2u];
emilmont 1:fdd22bb7aa52 839 T1 = pSrc16[(i1 * 2u) + 1u];
emilmont 1:fdd22bb7aa52 840
emilmont 1:fdd22bb7aa52 841 /* Read yd (real), xd(imag) input */
emilmont 1:fdd22bb7aa52 842 U0 = pSrc16[i3 * 2u];
emilmont 1:fdd22bb7aa52 843 U1 = pSrc16[(i3 * 2u) + 1u];
emilmont 1:fdd22bb7aa52 844
emilmont 1:fdd22bb7aa52 845
emilmont 1:fdd22bb7aa52 846 /* T0 = (yb + yd), T1 = (xb + xd) */
emilmont 1:fdd22bb7aa52 847 T0 = __SSAT(T0 + U0, 16);
emilmont 1:fdd22bb7aa52 848 T1 = __SSAT(T1 + U1, 16);
emilmont 1:fdd22bb7aa52 849
emilmont 1:fdd22bb7aa52 850 /* writing the butterfly processed i0 sample */
emilmont 1:fdd22bb7aa52 851
emilmont 1:fdd22bb7aa52 852 /* xa' = xa + xb + xc + xd */
emilmont 1:fdd22bb7aa52 853 /* ya' = ya + yb + yc + yd */
emilmont 1:fdd22bb7aa52 854 out1 = ((R0 >> 1u) + (T0 >> 1u)) >> 1u;
emilmont 1:fdd22bb7aa52 855 out2 = ((R1 >> 1u) + (T1 >> 1u)) >> 1u;
emilmont 1:fdd22bb7aa52 856
emilmont 1:fdd22bb7aa52 857 pSrc16[i0 * 2u] = out1;
emilmont 1:fdd22bb7aa52 858 pSrc16[(2u * i0) + 1u] = out2;
emilmont 1:fdd22bb7aa52 859
emilmont 1:fdd22bb7aa52 860 /* R0 = (ya + yc) - (yb + yd), R1 = (xa + xc) - (xb + xd) */
emilmont 1:fdd22bb7aa52 861 R0 = (R0 >> 1u) - (T0 >> 1u);
emilmont 1:fdd22bb7aa52 862 R1 = (R1 >> 1u) - (T1 >> 1u);
emilmont 1:fdd22bb7aa52 863
emilmont 1:fdd22bb7aa52 864 /* (ya-yb+yc-yd)* (si2) + (xa-xb+xc-xd)* co2 */
emilmont 1:fdd22bb7aa52 865 out1 = (short) ((Co2 * R0 + Si2 * R1) >> 16u);
emilmont 1:fdd22bb7aa52 866
emilmont 1:fdd22bb7aa52 867 /* (ya-yb+yc-yd)* co2 - (xa-xb+xc-xd)* (si2) */
emilmont 1:fdd22bb7aa52 868 out2 = (short) ((-Si2 * R0 + Co2 * R1) >> 16u);
emilmont 1:fdd22bb7aa52 869
emilmont 1:fdd22bb7aa52 870 /* Reading i0+3fftLen/4 */
emilmont 1:fdd22bb7aa52 871 /* Read yb (real), xb(imag) input */
emilmont 1:fdd22bb7aa52 872 T0 = pSrc16[i1 * 2u];
emilmont 1:fdd22bb7aa52 873 T1 = pSrc16[(i1 * 2u) + 1u];
emilmont 1:fdd22bb7aa52 874
emilmont 1:fdd22bb7aa52 875 /* writing the butterfly processed i0 + fftLen/4 sample */
emilmont 1:fdd22bb7aa52 876 /* xc' = (xa-xb+xc-xd)* co2 + (ya-yb+yc-yd)* (si2) */
emilmont 1:fdd22bb7aa52 877 /* yc' = (ya-yb+yc-yd)* co2 - (xa-xb+xc-xd)* (si2) */
emilmont 1:fdd22bb7aa52 878 pSrc16[i1 * 2u] = out1;
emilmont 1:fdd22bb7aa52 879 pSrc16[(i1 * 2u) + 1u] = out2;
emilmont 1:fdd22bb7aa52 880
emilmont 1:fdd22bb7aa52 881 /* Butterfly calculations */
emilmont 1:fdd22bb7aa52 882
emilmont 1:fdd22bb7aa52 883 /* Read yd (real), xd(imag) input */
emilmont 1:fdd22bb7aa52 884 U0 = pSrc16[i3 * 2u];
emilmont 1:fdd22bb7aa52 885 U1 = pSrc16[(i3 * 2u) + 1u];
emilmont 1:fdd22bb7aa52 886
emilmont 1:fdd22bb7aa52 887 /* T0 = yb-yd, T1 = xb-xd */
emilmont 1:fdd22bb7aa52 888 T0 = __SSAT(T0 - U0, 16);
emilmont 1:fdd22bb7aa52 889 T1 = __SSAT(T1 - U1, 16);
emilmont 1:fdd22bb7aa52 890
emilmont 1:fdd22bb7aa52 891 /* R0 = (ya-yc) + (xb- xd), R1 = (xa-xc) - (yb-yd)) */
emilmont 1:fdd22bb7aa52 892 R0 = (S0 >> 1u) - (T1 >> 1u);
emilmont 1:fdd22bb7aa52 893 R1 = (S1 >> 1u) + (T0 >> 1u);
emilmont 1:fdd22bb7aa52 894
emilmont 1:fdd22bb7aa52 895 /* S0 = (ya-yc) - (xb- xd), S1 = (xa-xc) + (yb-yd)) */
emilmont 1:fdd22bb7aa52 896 S0 = (S0 >> 1u) + (T1 >> 1u);
emilmont 1:fdd22bb7aa52 897 S1 = (S1 >> 1u) - (T0 >> 1u);
emilmont 1:fdd22bb7aa52 898
emilmont 1:fdd22bb7aa52 899 /* Butterfly process for the i0+fftLen/2 sample */
emilmont 1:fdd22bb7aa52 900 out1 = (short) ((Co1 * S0 + Si1 * S1) >> 16u);
emilmont 1:fdd22bb7aa52 901
emilmont 1:fdd22bb7aa52 902 out2 = (short) ((-Si1 * S0 + Co1 * S1) >> 16u);
emilmont 1:fdd22bb7aa52 903
emilmont 1:fdd22bb7aa52 904 /* xb' = (xa+yb-xc-yd)* co1 + (ya-xb-yc+xd)* (si1) */
emilmont 1:fdd22bb7aa52 905 /* yb' = (ya-xb-yc+xd)* co1 - (xa+yb-xc-yd)* (si1) */
emilmont 1:fdd22bb7aa52 906 pSrc16[i2 * 2u] = out1;
emilmont 1:fdd22bb7aa52 907 pSrc16[(i2 * 2u) + 1u] = out2;
emilmont 1:fdd22bb7aa52 908
emilmont 1:fdd22bb7aa52 909 /* Butterfly process for the i0+3fftLen/4 sample */
emilmont 1:fdd22bb7aa52 910 out1 = (short) ((Si3 * R1 + Co3 * R0) >> 16u);
emilmont 1:fdd22bb7aa52 911
emilmont 1:fdd22bb7aa52 912 out2 = (short) ((-Si3 * R0 + Co3 * R1) >> 16u);
emilmont 1:fdd22bb7aa52 913 /* xd' = (xa-yb-xc+yd)* Co3 + (ya+xb-yc-xd)* (si3) */
emilmont 1:fdd22bb7aa52 914 /* yd' = (ya+xb-yc-xd)* Co3 - (xa-yb-xc+yd)* (si3) */
emilmont 1:fdd22bb7aa52 915 pSrc16[i3 * 2u] = out1;
emilmont 1:fdd22bb7aa52 916 pSrc16[(i3 * 2u) + 1u] = out2;
emilmont 1:fdd22bb7aa52 917 }
emilmont 1:fdd22bb7aa52 918 }
emilmont 1:fdd22bb7aa52 919 /* Twiddle coefficients index modifier */
emilmont 1:fdd22bb7aa52 920 twidCoefModifier <<= 2u;
emilmont 1:fdd22bb7aa52 921 }
emilmont 1:fdd22bb7aa52 922 /* end of middle stage process */
emilmont 1:fdd22bb7aa52 923
emilmont 1:fdd22bb7aa52 924
emilmont 1:fdd22bb7aa52 925 /* data is in 10.6(q6) format for the 1024 point */
emilmont 1:fdd22bb7aa52 926 /* data is in 8.8(q8) format for the 256 point */
emilmont 1:fdd22bb7aa52 927 /* data is in 6.10(q10) format for the 64 point */
emilmont 1:fdd22bb7aa52 928 /* data is in 4.12(q12) format for the 16 point */
emilmont 1:fdd22bb7aa52 929
emilmont 1:fdd22bb7aa52 930 /* Initializations for the last stage */
emilmont 1:fdd22bb7aa52 931 n1 = n2;
emilmont 1:fdd22bb7aa52 932 n2 >>= 2u;
emilmont 1:fdd22bb7aa52 933
emilmont 1:fdd22bb7aa52 934 /* start of last stage process */
emilmont 1:fdd22bb7aa52 935
emilmont 1:fdd22bb7aa52 936 /* Butterfly implementation */
emilmont 1:fdd22bb7aa52 937 for (i0 = 0u; i0 <= (fftLen - n1); i0 += n1)
emilmont 1:fdd22bb7aa52 938 {
emilmont 1:fdd22bb7aa52 939 /* index calculation for the input as, */
emilmont 1:fdd22bb7aa52 940 /* pSrc16[i0 + 0], pSrc16[i0 + fftLen/4], pSrc16[i0 + fftLen/2], pSrc16[i0 + 3fftLen/4] */
emilmont 1:fdd22bb7aa52 941 i1 = i0 + n2;
emilmont 1:fdd22bb7aa52 942 i2 = i1 + n2;
emilmont 1:fdd22bb7aa52 943 i3 = i2 + n2;
emilmont 1:fdd22bb7aa52 944
emilmont 1:fdd22bb7aa52 945 /* Reading i0, i0+fftLen/2 inputs */
emilmont 1:fdd22bb7aa52 946 /* Read ya (real), xa(imag) input */
emilmont 1:fdd22bb7aa52 947 T0 = pSrc16[i0 * 2u];
emilmont 1:fdd22bb7aa52 948 T1 = pSrc16[(i0 * 2u) + 1u];
emilmont 1:fdd22bb7aa52 949
emilmont 1:fdd22bb7aa52 950 /* Read yc (real), xc(imag) input */
emilmont 1:fdd22bb7aa52 951 S0 = pSrc16[i2 * 2u];
emilmont 1:fdd22bb7aa52 952 S1 = pSrc16[(i2 * 2u) + 1u];
emilmont 1:fdd22bb7aa52 953
emilmont 1:fdd22bb7aa52 954 /* R0 = (ya + yc), R1 = (xa + xc) */
emilmont 1:fdd22bb7aa52 955 R0 = __SSAT(T0 + S0, 16u);
emilmont 1:fdd22bb7aa52 956 R1 = __SSAT(T1 + S1, 16u);
emilmont 1:fdd22bb7aa52 957
emilmont 1:fdd22bb7aa52 958 /* S0 = (ya - yc), S1 = (xa - xc) */
emilmont 1:fdd22bb7aa52 959 S0 = __SSAT(T0 - S0, 16u);
emilmont 1:fdd22bb7aa52 960 S1 = __SSAT(T1 - S1, 16u);
emilmont 1:fdd22bb7aa52 961
emilmont 1:fdd22bb7aa52 962 /* Reading i0+fftLen/4 , i0+3fftLen/4 inputs */
emilmont 1:fdd22bb7aa52 963 /* Read yb (real), xb(imag) input */
emilmont 1:fdd22bb7aa52 964 T0 = pSrc16[i1 * 2u];
emilmont 1:fdd22bb7aa52 965 T1 = pSrc16[(i1 * 2u) + 1u];
emilmont 1:fdd22bb7aa52 966 /* Read yd (real), xd(imag) input */
emilmont 1:fdd22bb7aa52 967 U0 = pSrc16[i3 * 2u];
emilmont 1:fdd22bb7aa52 968 U1 = pSrc16[(i3 * 2u) + 1u];
emilmont 1:fdd22bb7aa52 969
emilmont 1:fdd22bb7aa52 970 /* T0 = (yb + yd), T1 = (xb + xd)) */
emilmont 1:fdd22bb7aa52 971 T0 = __SSAT(T0 + U0, 16u);
emilmont 1:fdd22bb7aa52 972 T1 = __SSAT(T1 + U1, 16u);
emilmont 1:fdd22bb7aa52 973
emilmont 1:fdd22bb7aa52 974 /* writing the butterfly processed i0 sample */
emilmont 1:fdd22bb7aa52 975 /* xa' = xa + xb + xc + xd */
emilmont 1:fdd22bb7aa52 976 /* ya' = ya + yb + yc + yd */
emilmont 1:fdd22bb7aa52 977 pSrc16[i0 * 2u] = (R0 >> 1u) + (T0 >> 1u);
emilmont 1:fdd22bb7aa52 978 pSrc16[(i0 * 2u) + 1u] = (R1 >> 1u) + (T1 >> 1u);
emilmont 1:fdd22bb7aa52 979
emilmont 1:fdd22bb7aa52 980 /* R0 = (ya + yc) - (yb + yd), R1 = (xa + xc) - (xb + xd) */
emilmont 1:fdd22bb7aa52 981 R0 = (R0 >> 1u) - (T0 >> 1u);
emilmont 1:fdd22bb7aa52 982 R1 = (R1 >> 1u) - (T1 >> 1u);
emilmont 1:fdd22bb7aa52 983 /* Read yb (real), xb(imag) input */
emilmont 1:fdd22bb7aa52 984 T0 = pSrc16[i1 * 2u];
emilmont 1:fdd22bb7aa52 985 T1 = pSrc16[(i1 * 2u) + 1u];
emilmont 1:fdd22bb7aa52 986
emilmont 1:fdd22bb7aa52 987 /* writing the butterfly processed i0 + fftLen/4 sample */
emilmont 1:fdd22bb7aa52 988 /* xc' = (xa-xb+xc-xd) */
emilmont 1:fdd22bb7aa52 989 /* yc' = (ya-yb+yc-yd) */
emilmont 1:fdd22bb7aa52 990 pSrc16[i1 * 2u] = R0;
emilmont 1:fdd22bb7aa52 991 pSrc16[(i1 * 2u) + 1u] = R1;
emilmont 1:fdd22bb7aa52 992
emilmont 1:fdd22bb7aa52 993 /* Read yd (real), xd(imag) input */
emilmont 1:fdd22bb7aa52 994 U0 = pSrc16[i3 * 2u];
emilmont 1:fdd22bb7aa52 995 U1 = pSrc16[(i3 * 2u) + 1u];
emilmont 1:fdd22bb7aa52 996 /* T0 = (yb - yd), T1 = (xb - xd) */
emilmont 1:fdd22bb7aa52 997 T0 = __SSAT(T0 - U0, 16u);
emilmont 1:fdd22bb7aa52 998 T1 = __SSAT(T1 - U1, 16u);
emilmont 1:fdd22bb7aa52 999
emilmont 1:fdd22bb7aa52 1000 /* writing the butterfly processed i0 + fftLen/2 sample */
emilmont 1:fdd22bb7aa52 1001 /* xb' = (xa+yb-xc-yd) */
emilmont 1:fdd22bb7aa52 1002 /* yb' = (ya-xb-yc+xd) */
emilmont 1:fdd22bb7aa52 1003 pSrc16[i2 * 2u] = (S0 >> 1u) + (T1 >> 1u);
emilmont 1:fdd22bb7aa52 1004 pSrc16[(i2 * 2u) + 1u] = (S1 >> 1u) - (T0 >> 1u);
emilmont 1:fdd22bb7aa52 1005
emilmont 1:fdd22bb7aa52 1006 /* writing the butterfly processed i0 + 3fftLen/4 sample */
emilmont 1:fdd22bb7aa52 1007 /* xd' = (xa-yb-xc+yd) */
emilmont 1:fdd22bb7aa52 1008 /* yd' = (ya+xb-yc-xd) */
emilmont 1:fdd22bb7aa52 1009 pSrc16[i3 * 2u] = (S0 >> 1u) - (T1 >> 1u);
emilmont 1:fdd22bb7aa52 1010 pSrc16[(i3 * 2u) + 1u] = (S1 >> 1u) + (T0 >> 1u);
emilmont 1:fdd22bb7aa52 1011
emilmont 1:fdd22bb7aa52 1012 }
emilmont 1:fdd22bb7aa52 1013
emilmont 1:fdd22bb7aa52 1014 /* end of last stage process */
emilmont 1:fdd22bb7aa52 1015
emilmont 1:fdd22bb7aa52 1016 /* output is in 11.5(q5) format for the 1024 point */
emilmont 1:fdd22bb7aa52 1017 /* output is in 9.7(q7) format for the 256 point */
emilmont 1:fdd22bb7aa52 1018 /* output is in 7.9(q9) format for the 64 point */
emilmont 1:fdd22bb7aa52 1019 /* output is in 5.11(q11) format for the 16 point */
emilmont 1:fdd22bb7aa52 1020
mbed_official 3:7a284390b0ce 1021 #endif /* #ifndef ARM_MATH_CM0_FAMILY */
emilmont 1:fdd22bb7aa52 1022
emilmont 1:fdd22bb7aa52 1023 }
emilmont 1:fdd22bb7aa52 1024
emilmont 1:fdd22bb7aa52 1025
emilmont 1:fdd22bb7aa52 1026 /**
emilmont 1:fdd22bb7aa52 1027 * @brief Core function for the Q15 CIFFT butterfly process.
emilmont 1:fdd22bb7aa52 1028 * @param[in, out] *pSrc16 points to the in-place buffer of Q15 data type.
emilmont 1:fdd22bb7aa52 1029 * @param[in] fftLen length of the FFT.
emilmont 1:fdd22bb7aa52 1030 * @param[in] *pCoef16 points to twiddle coefficient buffer.
emilmont 1:fdd22bb7aa52 1031 * @param[in] twidCoefModifier twiddle coefficient modifier that supports different size FFTs with the same twiddle factor table.
emilmont 1:fdd22bb7aa52 1032 * @return none.
emilmont 1:fdd22bb7aa52 1033 */
emilmont 1:fdd22bb7aa52 1034
emilmont 1:fdd22bb7aa52 1035 /*
emilmont 1:fdd22bb7aa52 1036 * Radix-4 IFFT algorithm used is :
emilmont 1:fdd22bb7aa52 1037 *
emilmont 1:fdd22bb7aa52 1038 * CIFFT uses same twiddle coefficients as CFFT function
emilmont 1:fdd22bb7aa52 1039 * 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 1040 *
emilmont 1:fdd22bb7aa52 1041 *
emilmont 1:fdd22bb7aa52 1042 * IFFT is implemented with following changes in equations from FFT
emilmont 1:fdd22bb7aa52 1043 *
emilmont 1:fdd22bb7aa52 1044 * Input real and imaginary data:
emilmont 1:fdd22bb7aa52 1045 * x(n) = xa + j * ya
emilmont 1:fdd22bb7aa52 1046 * x(n+N/4 ) = xb + j * yb
emilmont 1:fdd22bb7aa52 1047 * x(n+N/2 ) = xc + j * yc
emilmont 1:fdd22bb7aa52 1048 * x(n+3N 4) = xd + j * yd
emilmont 1:fdd22bb7aa52 1049 *
emilmont 1:fdd22bb7aa52 1050 *
emilmont 1:fdd22bb7aa52 1051 * Output real and imaginary data:
emilmont 1:fdd22bb7aa52 1052 * x(4r) = xa'+ j * ya'
emilmont 1:fdd22bb7aa52 1053 * x(4r+1) = xb'+ j * yb'
emilmont 1:fdd22bb7aa52 1054 * x(4r+2) = xc'+ j * yc'
emilmont 1:fdd22bb7aa52 1055 * x(4r+3) = xd'+ j * yd'
emilmont 1:fdd22bb7aa52 1056 *
emilmont 1:fdd22bb7aa52 1057 *
emilmont 1:fdd22bb7aa52 1058 * Twiddle factors for radix-4 IFFT:
emilmont 1:fdd22bb7aa52 1059 * Wn = co1 + j * (si1)
emilmont 1:fdd22bb7aa52 1060 * W2n = co2 + j * (si2)
emilmont 1:fdd22bb7aa52 1061 * W3n = co3 + j * (si3)
emilmont 1:fdd22bb7aa52 1062
emilmont 1:fdd22bb7aa52 1063 * The real and imaginary output values for the radix-4 butterfly are
emilmont 1:fdd22bb7aa52 1064 * xa' = xa + xb + xc + xd
emilmont 1:fdd22bb7aa52 1065 * ya' = ya + yb + yc + yd
emilmont 1:fdd22bb7aa52 1066 * xb' = (xa-yb-xc+yd)* co1 - (ya+xb-yc-xd)* (si1)
emilmont 1:fdd22bb7aa52 1067 * yb' = (ya+xb-yc-xd)* co1 + (xa-yb-xc+yd)* (si1)
emilmont 1:fdd22bb7aa52 1068 * xc' = (xa-xb+xc-xd)* co2 - (ya-yb+yc-yd)* (si2)
emilmont 1:fdd22bb7aa52 1069 * yc' = (ya-yb+yc-yd)* co2 + (xa-xb+xc-xd)* (si2)
emilmont 1:fdd22bb7aa52 1070 * xd' = (xa+yb-xc-yd)* co3 - (ya-xb-yc+xd)* (si3)
emilmont 1:fdd22bb7aa52 1071 * yd' = (ya-xb-yc+xd)* co3 + (xa+yb-xc-yd)* (si3)
emilmont 1:fdd22bb7aa52 1072 *
emilmont 1:fdd22bb7aa52 1073 */
emilmont 1:fdd22bb7aa52 1074
emilmont 1:fdd22bb7aa52 1075 void arm_radix4_butterfly_inverse_q15(
emilmont 1:fdd22bb7aa52 1076 q15_t * pSrc16,
emilmont 1:fdd22bb7aa52 1077 uint32_t fftLen,
emilmont 1:fdd22bb7aa52 1078 q15_t * pCoef16,
emilmont 1:fdd22bb7aa52 1079 uint32_t twidCoefModifier)
emilmont 1:fdd22bb7aa52 1080 {
emilmont 1:fdd22bb7aa52 1081
mbed_official 3:7a284390b0ce 1082 #ifndef ARM_MATH_CM0_FAMILY
emilmont 1:fdd22bb7aa52 1083
emilmont 1:fdd22bb7aa52 1084 /* Run the below code for Cortex-M4 and Cortex-M3 */
emilmont 1:fdd22bb7aa52 1085
emilmont 1:fdd22bb7aa52 1086 q31_t R, S, T, U;
emilmont 1:fdd22bb7aa52 1087 q31_t C1, C2, C3, out1, out2;
emilmont 1:fdd22bb7aa52 1088 uint32_t n1, n2, ic, i0, i1, i2, i3, j, k;
emilmont 1:fdd22bb7aa52 1089 q15_t in;
emilmont 1:fdd22bb7aa52 1090
emilmont 1:fdd22bb7aa52 1091 q15_t *ptr1;
emilmont 1:fdd22bb7aa52 1092
emilmont 1:fdd22bb7aa52 1093
emilmont 1:fdd22bb7aa52 1094
emilmont 1:fdd22bb7aa52 1095 q31_t xaya, xbyb, xcyc, xdyd;
emilmont 1:fdd22bb7aa52 1096
emilmont 1:fdd22bb7aa52 1097 /* Total process is divided into three stages */
emilmont 1:fdd22bb7aa52 1098
emilmont 1:fdd22bb7aa52 1099 /* process first stage, middle stages, & last stage */
emilmont 1:fdd22bb7aa52 1100
emilmont 1:fdd22bb7aa52 1101 /* Initializations for the first stage */
emilmont 1:fdd22bb7aa52 1102 n2 = fftLen;
emilmont 1:fdd22bb7aa52 1103 n1 = n2;
emilmont 1:fdd22bb7aa52 1104
emilmont 1:fdd22bb7aa52 1105 /* n2 = fftLen/4 */
emilmont 1:fdd22bb7aa52 1106 n2 >>= 2u;
emilmont 1:fdd22bb7aa52 1107
emilmont 1:fdd22bb7aa52 1108 /* Index for twiddle coefficient */
emilmont 1:fdd22bb7aa52 1109 ic = 0u;
emilmont 1:fdd22bb7aa52 1110
emilmont 1:fdd22bb7aa52 1111 /* Index for input read and output write */
emilmont 1:fdd22bb7aa52 1112 i0 = 0u;
emilmont 1:fdd22bb7aa52 1113 j = n2;
emilmont 1:fdd22bb7aa52 1114
emilmont 1:fdd22bb7aa52 1115 /* Input is in 1.15(q15) format */
emilmont 1:fdd22bb7aa52 1116
emilmont 1:fdd22bb7aa52 1117 /* start of first stage process */
emilmont 1:fdd22bb7aa52 1118 do
emilmont 1:fdd22bb7aa52 1119 {
emilmont 1:fdd22bb7aa52 1120 /* Butterfly implementation */
emilmont 1:fdd22bb7aa52 1121
emilmont 1:fdd22bb7aa52 1122 /* index calculation for the input as, */
emilmont 1:fdd22bb7aa52 1123 /* pSrc16[i0 + 0], pSrc16[i0 + fftLen/4], pSrc16[i0 + fftLen/2], pSrc16[i0 + 3fftLen/4] */
emilmont 1:fdd22bb7aa52 1124 i1 = i0 + n2;
emilmont 1:fdd22bb7aa52 1125 i2 = i1 + n2;
emilmont 1:fdd22bb7aa52 1126 i3 = i2 + n2;
emilmont 1:fdd22bb7aa52 1127
emilmont 1:fdd22bb7aa52 1128 /* Reading i0, i0+fftLen/2 inputs */
emilmont 1:fdd22bb7aa52 1129 /* Read ya (real), xa(imag) input */
emilmont 1:fdd22bb7aa52 1130 T = _SIMD32_OFFSET(pSrc16 + (2u * i0));
emilmont 1:fdd22bb7aa52 1131 in = ((int16_t) (T & 0xFFFF)) >> 2;
emilmont 1:fdd22bb7aa52 1132 T = ((T >> 2) & 0xFFFF0000) | (in & 0xFFFF);
emilmont 1:fdd22bb7aa52 1133
emilmont 1:fdd22bb7aa52 1134 /* Read yc (real), xc(imag) input */
emilmont 1:fdd22bb7aa52 1135 S = _SIMD32_OFFSET(pSrc16 + (2u * i2));
emilmont 1:fdd22bb7aa52 1136 in = ((int16_t) (S & 0xFFFF)) >> 2;
emilmont 1:fdd22bb7aa52 1137 S = ((S >> 2) & 0xFFFF0000) | (in & 0xFFFF);
emilmont 1:fdd22bb7aa52 1138
emilmont 1:fdd22bb7aa52 1139 /* R = packed((ya + yc), (xa + xc) ) */
emilmont 1:fdd22bb7aa52 1140 R = __QADD16(T, S);
emilmont 1:fdd22bb7aa52 1141
emilmont 1:fdd22bb7aa52 1142 /* S = packed((ya - yc), (xa - xc) ) */
emilmont 1:fdd22bb7aa52 1143 S = __QSUB16(T, S);
emilmont 1:fdd22bb7aa52 1144
emilmont 1:fdd22bb7aa52 1145 /* Reading i0+fftLen/4 , i0+3fftLen/4 inputs */
emilmont 1:fdd22bb7aa52 1146 /* Read yb (real), xb(imag) input */
emilmont 1:fdd22bb7aa52 1147 T = _SIMD32_OFFSET(pSrc16 + (2u * i1));
emilmont 1:fdd22bb7aa52 1148 in = ((int16_t) (T & 0xFFFF)) >> 2;
emilmont 1:fdd22bb7aa52 1149 T = ((T >> 2) & 0xFFFF0000) | (in & 0xFFFF);
emilmont 1:fdd22bb7aa52 1150
emilmont 1:fdd22bb7aa52 1151 /* Read yd (real), xd(imag) input */
emilmont 1:fdd22bb7aa52 1152 U = _SIMD32_OFFSET(pSrc16 + (2u * i3));
emilmont 1:fdd22bb7aa52 1153 in = ((int16_t) (U & 0xFFFF)) >> 2;
emilmont 1:fdd22bb7aa52 1154 U = ((U >> 2) & 0xFFFF0000) | (in & 0xFFFF);
emilmont 1:fdd22bb7aa52 1155
emilmont 1:fdd22bb7aa52 1156 /* T = packed((yb + yd), (xb + xd) ) */
emilmont 1:fdd22bb7aa52 1157 T = __QADD16(T, U);
emilmont 1:fdd22bb7aa52 1158
emilmont 1:fdd22bb7aa52 1159 /* writing the butterfly processed i0 sample */
emilmont 1:fdd22bb7aa52 1160 /* xa' = xa + xb + xc + xd */
emilmont 1:fdd22bb7aa52 1161 /* ya' = ya + yb + yc + yd */
emilmont 1:fdd22bb7aa52 1162 _SIMD32_OFFSET(pSrc16 + (2u * i0)) = __SHADD16(R, T);
emilmont 1:fdd22bb7aa52 1163
emilmont 1:fdd22bb7aa52 1164 /* R = packed((ya + yc) - (yb + yd), (xa + xc)- (xb + xd)) */
emilmont 1:fdd22bb7aa52 1165 R = __QSUB16(R, T);
emilmont 1:fdd22bb7aa52 1166
emilmont 1:fdd22bb7aa52 1167 /* co2 & si2 are read from SIMD Coefficient pointer */
emilmont 1:fdd22bb7aa52 1168 C2 = _SIMD32_OFFSET(pCoef16 + (4u * ic));
emilmont 1:fdd22bb7aa52 1169
emilmont 1:fdd22bb7aa52 1170 #ifndef ARM_MATH_BIG_ENDIAN
emilmont 1:fdd22bb7aa52 1171
emilmont 1:fdd22bb7aa52 1172 /* xc' = (xa-xb+xc-xd)* co2 + (ya-yb+yc-yd)* (si2) */
emilmont 1:fdd22bb7aa52 1173 out1 = __SMUSD(C2, R) >> 16u;
emilmont 1:fdd22bb7aa52 1174 /* yc' = (ya-yb+yc-yd)* co2 - (xa-xb+xc-xd)* (si2) */
emilmont 1:fdd22bb7aa52 1175 out2 = __SMUADX(C2, R);
emilmont 1:fdd22bb7aa52 1176
emilmont 1:fdd22bb7aa52 1177 #else
emilmont 1:fdd22bb7aa52 1178
emilmont 1:fdd22bb7aa52 1179 /* xc' = (ya-yb+yc-yd)* co2 - (xa-xb+xc-xd)* (si2) */
emilmont 1:fdd22bb7aa52 1180 out1 = __SMUADX(C2, R) >> 16u;
emilmont 1:fdd22bb7aa52 1181 /* yc' = (xa-xb+xc-xd)* co2 + (ya-yb+yc-yd)* (si2) */
emilmont 1:fdd22bb7aa52 1182 out2 = __SMUSD(__QSUB16(0, C2), R);
emilmont 1:fdd22bb7aa52 1183
emilmont 1:fdd22bb7aa52 1184 #endif /* #ifndef ARM_MATH_BIG_ENDIAN */
emilmont 1:fdd22bb7aa52 1185
emilmont 1:fdd22bb7aa52 1186 /* Reading i0+fftLen/4 */
emilmont 1:fdd22bb7aa52 1187 /* T = packed(yb, xb) */
emilmont 1:fdd22bb7aa52 1188 T = _SIMD32_OFFSET(pSrc16 + (2u * i1));
emilmont 1:fdd22bb7aa52 1189 in = ((int16_t) (T & 0xFFFF)) >> 2;
emilmont 1:fdd22bb7aa52 1190 T = ((T >> 2) & 0xFFFF0000) | (in & 0xFFFF);
emilmont 1:fdd22bb7aa52 1191
emilmont 1:fdd22bb7aa52 1192 /* writing the butterfly processed i0 + fftLen/4 sample */
emilmont 1:fdd22bb7aa52 1193 /* writing output(xc', yc') in little endian format */
emilmont 1:fdd22bb7aa52 1194 _SIMD32_OFFSET(pSrc16 + (2u * i1)) =
emilmont 1:fdd22bb7aa52 1195 (q31_t) ((out2) & 0xFFFF0000) | (out1 & 0x0000FFFF);
emilmont 1:fdd22bb7aa52 1196
emilmont 1:fdd22bb7aa52 1197 /* Butterfly calculations */
emilmont 1:fdd22bb7aa52 1198 /* U = packed(yd, xd) */
emilmont 1:fdd22bb7aa52 1199 U = _SIMD32_OFFSET(pSrc16 + (2u * i3));
emilmont 1:fdd22bb7aa52 1200 in = ((int16_t) (U & 0xFFFF)) >> 2;
emilmont 1:fdd22bb7aa52 1201 U = ((U >> 2) & 0xFFFF0000) | (in & 0xFFFF);
emilmont 1:fdd22bb7aa52 1202
emilmont 1:fdd22bb7aa52 1203 /* T = packed(yb-yd, xb-xd) */
emilmont 1:fdd22bb7aa52 1204 T = __QSUB16(T, U);
emilmont 1:fdd22bb7aa52 1205
emilmont 1:fdd22bb7aa52 1206 #ifndef ARM_MATH_BIG_ENDIAN
emilmont 1:fdd22bb7aa52 1207
emilmont 1:fdd22bb7aa52 1208 /* R = packed((ya-yc) + (xb- xd) , (xa-xc) - (yb-yd)) */
emilmont 1:fdd22bb7aa52 1209 R = __QSAX(S, T);
emilmont 1:fdd22bb7aa52 1210 /* S = packed((ya-yc) + (xb- xd), (xa-xc) - (yb-yd)) */
emilmont 1:fdd22bb7aa52 1211 S = __QASX(S, T);
emilmont 1:fdd22bb7aa52 1212
emilmont 1:fdd22bb7aa52 1213 #else
emilmont 1:fdd22bb7aa52 1214
emilmont 1:fdd22bb7aa52 1215 /* R = packed((ya-yc) + (xb- xd) , (xa-xc) - (yb-yd)) */
emilmont 1:fdd22bb7aa52 1216 R = __QASX(S, T);
emilmont 1:fdd22bb7aa52 1217 /* S = packed((ya-yc) - (xb- xd), (xa-xc) + (yb-yd)) */
emilmont 1:fdd22bb7aa52 1218 S = __QSAX(S, T);
emilmont 1:fdd22bb7aa52 1219
emilmont 1:fdd22bb7aa52 1220 #endif /* #ifndef ARM_MATH_BIG_ENDIAN */
emilmont 1:fdd22bb7aa52 1221
emilmont 1:fdd22bb7aa52 1222 /* co1 & si1 are read from SIMD Coefficient pointer */
emilmont 1:fdd22bb7aa52 1223 C1 = _SIMD32_OFFSET(pCoef16 + (2u * ic));
emilmont 1:fdd22bb7aa52 1224 /* Butterfly process for the i0+fftLen/2 sample */
emilmont 1:fdd22bb7aa52 1225
emilmont 1:fdd22bb7aa52 1226 #ifndef ARM_MATH_BIG_ENDIAN
emilmont 1:fdd22bb7aa52 1227
emilmont 1:fdd22bb7aa52 1228 /* xb' = (xa+yb-xc-yd)* co1 + (ya-xb-yc+xd)* (si1) */
emilmont 1:fdd22bb7aa52 1229 out1 = __SMUSD(C1, S) >> 16u;
emilmont 1:fdd22bb7aa52 1230 /* yb' = (ya-xb-yc+xd)* co1 - (xa+yb-xc-yd)* (si1) */
emilmont 1:fdd22bb7aa52 1231 out2 = __SMUADX(C1, S);
emilmont 1:fdd22bb7aa52 1232
emilmont 1:fdd22bb7aa52 1233 #else
emilmont 1:fdd22bb7aa52 1234
emilmont 1:fdd22bb7aa52 1235 /* xb' = (ya-xb-yc+xd)* co1 - (xa+yb-xc-yd)* (si1) */
emilmont 1:fdd22bb7aa52 1236 out1 = __SMUADX(C1, S) >> 16u;
emilmont 1:fdd22bb7aa52 1237 /* yb' = (xa+yb-xc-yd)* co1 + (ya-xb-yc+xd)* (si1) */
emilmont 1:fdd22bb7aa52 1238 out2 = __SMUSD(__QSUB16(0, C1), S);
emilmont 1:fdd22bb7aa52 1239
emilmont 1:fdd22bb7aa52 1240 #endif /* #ifndef ARM_MATH_BIG_ENDIAN */
emilmont 1:fdd22bb7aa52 1241
emilmont 1:fdd22bb7aa52 1242 /* writing output(xb', yb') in little endian format */
emilmont 1:fdd22bb7aa52 1243 _SIMD32_OFFSET(pSrc16 + (2u * i2)) =
emilmont 1:fdd22bb7aa52 1244 ((out2) & 0xFFFF0000) | ((out1) & 0x0000FFFF);
emilmont 1:fdd22bb7aa52 1245
emilmont 1:fdd22bb7aa52 1246
emilmont 1:fdd22bb7aa52 1247 /* co3 & si3 are read from SIMD Coefficient pointer */
emilmont 1:fdd22bb7aa52 1248 C3 = _SIMD32_OFFSET(pCoef16 + (6u * ic));
emilmont 1:fdd22bb7aa52 1249 /* Butterfly process for the i0+3fftLen/4 sample */
emilmont 1:fdd22bb7aa52 1250
emilmont 1:fdd22bb7aa52 1251 #ifndef ARM_MATH_BIG_ENDIAN
emilmont 1:fdd22bb7aa52 1252
emilmont 1:fdd22bb7aa52 1253 /* xd' = (xa-yb-xc+yd)* co3 + (ya+xb-yc-xd)* (si3) */
emilmont 1:fdd22bb7aa52 1254 out1 = __SMUSD(C3, R) >> 16u;
emilmont 1:fdd22bb7aa52 1255 /* yd' = (ya+xb-yc-xd)* co3 - (xa-yb-xc+yd)* (si3) */
emilmont 1:fdd22bb7aa52 1256 out2 = __SMUADX(C3, R);
emilmont 1:fdd22bb7aa52 1257
emilmont 1:fdd22bb7aa52 1258 #else
emilmont 1:fdd22bb7aa52 1259
emilmont 1:fdd22bb7aa52 1260 /* xd' = (ya+xb-yc-xd)* co3 - (xa-yb-xc+yd)* (si3) */
emilmont 1:fdd22bb7aa52 1261 out1 = __SMUADX(C3, R) >> 16u;
emilmont 1:fdd22bb7aa52 1262 /* yd' = (xa-yb-xc+yd)* co3 + (ya+xb-yc-xd)* (si3) */
emilmont 1:fdd22bb7aa52 1263 out2 = __SMUSD(__QSUB16(0, C3), R);
emilmont 1:fdd22bb7aa52 1264
emilmont 1:fdd22bb7aa52 1265 #endif /* #ifndef ARM_MATH_BIG_ENDIAN */
emilmont 1:fdd22bb7aa52 1266
emilmont 1:fdd22bb7aa52 1267 /* writing output(xd', yd') in little endian format */
emilmont 1:fdd22bb7aa52 1268 _SIMD32_OFFSET(pSrc16 + (2u * i3)) =
emilmont 1:fdd22bb7aa52 1269 ((out2) & 0xFFFF0000) | (out1 & 0x0000FFFF);
emilmont 1:fdd22bb7aa52 1270
emilmont 1:fdd22bb7aa52 1271 /* Twiddle coefficients index modifier */
emilmont 1:fdd22bb7aa52 1272 ic = ic + twidCoefModifier;
emilmont 1:fdd22bb7aa52 1273
emilmont 1:fdd22bb7aa52 1274 /* Updating input index */
emilmont 1:fdd22bb7aa52 1275 i0 = i0 + 1u;
emilmont 1:fdd22bb7aa52 1276
emilmont 1:fdd22bb7aa52 1277 } while(--j);
emilmont 1:fdd22bb7aa52 1278 /* data is in 4.11(q11) format */
emilmont 1:fdd22bb7aa52 1279
emilmont 1:fdd22bb7aa52 1280 /* end of first stage process */
emilmont 1:fdd22bb7aa52 1281
emilmont 1:fdd22bb7aa52 1282
emilmont 1:fdd22bb7aa52 1283 /* start of middle stage process */
emilmont 1:fdd22bb7aa52 1284
emilmont 1:fdd22bb7aa52 1285 /* Twiddle coefficients index modifier */
emilmont 1:fdd22bb7aa52 1286 twidCoefModifier <<= 2u;
emilmont 1:fdd22bb7aa52 1287
emilmont 1:fdd22bb7aa52 1288 /* Calculation of Middle stage */
emilmont 1:fdd22bb7aa52 1289 for (k = fftLen / 4u; k > 4u; k >>= 2u)
emilmont 1:fdd22bb7aa52 1290 {
emilmont 1:fdd22bb7aa52 1291 /* Initializations for the middle stage */
emilmont 1:fdd22bb7aa52 1292 n1 = n2;
emilmont 1:fdd22bb7aa52 1293 n2 >>= 2u;
emilmont 1:fdd22bb7aa52 1294 ic = 0u;
emilmont 1:fdd22bb7aa52 1295
emilmont 1:fdd22bb7aa52 1296 for (j = 0u; j <= (n2 - 1u); j++)
emilmont 1:fdd22bb7aa52 1297 {
emilmont 1:fdd22bb7aa52 1298 /* index calculation for the coefficients */
emilmont 1:fdd22bb7aa52 1299 C1 = _SIMD32_OFFSET(pCoef16 + (2u * ic));
emilmont 1:fdd22bb7aa52 1300 C2 = _SIMD32_OFFSET(pCoef16 + (4u * ic));
emilmont 1:fdd22bb7aa52 1301 C3 = _SIMD32_OFFSET(pCoef16 + (6u * ic));
emilmont 1:fdd22bb7aa52 1302
emilmont 1:fdd22bb7aa52 1303 /* Twiddle coefficients index modifier */
emilmont 1:fdd22bb7aa52 1304 ic = ic + twidCoefModifier;
emilmont 1:fdd22bb7aa52 1305
emilmont 1:fdd22bb7aa52 1306 /* Butterfly implementation */
emilmont 1:fdd22bb7aa52 1307 for (i0 = j; i0 < fftLen; i0 += n1)
emilmont 1:fdd22bb7aa52 1308 {
emilmont 1:fdd22bb7aa52 1309 /* index calculation for the input as, */
emilmont 1:fdd22bb7aa52 1310 /* pSrc16[i0 + 0], pSrc16[i0 + fftLen/4], pSrc16[i0 + fftLen/2], pSrc16[i0 + 3fftLen/4] */
emilmont 1:fdd22bb7aa52 1311 i1 = i0 + n2;
emilmont 1:fdd22bb7aa52 1312 i2 = i1 + n2;
emilmont 1:fdd22bb7aa52 1313 i3 = i2 + n2;
emilmont 1:fdd22bb7aa52 1314
emilmont 1:fdd22bb7aa52 1315 /* Reading i0, i0+fftLen/2 inputs */
emilmont 1:fdd22bb7aa52 1316 /* Read ya (real), xa(imag) input */
emilmont 1:fdd22bb7aa52 1317 T = _SIMD32_OFFSET(pSrc16 + (2u * i0));
emilmont 1:fdd22bb7aa52 1318
emilmont 1:fdd22bb7aa52 1319 /* Read yc (real), xc(imag) input */
emilmont 1:fdd22bb7aa52 1320 S = _SIMD32_OFFSET(pSrc16 + (2u * i2));
emilmont 1:fdd22bb7aa52 1321
emilmont 1:fdd22bb7aa52 1322 /* R = packed( (ya + yc), (xa + xc)) */
emilmont 1:fdd22bb7aa52 1323 R = __QADD16(T, S);
emilmont 1:fdd22bb7aa52 1324
emilmont 1:fdd22bb7aa52 1325 /* S = packed((ya - yc), (xa - xc)) */
emilmont 1:fdd22bb7aa52 1326 S = __QSUB16(T, S);
emilmont 1:fdd22bb7aa52 1327
emilmont 1:fdd22bb7aa52 1328 /* Reading i0+fftLen/4 , i0+3fftLen/4 inputs */
emilmont 1:fdd22bb7aa52 1329 /* Read yb (real), xb(imag) input */
emilmont 1:fdd22bb7aa52 1330 T = _SIMD32_OFFSET(pSrc16 + (2u * i1));
emilmont 1:fdd22bb7aa52 1331
emilmont 1:fdd22bb7aa52 1332 /* Read yd (real), xd(imag) input */
emilmont 1:fdd22bb7aa52 1333 U = _SIMD32_OFFSET(pSrc16 + (2u * i3));
emilmont 1:fdd22bb7aa52 1334
emilmont 1:fdd22bb7aa52 1335 /* T = packed( (yb + yd), (xb + xd)) */
emilmont 1:fdd22bb7aa52 1336 T = __QADD16(T, U);
emilmont 1:fdd22bb7aa52 1337
emilmont 1:fdd22bb7aa52 1338 /* writing the butterfly processed i0 sample */
emilmont 1:fdd22bb7aa52 1339
emilmont 1:fdd22bb7aa52 1340 /* xa' = xa + xb + xc + xd */
emilmont 1:fdd22bb7aa52 1341 /* ya' = ya + yb + yc + yd */
emilmont 1:fdd22bb7aa52 1342 out1 = __SHADD16(R, T);
emilmont 1:fdd22bb7aa52 1343 in = ((int16_t) (out1 & 0xFFFF)) >> 1;
emilmont 1:fdd22bb7aa52 1344 out1 = ((out1 >> 1) & 0xFFFF0000) | (in & 0xFFFF);
emilmont 1:fdd22bb7aa52 1345 _SIMD32_OFFSET(pSrc16 + (2u * i0)) = out1;
emilmont 1:fdd22bb7aa52 1346
emilmont 1:fdd22bb7aa52 1347 /* R = packed( (ya + yc) - (yb + yd), (xa + xc) - (xb + xd)) */
emilmont 1:fdd22bb7aa52 1348 R = __SHSUB16(R, T);
emilmont 1:fdd22bb7aa52 1349
emilmont 1:fdd22bb7aa52 1350 #ifndef ARM_MATH_BIG_ENDIAN
emilmont 1:fdd22bb7aa52 1351
emilmont 1:fdd22bb7aa52 1352 /* (ya-yb+yc-yd)* (si2) + (xa-xb+xc-xd)* co2 */
emilmont 1:fdd22bb7aa52 1353 out1 = __SMUSD(C2, R) >> 16u;
emilmont 1:fdd22bb7aa52 1354
emilmont 1:fdd22bb7aa52 1355 /* (ya-yb+yc-yd)* co2 - (xa-xb+xc-xd)* (si2) */
emilmont 1:fdd22bb7aa52 1356 out2 = __SMUADX(C2, R);
emilmont 1:fdd22bb7aa52 1357
emilmont 1:fdd22bb7aa52 1358 #else
emilmont 1:fdd22bb7aa52 1359
emilmont 1:fdd22bb7aa52 1360 /* (ya-yb+yc-yd)* co2 - (xa-xb+xc-xd)* (si2) */
emilmont 1:fdd22bb7aa52 1361 out1 = __SMUADX(R, C2) >> 16u;
emilmont 1:fdd22bb7aa52 1362
emilmont 1:fdd22bb7aa52 1363 /* (ya-yb+yc-yd)* (si2) + (xa-xb+xc-xd)* co2 */
emilmont 1:fdd22bb7aa52 1364 out2 = __SMUSD(__QSUB16(0, C2), R);
emilmont 1:fdd22bb7aa52 1365
emilmont 1:fdd22bb7aa52 1366 #endif /* #ifndef ARM_MATH_BIG_ENDIAN */
emilmont 1:fdd22bb7aa52 1367
emilmont 1:fdd22bb7aa52 1368 /* Reading i0+3fftLen/4 */
emilmont 1:fdd22bb7aa52 1369 /* Read yb (real), xb(imag) input */
emilmont 1:fdd22bb7aa52 1370 T = _SIMD32_OFFSET(pSrc16 + (2u * i1));
emilmont 1:fdd22bb7aa52 1371
emilmont 1:fdd22bb7aa52 1372 /* writing the butterfly processed i0 + fftLen/4 sample */
emilmont 1:fdd22bb7aa52 1373 /* xc' = (xa-xb+xc-xd)* co2 + (ya-yb+yc-yd)* (si2) */
emilmont 1:fdd22bb7aa52 1374 /* yc' = (ya-yb+yc-yd)* co2 - (xa-xb+xc-xd)* (si2) */
emilmont 1:fdd22bb7aa52 1375 _SIMD32_OFFSET(pSrc16 + (2u * i1)) =
emilmont 1:fdd22bb7aa52 1376 ((out2) & 0xFFFF0000) | (out1 & 0x0000FFFF);
emilmont 1:fdd22bb7aa52 1377
emilmont 1:fdd22bb7aa52 1378 /* Butterfly calculations */
emilmont 1:fdd22bb7aa52 1379
emilmont 1:fdd22bb7aa52 1380 /* Read yd (real), xd(imag) input */
emilmont 1:fdd22bb7aa52 1381 U = _SIMD32_OFFSET(pSrc16 + (2u * i3));
emilmont 1:fdd22bb7aa52 1382
emilmont 1:fdd22bb7aa52 1383 /* T = packed(yb-yd, xb-xd) */
emilmont 1:fdd22bb7aa52 1384 T = __QSUB16(T, U);
emilmont 1:fdd22bb7aa52 1385
emilmont 1:fdd22bb7aa52 1386 #ifndef ARM_MATH_BIG_ENDIAN
emilmont 1:fdd22bb7aa52 1387
emilmont 1:fdd22bb7aa52 1388 /* R = packed((ya-yc) + (xb- xd) , (xa-xc) - (yb-yd)) */
emilmont 1:fdd22bb7aa52 1389 R = __SHSAX(S, T);
emilmont 1:fdd22bb7aa52 1390
emilmont 1:fdd22bb7aa52 1391 /* S = packed((ya-yc) - (xb- xd), (xa-xc) + (yb-yd)) */
emilmont 1:fdd22bb7aa52 1392 S = __SHASX(S, T);
emilmont 1:fdd22bb7aa52 1393
emilmont 1:fdd22bb7aa52 1394
emilmont 1:fdd22bb7aa52 1395 /* Butterfly process for the i0+fftLen/2 sample */
emilmont 1:fdd22bb7aa52 1396 out1 = __SMUSD(C1, S) >> 16u;
emilmont 1:fdd22bb7aa52 1397 out2 = __SMUADX(C1, S);
emilmont 1:fdd22bb7aa52 1398
emilmont 1:fdd22bb7aa52 1399 #else
emilmont 1:fdd22bb7aa52 1400
emilmont 1:fdd22bb7aa52 1401 /* R = packed((ya-yc) + (xb- xd) , (xa-xc) - (yb-yd)) */
emilmont 1:fdd22bb7aa52 1402 R = __SHASX(S, T);
emilmont 1:fdd22bb7aa52 1403
emilmont 1:fdd22bb7aa52 1404 /* S = packed((ya-yc) - (xb- xd), (xa-xc) + (yb-yd)) */
emilmont 1:fdd22bb7aa52 1405 S = __SHSAX(S, T);
emilmont 1:fdd22bb7aa52 1406
emilmont 1:fdd22bb7aa52 1407
emilmont 1:fdd22bb7aa52 1408 /* Butterfly process for the i0+fftLen/2 sample */
emilmont 1:fdd22bb7aa52 1409 out1 = __SMUADX(S, C1) >> 16u;
emilmont 1:fdd22bb7aa52 1410 out2 = __SMUSD(__QSUB16(0, C1), S);
emilmont 1:fdd22bb7aa52 1411
emilmont 1:fdd22bb7aa52 1412 #endif /* #ifndef ARM_MATH_BIG_ENDIAN */
emilmont 1:fdd22bb7aa52 1413
emilmont 1:fdd22bb7aa52 1414 /* xb' = (xa+yb-xc-yd)* co1 + (ya-xb-yc+xd)* (si1) */
emilmont 1:fdd22bb7aa52 1415 /* yb' = (ya-xb-yc+xd)* co1 - (xa+yb-xc-yd)* (si1) */
emilmont 1:fdd22bb7aa52 1416 _SIMD32_OFFSET(pSrc16 + (2u * i2)) =
emilmont 1:fdd22bb7aa52 1417 ((out2) & 0xFFFF0000) | (out1 & 0x0000FFFF);
emilmont 1:fdd22bb7aa52 1418
emilmont 1:fdd22bb7aa52 1419 /* Butterfly process for the i0+3fftLen/4 sample */
emilmont 1:fdd22bb7aa52 1420
emilmont 1:fdd22bb7aa52 1421 #ifndef ARM_MATH_BIG_ENDIAN
emilmont 1:fdd22bb7aa52 1422
emilmont 1:fdd22bb7aa52 1423 out1 = __SMUSD(C3, R) >> 16u;
emilmont 1:fdd22bb7aa52 1424 out2 = __SMUADX(C3, R);
emilmont 1:fdd22bb7aa52 1425
emilmont 1:fdd22bb7aa52 1426 #else
emilmont 1:fdd22bb7aa52 1427
emilmont 1:fdd22bb7aa52 1428 out1 = __SMUADX(C3, R) >> 16u;
emilmont 1:fdd22bb7aa52 1429 out2 = __SMUSD(__QSUB16(0, C3), R);
emilmont 1:fdd22bb7aa52 1430
emilmont 1:fdd22bb7aa52 1431 #endif /* #ifndef ARM_MATH_BIG_ENDIAN */
emilmont 1:fdd22bb7aa52 1432
emilmont 1:fdd22bb7aa52 1433 /* xd' = (xa-yb-xc+yd)* co3 + (ya+xb-yc-xd)* (si3) */
emilmont 1:fdd22bb7aa52 1434 /* yd' = (ya+xb-yc-xd)* co3 - (xa-yb-xc+yd)* (si3) */
emilmont 1:fdd22bb7aa52 1435 _SIMD32_OFFSET(pSrc16 + (2u * i3)) =
emilmont 1:fdd22bb7aa52 1436 ((out2) & 0xFFFF0000) | (out1 & 0x0000FFFF);
emilmont 1:fdd22bb7aa52 1437 }
emilmont 1:fdd22bb7aa52 1438 }
emilmont 1:fdd22bb7aa52 1439 /* Twiddle coefficients index modifier */
emilmont 1:fdd22bb7aa52 1440 twidCoefModifier <<= 2u;
emilmont 1:fdd22bb7aa52 1441 }
emilmont 1:fdd22bb7aa52 1442 /* end of middle stage process */
emilmont 1:fdd22bb7aa52 1443
emilmont 1:fdd22bb7aa52 1444 /* data is in 10.6(q6) format for the 1024 point */
emilmont 1:fdd22bb7aa52 1445 /* data is in 8.8(q8) format for the 256 point */
emilmont 1:fdd22bb7aa52 1446 /* data is in 6.10(q10) format for the 64 point */
emilmont 1:fdd22bb7aa52 1447 /* data is in 4.12(q12) format for the 16 point */
emilmont 1:fdd22bb7aa52 1448
emilmont 1:fdd22bb7aa52 1449 /* Initializations for the last stage */
emilmont 1:fdd22bb7aa52 1450 j = fftLen >> 2;
emilmont 1:fdd22bb7aa52 1451
emilmont 1:fdd22bb7aa52 1452 ptr1 = &pSrc16[0];
emilmont 1:fdd22bb7aa52 1453
emilmont 1:fdd22bb7aa52 1454 /* start of last stage process */
emilmont 1:fdd22bb7aa52 1455
emilmont 1:fdd22bb7aa52 1456 /* Butterfly implementation */
emilmont 1:fdd22bb7aa52 1457 do
emilmont 1:fdd22bb7aa52 1458 {
emilmont 1:fdd22bb7aa52 1459 /* Read xa (real), ya(imag) input */
emilmont 1:fdd22bb7aa52 1460 xaya = *__SIMD32(ptr1)++;
emilmont 1:fdd22bb7aa52 1461
emilmont 1:fdd22bb7aa52 1462 /* Read xb (real), yb(imag) input */
emilmont 1:fdd22bb7aa52 1463 xbyb = *__SIMD32(ptr1)++;
emilmont 1:fdd22bb7aa52 1464
emilmont 1:fdd22bb7aa52 1465 /* Read xc (real), yc(imag) input */
emilmont 1:fdd22bb7aa52 1466 xcyc = *__SIMD32(ptr1)++;
emilmont 1:fdd22bb7aa52 1467
emilmont 1:fdd22bb7aa52 1468 /* Read xd (real), yd(imag) input */
emilmont 1:fdd22bb7aa52 1469 xdyd = *__SIMD32(ptr1)++;
emilmont 1:fdd22bb7aa52 1470
emilmont 1:fdd22bb7aa52 1471 /* R = packed((ya + yc), (xa + xc)) */
emilmont 1:fdd22bb7aa52 1472 R = __QADD16(xaya, xcyc);
emilmont 1:fdd22bb7aa52 1473
emilmont 1:fdd22bb7aa52 1474 /* T = packed((yb + yd), (xb + xd)) */
emilmont 1:fdd22bb7aa52 1475 T = __QADD16(xbyb, xdyd);
emilmont 1:fdd22bb7aa52 1476
emilmont 1:fdd22bb7aa52 1477 /* pointer updation for writing */
emilmont 1:fdd22bb7aa52 1478 ptr1 = ptr1 - 8u;
emilmont 1:fdd22bb7aa52 1479
emilmont 1:fdd22bb7aa52 1480
emilmont 1:fdd22bb7aa52 1481 /* xa' = xa + xb + xc + xd */
emilmont 1:fdd22bb7aa52 1482 /* ya' = ya + yb + yc + yd */
emilmont 1:fdd22bb7aa52 1483 *__SIMD32(ptr1)++ = __SHADD16(R, T);
emilmont 1:fdd22bb7aa52 1484
emilmont 1:fdd22bb7aa52 1485 /* T = packed((yb + yd), (xb + xd)) */
emilmont 1:fdd22bb7aa52 1486 T = __QADD16(xbyb, xdyd);
emilmont 1:fdd22bb7aa52 1487
emilmont 1:fdd22bb7aa52 1488 /* xc' = (xa-xb+xc-xd) */
emilmont 1:fdd22bb7aa52 1489 /* yc' = (ya-yb+yc-yd) */
emilmont 1:fdd22bb7aa52 1490 *__SIMD32(ptr1)++ = __SHSUB16(R, T);
emilmont 1:fdd22bb7aa52 1491
emilmont 1:fdd22bb7aa52 1492 /* S = packed((ya - yc), (xa - xc)) */
emilmont 1:fdd22bb7aa52 1493 S = __QSUB16(xaya, xcyc);
emilmont 1:fdd22bb7aa52 1494
emilmont 1:fdd22bb7aa52 1495 /* Read yd (real), xd(imag) input */
emilmont 1:fdd22bb7aa52 1496 /* T = packed( (yb - yd), (xb - xd)) */
emilmont 1:fdd22bb7aa52 1497 U = __QSUB16(xbyb, xdyd);
emilmont 1:fdd22bb7aa52 1498
emilmont 1:fdd22bb7aa52 1499 #ifndef ARM_MATH_BIG_ENDIAN
emilmont 1:fdd22bb7aa52 1500
emilmont 1:fdd22bb7aa52 1501 /* xb' = (xa+yb-xc-yd) */
emilmont 1:fdd22bb7aa52 1502 /* yb' = (ya-xb-yc+xd) */
emilmont 1:fdd22bb7aa52 1503 *__SIMD32(ptr1)++ = __SHASX(S, U);
emilmont 1:fdd22bb7aa52 1504
emilmont 1:fdd22bb7aa52 1505
emilmont 1:fdd22bb7aa52 1506 /* xd' = (xa-yb-xc+yd) */
emilmont 1:fdd22bb7aa52 1507 /* yd' = (ya+xb-yc-xd) */
emilmont 1:fdd22bb7aa52 1508 *__SIMD32(ptr1)++ = __SHSAX(S, U);
emilmont 1:fdd22bb7aa52 1509
emilmont 1:fdd22bb7aa52 1510 #else
emilmont 1:fdd22bb7aa52 1511
emilmont 1:fdd22bb7aa52 1512 /* xb' = (xa+yb-xc-yd) */
emilmont 1:fdd22bb7aa52 1513 /* yb' = (ya-xb-yc+xd) */
emilmont 1:fdd22bb7aa52 1514 *__SIMD32(ptr1)++ = __SHSAX(S, U);
emilmont 1:fdd22bb7aa52 1515
emilmont 1:fdd22bb7aa52 1516
emilmont 1:fdd22bb7aa52 1517 /* xd' = (xa-yb-xc+yd) */
emilmont 1:fdd22bb7aa52 1518 /* yd' = (ya+xb-yc-xd) */
emilmont 1:fdd22bb7aa52 1519 *__SIMD32(ptr1)++ = __SHASX(S, U);
emilmont 1:fdd22bb7aa52 1520
emilmont 1:fdd22bb7aa52 1521
emilmont 1:fdd22bb7aa52 1522 #endif /* #ifndef ARM_MATH_BIG_ENDIAN */
emilmont 1:fdd22bb7aa52 1523
emilmont 1:fdd22bb7aa52 1524 } while(--j);
emilmont 1:fdd22bb7aa52 1525
emilmont 1:fdd22bb7aa52 1526 /* end of last stage process */
emilmont 1:fdd22bb7aa52 1527
emilmont 1:fdd22bb7aa52 1528 /* output is in 11.5(q5) format for the 1024 point */
emilmont 1:fdd22bb7aa52 1529 /* output is in 9.7(q7) format for the 256 point */
emilmont 1:fdd22bb7aa52 1530 /* output is in 7.9(q9) format for the 64 point */
emilmont 1:fdd22bb7aa52 1531 /* output is in 5.11(q11) format for the 16 point */
emilmont 1:fdd22bb7aa52 1532
emilmont 1:fdd22bb7aa52 1533
emilmont 1:fdd22bb7aa52 1534 #else
emilmont 1:fdd22bb7aa52 1535
emilmont 1:fdd22bb7aa52 1536 /* Run the below code for Cortex-M0 */
emilmont 1:fdd22bb7aa52 1537
emilmont 1:fdd22bb7aa52 1538 q15_t R0, R1, S0, S1, T0, T1, U0, U1;
emilmont 1:fdd22bb7aa52 1539 q15_t Co1, Si1, Co2, Si2, Co3, Si3, out1, out2;
emilmont 1:fdd22bb7aa52 1540 uint32_t n1, n2, ic, i0, i1, i2, i3, j, k;
emilmont 1:fdd22bb7aa52 1541
emilmont 1:fdd22bb7aa52 1542 /* Total process is divided into three stages */
emilmont 1:fdd22bb7aa52 1543
emilmont 1:fdd22bb7aa52 1544 /* process first stage, middle stages, & last stage */
emilmont 1:fdd22bb7aa52 1545
emilmont 1:fdd22bb7aa52 1546 /* Initializations for the first stage */
emilmont 1:fdd22bb7aa52 1547 n2 = fftLen;
emilmont 1:fdd22bb7aa52 1548 n1 = n2;
emilmont 1:fdd22bb7aa52 1549
emilmont 1:fdd22bb7aa52 1550 /* n2 = fftLen/4 */
emilmont 1:fdd22bb7aa52 1551 n2 >>= 2u;
emilmont 1:fdd22bb7aa52 1552
emilmont 1:fdd22bb7aa52 1553 /* Index for twiddle coefficient */
emilmont 1:fdd22bb7aa52 1554 ic = 0u;
emilmont 1:fdd22bb7aa52 1555
emilmont 1:fdd22bb7aa52 1556 /* Index for input read and output write */
emilmont 1:fdd22bb7aa52 1557 i0 = 0u;
emilmont 1:fdd22bb7aa52 1558
emilmont 1:fdd22bb7aa52 1559 j = n2;
emilmont 1:fdd22bb7aa52 1560
emilmont 1:fdd22bb7aa52 1561 /* Input is in 1.15(q15) format */
emilmont 1:fdd22bb7aa52 1562
emilmont 1:fdd22bb7aa52 1563 /* Start of first stage process */
emilmont 1:fdd22bb7aa52 1564 do
emilmont 1:fdd22bb7aa52 1565 {
emilmont 1:fdd22bb7aa52 1566 /* Butterfly implementation */
emilmont 1:fdd22bb7aa52 1567
emilmont 1:fdd22bb7aa52 1568 /* index calculation for the input as, */
emilmont 1:fdd22bb7aa52 1569 /* pSrc16[i0 + 0], pSrc16[i0 + fftLen/4], pSrc16[i0 + fftLen/2], pSrc16[i0 + 3fftLen/4] */
emilmont 1:fdd22bb7aa52 1570 i1 = i0 + n2;
emilmont 1:fdd22bb7aa52 1571 i2 = i1 + n2;
emilmont 1:fdd22bb7aa52 1572 i3 = i2 + n2;
emilmont 1:fdd22bb7aa52 1573
emilmont 1:fdd22bb7aa52 1574 /* Reading i0, i0+fftLen/2 inputs */
emilmont 1:fdd22bb7aa52 1575 /* input is down scale by 4 to avoid overflow */
emilmont 1:fdd22bb7aa52 1576 /* Read ya (real), xa(imag) input */
emilmont 1:fdd22bb7aa52 1577 T0 = pSrc16[i0 * 2u] >> 2u;
emilmont 1:fdd22bb7aa52 1578 T1 = pSrc16[(i0 * 2u) + 1u] >> 2u;
emilmont 1:fdd22bb7aa52 1579 /* input is down scale by 4 to avoid overflow */
emilmont 1:fdd22bb7aa52 1580 /* Read yc (real), xc(imag) input */
emilmont 1:fdd22bb7aa52 1581 S0 = pSrc16[i2 * 2u] >> 2u;
emilmont 1:fdd22bb7aa52 1582 S1 = pSrc16[(i2 * 2u) + 1u] >> 2u;
emilmont 1:fdd22bb7aa52 1583
emilmont 1:fdd22bb7aa52 1584 /* R0 = (ya + yc), R1 = (xa + xc) */
emilmont 1:fdd22bb7aa52 1585 R0 = __SSAT(T0 + S0, 16u);
emilmont 1:fdd22bb7aa52 1586 R1 = __SSAT(T1 + S1, 16u);
emilmont 1:fdd22bb7aa52 1587 /* S0 = (ya - yc), S1 = (xa - xc) */
emilmont 1:fdd22bb7aa52 1588 S0 = __SSAT(T0 - S0, 16u);
emilmont 1:fdd22bb7aa52 1589 S1 = __SSAT(T1 - S1, 16u);
emilmont 1:fdd22bb7aa52 1590
emilmont 1:fdd22bb7aa52 1591 /* Reading i0+fftLen/4 , i0+3fftLen/4 inputs */
emilmont 1:fdd22bb7aa52 1592 /* input is down scale by 4 to avoid overflow */
emilmont 1:fdd22bb7aa52 1593 /* Read yb (real), xb(imag) input */
emilmont 1:fdd22bb7aa52 1594 T0 = pSrc16[i1 * 2u] >> 2u;
emilmont 1:fdd22bb7aa52 1595 T1 = pSrc16[(i1 * 2u) + 1u] >> 2u;
emilmont 1:fdd22bb7aa52 1596 /* Read yd (real), xd(imag) input */
emilmont 1:fdd22bb7aa52 1597 /* input is down scale by 4 to avoid overflow */
emilmont 1:fdd22bb7aa52 1598 U0 = pSrc16[i3 * 2u] >> 2u;
emilmont 1:fdd22bb7aa52 1599 U1 = pSrc16[(i3 * 2u) + 1u] >> 2u;
emilmont 1:fdd22bb7aa52 1600
emilmont 1:fdd22bb7aa52 1601 /* T0 = (yb + yd), T1 = (xb + xd) */
emilmont 1:fdd22bb7aa52 1602 T0 = __SSAT(T0 + U0, 16u);
emilmont 1:fdd22bb7aa52 1603 T1 = __SSAT(T1 + U1, 16u);
emilmont 1:fdd22bb7aa52 1604
emilmont 1:fdd22bb7aa52 1605 /* writing the butterfly processed i0 sample */
emilmont 1:fdd22bb7aa52 1606 /* xa' = xa + xb + xc + xd */
emilmont 1:fdd22bb7aa52 1607 /* ya' = ya + yb + yc + yd */
emilmont 1:fdd22bb7aa52 1608 pSrc16[i0 * 2u] = (R0 >> 1u) + (T0 >> 1u);
emilmont 1:fdd22bb7aa52 1609 pSrc16[(i0 * 2u) + 1u] = (R1 >> 1u) + (T1 >> 1u);
emilmont 1:fdd22bb7aa52 1610
emilmont 1:fdd22bb7aa52 1611 /* R0 = (ya + yc) - (yb + yd), R1 = (xa + xc)- (xb + xd) */
emilmont 1:fdd22bb7aa52 1612 R0 = __SSAT(R0 - T0, 16u);
emilmont 1:fdd22bb7aa52 1613 R1 = __SSAT(R1 - T1, 16u);
emilmont 1:fdd22bb7aa52 1614 /* co2 & si2 are read from Coefficient pointer */
emilmont 1:fdd22bb7aa52 1615 Co2 = pCoef16[2u * ic * 2u];
emilmont 1:fdd22bb7aa52 1616 Si2 = pCoef16[(2u * ic * 2u) + 1u];
emilmont 1:fdd22bb7aa52 1617 /* xc' = (xa-xb+xc-xd)* co2 - (ya-yb+yc-yd)* (si2) */
emilmont 1:fdd22bb7aa52 1618 out1 = (short) ((Co2 * R0 - Si2 * R1) >> 16u);
emilmont 1:fdd22bb7aa52 1619 /* yc' = (ya-yb+yc-yd)* co2 + (xa-xb+xc-xd)* (si2) */
emilmont 1:fdd22bb7aa52 1620 out2 = (short) ((Si2 * R0 + Co2 * R1) >> 16u);
emilmont 1:fdd22bb7aa52 1621
emilmont 1:fdd22bb7aa52 1622 /* Reading i0+fftLen/4 */
emilmont 1:fdd22bb7aa52 1623 /* input is down scale by 4 to avoid overflow */
emilmont 1:fdd22bb7aa52 1624 /* T0 = yb, T1 = xb */
emilmont 1:fdd22bb7aa52 1625 T0 = pSrc16[i1 * 2u] >> 2u;
emilmont 1:fdd22bb7aa52 1626 T1 = pSrc16[(i1 * 2u) + 1u] >> 2u;
emilmont 1:fdd22bb7aa52 1627
emilmont 1:fdd22bb7aa52 1628 /* writing the butterfly processed i0 + fftLen/4 sample */
emilmont 1:fdd22bb7aa52 1629 /* writing output(xc', yc') in little endian format */
emilmont 1:fdd22bb7aa52 1630 pSrc16[i1 * 2u] = out1;
emilmont 1:fdd22bb7aa52 1631 pSrc16[(i1 * 2u) + 1u] = out2;
emilmont 1:fdd22bb7aa52 1632
emilmont 1:fdd22bb7aa52 1633 /* Butterfly calculations */
emilmont 1:fdd22bb7aa52 1634 /* input is down scale by 4 to avoid overflow */
emilmont 1:fdd22bb7aa52 1635 /* U0 = yd, U1 = xd) */
emilmont 1:fdd22bb7aa52 1636 U0 = pSrc16[i3 * 2u] >> 2u;
emilmont 1:fdd22bb7aa52 1637 U1 = pSrc16[(i3 * 2u) + 1u] >> 2u;
emilmont 1:fdd22bb7aa52 1638
emilmont 1:fdd22bb7aa52 1639 /* T0 = yb-yd, T1 = xb-xd) */
emilmont 1:fdd22bb7aa52 1640 T0 = __SSAT(T0 - U0, 16u);
emilmont 1:fdd22bb7aa52 1641 T1 = __SSAT(T1 - U1, 16u);
emilmont 1:fdd22bb7aa52 1642 /* R0 = (ya-yc) - (xb- xd) , R1 = (xa-xc) + (yb-yd) */
emilmont 1:fdd22bb7aa52 1643 R0 = (short) __SSAT((q31_t) (S0 + T1), 16);
emilmont 1:fdd22bb7aa52 1644 R1 = (short) __SSAT((q31_t) (S1 - T0), 16);
emilmont 1:fdd22bb7aa52 1645 /* S = (ya-yc) + (xb- xd), S1 = (xa-xc) - (yb-yd) */
emilmont 1:fdd22bb7aa52 1646 S0 = (short) __SSAT((q31_t) (S0 - T1), 16);
emilmont 1:fdd22bb7aa52 1647 S1 = (short) __SSAT((q31_t) (S1 + T0), 16);
emilmont 1:fdd22bb7aa52 1648
emilmont 1:fdd22bb7aa52 1649 /* co1 & si1 are read from Coefficient pointer */
emilmont 1:fdd22bb7aa52 1650 Co1 = pCoef16[ic * 2u];
emilmont 1:fdd22bb7aa52 1651 Si1 = pCoef16[(ic * 2u) + 1u];
emilmont 1:fdd22bb7aa52 1652 /* Butterfly process for the i0+fftLen/2 sample */
emilmont 1:fdd22bb7aa52 1653 /* xb' = (xa-yb-xc+yd)* co1 - (ya+xb-yc-xd)* (si1) */
emilmont 1:fdd22bb7aa52 1654 out1 = (short) ((Co1 * S0 - Si1 * S1) >> 16u);
emilmont 1:fdd22bb7aa52 1655 /* yb' = (ya+xb-yc-xd)* co1 + (xa-yb-xc+yd)* (si1) */
emilmont 1:fdd22bb7aa52 1656 out2 = (short) ((Si1 * S0 + Co1 * S1) >> 16u);
emilmont 1:fdd22bb7aa52 1657 /* writing output(xb', yb') in little endian format */
emilmont 1:fdd22bb7aa52 1658 pSrc16[i2 * 2u] = out1;
emilmont 1:fdd22bb7aa52 1659 pSrc16[(i2 * 2u) + 1u] = out2;
emilmont 1:fdd22bb7aa52 1660
emilmont 1:fdd22bb7aa52 1661 /* Co3 & si3 are read from Coefficient pointer */
emilmont 1:fdd22bb7aa52 1662 Co3 = pCoef16[3u * ic * 2u];
emilmont 1:fdd22bb7aa52 1663 Si3 = pCoef16[(3u * ic * 2u) + 1u];
emilmont 1:fdd22bb7aa52 1664 /* Butterfly process for the i0+3fftLen/4 sample */
emilmont 1:fdd22bb7aa52 1665 /* xd' = (xa+yb-xc-yd)* Co3 - (ya-xb-yc+xd)* (si3) */
emilmont 1:fdd22bb7aa52 1666 out1 = (short) ((Co3 * R0 - Si3 * R1) >> 16u);
emilmont 1:fdd22bb7aa52 1667 /* yd' = (ya-xb-yc+xd)* Co3 + (xa+yb-xc-yd)* (si3) */
emilmont 1:fdd22bb7aa52 1668 out2 = (short) ((Si3 * R0 + Co3 * R1) >> 16u);
emilmont 1:fdd22bb7aa52 1669 /* writing output(xd', yd') in little endian format */
emilmont 1:fdd22bb7aa52 1670 pSrc16[i3 * 2u] = out1;
emilmont 1:fdd22bb7aa52 1671 pSrc16[(i3 * 2u) + 1u] = out2;
emilmont 1:fdd22bb7aa52 1672
emilmont 1:fdd22bb7aa52 1673 /* Twiddle coefficients index modifier */
emilmont 1:fdd22bb7aa52 1674 ic = ic + twidCoefModifier;
emilmont 1:fdd22bb7aa52 1675
emilmont 1:fdd22bb7aa52 1676 /* Updating input index */
emilmont 1:fdd22bb7aa52 1677 i0 = i0 + 1u;
emilmont 1:fdd22bb7aa52 1678
emilmont 1:fdd22bb7aa52 1679 } while(--j);
emilmont 1:fdd22bb7aa52 1680
emilmont 1:fdd22bb7aa52 1681 /* End of first stage process */
emilmont 1:fdd22bb7aa52 1682
emilmont 1:fdd22bb7aa52 1683 /* data is in 4.11(q11) format */
emilmont 1:fdd22bb7aa52 1684
emilmont 1:fdd22bb7aa52 1685
emilmont 1:fdd22bb7aa52 1686 /* Start of Middle stage process */
emilmont 1:fdd22bb7aa52 1687
emilmont 1:fdd22bb7aa52 1688 /* Twiddle coefficients index modifier */
emilmont 1:fdd22bb7aa52 1689 twidCoefModifier <<= 2u;
emilmont 1:fdd22bb7aa52 1690
emilmont 1:fdd22bb7aa52 1691 /* Calculation of Middle stage */
emilmont 1:fdd22bb7aa52 1692 for (k = fftLen / 4u; k > 4u; k >>= 2u)
emilmont 1:fdd22bb7aa52 1693 {
emilmont 1:fdd22bb7aa52 1694 /* Initializations for the middle stage */
emilmont 1:fdd22bb7aa52 1695 n1 = n2;
emilmont 1:fdd22bb7aa52 1696 n2 >>= 2u;
emilmont 1:fdd22bb7aa52 1697 ic = 0u;
emilmont 1:fdd22bb7aa52 1698
emilmont 1:fdd22bb7aa52 1699 for (j = 0u; j <= (n2 - 1u); j++)
emilmont 1:fdd22bb7aa52 1700 {
emilmont 1:fdd22bb7aa52 1701 /* index calculation for the coefficients */
emilmont 1:fdd22bb7aa52 1702 Co1 = pCoef16[ic * 2u];
emilmont 1:fdd22bb7aa52 1703 Si1 = pCoef16[(ic * 2u) + 1u];
emilmont 1:fdd22bb7aa52 1704 Co2 = pCoef16[2u * ic * 2u];
emilmont 1:fdd22bb7aa52 1705 Si2 = pCoef16[2u * ic * 2u + 1u];
emilmont 1:fdd22bb7aa52 1706 Co3 = pCoef16[3u * ic * 2u];
emilmont 1:fdd22bb7aa52 1707 Si3 = pCoef16[(3u * ic * 2u) + 1u];
emilmont 1:fdd22bb7aa52 1708
emilmont 1:fdd22bb7aa52 1709 /* Twiddle coefficients index modifier */
emilmont 1:fdd22bb7aa52 1710 ic = ic + twidCoefModifier;
emilmont 1:fdd22bb7aa52 1711
emilmont 1:fdd22bb7aa52 1712 /* Butterfly implementation */
emilmont 1:fdd22bb7aa52 1713 for (i0 = j; i0 < fftLen; i0 += n1)
emilmont 1:fdd22bb7aa52 1714 {
emilmont 1:fdd22bb7aa52 1715 /* index calculation for the input as, */
emilmont 1:fdd22bb7aa52 1716 /* pSrc16[i0 + 0], pSrc16[i0 + fftLen/4], pSrc16[i0 + fftLen/2], pSrc16[i0 + 3fftLen/4] */
emilmont 1:fdd22bb7aa52 1717 i1 = i0 + n2;
emilmont 1:fdd22bb7aa52 1718 i2 = i1 + n2;
emilmont 1:fdd22bb7aa52 1719 i3 = i2 + n2;
emilmont 1:fdd22bb7aa52 1720
emilmont 1:fdd22bb7aa52 1721 /* Reading i0, i0+fftLen/2 inputs */
emilmont 1:fdd22bb7aa52 1722 /* Read ya (real), xa(imag) input */
emilmont 1:fdd22bb7aa52 1723 T0 = pSrc16[i0 * 2u];
emilmont 1:fdd22bb7aa52 1724 T1 = pSrc16[(i0 * 2u) + 1u];
emilmont 1:fdd22bb7aa52 1725
emilmont 1:fdd22bb7aa52 1726 /* Read yc (real), xc(imag) input */
emilmont 1:fdd22bb7aa52 1727 S0 = pSrc16[i2 * 2u];
emilmont 1:fdd22bb7aa52 1728 S1 = pSrc16[(i2 * 2u) + 1u];
emilmont 1:fdd22bb7aa52 1729
emilmont 1:fdd22bb7aa52 1730
emilmont 1:fdd22bb7aa52 1731 /* R0 = (ya + yc), R1 = (xa + xc) */
emilmont 1:fdd22bb7aa52 1732 R0 = __SSAT(T0 + S0, 16u);
emilmont 1:fdd22bb7aa52 1733 R1 = __SSAT(T1 + S1, 16u);
emilmont 1:fdd22bb7aa52 1734 /* S0 = (ya - yc), S1 = (xa - xc) */
emilmont 1:fdd22bb7aa52 1735 S0 = __SSAT(T0 - S0, 16u);
emilmont 1:fdd22bb7aa52 1736 S1 = __SSAT(T1 - S1, 16u);
emilmont 1:fdd22bb7aa52 1737
emilmont 1:fdd22bb7aa52 1738 /* Reading i0+fftLen/4 , i0+3fftLen/4 inputs */
emilmont 1:fdd22bb7aa52 1739 /* Read yb (real), xb(imag) input */
emilmont 1:fdd22bb7aa52 1740 T0 = pSrc16[i1 * 2u];
emilmont 1:fdd22bb7aa52 1741 T1 = pSrc16[(i1 * 2u) + 1u];
emilmont 1:fdd22bb7aa52 1742
emilmont 1:fdd22bb7aa52 1743 /* Read yd (real), xd(imag) input */
emilmont 1:fdd22bb7aa52 1744 U0 = pSrc16[i3 * 2u];
emilmont 1:fdd22bb7aa52 1745 U1 = pSrc16[(i3 * 2u) + 1u];
emilmont 1:fdd22bb7aa52 1746
emilmont 1:fdd22bb7aa52 1747 /* T0 = (yb + yd), T1 = (xb + xd) */
emilmont 1:fdd22bb7aa52 1748 T0 = __SSAT(T0 + U0, 16u);
emilmont 1:fdd22bb7aa52 1749 T1 = __SSAT(T1 + U1, 16u);
emilmont 1:fdd22bb7aa52 1750
emilmont 1:fdd22bb7aa52 1751 /* writing the butterfly processed i0 sample */
emilmont 1:fdd22bb7aa52 1752 /* xa' = xa + xb + xc + xd */
emilmont 1:fdd22bb7aa52 1753 /* ya' = ya + yb + yc + yd */
emilmont 1:fdd22bb7aa52 1754 pSrc16[i0 * 2u] = ((R0 >> 1u) + (T0 >> 1u)) >> 1u;
emilmont 1:fdd22bb7aa52 1755 pSrc16[(i0 * 2u) + 1u] = ((R1 >> 1u) + (T1 >> 1u)) >> 1u;
emilmont 1:fdd22bb7aa52 1756
emilmont 1:fdd22bb7aa52 1757 /* R0 = (ya + yc) - (yb + yd), R1 = (xa + xc) - (xb + xd) */
emilmont 1:fdd22bb7aa52 1758 R0 = (R0 >> 1u) - (T0 >> 1u);
emilmont 1:fdd22bb7aa52 1759 R1 = (R1 >> 1u) - (T1 >> 1u);
emilmont 1:fdd22bb7aa52 1760
emilmont 1:fdd22bb7aa52 1761 /* (ya-yb+yc-yd)* (si2) - (xa-xb+xc-xd)* co2 */
emilmont 1:fdd22bb7aa52 1762 out1 = (short) ((Co2 * R0 - Si2 * R1) >> 16);
emilmont 1:fdd22bb7aa52 1763 /* (ya-yb+yc-yd)* co2 + (xa-xb+xc-xd)* (si2) */
emilmont 1:fdd22bb7aa52 1764 out2 = (short) ((Si2 * R0 + Co2 * R1) >> 16);
emilmont 1:fdd22bb7aa52 1765
emilmont 1:fdd22bb7aa52 1766 /* Reading i0+3fftLen/4 */
emilmont 1:fdd22bb7aa52 1767 /* Read yb (real), xb(imag) input */
emilmont 1:fdd22bb7aa52 1768 T0 = pSrc16[i1 * 2u];
emilmont 1:fdd22bb7aa52 1769 T1 = pSrc16[(i1 * 2u) + 1u];
emilmont 1:fdd22bb7aa52 1770
emilmont 1:fdd22bb7aa52 1771 /* writing the butterfly processed i0 + fftLen/4 sample */
emilmont 1:fdd22bb7aa52 1772 /* xc' = (xa-xb+xc-xd)* co2 - (ya-yb+yc-yd)* (si2) */
emilmont 1:fdd22bb7aa52 1773 /* yc' = (ya-yb+yc-yd)* co2 + (xa-xb+xc-xd)* (si2) */
emilmont 1:fdd22bb7aa52 1774 pSrc16[i1 * 2u] = out1;
emilmont 1:fdd22bb7aa52 1775 pSrc16[(i1 * 2u) + 1u] = out2;
emilmont 1:fdd22bb7aa52 1776
emilmont 1:fdd22bb7aa52 1777 /* Butterfly calculations */
emilmont 1:fdd22bb7aa52 1778 /* Read yd (real), xd(imag) input */
emilmont 1:fdd22bb7aa52 1779 U0 = pSrc16[i3 * 2u];
emilmont 1:fdd22bb7aa52 1780 U1 = pSrc16[(i3 * 2u) + 1u];
emilmont 1:fdd22bb7aa52 1781
emilmont 1:fdd22bb7aa52 1782 /* T0 = yb-yd, T1 = xb-xd) */
emilmont 1:fdd22bb7aa52 1783 T0 = __SSAT(T0 - U0, 16u);
emilmont 1:fdd22bb7aa52 1784 T1 = __SSAT(T1 - U1, 16u);
emilmont 1:fdd22bb7aa52 1785
emilmont 1:fdd22bb7aa52 1786 /* R0 = (ya-yc) - (xb- xd) , R1 = (xa-xc) + (yb-yd) */
emilmont 1:fdd22bb7aa52 1787 R0 = (S0 >> 1u) + (T1 >> 1u);
emilmont 1:fdd22bb7aa52 1788 R1 = (S1 >> 1u) - (T0 >> 1u);
emilmont 1:fdd22bb7aa52 1789
emilmont 1:fdd22bb7aa52 1790 /* S1 = (ya-yc) + (xb- xd), S1 = (xa-xc) - (yb-yd) */
emilmont 1:fdd22bb7aa52 1791 S0 = (S0 >> 1u) - (T1 >> 1u);
emilmont 1:fdd22bb7aa52 1792 S1 = (S1 >> 1u) + (T0 >> 1u);
emilmont 1:fdd22bb7aa52 1793
emilmont 1:fdd22bb7aa52 1794 /* Butterfly process for the i0+fftLen/2 sample */
emilmont 1:fdd22bb7aa52 1795 out1 = (short) ((Co1 * S0 - Si1 * S1) >> 16u);
emilmont 1:fdd22bb7aa52 1796 out2 = (short) ((Si1 * S0 + Co1 * S1) >> 16u);
emilmont 1:fdd22bb7aa52 1797 /* xb' = (xa-yb-xc+yd)* co1 - (ya+xb-yc-xd)* (si1) */
emilmont 1:fdd22bb7aa52 1798 /* yb' = (ya+xb-yc-xd)* co1 + (xa-yb-xc+yd)* (si1) */
emilmont 1:fdd22bb7aa52 1799 pSrc16[i2 * 2u] = out1;
emilmont 1:fdd22bb7aa52 1800 pSrc16[(i2 * 2u) + 1u] = out2;
emilmont 1:fdd22bb7aa52 1801
emilmont 1:fdd22bb7aa52 1802 /* Butterfly process for the i0+3fftLen/4 sample */
emilmont 1:fdd22bb7aa52 1803 out1 = (short) ((Co3 * R0 - Si3 * R1) >> 16u);
emilmont 1:fdd22bb7aa52 1804
emilmont 1:fdd22bb7aa52 1805 out2 = (short) ((Si3 * R0 + Co3 * R1) >> 16u);
emilmont 1:fdd22bb7aa52 1806 /* xd' = (xa+yb-xc-yd)* Co3 - (ya-xb-yc+xd)* (si3) */
emilmont 1:fdd22bb7aa52 1807 /* yd' = (ya-xb-yc+xd)* Co3 + (xa+yb-xc-yd)* (si3) */
emilmont 1:fdd22bb7aa52 1808 pSrc16[i3 * 2u] = out1;
emilmont 1:fdd22bb7aa52 1809 pSrc16[(i3 * 2u) + 1u] = out2;
emilmont 1:fdd22bb7aa52 1810
emilmont 1:fdd22bb7aa52 1811
emilmont 1:fdd22bb7aa52 1812 }
emilmont 1:fdd22bb7aa52 1813 }
emilmont 1:fdd22bb7aa52 1814 /* Twiddle coefficients index modifier */
emilmont 1:fdd22bb7aa52 1815 twidCoefModifier <<= 2u;
emilmont 1:fdd22bb7aa52 1816 }
emilmont 1:fdd22bb7aa52 1817 /* End of Middle stages process */
emilmont 1:fdd22bb7aa52 1818
emilmont 1:fdd22bb7aa52 1819
emilmont 1:fdd22bb7aa52 1820 /* data is in 10.6(q6) format for the 1024 point */
emilmont 1:fdd22bb7aa52 1821 /* data is in 8.8(q8) format for the 256 point */
emilmont 1:fdd22bb7aa52 1822 /* data is in 6.10(q10) format for the 64 point */
emilmont 1:fdd22bb7aa52 1823 /* data is in 4.12(q12) format for the 16 point */
emilmont 1:fdd22bb7aa52 1824
emilmont 1:fdd22bb7aa52 1825 /* start of last stage process */
emilmont 1:fdd22bb7aa52 1826
emilmont 1:fdd22bb7aa52 1827
emilmont 1:fdd22bb7aa52 1828 /* Initializations for the last stage */
emilmont 1:fdd22bb7aa52 1829 n1 = n2;
emilmont 1:fdd22bb7aa52 1830 n2 >>= 2u;
emilmont 1:fdd22bb7aa52 1831
emilmont 1:fdd22bb7aa52 1832 /* Butterfly implementation */
emilmont 1:fdd22bb7aa52 1833 for (i0 = 0u; i0 <= (fftLen - n1); i0 += n1)
emilmont 1:fdd22bb7aa52 1834 {
emilmont 1:fdd22bb7aa52 1835 /* index calculation for the input as, */
emilmont 1:fdd22bb7aa52 1836 /* pSrc16[i0 + 0], pSrc16[i0 + fftLen/4], pSrc16[i0 + fftLen/2], pSrc16[i0 + 3fftLen/4] */
emilmont 1:fdd22bb7aa52 1837 i1 = i0 + n2;
emilmont 1:fdd22bb7aa52 1838 i2 = i1 + n2;
emilmont 1:fdd22bb7aa52 1839 i3 = i2 + n2;
emilmont 1:fdd22bb7aa52 1840
emilmont 1:fdd22bb7aa52 1841 /* Reading i0, i0+fftLen/2 inputs */
emilmont 1:fdd22bb7aa52 1842 /* Read ya (real), xa(imag) input */
emilmont 1:fdd22bb7aa52 1843 T0 = pSrc16[i0 * 2u];
emilmont 1:fdd22bb7aa52 1844 T1 = pSrc16[(i0 * 2u) + 1u];
emilmont 1:fdd22bb7aa52 1845 /* Read yc (real), xc(imag) input */
emilmont 1:fdd22bb7aa52 1846 S0 = pSrc16[i2 * 2u];
emilmont 1:fdd22bb7aa52 1847 S1 = pSrc16[(i2 * 2u) + 1u];
emilmont 1:fdd22bb7aa52 1848
emilmont 1:fdd22bb7aa52 1849 /* R0 = (ya + yc), R1 = (xa + xc) */
emilmont 1:fdd22bb7aa52 1850 R0 = __SSAT(T0 + S0, 16u);
emilmont 1:fdd22bb7aa52 1851 R1 = __SSAT(T1 + S1, 16u);
emilmont 1:fdd22bb7aa52 1852 /* S0 = (ya - yc), S1 = (xa - xc) */
emilmont 1:fdd22bb7aa52 1853 S0 = __SSAT(T0 - S0, 16u);
emilmont 1:fdd22bb7aa52 1854 S1 = __SSAT(T1 - S1, 16u);
emilmont 1:fdd22bb7aa52 1855
emilmont 1:fdd22bb7aa52 1856 /* Reading i0+fftLen/4 , i0+3fftLen/4 inputs */
emilmont 1:fdd22bb7aa52 1857 /* Read yb (real), xb(imag) input */
emilmont 1:fdd22bb7aa52 1858 T0 = pSrc16[i1 * 2u];
emilmont 1:fdd22bb7aa52 1859 T1 = pSrc16[(i1 * 2u) + 1u];
emilmont 1:fdd22bb7aa52 1860 /* Read yd (real), xd(imag) input */
emilmont 1:fdd22bb7aa52 1861 U0 = pSrc16[i3 * 2u];
emilmont 1:fdd22bb7aa52 1862 U1 = pSrc16[(i3 * 2u) + 1u];
emilmont 1:fdd22bb7aa52 1863
emilmont 1:fdd22bb7aa52 1864 /* T0 = (yb + yd), T1 = (xb + xd) */
emilmont 1:fdd22bb7aa52 1865 T0 = __SSAT(T0 + U0, 16u);
emilmont 1:fdd22bb7aa52 1866 T1 = __SSAT(T1 + U1, 16u);
emilmont 1:fdd22bb7aa52 1867
emilmont 1:fdd22bb7aa52 1868 /* writing the butterfly processed i0 sample */
emilmont 1:fdd22bb7aa52 1869 /* xa' = xa + xb + xc + xd */
emilmont 1:fdd22bb7aa52 1870 /* ya' = ya + yb + yc + yd */
emilmont 1:fdd22bb7aa52 1871 pSrc16[i0 * 2u] = (R0 >> 1u) + (T0 >> 1u);
emilmont 1:fdd22bb7aa52 1872 pSrc16[(i0 * 2u) + 1u] = (R1 >> 1u) + (T1 >> 1u);
emilmont 1:fdd22bb7aa52 1873
emilmont 1:fdd22bb7aa52 1874 /* R0 = (ya + yc) - (yb + yd), R1 = (xa + xc) - (xb + xd) */
emilmont 1:fdd22bb7aa52 1875 R0 = (R0 >> 1u) - (T0 >> 1u);
emilmont 1:fdd22bb7aa52 1876 R1 = (R1 >> 1u) - (T1 >> 1u);
emilmont 1:fdd22bb7aa52 1877
emilmont 1:fdd22bb7aa52 1878 /* Read yb (real), xb(imag) input */
emilmont 1:fdd22bb7aa52 1879 T0 = pSrc16[i1 * 2u];
emilmont 1:fdd22bb7aa52 1880 T1 = pSrc16[(i1 * 2u) + 1u];
emilmont 1:fdd22bb7aa52 1881
emilmont 1:fdd22bb7aa52 1882 /* writing the butterfly processed i0 + fftLen/4 sample */
emilmont 1:fdd22bb7aa52 1883 /* xc' = (xa-xb+xc-xd) */
emilmont 1:fdd22bb7aa52 1884 /* yc' = (ya-yb+yc-yd) */
emilmont 1:fdd22bb7aa52 1885 pSrc16[i1 * 2u] = R0;
emilmont 1:fdd22bb7aa52 1886 pSrc16[(i1 * 2u) + 1u] = R1;
emilmont 1:fdd22bb7aa52 1887
emilmont 1:fdd22bb7aa52 1888 /* Read yd (real), xd(imag) input */
emilmont 1:fdd22bb7aa52 1889 U0 = pSrc16[i3 * 2u];
emilmont 1:fdd22bb7aa52 1890 U1 = pSrc16[(i3 * 2u) + 1u];
emilmont 1:fdd22bb7aa52 1891 /* T0 = (yb - yd), T1 = (xb - xd) */
emilmont 1:fdd22bb7aa52 1892 T0 = __SSAT(T0 - U0, 16u);
emilmont 1:fdd22bb7aa52 1893 T1 = __SSAT(T1 - U1, 16u);
emilmont 1:fdd22bb7aa52 1894
emilmont 1:fdd22bb7aa52 1895 /* writing the butterfly processed i0 + fftLen/2 sample */
emilmont 1:fdd22bb7aa52 1896 /* xb' = (xa-yb-xc+yd) */
emilmont 1:fdd22bb7aa52 1897 /* yb' = (ya+xb-yc-xd) */
emilmont 1:fdd22bb7aa52 1898 pSrc16[i2 * 2u] = (S0 >> 1u) - (T1 >> 1u);
emilmont 1:fdd22bb7aa52 1899 pSrc16[(i2 * 2u) + 1u] = (S1 >> 1u) + (T0 >> 1u);
emilmont 1:fdd22bb7aa52 1900
emilmont 1:fdd22bb7aa52 1901
emilmont 1:fdd22bb7aa52 1902 /* writing the butterfly processed i0 + 3fftLen/4 sample */
emilmont 1:fdd22bb7aa52 1903 /* xd' = (xa+yb-xc-yd) */
emilmont 1:fdd22bb7aa52 1904 /* yd' = (ya-xb-yc+xd) */
emilmont 1:fdd22bb7aa52 1905 pSrc16[i3 * 2u] = (S0 >> 1u) + (T1 >> 1u);
emilmont 1:fdd22bb7aa52 1906 pSrc16[(i3 * 2u) + 1u] = (S1 >> 1u) - (T0 >> 1u);
emilmont 1:fdd22bb7aa52 1907 }
emilmont 1:fdd22bb7aa52 1908 /* end of last stage process */
emilmont 1:fdd22bb7aa52 1909
emilmont 1:fdd22bb7aa52 1910 /* output is in 11.5(q5) format for the 1024 point */
emilmont 1:fdd22bb7aa52 1911 /* output is in 9.7(q7) format for the 256 point */
emilmont 1:fdd22bb7aa52 1912 /* output is in 7.9(q9) format for the 64 point */
emilmont 1:fdd22bb7aa52 1913 /* output is in 5.11(q11) format for the 16 point */
emilmont 1:fdd22bb7aa52 1914
mbed_official 3:7a284390b0ce 1915 #endif /* #ifndef ARM_MATH_CM0_FAMILY */
emilmont 1:fdd22bb7aa52 1916
emilmont 1:fdd22bb7aa52 1917 }