огшор
Dependencies: CMSIS_DSP_401 mbed
Fork of fir_f32 by
arm_cfft_f32.c@1:9b1df0b2507d, 2014-10-03 (annotated)
- Committer:
- Sergeev
- Date:
- Fri Oct 03 14:35:00 2014 +0000
- Revision:
- 1:9b1df0b2507d
- Child:
- 3:1f56b8f439a1
Ura
Who changed what in which revision?
User | Revision | Line number | New contents of line |
---|---|---|---|
Sergeev | 1:9b1df0b2507d | 1 | /* ---------------------------------------------------------------------- |
Sergeev | 1:9b1df0b2507d | 2 | * Copyright (C) 2010-2013 ARM Limited. All rights reserved. |
Sergeev | 1:9b1df0b2507d | 3 | * |
Sergeev | 1:9b1df0b2507d | 4 | * $Date: 17. January 2013 |
Sergeev | 1:9b1df0b2507d | 5 | * $Revision: V1.4.1 |
Sergeev | 1:9b1df0b2507d | 6 | * |
Sergeev | 1:9b1df0b2507d | 7 | * Project: CMSIS DSP Library |
Sergeev | 1:9b1df0b2507d | 8 | * Title: arm_cfft_f32.c |
Sergeev | 1:9b1df0b2507d | 9 | * |
Sergeev | 1:9b1df0b2507d | 10 | * Description: Combined Radix Decimation in Frequency CFFT Floating point processing function |
Sergeev | 1:9b1df0b2507d | 11 | * |
Sergeev | 1:9b1df0b2507d | 12 | * Target Processor: Cortex-M4/Cortex-M3/Cortex-M0 |
Sergeev | 1:9b1df0b2507d | 13 | * |
Sergeev | 1:9b1df0b2507d | 14 | * Redistribution and use in source and binary forms, with or without |
Sergeev | 1:9b1df0b2507d | 15 | * modification, are permitted provided that the following conditions |
Sergeev | 1:9b1df0b2507d | 16 | * are met: |
Sergeev | 1:9b1df0b2507d | 17 | * - Redistributions of source code must retain the above copyright |
Sergeev | 1:9b1df0b2507d | 18 | * notice, this list of conditions and the following disclaimer. |
Sergeev | 1:9b1df0b2507d | 19 | * - Redistributions in binary form must reproduce the above copyright |
Sergeev | 1:9b1df0b2507d | 20 | * notice, this list of conditions and the following disclaimer in |
Sergeev | 1:9b1df0b2507d | 21 | * the documentation and/or other materials provided with the |
Sergeev | 1:9b1df0b2507d | 22 | * distribution. |
Sergeev | 1:9b1df0b2507d | 23 | * - Neither the name of ARM LIMITED nor the names of its contributors |
Sergeev | 1:9b1df0b2507d | 24 | * may be used to endorse or promote products derived from this |
Sergeev | 1:9b1df0b2507d | 25 | * software without specific prior written permission. |
Sergeev | 1:9b1df0b2507d | 26 | * |
Sergeev | 1:9b1df0b2507d | 27 | * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS |
Sergeev | 1:9b1df0b2507d | 28 | * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT |
Sergeev | 1:9b1df0b2507d | 29 | * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS |
Sergeev | 1:9b1df0b2507d | 30 | * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE |
Sergeev | 1:9b1df0b2507d | 31 | * COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, |
Sergeev | 1:9b1df0b2507d | 32 | * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, |
Sergeev | 1:9b1df0b2507d | 33 | * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; |
Sergeev | 1:9b1df0b2507d | 34 | * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER |
Sergeev | 1:9b1df0b2507d | 35 | * CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT |
Sergeev | 1:9b1df0b2507d | 36 | * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN |
Sergeev | 1:9b1df0b2507d | 37 | * ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE |
Sergeev | 1:9b1df0b2507d | 38 | * POSSIBILITY OF SUCH DAMAGE. |
Sergeev | 1:9b1df0b2507d | 39 | * -------------------------------------------------------------------- */ |
Sergeev | 1:9b1df0b2507d | 40 | |
Sergeev | 1:9b1df0b2507d | 41 | |
Sergeev | 1:9b1df0b2507d | 42 | #include "arm_math.h" |
Sergeev | 1:9b1df0b2507d | 43 | #include "arm_common_tables.h" |
Sergeev | 1:9b1df0b2507d | 44 | |
Sergeev | 1:9b1df0b2507d | 45 | extern void arm_radix8_butterfly_f32( |
Sergeev | 1:9b1df0b2507d | 46 | float32_t * pSrc, |
Sergeev | 1:9b1df0b2507d | 47 | uint16_t fftLen, |
Sergeev | 1:9b1df0b2507d | 48 | const float32_t * pCoef, |
Sergeev | 1:9b1df0b2507d | 49 | uint16_t twidCoefModifier); |
Sergeev | 1:9b1df0b2507d | 50 | |
Sergeev | 1:9b1df0b2507d | 51 | |
Sergeev | 1:9b1df0b2507d | 52 | void arm_bitreversal_32( |
Sergeev | 1:9b1df0b2507d | 53 | uint32_t * pSrc, |
Sergeev | 1:9b1df0b2507d | 54 | const uint16_t bitRevLen, |
Sergeev | 1:9b1df0b2507d | 55 | const uint16_t * pBitRevTable){ |
Sergeev | 1:9b1df0b2507d | 56 | float32_t pSrc1[1024]; |
Sergeev | 1:9b1df0b2507d | 57 | for (int i =0; i<bitRevLen; i++) |
Sergeev | 1:9b1df0b2507d | 58 | { |
Sergeev | 1:9b1df0b2507d | 59 | pSrc1[i]=(float32_t)pSrc[i]; |
Sergeev | 1:9b1df0b2507d | 60 | } |
Sergeev | 1:9b1df0b2507d | 61 | arm_bitreversal_f32(pSrc1, bitRevLen, 1, pBitRevTable); |
Sergeev | 1:9b1df0b2507d | 62 | } |
Sergeev | 1:9b1df0b2507d | 63 | |
Sergeev | 1:9b1df0b2507d | 64 | /** |
Sergeev | 1:9b1df0b2507d | 65 | * @ingroup groupTransforms |
Sergeev | 1:9b1df0b2507d | 66 | */ |
Sergeev | 1:9b1df0b2507d | 67 | |
Sergeev | 1:9b1df0b2507d | 68 | /** |
Sergeev | 1:9b1df0b2507d | 69 | * @defgroup ComplexFFT Complex FFT Functions |
Sergeev | 1:9b1df0b2507d | 70 | * |
Sergeev | 1:9b1df0b2507d | 71 | * \par |
Sergeev | 1:9b1df0b2507d | 72 | * The Fast Fourier Transform (FFT) is an efficient algorithm for computing the |
Sergeev | 1:9b1df0b2507d | 73 | * Discrete Fourier Transform (DFT). The FFT can be orders of magnitude faster |
Sergeev | 1:9b1df0b2507d | 74 | * than the DFT, especially for long lengths. |
Sergeev | 1:9b1df0b2507d | 75 | * The algorithms described in this section |
Sergeev | 1:9b1df0b2507d | 76 | * operate on complex data. A separate set of functions is devoted to handling |
Sergeev | 1:9b1df0b2507d | 77 | * of real sequences. |
Sergeev | 1:9b1df0b2507d | 78 | * \par |
Sergeev | 1:9b1df0b2507d | 79 | * There are separate algorithms for handling floating-point, Q15, and Q31 data |
Sergeev | 1:9b1df0b2507d | 80 | * types. The algorithms available for each data type are described next. |
Sergeev | 1:9b1df0b2507d | 81 | * \par |
Sergeev | 1:9b1df0b2507d | 82 | * The FFT functions operate in-place. That is, the array holding the input data |
Sergeev | 1:9b1df0b2507d | 83 | * will also be used to hold the corresponding result. The input data is complex |
Sergeev | 1:9b1df0b2507d | 84 | * and contains <code>2*fftLen</code> interleaved values as shown below. |
Sergeev | 1:9b1df0b2507d | 85 | * <pre> {real[0], imag[0], real[1], imag[1],..} </pre> |
Sergeev | 1:9b1df0b2507d | 86 | * The FFT result will be contained in the same array and the frequency domain |
Sergeev | 1:9b1df0b2507d | 87 | * values will have the same interleaving. |
Sergeev | 1:9b1df0b2507d | 88 | * |
Sergeev | 1:9b1df0b2507d | 89 | * \par Floating-point |
Sergeev | 1:9b1df0b2507d | 90 | * The floating-point complex FFT uses a mixed-radix algorithm. Multiple radix-8 |
Sergeev | 1:9b1df0b2507d | 91 | * stages are performed along with a single radix-2 or radix-4 stage, as needed. |
Sergeev | 1:9b1df0b2507d | 92 | * The algorithm supports lengths of [16, 32, 64, ..., 4096] and each length uses |
Sergeev | 1:9b1df0b2507d | 93 | * a different twiddle factor table. |
Sergeev | 1:9b1df0b2507d | 94 | * \par |
Sergeev | 1:9b1df0b2507d | 95 | * The function uses the standard FFT definition and output values may grow by a |
Sergeev | 1:9b1df0b2507d | 96 | * factor of <code>fftLen</code> when computing the forward transform. The |
Sergeev | 1:9b1df0b2507d | 97 | * inverse transform includes a scale of <code>1/fftLen</code> as part of the |
Sergeev | 1:9b1df0b2507d | 98 | * calculation and this matches the textbook definition of the inverse FFT. |
Sergeev | 1:9b1df0b2507d | 99 | * \par |
Sergeev | 1:9b1df0b2507d | 100 | * Preinitialized data structures containing twiddle factors and bit reversal |
Sergeev | 1:9b1df0b2507d | 101 | * tables are provided and defined in <code>arm_const_structs.h</code>. Include |
Sergeev | 1:9b1df0b2507d | 102 | * this header in your function and then pass one of the constant structures as |
Sergeev | 1:9b1df0b2507d | 103 | * an argument to arm_cfft_f32. For example: |
Sergeev | 1:9b1df0b2507d | 104 | * \par |
Sergeev | 1:9b1df0b2507d | 105 | * <code>arm_cfft_f32(arm_cfft_sR_f32_len64, pSrc, 1, 1)</code> |
Sergeev | 1:9b1df0b2507d | 106 | * \par |
Sergeev | 1:9b1df0b2507d | 107 | * computes a 64-point inverse complex FFT including bit reversal. |
Sergeev | 1:9b1df0b2507d | 108 | * The data structures are treated as constant data and not modified during the |
Sergeev | 1:9b1df0b2507d | 109 | * calculation. The same data structure can be reused for multiple transforms |
Sergeev | 1:9b1df0b2507d | 110 | * including mixing forward and inverse transforms. |
Sergeev | 1:9b1df0b2507d | 111 | * \par |
Sergeev | 1:9b1df0b2507d | 112 | * Earlier releases of the library provided separate radix-2 and radix-4 |
Sergeev | 1:9b1df0b2507d | 113 | * algorithms that operated on floating-point data. These functions are still |
Sergeev | 1:9b1df0b2507d | 114 | * provided but are deprecated. The older functions are slower and less general |
Sergeev | 1:9b1df0b2507d | 115 | * than the new functions. |
Sergeev | 1:9b1df0b2507d | 116 | * \par |
Sergeev | 1:9b1df0b2507d | 117 | * An example of initialization of the constants for the arm_cfft_f32 function follows: |
Sergeev | 1:9b1df0b2507d | 118 | * \par |
Sergeev | 1:9b1df0b2507d | 119 | * const static arm_cfft_instance_f32 *S; |
Sergeev | 1:9b1df0b2507d | 120 | * ... |
Sergeev | 1:9b1df0b2507d | 121 | * switch (length) { |
Sergeev | 1:9b1df0b2507d | 122 | * case 16: |
Sergeev | 1:9b1df0b2507d | 123 | * S = & arm_cfft_sR_f32_len16; |
Sergeev | 1:9b1df0b2507d | 124 | * break; |
Sergeev | 1:9b1df0b2507d | 125 | * case 32: |
Sergeev | 1:9b1df0b2507d | 126 | * S = & arm_cfft_sR_f32_len32; |
Sergeev | 1:9b1df0b2507d | 127 | * break; |
Sergeev | 1:9b1df0b2507d | 128 | * case 64: |
Sergeev | 1:9b1df0b2507d | 129 | * S = & arm_cfft_sR_f32_len64; |
Sergeev | 1:9b1df0b2507d | 130 | * break; |
Sergeev | 1:9b1df0b2507d | 131 | * case 128: |
Sergeev | 1:9b1df0b2507d | 132 | * S = & arm_cfft_sR_f32_len128; |
Sergeev | 1:9b1df0b2507d | 133 | * break; |
Sergeev | 1:9b1df0b2507d | 134 | * case 256: |
Sergeev | 1:9b1df0b2507d | 135 | * S = & arm_cfft_sR_f32_len256; |
Sergeev | 1:9b1df0b2507d | 136 | * break; |
Sergeev | 1:9b1df0b2507d | 137 | * case 512: |
Sergeev | 1:9b1df0b2507d | 138 | * S = & arm_cfft_sR_f32_len512; |
Sergeev | 1:9b1df0b2507d | 139 | * break; |
Sergeev | 1:9b1df0b2507d | 140 | * case 1024: |
Sergeev | 1:9b1df0b2507d | 141 | * S = & arm_cfft_sR_f32_len1024; |
Sergeev | 1:9b1df0b2507d | 142 | * break; |
Sergeev | 1:9b1df0b2507d | 143 | * case 2048: |
Sergeev | 1:9b1df0b2507d | 144 | * S = & arm_cfft_sR_f32_len2048; |
Sergeev | 1:9b1df0b2507d | 145 | * break; |
Sergeev | 1:9b1df0b2507d | 146 | * case 4096: |
Sergeev | 1:9b1df0b2507d | 147 | * S = & arm_cfft_sR_f32_len4096; |
Sergeev | 1:9b1df0b2507d | 148 | * break; |
Sergeev | 1:9b1df0b2507d | 149 | * } |
Sergeev | 1:9b1df0b2507d | 150 | * \par Q15 and Q31 |
Sergeev | 1:9b1df0b2507d | 151 | * The library provides radix-2 and radix-4 FFT algorithms for fixed-point data. The |
Sergeev | 1:9b1df0b2507d | 152 | * radix-2 algorithm supports lengths of [16, 32, 64, ..., 4096]. The radix-4 |
Sergeev | 1:9b1df0b2507d | 153 | * algorithm supports lengths of [16, 64, 256, ..., 4096]. When possible, you |
Sergeev | 1:9b1df0b2507d | 154 | * should use the radix-4 algorithm since it is faster than the radix-2 of the |
Sergeev | 1:9b1df0b2507d | 155 | * same length. |
Sergeev | 1:9b1df0b2507d | 156 | * \par |
Sergeev | 1:9b1df0b2507d | 157 | * The forward FFTs include scaling in order to prevent results from overflowing. |
Sergeev | 1:9b1df0b2507d | 158 | * Intermediate results are scaled down during each butterfly stage. In the |
Sergeev | 1:9b1df0b2507d | 159 | * radix-2 algorithm, a scale of 0.5 is applied during each butterfly. In the |
Sergeev | 1:9b1df0b2507d | 160 | * radix-4 algorithm, a scale of 0.25 is applied. The scaling applies to both |
Sergeev | 1:9b1df0b2507d | 161 | * the forward and the inverse FFTs. Thus the forward FFT contains an additional |
Sergeev | 1:9b1df0b2507d | 162 | * scale factor of <code>1/fftLen</code> as compared to the standard textbook |
Sergeev | 1:9b1df0b2507d | 163 | * definition of the FFT. The inverse FFT also scales down during each butterfly |
Sergeev | 1:9b1df0b2507d | 164 | * stage and this corresponds to the standard textbook definition. |
Sergeev | 1:9b1df0b2507d | 165 | * \par |
Sergeev | 1:9b1df0b2507d | 166 | * A separate instance structure must be defined for each transform used but |
Sergeev | 1:9b1df0b2507d | 167 | * twiddle factor and bit reversal tables can be reused. |
Sergeev | 1:9b1df0b2507d | 168 | * \par |
Sergeev | 1:9b1df0b2507d | 169 | * There is also an associated initialization function for each data type. |
Sergeev | 1:9b1df0b2507d | 170 | * The initialization function performs the following operations: |
Sergeev | 1:9b1df0b2507d | 171 | * - Sets the values of the internal structure fields. |
Sergeev | 1:9b1df0b2507d | 172 | * - Initializes twiddle factor table and bit reversal table pointers. |
Sergeev | 1:9b1df0b2507d | 173 | * \par |
Sergeev | 1:9b1df0b2507d | 174 | * Use of the initialization function is optional. |
Sergeev | 1:9b1df0b2507d | 175 | * However, if the initialization function is used, then the instance structure |
Sergeev | 1:9b1df0b2507d | 176 | * cannot be placed into a const data section. To place an instance structure |
Sergeev | 1:9b1df0b2507d | 177 | * into a const data section, the instance structure should be manually |
Sergeev | 1:9b1df0b2507d | 178 | * initialized as follows: |
Sergeev | 1:9b1df0b2507d | 179 | * <pre> |
Sergeev | 1:9b1df0b2507d | 180 | *arm_cfft_radix2_instance_q31 S = {fftLen, ifftFlag, bitReverseFlag, pTwiddle, pBitRevTable, twidCoefModifier, bitRevFactor}; |
Sergeev | 1:9b1df0b2507d | 181 | *arm_cfft_radix2_instance_q15 S = {fftLen, ifftFlag, bitReverseFlag, pTwiddle, pBitRevTable, twidCoefModifier, bitRevFactor}; |
Sergeev | 1:9b1df0b2507d | 182 | *arm_cfft_radix4_instance_q31 S = {fftLen, ifftFlag, bitReverseFlag, pTwiddle, pBitRevTable, twidCoefModifier, bitRevFactor}; |
Sergeev | 1:9b1df0b2507d | 183 | *arm_cfft_radix4_instance_q15 S = {fftLen, ifftFlag, bitReverseFlag, pTwiddle, pBitRevTable, twidCoefModifier, bitRevFactor}; |
Sergeev | 1:9b1df0b2507d | 184 | *arm_cfft_instance_f32 S = {fftLen, pTwiddle, pBitRevTable, bitRevLength}; |
Sergeev | 1:9b1df0b2507d | 185 | * </pre> |
Sergeev | 1:9b1df0b2507d | 186 | * \par |
Sergeev | 1:9b1df0b2507d | 187 | * where <code>fftLen</code> length of CFFT/CIFFT; <code>ifftFlag</code> Flag for |
Sergeev | 1:9b1df0b2507d | 188 | * selection of forward or inverse transform. When ifftFlag is set the inverse |
Sergeev | 1:9b1df0b2507d | 189 | * transform is calculated. |
Sergeev | 1:9b1df0b2507d | 190 | * <code>bitReverseFlag</code> Flag for selection of output order (Set bitReverseFlag to output in normal order otherwise output in bit reversed order); |
Sergeev | 1:9b1df0b2507d | 191 | * <code>pTwiddle</code>points to array of twiddle coefficients; <code>pBitRevTable</code> points to the bit reversal table. |
Sergeev | 1:9b1df0b2507d | 192 | * <code>twidCoefModifier</code> modifier for twiddle factor table which supports all FFT lengths with same table; |
Sergeev | 1:9b1df0b2507d | 193 | * <code>pBitRevTable</code> modifier for bit reversal table which supports all FFT lengths with same table. |
Sergeev | 1:9b1df0b2507d | 194 | * <code>onebyfftLen</code> value of 1/fftLen to calculate CIFFT; |
Sergeev | 1:9b1df0b2507d | 195 | * \par |
Sergeev | 1:9b1df0b2507d | 196 | * The Q15 and Q31 FFT functions use a large bit reversal and twiddle factor |
Sergeev | 1:9b1df0b2507d | 197 | * table. The tables are defined for the maximum length transform and a subset |
Sergeev | 1:9b1df0b2507d | 198 | * of the coefficients are used in shorter transforms. |
Sergeev | 1:9b1df0b2507d | 199 | * |
Sergeev | 1:9b1df0b2507d | 200 | */ |
Sergeev | 1:9b1df0b2507d | 201 | |
Sergeev | 1:9b1df0b2507d | 202 | void arm_cfft_radix8by2_f32( arm_cfft_instance_f32 * S, float32_t * p1) |
Sergeev | 1:9b1df0b2507d | 203 | { |
Sergeev | 1:9b1df0b2507d | 204 | uint32_t L = S->fftLen; |
Sergeev | 1:9b1df0b2507d | 205 | float32_t * pCol1, * pCol2, * pMid1, * pMid2; |
Sergeev | 1:9b1df0b2507d | 206 | float32_t * p2 = p1 + L; |
Sergeev | 1:9b1df0b2507d | 207 | const float32_t * tw = (float32_t *) S->pTwiddle; |
Sergeev | 1:9b1df0b2507d | 208 | float32_t t1[4], t2[4], t3[4], t4[4], twR, twI; |
Sergeev | 1:9b1df0b2507d | 209 | float32_t m0, m1, m2, m3; |
Sergeev | 1:9b1df0b2507d | 210 | uint32_t l; |
Sergeev | 1:9b1df0b2507d | 211 | |
Sergeev | 1:9b1df0b2507d | 212 | pCol1 = p1; |
Sergeev | 1:9b1df0b2507d | 213 | pCol2 = p2; |
Sergeev | 1:9b1df0b2507d | 214 | |
Sergeev | 1:9b1df0b2507d | 215 | // Define new length |
Sergeev | 1:9b1df0b2507d | 216 | L >>= 1; |
Sergeev | 1:9b1df0b2507d | 217 | // Initialize mid pointers |
Sergeev | 1:9b1df0b2507d | 218 | pMid1 = p1 + L; |
Sergeev | 1:9b1df0b2507d | 219 | pMid2 = p2 + L; |
Sergeev | 1:9b1df0b2507d | 220 | |
Sergeev | 1:9b1df0b2507d | 221 | // do two dot Fourier transform |
Sergeev | 1:9b1df0b2507d | 222 | for ( l = L >> 2; l > 0; l-- ) |
Sergeev | 1:9b1df0b2507d | 223 | { |
Sergeev | 1:9b1df0b2507d | 224 | t1[0] = p1[0]; |
Sergeev | 1:9b1df0b2507d | 225 | t1[1] = p1[1]; |
Sergeev | 1:9b1df0b2507d | 226 | t1[2] = p1[2]; |
Sergeev | 1:9b1df0b2507d | 227 | t1[3] = p1[3]; |
Sergeev | 1:9b1df0b2507d | 228 | |
Sergeev | 1:9b1df0b2507d | 229 | t2[0] = p2[0]; |
Sergeev | 1:9b1df0b2507d | 230 | t2[1] = p2[1]; |
Sergeev | 1:9b1df0b2507d | 231 | t2[2] = p2[2]; |
Sergeev | 1:9b1df0b2507d | 232 | t2[3] = p2[3]; |
Sergeev | 1:9b1df0b2507d | 233 | |
Sergeev | 1:9b1df0b2507d | 234 | t3[0] = pMid1[0]; |
Sergeev | 1:9b1df0b2507d | 235 | t3[1] = pMid1[1]; |
Sergeev | 1:9b1df0b2507d | 236 | t3[2] = pMid1[2]; |
Sergeev | 1:9b1df0b2507d | 237 | t3[3] = pMid1[3]; |
Sergeev | 1:9b1df0b2507d | 238 | |
Sergeev | 1:9b1df0b2507d | 239 | t4[0] = pMid2[0]; |
Sergeev | 1:9b1df0b2507d | 240 | t4[1] = pMid2[1]; |
Sergeev | 1:9b1df0b2507d | 241 | t4[2] = pMid2[2]; |
Sergeev | 1:9b1df0b2507d | 242 | t4[3] = pMid2[3]; |
Sergeev | 1:9b1df0b2507d | 243 | |
Sergeev | 1:9b1df0b2507d | 244 | *p1++ = t1[0] + t2[0]; |
Sergeev | 1:9b1df0b2507d | 245 | *p1++ = t1[1] + t2[1]; |
Sergeev | 1:9b1df0b2507d | 246 | *p1++ = t1[2] + t2[2]; |
Sergeev | 1:9b1df0b2507d | 247 | *p1++ = t1[3] + t2[3]; // col 1 |
Sergeev | 1:9b1df0b2507d | 248 | |
Sergeev | 1:9b1df0b2507d | 249 | t2[0] = t1[0] - t2[0]; |
Sergeev | 1:9b1df0b2507d | 250 | t2[1] = t1[1] - t2[1]; |
Sergeev | 1:9b1df0b2507d | 251 | t2[2] = t1[2] - t2[2]; |
Sergeev | 1:9b1df0b2507d | 252 | t2[3] = t1[3] - t2[3]; // for col 2 |
Sergeev | 1:9b1df0b2507d | 253 | |
Sergeev | 1:9b1df0b2507d | 254 | *pMid1++ = t3[0] + t4[0]; |
Sergeev | 1:9b1df0b2507d | 255 | *pMid1++ = t3[1] + t4[1]; |
Sergeev | 1:9b1df0b2507d | 256 | *pMid1++ = t3[2] + t4[2]; |
Sergeev | 1:9b1df0b2507d | 257 | *pMid1++ = t3[3] + t4[3]; // col 1 |
Sergeev | 1:9b1df0b2507d | 258 | |
Sergeev | 1:9b1df0b2507d | 259 | t4[0] = t4[0] - t3[0]; |
Sergeev | 1:9b1df0b2507d | 260 | t4[1] = t4[1] - t3[1]; |
Sergeev | 1:9b1df0b2507d | 261 | t4[2] = t4[2] - t3[2]; |
Sergeev | 1:9b1df0b2507d | 262 | t4[3] = t4[3] - t3[3]; // for col 2 |
Sergeev | 1:9b1df0b2507d | 263 | |
Sergeev | 1:9b1df0b2507d | 264 | twR = *tw++; |
Sergeev | 1:9b1df0b2507d | 265 | twI = *tw++; |
Sergeev | 1:9b1df0b2507d | 266 | |
Sergeev | 1:9b1df0b2507d | 267 | // multiply by twiddle factors |
Sergeev | 1:9b1df0b2507d | 268 | m0 = t2[0] * twR; |
Sergeev | 1:9b1df0b2507d | 269 | m1 = t2[1] * twI; |
Sergeev | 1:9b1df0b2507d | 270 | m2 = t2[1] * twR; |
Sergeev | 1:9b1df0b2507d | 271 | m3 = t2[0] * twI; |
Sergeev | 1:9b1df0b2507d | 272 | |
Sergeev | 1:9b1df0b2507d | 273 | // R = R * Tr - I * Ti |
Sergeev | 1:9b1df0b2507d | 274 | *p2++ = m0 + m1; |
Sergeev | 1:9b1df0b2507d | 275 | // I = I * Tr + R * Ti |
Sergeev | 1:9b1df0b2507d | 276 | *p2++ = m2 - m3; |
Sergeev | 1:9b1df0b2507d | 277 | |
Sergeev | 1:9b1df0b2507d | 278 | // use vertical symmetry |
Sergeev | 1:9b1df0b2507d | 279 | // 0.9988 - 0.0491i <==> -0.0491 - 0.9988i |
Sergeev | 1:9b1df0b2507d | 280 | m0 = t4[0] * twI; |
Sergeev | 1:9b1df0b2507d | 281 | m1 = t4[1] * twR; |
Sergeev | 1:9b1df0b2507d | 282 | m2 = t4[1] * twI; |
Sergeev | 1:9b1df0b2507d | 283 | m3 = t4[0] * twR; |
Sergeev | 1:9b1df0b2507d | 284 | |
Sergeev | 1:9b1df0b2507d | 285 | *pMid2++ = m0 - m1; |
Sergeev | 1:9b1df0b2507d | 286 | *pMid2++ = m2 + m3; |
Sergeev | 1:9b1df0b2507d | 287 | |
Sergeev | 1:9b1df0b2507d | 288 | twR = *tw++; |
Sergeev | 1:9b1df0b2507d | 289 | twI = *tw++; |
Sergeev | 1:9b1df0b2507d | 290 | |
Sergeev | 1:9b1df0b2507d | 291 | m0 = t2[2] * twR; |
Sergeev | 1:9b1df0b2507d | 292 | m1 = t2[3] * twI; |
Sergeev | 1:9b1df0b2507d | 293 | m2 = t2[3] * twR; |
Sergeev | 1:9b1df0b2507d | 294 | m3 = t2[2] * twI; |
Sergeev | 1:9b1df0b2507d | 295 | |
Sergeev | 1:9b1df0b2507d | 296 | *p2++ = m0 + m1; |
Sergeev | 1:9b1df0b2507d | 297 | *p2++ = m2 - m3; |
Sergeev | 1:9b1df0b2507d | 298 | |
Sergeev | 1:9b1df0b2507d | 299 | m0 = t4[2] * twI; |
Sergeev | 1:9b1df0b2507d | 300 | m1 = t4[3] * twR; |
Sergeev | 1:9b1df0b2507d | 301 | m2 = t4[3] * twI; |
Sergeev | 1:9b1df0b2507d | 302 | m3 = t4[2] * twR; |
Sergeev | 1:9b1df0b2507d | 303 | |
Sergeev | 1:9b1df0b2507d | 304 | *pMid2++ = m0 - m1; |
Sergeev | 1:9b1df0b2507d | 305 | *pMid2++ = m2 + m3; |
Sergeev | 1:9b1df0b2507d | 306 | } |
Sergeev | 1:9b1df0b2507d | 307 | |
Sergeev | 1:9b1df0b2507d | 308 | // first col |
Sergeev | 1:9b1df0b2507d | 309 | arm_radix8_butterfly_f32( pCol1, L, (float32_t *) S->pTwiddle, 2u); |
Sergeev | 1:9b1df0b2507d | 310 | // second col |
Sergeev | 1:9b1df0b2507d | 311 | arm_radix8_butterfly_f32( pCol2, L, (float32_t *) S->pTwiddle, 2u); |
Sergeev | 1:9b1df0b2507d | 312 | |
Sergeev | 1:9b1df0b2507d | 313 | } |
Sergeev | 1:9b1df0b2507d | 314 | |
Sergeev | 1:9b1df0b2507d | 315 | void arm_cfft_radix8by4_f32( arm_cfft_instance_f32 * S, float32_t * p1) |
Sergeev | 1:9b1df0b2507d | 316 | { |
Sergeev | 1:9b1df0b2507d | 317 | uint32_t L = S->fftLen >> 1; |
Sergeev | 1:9b1df0b2507d | 318 | float32_t * pCol1, *pCol2, *pCol3, *pCol4, *pEnd1, *pEnd2, *pEnd3, *pEnd4; |
Sergeev | 1:9b1df0b2507d | 319 | const float32_t *tw2, *tw3, *tw4; |
Sergeev | 1:9b1df0b2507d | 320 | float32_t * p2 = p1 + L; |
Sergeev | 1:9b1df0b2507d | 321 | float32_t * p3 = p2 + L; |
Sergeev | 1:9b1df0b2507d | 322 | float32_t * p4 = p3 + L; |
Sergeev | 1:9b1df0b2507d | 323 | float32_t t2[4], t3[4], t4[4], twR, twI; |
Sergeev | 1:9b1df0b2507d | 324 | float32_t p1ap3_0, p1sp3_0, p1ap3_1, p1sp3_1; |
Sergeev | 1:9b1df0b2507d | 325 | float32_t m0, m1, m2, m3; |
Sergeev | 1:9b1df0b2507d | 326 | uint32_t l, twMod2, twMod3, twMod4; |
Sergeev | 1:9b1df0b2507d | 327 | |
Sergeev | 1:9b1df0b2507d | 328 | pCol1 = p1; // points to real values by default |
Sergeev | 1:9b1df0b2507d | 329 | pCol2 = p2; |
Sergeev | 1:9b1df0b2507d | 330 | pCol3 = p3; |
Sergeev | 1:9b1df0b2507d | 331 | pCol4 = p4; |
Sergeev | 1:9b1df0b2507d | 332 | pEnd1 = p2 - 1; // points to imaginary values by default |
Sergeev | 1:9b1df0b2507d | 333 | pEnd2 = p3 - 1; |
Sergeev | 1:9b1df0b2507d | 334 | pEnd3 = p4 - 1; |
Sergeev | 1:9b1df0b2507d | 335 | pEnd4 = pEnd3 + L; |
Sergeev | 1:9b1df0b2507d | 336 | |
Sergeev | 1:9b1df0b2507d | 337 | tw2 = tw3 = tw4 = (float32_t *) S->pTwiddle; |
Sergeev | 1:9b1df0b2507d | 338 | |
Sergeev | 1:9b1df0b2507d | 339 | L >>= 1; |
Sergeev | 1:9b1df0b2507d | 340 | |
Sergeev | 1:9b1df0b2507d | 341 | // do four dot Fourier transform |
Sergeev | 1:9b1df0b2507d | 342 | |
Sergeev | 1:9b1df0b2507d | 343 | twMod2 = 2; |
Sergeev | 1:9b1df0b2507d | 344 | twMod3 = 4; |
Sergeev | 1:9b1df0b2507d | 345 | twMod4 = 6; |
Sergeev | 1:9b1df0b2507d | 346 | |
Sergeev | 1:9b1df0b2507d | 347 | // TOP |
Sergeev | 1:9b1df0b2507d | 348 | p1ap3_0 = p1[0] + p3[0]; |
Sergeev | 1:9b1df0b2507d | 349 | p1sp3_0 = p1[0] - p3[0]; |
Sergeev | 1:9b1df0b2507d | 350 | p1ap3_1 = p1[1] + p3[1]; |
Sergeev | 1:9b1df0b2507d | 351 | p1sp3_1 = p1[1] - p3[1]; |
Sergeev | 1:9b1df0b2507d | 352 | |
Sergeev | 1:9b1df0b2507d | 353 | // col 2 |
Sergeev | 1:9b1df0b2507d | 354 | t2[0] = p1sp3_0 + p2[1] - p4[1]; |
Sergeev | 1:9b1df0b2507d | 355 | t2[1] = p1sp3_1 - p2[0] + p4[0]; |
Sergeev | 1:9b1df0b2507d | 356 | // col 3 |
Sergeev | 1:9b1df0b2507d | 357 | t3[0] = p1ap3_0 - p2[0] - p4[0]; |
Sergeev | 1:9b1df0b2507d | 358 | t3[1] = p1ap3_1 - p2[1] - p4[1]; |
Sergeev | 1:9b1df0b2507d | 359 | // col 4 |
Sergeev | 1:9b1df0b2507d | 360 | t4[0] = p1sp3_0 - p2[1] + p4[1]; |
Sergeev | 1:9b1df0b2507d | 361 | t4[1] = p1sp3_1 + p2[0] - p4[0]; |
Sergeev | 1:9b1df0b2507d | 362 | // col 1 |
Sergeev | 1:9b1df0b2507d | 363 | *p1++ = p1ap3_0 + p2[0] + p4[0]; |
Sergeev | 1:9b1df0b2507d | 364 | *p1++ = p1ap3_1 + p2[1] + p4[1]; |
Sergeev | 1:9b1df0b2507d | 365 | |
Sergeev | 1:9b1df0b2507d | 366 | // Twiddle factors are ones |
Sergeev | 1:9b1df0b2507d | 367 | *p2++ = t2[0]; |
Sergeev | 1:9b1df0b2507d | 368 | *p2++ = t2[1]; |
Sergeev | 1:9b1df0b2507d | 369 | *p3++ = t3[0]; |
Sergeev | 1:9b1df0b2507d | 370 | *p3++ = t3[1]; |
Sergeev | 1:9b1df0b2507d | 371 | *p4++ = t4[0]; |
Sergeev | 1:9b1df0b2507d | 372 | *p4++ = t4[1]; |
Sergeev | 1:9b1df0b2507d | 373 | |
Sergeev | 1:9b1df0b2507d | 374 | tw2 += twMod2; |
Sergeev | 1:9b1df0b2507d | 375 | tw3 += twMod3; |
Sergeev | 1:9b1df0b2507d | 376 | tw4 += twMod4; |
Sergeev | 1:9b1df0b2507d | 377 | |
Sergeev | 1:9b1df0b2507d | 378 | for (l = (L - 2) >> 1; l > 0; l-- ) |
Sergeev | 1:9b1df0b2507d | 379 | { |
Sergeev | 1:9b1df0b2507d | 380 | |
Sergeev | 1:9b1df0b2507d | 381 | // TOP |
Sergeev | 1:9b1df0b2507d | 382 | p1ap3_0 = p1[0] + p3[0]; |
Sergeev | 1:9b1df0b2507d | 383 | p1sp3_0 = p1[0] - p3[0]; |
Sergeev | 1:9b1df0b2507d | 384 | p1ap3_1 = p1[1] + p3[1]; |
Sergeev | 1:9b1df0b2507d | 385 | p1sp3_1 = p1[1] - p3[1]; |
Sergeev | 1:9b1df0b2507d | 386 | // col 2 |
Sergeev | 1:9b1df0b2507d | 387 | t2[0] = p1sp3_0 + p2[1] - p4[1]; |
Sergeev | 1:9b1df0b2507d | 388 | t2[1] = p1sp3_1 - p2[0] + p4[0]; |
Sergeev | 1:9b1df0b2507d | 389 | // col 3 |
Sergeev | 1:9b1df0b2507d | 390 | t3[0] = p1ap3_0 - p2[0] - p4[0]; |
Sergeev | 1:9b1df0b2507d | 391 | t3[1] = p1ap3_1 - p2[1] - p4[1]; |
Sergeev | 1:9b1df0b2507d | 392 | // col 4 |
Sergeev | 1:9b1df0b2507d | 393 | t4[0] = p1sp3_0 - p2[1] + p4[1]; |
Sergeev | 1:9b1df0b2507d | 394 | t4[1] = p1sp3_1 + p2[0] - p4[0]; |
Sergeev | 1:9b1df0b2507d | 395 | // col 1 - top |
Sergeev | 1:9b1df0b2507d | 396 | *p1++ = p1ap3_0 + p2[0] + p4[0]; |
Sergeev | 1:9b1df0b2507d | 397 | *p1++ = p1ap3_1 + p2[1] + p4[1]; |
Sergeev | 1:9b1df0b2507d | 398 | |
Sergeev | 1:9b1df0b2507d | 399 | // BOTTOM |
Sergeev | 1:9b1df0b2507d | 400 | p1ap3_1 = pEnd1[-1] + pEnd3[-1]; |
Sergeev | 1:9b1df0b2507d | 401 | p1sp3_1 = pEnd1[-1] - pEnd3[-1]; |
Sergeev | 1:9b1df0b2507d | 402 | p1ap3_0 = pEnd1[0] + pEnd3[0]; |
Sergeev | 1:9b1df0b2507d | 403 | p1sp3_0 = pEnd1[0] - pEnd3[0]; |
Sergeev | 1:9b1df0b2507d | 404 | // col 2 |
Sergeev | 1:9b1df0b2507d | 405 | t2[2] = pEnd2[0] - pEnd4[0] + p1sp3_1; |
Sergeev | 1:9b1df0b2507d | 406 | t2[3] = pEnd1[0] - pEnd3[0] - pEnd2[-1] + pEnd4[-1]; |
Sergeev | 1:9b1df0b2507d | 407 | // col 3 |
Sergeev | 1:9b1df0b2507d | 408 | t3[2] = p1ap3_1 - pEnd2[-1] - pEnd4[-1]; |
Sergeev | 1:9b1df0b2507d | 409 | t3[3] = p1ap3_0 - pEnd2[0] - pEnd4[0]; |
Sergeev | 1:9b1df0b2507d | 410 | // col 4 |
Sergeev | 1:9b1df0b2507d | 411 | t4[2] = pEnd2[0] - pEnd4[0] - p1sp3_1; |
Sergeev | 1:9b1df0b2507d | 412 | t4[3] = pEnd4[-1] - pEnd2[-1] - p1sp3_0; |
Sergeev | 1:9b1df0b2507d | 413 | // col 1 - Bottom |
Sergeev | 1:9b1df0b2507d | 414 | *pEnd1-- = p1ap3_0 + pEnd2[0] + pEnd4[0]; |
Sergeev | 1:9b1df0b2507d | 415 | *pEnd1-- = p1ap3_1 + pEnd2[-1] + pEnd4[-1]; |
Sergeev | 1:9b1df0b2507d | 416 | |
Sergeev | 1:9b1df0b2507d | 417 | // COL 2 |
Sergeev | 1:9b1df0b2507d | 418 | // read twiddle factors |
Sergeev | 1:9b1df0b2507d | 419 | twR = *tw2++; |
Sergeev | 1:9b1df0b2507d | 420 | twI = *tw2++; |
Sergeev | 1:9b1df0b2507d | 421 | // multiply by twiddle factors |
Sergeev | 1:9b1df0b2507d | 422 | // let Z1 = a + i(b), Z2 = c + i(d) |
Sergeev | 1:9b1df0b2507d | 423 | // => Z1 * Z2 = (a*c - b*d) + i(b*c + a*d) |
Sergeev | 1:9b1df0b2507d | 424 | // Top |
Sergeev | 1:9b1df0b2507d | 425 | m0 = t2[0] * twR; |
Sergeev | 1:9b1df0b2507d | 426 | m1 = t2[1] * twI; |
Sergeev | 1:9b1df0b2507d | 427 | m2 = t2[1] * twR; |
Sergeev | 1:9b1df0b2507d | 428 | m3 = t2[0] * twI; |
Sergeev | 1:9b1df0b2507d | 429 | |
Sergeev | 1:9b1df0b2507d | 430 | *p2++ = m0 + m1; |
Sergeev | 1:9b1df0b2507d | 431 | *p2++ = m2 - m3; |
Sergeev | 1:9b1df0b2507d | 432 | // use vertical symmetry col 2 |
Sergeev | 1:9b1df0b2507d | 433 | // 0.9997 - 0.0245i <==> 0.0245 - 0.9997i |
Sergeev | 1:9b1df0b2507d | 434 | // Bottom |
Sergeev | 1:9b1df0b2507d | 435 | m0 = t2[3] * twI; |
Sergeev | 1:9b1df0b2507d | 436 | m1 = t2[2] * twR; |
Sergeev | 1:9b1df0b2507d | 437 | m2 = t2[2] * twI; |
Sergeev | 1:9b1df0b2507d | 438 | m3 = t2[3] * twR; |
Sergeev | 1:9b1df0b2507d | 439 | |
Sergeev | 1:9b1df0b2507d | 440 | *pEnd2-- = m0 - m1; |
Sergeev | 1:9b1df0b2507d | 441 | *pEnd2-- = m2 + m3; |
Sergeev | 1:9b1df0b2507d | 442 | |
Sergeev | 1:9b1df0b2507d | 443 | // COL 3 |
Sergeev | 1:9b1df0b2507d | 444 | twR = tw3[0]; |
Sergeev | 1:9b1df0b2507d | 445 | twI = tw3[1]; |
Sergeev | 1:9b1df0b2507d | 446 | tw3 += twMod3; |
Sergeev | 1:9b1df0b2507d | 447 | // Top |
Sergeev | 1:9b1df0b2507d | 448 | m0 = t3[0] * twR; |
Sergeev | 1:9b1df0b2507d | 449 | m1 = t3[1] * twI; |
Sergeev | 1:9b1df0b2507d | 450 | m2 = t3[1] * twR; |
Sergeev | 1:9b1df0b2507d | 451 | m3 = t3[0] * twI; |
Sergeev | 1:9b1df0b2507d | 452 | |
Sergeev | 1:9b1df0b2507d | 453 | *p3++ = m0 + m1; |
Sergeev | 1:9b1df0b2507d | 454 | *p3++ = m2 - m3; |
Sergeev | 1:9b1df0b2507d | 455 | // use vertical symmetry col 3 |
Sergeev | 1:9b1df0b2507d | 456 | // 0.9988 - 0.0491i <==> -0.9988 - 0.0491i |
Sergeev | 1:9b1df0b2507d | 457 | // Bottom |
Sergeev | 1:9b1df0b2507d | 458 | m0 = -t3[3] * twR; |
Sergeev | 1:9b1df0b2507d | 459 | m1 = t3[2] * twI; |
Sergeev | 1:9b1df0b2507d | 460 | m2 = t3[2] * twR; |
Sergeev | 1:9b1df0b2507d | 461 | m3 = t3[3] * twI; |
Sergeev | 1:9b1df0b2507d | 462 | |
Sergeev | 1:9b1df0b2507d | 463 | *pEnd3-- = m0 - m1; |
Sergeev | 1:9b1df0b2507d | 464 | *pEnd3-- = m3 - m2; |
Sergeev | 1:9b1df0b2507d | 465 | |
Sergeev | 1:9b1df0b2507d | 466 | // COL 4 |
Sergeev | 1:9b1df0b2507d | 467 | twR = tw4[0]; |
Sergeev | 1:9b1df0b2507d | 468 | twI = tw4[1]; |
Sergeev | 1:9b1df0b2507d | 469 | tw4 += twMod4; |
Sergeev | 1:9b1df0b2507d | 470 | // Top |
Sergeev | 1:9b1df0b2507d | 471 | m0 = t4[0] * twR; |
Sergeev | 1:9b1df0b2507d | 472 | m1 = t4[1] * twI; |
Sergeev | 1:9b1df0b2507d | 473 | m2 = t4[1] * twR; |
Sergeev | 1:9b1df0b2507d | 474 | m3 = t4[0] * twI; |
Sergeev | 1:9b1df0b2507d | 475 | |
Sergeev | 1:9b1df0b2507d | 476 | *p4++ = m0 + m1; |
Sergeev | 1:9b1df0b2507d | 477 | *p4++ = m2 - m3; |
Sergeev | 1:9b1df0b2507d | 478 | // use vertical symmetry col 4 |
Sergeev | 1:9b1df0b2507d | 479 | // 0.9973 - 0.0736i <==> -0.0736 + 0.9973i |
Sergeev | 1:9b1df0b2507d | 480 | // Bottom |
Sergeev | 1:9b1df0b2507d | 481 | m0 = t4[3] * twI; |
Sergeev | 1:9b1df0b2507d | 482 | m1 = t4[2] * twR; |
Sergeev | 1:9b1df0b2507d | 483 | m2 = t4[2] * twI; |
Sergeev | 1:9b1df0b2507d | 484 | m3 = t4[3] * twR; |
Sergeev | 1:9b1df0b2507d | 485 | |
Sergeev | 1:9b1df0b2507d | 486 | *pEnd4-- = m0 - m1; |
Sergeev | 1:9b1df0b2507d | 487 | *pEnd4-- = m2 + m3; |
Sergeev | 1:9b1df0b2507d | 488 | } |
Sergeev | 1:9b1df0b2507d | 489 | |
Sergeev | 1:9b1df0b2507d | 490 | //MIDDLE |
Sergeev | 1:9b1df0b2507d | 491 | // Twiddle factors are |
Sergeev | 1:9b1df0b2507d | 492 | // 1.0000 0.7071-0.7071i -1.0000i -0.7071-0.7071i |
Sergeev | 1:9b1df0b2507d | 493 | p1ap3_0 = p1[0] + p3[0]; |
Sergeev | 1:9b1df0b2507d | 494 | p1sp3_0 = p1[0] - p3[0]; |
Sergeev | 1:9b1df0b2507d | 495 | p1ap3_1 = p1[1] + p3[1]; |
Sergeev | 1:9b1df0b2507d | 496 | p1sp3_1 = p1[1] - p3[1]; |
Sergeev | 1:9b1df0b2507d | 497 | |
Sergeev | 1:9b1df0b2507d | 498 | // col 2 |
Sergeev | 1:9b1df0b2507d | 499 | t2[0] = p1sp3_0 + p2[1] - p4[1]; |
Sergeev | 1:9b1df0b2507d | 500 | t2[1] = p1sp3_1 - p2[0] + p4[0]; |
Sergeev | 1:9b1df0b2507d | 501 | // col 3 |
Sergeev | 1:9b1df0b2507d | 502 | t3[0] = p1ap3_0 - p2[0] - p4[0]; |
Sergeev | 1:9b1df0b2507d | 503 | t3[1] = p1ap3_1 - p2[1] - p4[1]; |
Sergeev | 1:9b1df0b2507d | 504 | // col 4 |
Sergeev | 1:9b1df0b2507d | 505 | t4[0] = p1sp3_0 - p2[1] + p4[1]; |
Sergeev | 1:9b1df0b2507d | 506 | t4[1] = p1sp3_1 + p2[0] - p4[0]; |
Sergeev | 1:9b1df0b2507d | 507 | // col 1 - Top |
Sergeev | 1:9b1df0b2507d | 508 | *p1++ = p1ap3_0 + p2[0] + p4[0]; |
Sergeev | 1:9b1df0b2507d | 509 | *p1++ = p1ap3_1 + p2[1] + p4[1]; |
Sergeev | 1:9b1df0b2507d | 510 | |
Sergeev | 1:9b1df0b2507d | 511 | // COL 2 |
Sergeev | 1:9b1df0b2507d | 512 | twR = tw2[0]; |
Sergeev | 1:9b1df0b2507d | 513 | twI = tw2[1]; |
Sergeev | 1:9b1df0b2507d | 514 | |
Sergeev | 1:9b1df0b2507d | 515 | m0 = t2[0] * twR; |
Sergeev | 1:9b1df0b2507d | 516 | m1 = t2[1] * twI; |
Sergeev | 1:9b1df0b2507d | 517 | m2 = t2[1] * twR; |
Sergeev | 1:9b1df0b2507d | 518 | m3 = t2[0] * twI; |
Sergeev | 1:9b1df0b2507d | 519 | |
Sergeev | 1:9b1df0b2507d | 520 | *p2++ = m0 + m1; |
Sergeev | 1:9b1df0b2507d | 521 | *p2++ = m2 - m3; |
Sergeev | 1:9b1df0b2507d | 522 | // COL 3 |
Sergeev | 1:9b1df0b2507d | 523 | twR = tw3[0]; |
Sergeev | 1:9b1df0b2507d | 524 | twI = tw3[1]; |
Sergeev | 1:9b1df0b2507d | 525 | |
Sergeev | 1:9b1df0b2507d | 526 | m0 = t3[0] * twR; |
Sergeev | 1:9b1df0b2507d | 527 | m1 = t3[1] * twI; |
Sergeev | 1:9b1df0b2507d | 528 | m2 = t3[1] * twR; |
Sergeev | 1:9b1df0b2507d | 529 | m3 = t3[0] * twI; |
Sergeev | 1:9b1df0b2507d | 530 | |
Sergeev | 1:9b1df0b2507d | 531 | *p3++ = m0 + m1; |
Sergeev | 1:9b1df0b2507d | 532 | *p3++ = m2 - m3; |
Sergeev | 1:9b1df0b2507d | 533 | // COL 4 |
Sergeev | 1:9b1df0b2507d | 534 | twR = tw4[0]; |
Sergeev | 1:9b1df0b2507d | 535 | twI = tw4[1]; |
Sergeev | 1:9b1df0b2507d | 536 | |
Sergeev | 1:9b1df0b2507d | 537 | m0 = t4[0] * twR; |
Sergeev | 1:9b1df0b2507d | 538 | m1 = t4[1] * twI; |
Sergeev | 1:9b1df0b2507d | 539 | m2 = t4[1] * twR; |
Sergeev | 1:9b1df0b2507d | 540 | m3 = t4[0] * twI; |
Sergeev | 1:9b1df0b2507d | 541 | |
Sergeev | 1:9b1df0b2507d | 542 | *p4++ = m0 + m1; |
Sergeev | 1:9b1df0b2507d | 543 | *p4++ = m2 - m3; |
Sergeev | 1:9b1df0b2507d | 544 | |
Sergeev | 1:9b1df0b2507d | 545 | // first col |
Sergeev | 1:9b1df0b2507d | 546 | arm_radix8_butterfly_f32( pCol1, L, (float32_t *) S->pTwiddle, 4u); |
Sergeev | 1:9b1df0b2507d | 547 | // second col |
Sergeev | 1:9b1df0b2507d | 548 | arm_radix8_butterfly_f32( pCol2, L, (float32_t *) S->pTwiddle, 4u); |
Sergeev | 1:9b1df0b2507d | 549 | // third col |
Sergeev | 1:9b1df0b2507d | 550 | arm_radix8_butterfly_f32( pCol3, L, (float32_t *) S->pTwiddle, 4u); |
Sergeev | 1:9b1df0b2507d | 551 | // fourth col |
Sergeev | 1:9b1df0b2507d | 552 | arm_radix8_butterfly_f32( pCol4, L, (float32_t *) S->pTwiddle, 4u); |
Sergeev | 1:9b1df0b2507d | 553 | |
Sergeev | 1:9b1df0b2507d | 554 | } |
Sergeev | 1:9b1df0b2507d | 555 | |
Sergeev | 1:9b1df0b2507d | 556 | /** |
Sergeev | 1:9b1df0b2507d | 557 | * @addtogroup ComplexFFT |
Sergeev | 1:9b1df0b2507d | 558 | * @{ |
Sergeev | 1:9b1df0b2507d | 559 | */ |
Sergeev | 1:9b1df0b2507d | 560 | |
Sergeev | 1:9b1df0b2507d | 561 | /** |
Sergeev | 1:9b1df0b2507d | 562 | * @details |
Sergeev | 1:9b1df0b2507d | 563 | * @brief Processing function for the floating-point complex FFT. |
Sergeev | 1:9b1df0b2507d | 564 | * @param[in] *S points to an instance of the floating-point CFFT structure. |
Sergeev | 1:9b1df0b2507d | 565 | * @param[in, out] *p1 points to the complex data buffer of size <code>2*fftLen</code>. Processing occurs in-place. |
Sergeev | 1:9b1df0b2507d | 566 | * @param[in] ifftFlag flag that selects forward (ifftFlag=0) or inverse (ifftFlag=1) transform. |
Sergeev | 1:9b1df0b2507d | 567 | * @param[in] bitReverseFlag flag that enables (bitReverseFlag=1) or disables (bitReverseFlag=0) bit reversal of output. |
Sergeev | 1:9b1df0b2507d | 568 | * @return none. |
Sergeev | 1:9b1df0b2507d | 569 | */ |
Sergeev | 1:9b1df0b2507d | 570 | |
Sergeev | 1:9b1df0b2507d | 571 | void arm_cfft_f32( |
Sergeev | 1:9b1df0b2507d | 572 | const arm_cfft_instance_f32 * S, |
Sergeev | 1:9b1df0b2507d | 573 | float32_t * p1, |
Sergeev | 1:9b1df0b2507d | 574 | uint8_t ifftFlag, |
Sergeev | 1:9b1df0b2507d | 575 | uint8_t bitReverseFlag) |
Sergeev | 1:9b1df0b2507d | 576 | { |
Sergeev | 1:9b1df0b2507d | 577 | |
Sergeev | 1:9b1df0b2507d | 578 | uint32_t L = S->fftLen, l; |
Sergeev | 1:9b1df0b2507d | 579 | float32_t invL, * pSrc; |
Sergeev | 1:9b1df0b2507d | 580 | |
Sergeev | 1:9b1df0b2507d | 581 | if(ifftFlag == 1u) |
Sergeev | 1:9b1df0b2507d | 582 | { |
Sergeev | 1:9b1df0b2507d | 583 | /* Conjugate input data */ |
Sergeev | 1:9b1df0b2507d | 584 | pSrc = p1 + 1; |
Sergeev | 1:9b1df0b2507d | 585 | for(l=0; l<L; l++) { |
Sergeev | 1:9b1df0b2507d | 586 | *pSrc = -*pSrc; |
Sergeev | 1:9b1df0b2507d | 587 | pSrc += 2; |
Sergeev | 1:9b1df0b2507d | 588 | } |
Sergeev | 1:9b1df0b2507d | 589 | } |
Sergeev | 1:9b1df0b2507d | 590 | |
Sergeev | 1:9b1df0b2507d | 591 | switch (L) { |
Sergeev | 1:9b1df0b2507d | 592 | case 16: |
Sergeev | 1:9b1df0b2507d | 593 | case 128: |
Sergeev | 1:9b1df0b2507d | 594 | case 1024: |
Sergeev | 1:9b1df0b2507d | 595 | arm_cfft_radix8by2_f32 ( (arm_cfft_instance_f32 *) S, p1); |
Sergeev | 1:9b1df0b2507d | 596 | break; |
Sergeev | 1:9b1df0b2507d | 597 | case 32: |
Sergeev | 1:9b1df0b2507d | 598 | case 256: |
Sergeev | 1:9b1df0b2507d | 599 | case 2048: |
Sergeev | 1:9b1df0b2507d | 600 | arm_cfft_radix8by4_f32 ( (arm_cfft_instance_f32 *) S, p1); |
Sergeev | 1:9b1df0b2507d | 601 | break; |
Sergeev | 1:9b1df0b2507d | 602 | case 64: |
Sergeev | 1:9b1df0b2507d | 603 | case 512: |
Sergeev | 1:9b1df0b2507d | 604 | case 4096: |
Sergeev | 1:9b1df0b2507d | 605 | arm_radix8_butterfly_f32( p1, L, (float32_t *) S->pTwiddle, 1); |
Sergeev | 1:9b1df0b2507d | 606 | break; |
Sergeev | 1:9b1df0b2507d | 607 | } |
Sergeev | 1:9b1df0b2507d | 608 | |
Sergeev | 1:9b1df0b2507d | 609 | if( bitReverseFlag ) |
Sergeev | 1:9b1df0b2507d | 610 | arm_bitreversal_32((uint32_t*)p1,S->bitRevLength,S->pBitRevTable); |
Sergeev | 1:9b1df0b2507d | 611 | |
Sergeev | 1:9b1df0b2507d | 612 | if(ifftFlag == 1u) |
Sergeev | 1:9b1df0b2507d | 613 | { |
Sergeev | 1:9b1df0b2507d | 614 | invL = 1.0f/(float32_t)L; |
Sergeev | 1:9b1df0b2507d | 615 | /* Conjugate and scale output data */ |
Sergeev | 1:9b1df0b2507d | 616 | pSrc = p1; |
Sergeev | 1:9b1df0b2507d | 617 | for(l=0; l<L; l++) { |
Sergeev | 1:9b1df0b2507d | 618 | *pSrc++ *= invL ; |
Sergeev | 1:9b1df0b2507d | 619 | *pSrc = -(*pSrc) * invL; |
Sergeev | 1:9b1df0b2507d | 620 | pSrc++; |
Sergeev | 1:9b1df0b2507d | 621 | } |
Sergeev | 1:9b1df0b2507d | 622 | } |
Sergeev | 1:9b1df0b2507d | 623 | } |