thomas/CelloEffects
1
0 Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

init

Browse Source
This commit is contained in:
thomas committed 2026-10-18 14:06:40 +02:00
commit 2c2202e60d
2491 files changed
+1295990
Download Patch File Download Diff File Expand all files Collapse all files

No files matched your search

Drivers/CMSIS/DSP/Source/TransformFunctions/arm_bitreversal2.c
+134
View File
@@ -0,0 +1,134 @@
/* ----------------------------------------------------------------------
* Project: CMSIS DSP Library
* Title: arm_bitreversal2.c
* Description: Bitreversal functions
*
* $Date: 23 April 2021
* $Revision: V1.9.0
*
* Target Processor: Cortex-M and Cortex-A cores
* -------------------------------------------------------------------- */
/*
* Copyright (C) 2019 ARM Limited or its affiliates. All rights reserved.
*
* SPDX-License-Identifier: Apache-2.0
*
* Licensed under the Apache License, Version 2.0 (the License); you may
* not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an AS IS BASIS, WITHOUT
* WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
#include "dsp/transform_functions.h"
#include "arm_common_tables.h"
/**
@brief In-place 64 bit reversal function.
@param[in,out] pSrc points to in-place buffer of unknown 64-bit data type
@param[in] bitRevLen bit reversal table length
@param[in] pBitRevTab points to bit reversal table
@return none
*/
void arm_bitreversal_64(
uint64_t *pSrc,
const uint16_t bitRevLen,
const uint16_t *pBitRevTab)
{
uint64_t a, b, i, tmp;
for (i = 0; i < bitRevLen; )
{
a = pBitRevTab[i ] >> 2;
b = pBitRevTab[i + 1] >> 2;
//real
tmp = pSrc[a];
pSrc[a] = pSrc[b];
pSrc[b] = tmp;
//complex
tmp = pSrc[a+1];
pSrc[a+1] = pSrc[b+1];
pSrc[b+1] = tmp;
i += 2;
}
}
/**
@brief In-place 32 bit reversal function.
@param[in,out] pSrc points to in-place buffer of unknown 32-bit data type
@param[in] bitRevLen bit reversal table length
@param[in] pBitRevTab points to bit reversal table
@return none
*/
void arm_bitreversal_32(
uint32_t *pSrc,
const uint16_t bitRevLen,
const uint16_t *pBitRevTab)
{
uint32_t a, b, i, tmp;
for (i = 0; i < bitRevLen; )
{
a = pBitRevTab[i ] >> 2;
b = pBitRevTab[i + 1] >> 2;
//real
tmp = pSrc[a];
pSrc[a] = pSrc[b];
pSrc[b] = tmp;
//complex
tmp = pSrc[a+1];
pSrc[a+1] = pSrc[b+1];
pSrc[b+1] = tmp;
i += 2;
}
}
/**
@brief In-place 16 bit reversal function.
@param[in,out] pSrc points to in-place buffer of unknown 16-bit data type
@param[in] bitRevLen bit reversal table length
@param[in] pBitRevTab points to bit reversal table
@return none
*/
void arm_bitreversal_16(
uint16_t *pSrc,
const uint16_t bitRevLen,
const uint16_t *pBitRevTab)
{
uint16_t a, b, i, tmp;
for (i = 0; i < bitRevLen; )
{
a = pBitRevTab[i ] >> 2;
b = pBitRevTab[i + 1] >> 2;
//real
tmp = pSrc[a];
pSrc[a] = pSrc[b];
pSrc[b] = tmp;
//complex
tmp = pSrc[a+1];
pSrc[a+1] = pSrc[b+1];
pSrc[b+1] = tmp;
i += 2;
}
}
Drivers/CMSIS/DSP/Source/TransformFunctions/arm_cfft_f32.c
+1192
View File
File diff suppressed because it is too large. Load diff
Drivers/CMSIS/DSP/Source/TransformFunctions/arm_cfft_init_f32.c
+358
View File
@@ -0,0 +1,358 @@
/* ----------------------------------------------------------------------
* Project: CMSIS DSP Library
* Title: arm_cfft_init_f32.c
* Description: Initialization function for cfft f32 instance
*
* $Date: 23 April 2021
* $Revision: V1.9.0
*
* Target Processor: Cortex-M and Cortex-A cores
* -------------------------------------------------------------------- */
/*
* Copyright (C) 2010-2021 ARM Limited or its affiliates. All rights reserved.
*
* SPDX-License-Identifier: Apache-2.0
*
* Licensed under the Apache License, Version 2.0 (the License); you may
* not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an AS IS BASIS, WITHOUT
* WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
#define FFTINIT(EXT,SIZE) \
S->bitRevLength = arm_cfft_sR_##EXT##_len##SIZE.bitRevLength; \
S->pBitRevTable = arm_cfft_sR_##EXT##_len##SIZE.pBitRevTable; \
S->pTwiddle = arm_cfft_sR_##EXT##_len##SIZE.pTwiddle;
/**
@addtogroup ComplexFFT
@{
*/
/**
@brief Initialization function for the cfft f32 function
@param[in,out] S points to an instance of the floating-point CFFT structure
@param[in] fftLen fft length (number of complex samples)
@return execution status
- \ref ARM_MATH_SUCCESS : Operation successful
- \ref ARM_MATH_ARGUMENT_ERROR : an error is detected
@par Use of this function is mandatory only for the MVE version of the FFT.
Other versions can still initialize directly the data structure using
variables declared in arm_const_structs.h
*/
#include "dsp/transform_functions.h"
#include "arm_common_tables.h"
#include "arm_const_structs.h"
#if defined(ARM_MATH_MVEF) && !defined(ARM_MATH_AUTOVECTORIZE)
#include "arm_vec_fft.h"
#include "arm_mve_tables.h"
arm_status arm_cfft_radix4by2_rearrange_twiddles_f32(arm_cfft_instance_f32 *S, int twidCoefModifier)
{
switch (S->fftLen >> (twidCoefModifier - 1)) {
#if !defined(ARM_DSP_CONFIG_TABLES) || defined(ARM_ALL_FFT_TABLES) \
|| defined(ARM_TABLE_TWIDDLECOEF_F32_4096)
case 4096U:
S->rearranged_twiddle_tab_stride1_arr = rearranged_twiddle_tab_stride1_arr_4096_f32;
S->rearranged_twiddle_stride1 = rearranged_twiddle_stride1_4096_f32;
S->rearranged_twiddle_tab_stride2_arr = rearranged_twiddle_tab_stride2_arr_4096_f32;
S->rearranged_twiddle_stride2 = rearranged_twiddle_stride2_4096_f32;
S->rearranged_twiddle_tab_stride3_arr = rearranged_twiddle_tab_stride3_arr_4096_f32;
S->rearranged_twiddle_stride3 = rearranged_twiddle_stride3_4096_f32;
break;
#endif
#if !defined(ARM_DSP_CONFIG_TABLES) || defined(ARM_ALL_FFT_TABLES) \
|| defined(ARM_TABLE_TWIDDLECOEF_F32_1024) || defined(ARM_TABLE_TWIDDLECOEF_F32_2048)
case 1024U:
S->rearranged_twiddle_tab_stride1_arr = rearranged_twiddle_tab_stride1_arr_1024_f32;
S->rearranged_twiddle_stride1 = rearranged_twiddle_stride1_1024_f32;
S->rearranged_twiddle_tab_stride2_arr = rearranged_twiddle_tab_stride2_arr_1024_f32;
S->rearranged_twiddle_stride2 = rearranged_twiddle_stride2_1024_f32;
S->rearranged_twiddle_tab_stride3_arr = rearranged_twiddle_tab_stride3_arr_1024_f32;
S->rearranged_twiddle_stride3 = rearranged_twiddle_stride3_1024_f32;
break;
#endif
#if !defined(ARM_DSP_CONFIG_TABLES) || defined(ARM_ALL_FFT_TABLES) \
|| defined(ARM_TABLE_TWIDDLECOEF_F32_256) || defined(ARM_TABLE_TWIDDLECOEF_F32_512)
case 256U:
S->rearranged_twiddle_tab_stride1_arr = rearranged_twiddle_tab_stride1_arr_256_f32;
S->rearranged_twiddle_stride1 = rearranged_twiddle_stride1_256_f32;
S->rearranged_twiddle_tab_stride2_arr = rearranged_twiddle_tab_stride2_arr_256_f32;
S->rearranged_twiddle_stride2 = rearranged_twiddle_stride2_256_f32;
S->rearranged_twiddle_tab_stride3_arr = rearranged_twiddle_tab_stride3_arr_256_f32;
S->rearranged_twiddle_stride3 = rearranged_twiddle_stride3_256_f32;
break;
#endif
#if !defined(ARM_DSP_CONFIG_TABLES) || defined(ARM_ALL_FFT_TABLES) \
|| defined(ARM_TABLE_TWIDDLECOEF_F32_64) || defined(ARM_TABLE_TWIDDLECOEF_F32_128)
case 64U:
S->rearranged_twiddle_tab_stride1_arr = rearranged_twiddle_tab_stride1_arr_64_f32;
S->rearranged_twiddle_stride1 = rearranged_twiddle_stride1_64_f32;
S->rearranged_twiddle_tab_stride2_arr = rearranged_twiddle_tab_stride2_arr_64_f32;
S->rearranged_twiddle_stride2 = rearranged_twiddle_stride2_64_f32;
S->rearranged_twiddle_tab_stride3_arr = rearranged_twiddle_tab_stride3_arr_64_f32;
S->rearranged_twiddle_stride3 = rearranged_twiddle_stride3_64_f32;
break;
#endif
#if !defined(ARM_DSP_CONFIG_TABLES) || defined(ARM_ALL_FFT_TABLES) \
|| defined(ARM_TABLE_TWIDDLECOEF_F32_16) || defined(ARM_TABLE_TWIDDLECOEF_F32_32)
case 16U:
S->rearranged_twiddle_tab_stride1_arr = rearranged_twiddle_tab_stride1_arr_16_f32;
S->rearranged_twiddle_stride1 = rearranged_twiddle_stride1_16_f32;
S->rearranged_twiddle_tab_stride2_arr = rearranged_twiddle_tab_stride2_arr_16_f32;
S->rearranged_twiddle_stride2 = rearranged_twiddle_stride2_16_f32;
S->rearranged_twiddle_tab_stride3_arr = rearranged_twiddle_tab_stride3_arr_16_f32;
S->rearranged_twiddle_stride3 = rearranged_twiddle_stride3_16_f32;
break;
#endif
default:
return(ARM_MATH_ARGUMENT_ERROR);
break;
/* invalid sizes already filtered */
}
return(ARM_MATH_SUCCESS);
}
arm_status arm_cfft_init_f32(
arm_cfft_instance_f32 * S,
uint16_t fftLen)
{
/* Initialise the default arm status */
arm_status status = ARM_MATH_SUCCESS;
/* Initialise the FFT length */
S->fftLen = fftLen;
/* Initialise the Twiddle coefficient pointer */
S->pTwiddle = NULL;
/* Initializations of Instance structure depending on the FFT length */
switch (S->fftLen) {
/* Initializations of structure parameters for 4096 point FFT */
#if !defined(ARM_DSP_CONFIG_TABLES) || defined(ARM_ALL_FFT_TABLES) || (defined(ARM_TABLE_BITREVIDX_FXT_4096) && defined(ARM_TABLE_TWIDDLECOEF_F32_4096))
case 4096U:
/* Initialise the bit reversal table modifier */
S->bitRevLength = ARMBITREVINDEXTABLE_FIXED_4096_TABLE_LENGTH;
S->pBitRevTable = (uint16_t *)armBitRevIndexTable_fixed_4096;
S->pTwiddle = (float32_t *)twiddleCoef_4096;
status=arm_cfft_radix4by2_rearrange_twiddles_f32(S, 1);
break;
#endif
#if !defined(ARM_DSP_CONFIG_TABLES) || defined(ARM_ALL_FFT_TABLES) || (defined(ARM_TABLE_BITREVIDX_FXT_2048) && defined(ARM_TABLE_TWIDDLECOEF_F32_2048))
/* Initializations of structure parameters for 2048 point FFT */
case 2048U:
/* Initialise the bit reversal table modifier */
S->bitRevLength = ARMBITREVINDEXTABLE_FIXED_2048_TABLE_LENGTH;
S->pBitRevTable = (uint16_t *)armBitRevIndexTable_fixed_2048;
S->pTwiddle = (float32_t *)twiddleCoef_2048;
status=arm_cfft_radix4by2_rearrange_twiddles_f32(S, 2);
break;
#endif
#if !defined(ARM_DSP_CONFIG_TABLES) || defined(ARM_ALL_FFT_TABLES) || (defined(ARM_TABLE_BITREVIDX_FXT_1024) && defined(ARM_TABLE_TWIDDLECOEF_F32_1024))
/* Initializations of structure parameters for 1024 point FFT */
case 1024U:
/* Initialise the bit reversal table modifier */
S->bitRevLength = ARMBITREVINDEXTABLE_FIXED_1024_TABLE_LENGTH;
S->pBitRevTable = (uint16_t *)armBitRevIndexTable_fixed_1024;
S->pTwiddle = (float32_t *)twiddleCoef_1024;
status=arm_cfft_radix4by2_rearrange_twiddles_f32(S, 1);
break;
#endif
#if !defined(ARM_DSP_CONFIG_TABLES) || defined(ARM_ALL_FFT_TABLES) || (defined(ARM_TABLE_BITREVIDX_FXT_512) && defined(ARM_TABLE_TWIDDLECOEF_F32_512))
/* Initializations of structure parameters for 512 point FFT */
case 512U:
/* Initialise the bit reversal table modifier */
S->bitRevLength = ARMBITREVINDEXTABLE_FIXED_512_TABLE_LENGTH;
S->pBitRevTable = (uint16_t *)armBitRevIndexTable_fixed_512;
S->pTwiddle = (float32_t *)twiddleCoef_512;
status=arm_cfft_radix4by2_rearrange_twiddles_f32(S, 2);
break;
#endif
#if !defined(ARM_DSP_CONFIG_TABLES) || defined(ARM_ALL_FFT_TABLES) || (defined(ARM_TABLE_BITREVIDX_FXT_256) && defined(ARM_TABLE_TWIDDLECOEF_F32_256))
case 256U:
S->bitRevLength = ARMBITREVINDEXTABLE_FIXED_256_TABLE_LENGTH;
S->pBitRevTable = (uint16_t *)armBitRevIndexTable_fixed_256;
S->pTwiddle = (float32_t *)twiddleCoef_256;
status=arm_cfft_radix4by2_rearrange_twiddles_f32(S, 1);
break;
#endif
#if !defined(ARM_DSP_CONFIG_TABLES) || defined(ARM_ALL_FFT_TABLES) || (defined(ARM_TABLE_BITREVIDX_FXT_128) && defined(ARM_TABLE_TWIDDLECOEF_F32_128))
case 128U:
S->bitRevLength = ARMBITREVINDEXTABLE_FIXED_128_TABLE_LENGTH;
S->pBitRevTable = (uint16_t *)armBitRevIndexTable_fixed_128;
S->pTwiddle = (float32_t *)twiddleCoef_128;
status=arm_cfft_radix4by2_rearrange_twiddles_f32(S, 2);
break;
#endif
#if !defined(ARM_DSP_CONFIG_TABLES) || defined(ARM_ALL_FFT_TABLES) || (defined(ARM_TABLE_BITREVIDX_FXT_64) && defined(ARM_TABLE_TWIDDLECOEF_F32_64))
case 64U:
S->bitRevLength = ARMBITREVINDEXTABLE_FIXED_64_TABLE_LENGTH;
S->pBitRevTable = (uint16_t *)armBitRevIndexTable_fixed_64;
S->pTwiddle = (float32_t *)twiddleCoef_64;
status=arm_cfft_radix4by2_rearrange_twiddles_f32(S, 1);
break;
#endif
#if !defined(ARM_DSP_CONFIG_TABLES) || defined(ARM_ALL_FFT_TABLES) || (defined(ARM_TABLE_BITREVIDX_FXT_32) && defined(ARM_TABLE_TWIDDLECOEF_F32_32))
case 32U:
S->bitRevLength = ARMBITREVINDEXTABLE_FIXED_32_TABLE_LENGTH;
S->pBitRevTable = (uint16_t *)armBitRevIndexTable_fixed_32;
S->pTwiddle = (float32_t *)twiddleCoef_32;
status=arm_cfft_radix4by2_rearrange_twiddles_f32(S, 2);
break;
#endif
#if !defined(ARM_DSP_CONFIG_TABLES) || defined(ARM_ALL_FFT_TABLES) || (defined(ARM_TABLE_BITREVIDX_FXT_16) && defined(ARM_TABLE_TWIDDLECOEF_F32_16))
case 16U:
/* Initializations of structure parameters for 16 point FFT */
S->bitRevLength = ARMBITREVINDEXTABLE_FIXED_16_TABLE_LENGTH;
S->pBitRevTable = (uint16_t *)armBitRevIndexTable_fixed_16;
S->pTwiddle = (float32_t *)twiddleCoef_16;
status=arm_cfft_radix4by2_rearrange_twiddles_f32(S, 1);
break;
#endif
default:
/* Reporting argument error if fftSize is not valid value */
status = ARM_MATH_ARGUMENT_ERROR;
break;
}
return (status);
}
#else
arm_status arm_cfft_init_f32(
arm_cfft_instance_f32 * S,
uint16_t fftLen)
{
/* Initialise the default arm status */
arm_status status = ARM_MATH_SUCCESS;
/* Initialise the FFT length */
S->fftLen = fftLen;
/* Initialise the Twiddle coefficient pointer */
S->pTwiddle = NULL;
/* Initializations of Instance structure depending on the FFT length */
switch (S->fftLen) {
#if !defined(ARM_DSP_CONFIG_TABLES) || defined(ARM_ALL_FFT_TABLES) || (defined(ARM_TABLE_TWIDDLECOEF_F32_4096) && defined(ARM_TABLE_BITREVIDX_FLT_4096))
/* Initializations of structure parameters for 4096 point FFT */
case 4096U:
/* Initialise the bit reversal table modifier */
FFTINIT(f32,4096);
break;
#endif
#if !defined(ARM_DSP_CONFIG_TABLES) || defined(ARM_ALL_FFT_TABLES) || (defined(ARM_TABLE_TWIDDLECOEF_F32_2048) && defined(ARM_TABLE_BITREVIDX_FLT_2048))
/* Initializations of structure parameters for 2048 point FFT */
case 2048U:
/* Initialise the bit reversal table modifier */
FFTINIT(f32,2048);
break;
#endif
#if !defined(ARM_DSP_CONFIG_TABLES) || defined(ARM_ALL_FFT_TABLES) || (defined(ARM_TABLE_TWIDDLECOEF_F32_1024) && defined(ARM_TABLE_BITREVIDX_FLT_1024))
/* Initializations of structure parameters for 1024 point FFT */
case 1024U:
/* Initialise the bit reversal table modifier */
FFTINIT(f32,1024);
break;
#endif
#if !defined(ARM_DSP_CONFIG_TABLES) || defined(ARM_ALL_FFT_TABLES) || (defined(ARM_TABLE_TWIDDLECOEF_F32_512) && defined(ARM_TABLE_BITREVIDX_FLT_512))
/* Initializations of structure parameters for 512 point FFT */
case 512U:
/* Initialise the bit reversal table modifier */
FFTINIT(f32,512);
break;
#endif
#if !defined(ARM_DSP_CONFIG_TABLES) || defined(ARM_ALL_FFT_TABLES) || (defined(ARM_TABLE_TWIDDLECOEF_F32_256) && defined(ARM_TABLE_BITREVIDX_FLT_256))
case 256U:
FFTINIT(f32,256);
break;
#endif
#if !defined(ARM_DSP_CONFIG_TABLES) || defined(ARM_ALL_FFT_TABLES) || (defined(ARM_TABLE_TWIDDLECOEF_F32_128) && defined(ARM_TABLE_BITREVIDX_FLT_128))
case 128U:
FFTINIT(f32,128);
break;
#endif
#if !defined(ARM_DSP_CONFIG_TABLES) || defined(ARM_ALL_FFT_TABLES) || (defined(ARM_TABLE_TWIDDLECOEF_F32_64) && defined(ARM_TABLE_BITREVIDX_FLT_64))
case 64U:
FFTINIT(f32,64);
break;
#endif
#if !defined(ARM_DSP_CONFIG_TABLES) || defined(ARM_ALL_FFT_TABLES) || (defined(ARM_TABLE_TWIDDLECOEF_F32_32) && defined(ARM_TABLE_BITREVIDX_FLT_32))
case 32U:
FFTINIT(f32,32);
break;
#endif
#if !defined(ARM_DSP_CONFIG_TABLES) || defined(ARM_ALL_FFT_TABLES) || (defined(ARM_TABLE_TWIDDLECOEF_F32_16) && defined(ARM_TABLE_BITREVIDX_FLT_16))
case 16U:
/* Initializations of structure parameters for 16 point FFT */
FFTINIT(f32,16);
break;
#endif
default:
/* Reporting argument error if fftSize is not valid value */
status = ARM_MATH_ARGUMENT_ERROR;
break;
}
return (status);
}
#endif /* defined(ARM_MATH_MVEF) && !defined(ARM_MATH_AUTOVECTORIZE) */
/**
@} end of ComplexFFT group
*/
Drivers/CMSIS/DSP/Source/TransformFunctions/arm_cfft_radix8_f32.c
+285
View File
@@ -0,0 +1,285 @@
/* ----------------------------------------------------------------------
* Project: CMSIS DSP Library
* Title: arm_cfft_radix8_f32.c
* Description: Radix-8 Decimation in Frequency CFFT & CIFFT Floating point processing function
*
* $Date: 23 April 2021
* $Revision: V1.9.0
*
* Target Processor: Cortex-M and Cortex-A cores
* -------------------------------------------------------------------- */
/*
* Copyright (C) 2010-2021 ARM Limited or its affiliates. All rights reserved.
*
* SPDX-License-Identifier: Apache-2.0
*
* Licensed under the Apache License, Version 2.0 (the License); you may
* not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an AS IS BASIS, WITHOUT
* WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
#include "dsp/transform_functions.h"
/* ----------------------------------------------------------------------
* Internal helper function used by the FFTs
* -------------------------------------------------------------------- */
/**
brief Core function for the floating-point CFFT butterfly process.
param[in,out] pSrc points to the in-place buffer of floating-point data type.
param[in] fftLen length of the FFT.
param[in] pCoef points to the twiddle coefficient buffer.
param[in] twidCoefModifier twiddle coefficient modifier that supports different size FFTs with the same twiddle factor table.
return none
*/
void arm_radix8_butterfly_f32(
float32_t * pSrc,
uint16_t fftLen,
const float32_t * pCoef,
uint16_t twidCoefModifier)
{
uint32_t ia1, ia2, ia3, ia4, ia5, ia6, ia7;
uint32_t i1, i2, i3, i4, i5, i6, i7, i8;
uint32_t id;
uint32_t n1, n2, j;
float32_t r1, r2, r3, r4, r5, r6, r7, r8;
float32_t t1, t2;
float32_t s1, s2, s3, s4, s5, s6, s7, s8;
float32_t p1, p2, p3, p4;
float32_t co2, co3, co4, co5, co6, co7, co8;
float32_t si2, si3, si4, si5, si6, si7, si8;
const float32_t C81 = 0.70710678118f;
n2 = fftLen;
do
{
n1 = n2;
n2 = n2 >> 3;
i1 = 0;
do
{
i2 = i1 + n2;
i3 = i2 + n2;
i4 = i3 + n2;
i5 = i4 + n2;
i6 = i5 + n2;
i7 = i6 + n2;
i8 = i7 + n2;
r1 = pSrc[2 * i1] + pSrc[2 * i5];
r5 = pSrc[2 * i1] - pSrc[2 * i5];
r2 = pSrc[2 * i2] + pSrc[2 * i6];
r6 = pSrc[2 * i2] - pSrc[2 * i6];
r3 = pSrc[2 * i3] + pSrc[2 * i7];
r7 = pSrc[2 * i3] - pSrc[2 * i7];
r4 = pSrc[2 * i4] + pSrc[2 * i8];
r8 = pSrc[2 * i4] - pSrc[2 * i8];
t1 = r1 - r3;
r1 = r1 + r3;
r3 = r2 - r4;
r2 = r2 + r4;
pSrc[2 * i1] = r1 + r2;
pSrc[2 * i5] = r1 - r2;
r1 = pSrc[2 * i1 + 1] + pSrc[2 * i5 + 1];
s5 = pSrc[2 * i1 + 1] - pSrc[2 * i5 + 1];
r2 = pSrc[2 * i2 + 1] + pSrc[2 * i6 + 1];
s6 = pSrc[2 * i2 + 1] - pSrc[2 * i6 + 1];
s3 = pSrc[2 * i3 + 1] + pSrc[2 * i7 + 1];
s7 = pSrc[2 * i3 + 1] - pSrc[2 * i7 + 1];
r4 = pSrc[2 * i4 + 1] + pSrc[2 * i8 + 1];
s8 = pSrc[2 * i4 + 1] - pSrc[2 * i8 + 1];
t2 = r1 - s3;
r1 = r1 + s3;
s3 = r2 - r4;
r2 = r2 + r4;
pSrc[2 * i1 + 1] = r1 + r2;
pSrc[2 * i5 + 1] = r1 - r2;
pSrc[2 * i3] = t1 + s3;
pSrc[2 * i7] = t1 - s3;
pSrc[2 * i3 + 1] = t2 - r3;
pSrc[2 * i7 + 1] = t2 + r3;
r1 = (r6 - r8) * C81;
r6 = (r6 + r8) * C81;
r2 = (s6 - s8) * C81;
s6 = (s6 + s8) * C81;
t1 = r5 - r1;
r5 = r5 + r1;
r8 = r7 - r6;
r7 = r7 + r6;
t2 = s5 - r2;
s5 = s5 + r2;
s8 = s7 - s6;
s7 = s7 + s6;
pSrc[2 * i2] = r5 + s7;
pSrc[2 * i8] = r5 - s7;
pSrc[2 * i6] = t1 + s8;
pSrc[2 * i4] = t1 - s8;
pSrc[2 * i2 + 1] = s5 - r7;
pSrc[2 * i8 + 1] = s5 + r7;
pSrc[2 * i6 + 1] = t2 - r8;
pSrc[2 * i4 + 1] = t2 + r8;
i1 += n1;
} while (i1 < fftLen);
if (n2 < 8)
break;
ia1 = 0;
j = 1;
do
{
/* index calculation for the coefficients */
id = ia1 + twidCoefModifier;
ia1 = id;
ia2 = ia1 + id;
ia3 = ia2 + id;
ia4 = ia3 + id;
ia5 = ia4 + id;
ia6 = ia5 + id;
ia7 = ia6 + id;
co2 = pCoef[2 * ia1];
co3 = pCoef[2 * ia2];
co4 = pCoef[2 * ia3];
co5 = pCoef[2 * ia4];
co6 = pCoef[2 * ia5];
co7 = pCoef[2 * ia6];
co8 = pCoef[2 * ia7];
si2 = pCoef[2 * ia1 + 1];
si3 = pCoef[2 * ia2 + 1];
si4 = pCoef[2 * ia3 + 1];
si5 = pCoef[2 * ia4 + 1];
si6 = pCoef[2 * ia5 + 1];
si7 = pCoef[2 * ia6 + 1];
si8 = pCoef[2 * ia7 + 1];
i1 = j;
do
{
/* index calculation for the input */
i2 = i1 + n2;
i3 = i2 + n2;
i4 = i3 + n2;
i5 = i4 + n2;
i6 = i5 + n2;
i7 = i6 + n2;
i8 = i7 + n2;
r1 = pSrc[2 * i1] + pSrc[2 * i5];
r5 = pSrc[2 * i1] - pSrc[2 * i5];
r2 = pSrc[2 * i2] + pSrc[2 * i6];
r6 = pSrc[2 * i2] - pSrc[2 * i6];
r3 = pSrc[2 * i3] + pSrc[2 * i7];
r7 = pSrc[2 * i3] - pSrc[2 * i7];
r4 = pSrc[2 * i4] + pSrc[2 * i8];
r8 = pSrc[2 * i4] - pSrc[2 * i8];
t1 = r1 - r3;
r1 = r1 + r3;
r3 = r2 - r4;
r2 = r2 + r4;
pSrc[2 * i1] = r1 + r2;
r2 = r1 - r2;
s1 = pSrc[2 * i1 + 1] + pSrc[2 * i5 + 1];
s5 = pSrc[2 * i1 + 1] - pSrc[2 * i5 + 1];
s2 = pSrc[2 * i2 + 1] + pSrc[2 * i6 + 1];
s6 = pSrc[2 * i2 + 1] - pSrc[2 * i6 + 1];
s3 = pSrc[2 * i3 + 1] + pSrc[2 * i7 + 1];
s7 = pSrc[2 * i3 + 1] - pSrc[2 * i7 + 1];
s4 = pSrc[2 * i4 + 1] + pSrc[2 * i8 + 1];
s8 = pSrc[2 * i4 + 1] - pSrc[2 * i8 + 1];
t2 = s1 - s3;
s1 = s1 + s3;
s3 = s2 - s4;
s2 = s2 + s4;
r1 = t1 + s3;
t1 = t1 - s3;
pSrc[2 * i1 + 1] = s1 + s2;
s2 = s1 - s2;
s1 = t2 - r3;
t2 = t2 + r3;
p1 = co5 * r2;
p2 = si5 * s2;
p3 = co5 * s2;
p4 = si5 * r2;
pSrc[2 * i5] = p1 + p2;
pSrc[2 * i5 + 1] = p3 - p4;
p1 = co3 * r1;
p2 = si3 * s1;
p3 = co3 * s1;
p4 = si3 * r1;
pSrc[2 * i3] = p1 + p2;
pSrc[2 * i3 + 1] = p3 - p4;
p1 = co7 * t1;
p2 = si7 * t2;
p3 = co7 * t2;
p4 = si7 * t1;
pSrc[2 * i7] = p1 + p2;
pSrc[2 * i7 + 1] = p3 - p4;
r1 = (r6 - r8) * C81;
r6 = (r6 + r8) * C81;
s1 = (s6 - s8) * C81;
s6 = (s6 + s8) * C81;
t1 = r5 - r1;
r5 = r5 + r1;
r8 = r7 - r6;
r7 = r7 + r6;
t2 = s5 - s1;
s5 = s5 + s1;
s8 = s7 - s6;
s7 = s7 + s6;
r1 = r5 + s7;
r5 = r5 - s7;
r6 = t1 + s8;
t1 = t1 - s8;
s1 = s5 - r7;
s5 = s5 + r7;
s6 = t2 - r8;
t2 = t2 + r8;
p1 = co2 * r1;
p2 = si2 * s1;
p3 = co2 * s1;
p4 = si2 * r1;
pSrc[2 * i2] = p1 + p2;
pSrc[2 * i2 + 1] = p3 - p4;
p1 = co8 * r5;
p2 = si8 * s5;
p3 = co8 * s5;
p4 = si8 * r5;
pSrc[2 * i8] = p1 + p2;
pSrc[2 * i8 + 1] = p3 - p4;
p1 = co6 * r6;
p2 = si6 * s6;
p3 = co6 * s6;
p4 = si6 * r6;
pSrc[2 * i6] = p1 + p2;
pSrc[2 * i6 + 1] = p3 - p4;
p1 = co4 * t1;
p2 = si4 * t2;
p3 = co4 * t2;
p4 = si4 * t1;
pSrc[2 * i4] = p1 + p2;
pSrc[2 * i4 + 1] = p3 - p4;
i1 += n1;
} while (i1 < fftLen);
j++;
} while (j < n2);
twidCoefModifier <<= 3;
} while (n2 > 7);
}
Drivers/CMSIS/DSP/Source/TransformFunctions/arm_rfft_fast_f32.c
+603
View File
@@ -0,0 +1,603 @@
/* ----------------------------------------------------------------------
* Project: CMSIS DSP Library
* Title: arm_rfft_fast_f32.c
* Description: RFFT & RIFFT Floating point process function
*
* $Date: 23 April 2021
* $Revision: V1.9.0
*
* Target Processor: Cortex-M and Cortex-A cores
* -------------------------------------------------------------------- */
/*
* Copyright (C) 2010-2021 ARM Limited or its affiliates. All rights reserved.
*
* SPDX-License-Identifier: Apache-2.0
*
* Licensed under the Apache License, Version 2.0 (the License); you may
* not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an AS IS BASIS, WITHOUT
* WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
#include "dsp/transform_functions.h"
#if defined(ARM_MATH_MVEF) && !defined(ARM_MATH_AUTOVECTORIZE)
void stage_rfft_f32(
const arm_rfft_fast_instance_f32 * S,
float32_t * p,
float32_t * pOut)
{
int32_t k; /* Loop Counter */
float32_t twR, twI; /* RFFT Twiddle coefficients */
const float32_t * pCoeff = S->pTwiddleRFFT; /* Points to RFFT Twiddle factors */
float32_t *pA = p; /* increasing pointer */
float32_t *pB = p; /* decreasing pointer */
float32_t xAR, xAI, xBR, xBI; /* temporary variables */
float32_t t1a, t1b; /* temporary variables */
float32_t p0, p1, p2, p3; /* temporary variables */
float32x4x2_t tw,xA,xB;
float32x4x2_t tmp1, tmp2, res;
uint32x4_t vecStridesFwd, vecStridesBkwd;
vecStridesFwd = vidupq_u32((uint32_t)0, 2);
vecStridesBkwd = -vecStridesFwd;
int blockCnt;
k = (S->Sint).fftLen - 1;
/* Pack first and last sample of the frequency domain together */
xBR = pB[0];
xBI = pB[1];
xAR = pA[0];
xAI = pA[1];
twR = *pCoeff++ ;
twI = *pCoeff++ ;
// U1 = XA(1) + XB(1); % It is real
t1a = xBR + xAR ;
// U2 = XB(1) - XA(1); % It is imaginary
t1b = xBI + xAI ;
// real(tw * (xB - xA)) = twR * (xBR - xAR) - twI * (xBI - xAI);
// imag(tw * (xB - xA)) = twI * (xBR - xAR) + twR * (xBI - xAI);
*pOut++ = 0.5f * ( t1a + t1b );
*pOut++ = 0.5f * ( t1a - t1b );
// XA(1) = 1/2*( U1 - imag(U2) + i*( U1 +imag(U2) ));
pB = p + 2*k;
pA += 2;
blockCnt = k >> 2;
while (blockCnt > 0)
{
/*
function X = my_split_rfft(X, ifftFlag)
% X is a series of real numbers
L = length(X);
XC = X(1:2:end) +i*X(2:2:end);
XA = fft(XC);
XB = conj(XA([1 end:-1:2]));
TW = i*exp(-2*pi*i*[0:L/2-1]/L).';
for l = 2:L/2
XA(l) = 1/2 * (XA(l) + XB(l) + TW(l) * (XB(l) - XA(l)));
end
XA(1) = 1/2* (XA(1) + XB(1) + TW(1) * (XB(1) - XA(1))) + i*( 1/2*( XA(1) + XB(1) + i*( XA(1) - XB(1))));
X = XA;
*/
xA = vld2q_f32(pA);
pA += 8;
xB = vld2q_f32(pB);
xB.val[0] = vldrwq_gather_shifted_offset_f32(pB, vecStridesBkwd);
xB.val[1] = vldrwq_gather_shifted_offset_f32(&pB[1], vecStridesBkwd);
xB.val[1] = vnegq_f32(xB.val[1]);
pB -= 8;
tw = vld2q_f32(pCoeff);
pCoeff += 8;
tmp1.val[0] = vaddq_f32(xA.val[0],xB.val[0]);
tmp1.val[1] = vaddq_f32(xA.val[1],xB.val[1]);
tmp2.val[0] = vsubq_f32(xB.val[0],xA.val[0]);
tmp2.val[1] = vsubq_f32(xB.val[1],xA.val[1]);
res.val[0] = vmulq(tw.val[0], tmp2.val[0]);
res.val[0] = vfmsq(res.val[0],tw.val[1], tmp2.val[1]);
res.val[1] = vmulq(tw.val[0], tmp2.val[1]);
res.val[1] = vfmaq(res.val[1], tw.val[1], tmp2.val[0]);
res.val[0] = vaddq_f32(res.val[0],tmp1.val[0] );
res.val[1] = vaddq_f32(res.val[1],tmp1.val[1] );
res.val[0] = vmulq_n_f32(res.val[0], 0.5f);
res.val[1] = vmulq_n_f32(res.val[1], 0.5f);
vst2q_f32(pOut, res);
pOut += 8;
blockCnt--;
}
blockCnt = k & 3;
while (blockCnt > 0)
{
/*
function X = my_split_rfft(X, ifftFlag)
% X is a series of real numbers
L = length(X);
XC = X(1:2:end) +i*X(2:2:end);
XA = fft(XC);
XB = conj(XA([1 end:-1:2]));
TW = i*exp(-2*pi*i*[0:L/2-1]/L).';
for l = 2:L/2
XA(l) = 1/2 * (XA(l) + XB(l) + TW(l) * (XB(l) - XA(l)));
end
XA(1) = 1/2* (XA(1) + XB(1) + TW(1) * (XB(1) - XA(1))) + i*( 1/2*( XA(1) + XB(1) + i*( XA(1) - XB(1))));
X = XA;
*/
xBI = pB[1];
xBR = pB[0];
xAR = pA[0];
xAI = pA[1];
twR = *pCoeff++;
twI = *pCoeff++;
t1a = xBR - xAR ;
t1b = xBI + xAI ;
// real(tw * (xB - xA)) = twR * (xBR - xAR) - twI * (xBI - xAI);
// imag(tw * (xB - xA)) = twI * (xBR - xAR) + twR * (xBI - xAI);
p0 = twR * t1a;
p1 = twI * t1a;
p2 = twR * t1b;
p3 = twI * t1b;
*pOut++ = 0.5f * (xAR + xBR + p0 + p3 ); //xAR
*pOut++ = 0.5f * (xAI - xBI + p1 - p2 ); //xAI
pA += 2;
pB -= 2;
blockCnt--;
}
}
/* Prepares data for inverse cfft */
void merge_rfft_f32(
const arm_rfft_fast_instance_f32 * S,
float32_t * p,
float32_t * pOut)
{
int32_t k; /* Loop Counter */
float32_t twR, twI; /* RFFT Twiddle coefficients */
const float32_t *pCoeff = S->pTwiddleRFFT; /* Points to RFFT Twiddle factors */
float32_t *pA = p; /* increasing pointer */
float32_t *pB = p; /* decreasing pointer */
float32_t xAR, xAI, xBR, xBI; /* temporary variables */
float32_t t1a, t1b, r, s, t, u; /* temporary variables */
float32x4x2_t tw,xA,xB;
float32x4x2_t tmp1, tmp2, res;
uint32x4_t vecStridesFwd, vecStridesBkwd;
vecStridesFwd = vidupq_u32((uint32_t)0, 2);
vecStridesBkwd = -vecStridesFwd;
int blockCnt;
k = (S->Sint).fftLen - 1;
xAR = pA[0];
xAI = pA[1];
pCoeff += 2 ;
*pOut++ = 0.5f * ( xAR + xAI );
*pOut++ = 0.5f * ( xAR - xAI );
pB = p + 2*k ;
pA += 2 ;
blockCnt = k >> 2;
while (blockCnt > 0)
{
/* G is half of the frequency complex spectrum */
//for k = 2:N
// Xk(k) = 1/2 * (G(k) + conj(G(N-k+2)) + Tw(k)*( G(k) - conj(G(N-k+2))));
xA = vld2q_f32(pA);
pA += 8;
xB = vld2q_f32(pB);
xB.val[0] = vldrwq_gather_shifted_offset_f32(pB, vecStridesBkwd);
xB.val[1] = vldrwq_gather_shifted_offset_f32(&pB[1], vecStridesBkwd);
xB.val[1] = vnegq_f32(xB.val[1]);
pB -= 8;
tw = vld2q_f32(pCoeff);
tw.val[1] = vnegq_f32(tw.val[1]);
pCoeff += 8;
tmp1.val[0] = vaddq_f32(xA.val[0],xB.val[0]);
tmp1.val[1] = vaddq_f32(xA.val[1],xB.val[1]);
tmp2.val[0] = vsubq_f32(xB.val[0],xA.val[0]);
tmp2.val[1] = vsubq_f32(xB.val[1],xA.val[1]);
res.val[0] = vmulq(tw.val[0], tmp2.val[0]);
res.val[0] = vfmsq(res.val[0],tw.val[1], tmp2.val[1]);
res.val[1] = vmulq(tw.val[0], tmp2.val[1]);
res.val[1] = vfmaq(res.val[1], tw.val[1], tmp2.val[0]);
res.val[0] = vaddq_f32(res.val[0],tmp1.val[0] );
res.val[1] = vaddq_f32(res.val[1],tmp1.val[1] );
res.val[0] = vmulq_n_f32(res.val[0], 0.5f);
res.val[1] = vmulq_n_f32(res.val[1], 0.5f);
vst2q_f32(pOut, res);
pOut += 8;
blockCnt--;
}
blockCnt = k & 3;
while (blockCnt > 0)
{
/* G is half of the frequency complex spectrum */
//for k = 2:N
// Xk(k) = 1/2 * (G(k) + conj(G(N-k+2)) + Tw(k)*( G(k) - conj(G(N-k+2))));
xBI = pB[1] ;
xBR = pB[0] ;
xAR = pA[0];
xAI = pA[1];
twR = *pCoeff++;
twI = *pCoeff++;
t1a = xAR - xBR ;
t1b = xAI + xBI ;
r = twR * t1a;
s = twI * t1b;
t = twI * t1a;
u = twR * t1b;
// real(tw * (xA - xB)) = twR * (xAR - xBR) - twI * (xAI - xBI);
// imag(tw * (xA - xB)) = twI * (xAR - xBR) + twR * (xAI - xBI);
*pOut++ = 0.5f * (xAR + xBR - r - s ); //xAR
*pOut++ = 0.5f * (xAI - xBI + t - u ); //xAI
pA += 2;
pB -= 2;
blockCnt--;
}
}
#else
void stage_rfft_f32(
const arm_rfft_fast_instance_f32 * S,
float32_t * p,
float32_t * pOut)
{
int32_t k; /* Loop Counter */
float32_t twR, twI; /* RFFT Twiddle coefficients */
const float32_t * pCoeff = S->pTwiddleRFFT; /* Points to RFFT Twiddle factors */
float32_t *pA = p; /* increasing pointer */
float32_t *pB = p; /* decreasing pointer */
float32_t xAR, xAI, xBR, xBI; /* temporary variables */
float32_t t1a, t1b; /* temporary variables */
float32_t p0, p1, p2, p3; /* temporary variables */
k = (S->Sint).fftLen - 1;
/* Pack first and last sample of the frequency domain together */
xBR = pB[0];
xBI = pB[1];
xAR = pA[0];
xAI = pA[1];
twR = *pCoeff++ ;
twI = *pCoeff++ ;
// U1 = XA(1) + XB(1); % It is real
t1a = xBR + xAR ;
// U2 = XB(1) - XA(1); % It is imaginary
t1b = xBI + xAI ;
// real(tw * (xB - xA)) = twR * (xBR - xAR) - twI * (xBI - xAI);
// imag(tw * (xB - xA)) = twI * (xBR - xAR) + twR * (xBI - xAI);
*pOut++ = 0.5f * ( t1a + t1b );
*pOut++ = 0.5f * ( t1a - t1b );
// XA(1) = 1/2*( U1 - imag(U2) + i*( U1 +imag(U2) ));
pB = p + 2*k;
pA += 2;
do
{
/*
function X = my_split_rfft(X, ifftFlag)
% X is a series of real numbers
L = length(X);
XC = X(1:2:end) +i*X(2:2:end);
XA = fft(XC);
XB = conj(XA([1 end:-1:2]));
TW = i*exp(-2*pi*i*[0:L/2-1]/L).';
for l = 2:L/2
XA(l) = 1/2 * (XA(l) + XB(l) + TW(l) * (XB(l) - XA(l)));
end
XA(1) = 1/2* (XA(1) + XB(1) + TW(1) * (XB(1) - XA(1))) + i*( 1/2*( XA(1) + XB(1) + i*( XA(1) - XB(1))));
X = XA;
*/
xBI = pB[1];
xBR = pB[0];
xAR = pA[0];
xAI = pA[1];
twR = *pCoeff++;
twI = *pCoeff++;
t1a = xBR - xAR ;
t1b = xBI + xAI ;
// real(tw * (xB - xA)) = twR * (xBR - xAR) - twI * (xBI - xAI);
// imag(tw * (xB - xA)) = twI * (xBR - xAR) + twR * (xBI - xAI);
p0 = twR * t1a;
p1 = twI * t1a;
p2 = twR * t1b;
p3 = twI * t1b;
*pOut++ = 0.5f * (xAR + xBR + p0 + p3 ); //xAR
*pOut++ = 0.5f * (xAI - xBI + p1 - p2 ); //xAI
pA += 2;
pB -= 2;
k--;
} while (k > 0);
}
/* Prepares data for inverse cfft */
void merge_rfft_f32(
const arm_rfft_fast_instance_f32 * S,
float32_t * p,
float32_t * pOut)
{
int32_t k; /* Loop Counter */
float32_t twR, twI; /* RFFT Twiddle coefficients */
const float32_t *pCoeff = S->pTwiddleRFFT; /* Points to RFFT Twiddle factors */
float32_t *pA = p; /* increasing pointer */
float32_t *pB = p; /* decreasing pointer */
float32_t xAR, xAI, xBR, xBI; /* temporary variables */
float32_t t1a, t1b, r, s, t, u; /* temporary variables */
k = (S->Sint).fftLen - 1;
xAR = pA[0];
xAI = pA[1];
pCoeff += 2 ;
*pOut++ = 0.5f * ( xAR + xAI );
*pOut++ = 0.5f * ( xAR - xAI );
pB = p + 2*k ;
pA += 2 ;
while (k > 0)
{
/* G is half of the frequency complex spectrum */
//for k = 2:N
// Xk(k) = 1/2 * (G(k) + conj(G(N-k+2)) + Tw(k)*( G(k) - conj(G(N-k+2))));
xBI = pB[1] ;
xBR = pB[0] ;
xAR = pA[0];
xAI = pA[1];
twR = *pCoeff++;
twI = *pCoeff++;
t1a = xAR - xBR ;
t1b = xAI + xBI ;
r = twR * t1a;
s = twI * t1b;
t = twI * t1a;
u = twR * t1b;
// real(tw * (xA - xB)) = twR * (xAR - xBR) - twI * (xAI - xBI);
// imag(tw * (xA - xB)) = twI * (xAR - xBR) + twR * (xAI - xBI);
*pOut++ = 0.5f * (xAR + xBR - r - s ); //xAR
*pOut++ = 0.5f * (xAI - xBI + t - u ); //xAI
pA += 2;
pB -= 2;
k--;
}
}
#endif /* defined(ARM_MATH_MVEF) && !defined(ARM_MATH_AUTOVECTORIZE) */
/**
@ingroup groupTransforms
*/
/**
@defgroup RealFFT Real FFT Functions
@par
The CMSIS DSP library includes specialized algorithms for computing the
FFT of real data sequences. The FFT is defined over complex data but
in many applications the input is real. Real FFT algorithms take advantage
of the symmetry properties of the FFT and have a speed advantage over complex
algorithms of the same length.
@par
The Fast RFFT algorithm relays on the mixed radix CFFT that save processor usage.
@par
The real length N forward FFT of a sequence is computed using the steps shown below.
@par
\image html RFFT.gif "Real Fast Fourier Transform"
@par
The real sequence is initially treated as if it were complex to perform a CFFT.
Later, a processing stage reshapes the data to obtain half of the frequency spectrum
in complex format. Except the first complex number that contains the two real numbers
X[0] and X[N/2] all the data is complex. In other words, the first complex sample
contains two real values packed.
@par
The input for the inverse RFFT should keep the same format as the output of the
forward RFFT. A first processing stage pre-process the data to later perform an
inverse CFFT.
@par
\image html RIFFT.gif "Real Inverse Fast Fourier Transform"
@par
The algorithms for floating-point, Q15, and Q31 data are slightly different
and we describe each algorithm in turn.
@par Floating-point
The main functions are \ref arm_rfft_fast_f32() and \ref arm_rfft_fast_init_f32().
The older functions \ref arm_rfft_f32() and \ref arm_rfft_init_f32() have been deprecated
but are still documented.
@par
The FFT of a real N-point sequence has even symmetry in the frequency domain.
The second half of the data equals the conjugate of the first half flipped in frequency.
Looking at the data, we see that we can uniquely represent the FFT using only N/2 complex numbers.
These are packed into the output array in alternating real and imaginary components:
@par
X = { real[0], imag[0], real[1], imag[1], real[2], imag[2] ...
real[(N/2)-1], imag[(N/2)-1 }
@par
It happens that the first complex number (real[0], imag[0]) is actually
all real. real[0] represents the DC offset, and imag[0] should be 0.
(real[1], imag[1]) is the fundamental frequency, (real[2], imag[2]) is
the first harmonic and so on.
@par
The real FFT functions pack the frequency domain data in this fashion.
The forward transform outputs the data in this form and the inverse
transform expects input data in this form. The function always performs
the needed bitreversal so that the input and output data is always in
normal order. The functions support lengths of [32, 64, 128, ..., 4096]
samples.
@par Q15 and Q31
The real algorithms are defined in a similar manner and utilize N/2 complex
transforms behind the scenes.
@par
The complex transforms used internally include scaling to prevent fixed-point
overflows. The overall scaling equals 1/(fftLen/2).
Due to the use of complex transform internally, the source buffer is
modified by the rfft.
@par
A separate instance structure must be defined for each transform used but
twiddle factor and bit reversal tables can be reused.
@par
There is also an associated initialization function for each data type.
The initialization function performs the following operations:
- Sets the values of the internal structure fields.
- Initializes twiddle factor table and bit reversal table pointers.
- Initializes the internal complex FFT data structure.
@par
Use of the initialization function is optional **except for MVE versions where it is mandatory**.
If you don't use the initialization functions, then the structures should be initialized with code
similar to the one below:
<pre>
arm_rfft_instance_q31 S = {fftLenReal, fftLenBy2, ifftFlagR, bitReverseFlagR, twidCoefRModifier, pTwiddleAReal, pTwiddleBReal, pCfft};
arm_rfft_instance_q15 S = {fftLenReal, fftLenBy2, ifftFlagR, bitReverseFlagR, twidCoefRModifier, pTwiddleAReal, pTwiddleBReal, pCfft};
</pre>
where <code>fftLenReal</code> is the length of the real transform;
<code>fftLenBy2</code> length of the internal complex transform (fftLenReal/2).
<code>ifftFlagR</code> Selects forward (=0) or inverse (=1) transform.
<code>bitReverseFlagR</code> Selects bit reversed output (=0) or normal order
output (=1).
<code>twidCoefRModifier</code> stride modifier for the twiddle factor table.
The value is based on the FFT length;
<code>pTwiddleAReal</code>points to the A array of twiddle coefficients;
<code>pTwiddleBReal</code>points to the B array of twiddle coefficients;
<code>pCfft</code> points to the CFFT Instance structure. The CFFT structure
must also be initialized.
@par
Note that with MVE versions you can't initialize instance structures directly and **must
use the initialization function**.
*/
/**
@addtogroup RealFFT
@{
*/
/**
@brief Processing function for the floating-point real FFT.
@param[in] S points to an arm_rfft_fast_instance_f32 structure
@param[in] p points to input buffer (Source buffer is modified by this function.)
@param[in] pOut points to output buffer
@param[in] ifftFlag
- value = 0: RFFT
- value = 1: RIFFT
@return none
*/
void arm_rfft_fast_f32(
const arm_rfft_fast_instance_f32 * S,
float32_t * p,
float32_t * pOut,
uint8_t ifftFlag)
{
const arm_cfft_instance_f32 * Sint = &(S->Sint);
/* Calculation of Real FFT */
if (ifftFlag)
{
/* Real FFT compression */
merge_rfft_f32(S, p, pOut);
/* Complex radix-4 IFFT process */
arm_cfft_f32( Sint, pOut, ifftFlag, 1);
}
else
{
/* Calculation of RFFT of input */
arm_cfft_f32( Sint, p, ifftFlag, 1);
/* Real FFT extraction */
stage_rfft_f32(S, p, pOut);
}
}
/**
* @} end of RealFFT group
*/
Drivers/CMSIS/DSP/Source/TransformFunctions/arm_rfft_fast_init_f32.c
+352
View File
@@ -0,0 +1,352 @@
/* ----------------------------------------------------------------------
* Project: CMSIS DSP Library
* Title: arm_rfft_fast_init_f32.c
* Description: Split Radix Decimation in Frequency CFFT Floating point processing function
*
* $Date: 23 April 2021
* $Revision: V1.9.0
*
* Target Processor: Cortex-M and Cortex-A cores
* -------------------------------------------------------------------- */
/*
* Copyright (C) 2010-2021 ARM Limited or its affiliates. All rights reserved.
*
* SPDX-License-Identifier: Apache-2.0
*
* Licensed under the Apache License, Version 2.0 (the License); you may
* not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an AS IS BASIS, WITHOUT
* WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
#include "dsp/transform_functions.h"
#include "arm_common_tables.h"
/**
@ingroup groupTransforms
*/
/**
@addtogroup RealFFT
@{
*/
#if !defined(ARM_DSP_CONFIG_TABLES) || defined(ARM_ALL_FFT_TABLES) || (defined(ARM_TABLE_TWIDDLECOEF_F32_16) && defined(ARM_TABLE_BITREVIDX_FLT_16) && defined(ARM_TABLE_TWIDDLECOEF_RFFT_F32_32))
/**
@private
@brief Initialization function for the 32pt floating-point real FFT.
@param[in,out] S points to an arm_rfft_fast_instance_f32 structure
@return execution status
- \ref ARM_MATH_SUCCESS : Operation successful
- \ref ARM_MATH_ARGUMENT_ERROR : an error is detected
*/
static arm_status arm_rfft_32_fast_init_f32( arm_rfft_fast_instance_f32 * S ) {
arm_status status;
if( !S ) return ARM_MATH_ARGUMENT_ERROR;
status=arm_cfft_init_f32(&(S->Sint),16);
if (status != ARM_MATH_SUCCESS)
{
return(status);
}
S->fftLenRFFT = 32U;
S->pTwiddleRFFT = (float32_t *) twiddleCoef_rfft_32;
return ARM_MATH_SUCCESS;
}
#endif
#if !defined(ARM_DSP_CONFIG_TABLES) || defined(ARM_ALL_FFT_TABLES) || (defined(ARM_TABLE_TWIDDLECOEF_F32_32) && defined(ARM_TABLE_BITREVIDX_FLT_32) && defined(ARM_TABLE_TWIDDLECOEF_RFFT_F32_64))
/**
@private
@brief Initialization function for the 64pt floating-point real FFT.
@param[in,out] S points to an arm_rfft_fast_instance_f32 structure
@return execution status
- \ref ARM_MATH_SUCCESS : Operation successful
- \ref ARM_MATH_ARGUMENT_ERROR : an error is detected
*/
static arm_status arm_rfft_64_fast_init_f32( arm_rfft_fast_instance_f32 * S ) {
arm_status status;
if( !S ) return ARM_MATH_ARGUMENT_ERROR;
status=arm_cfft_init_f32(&(S->Sint),32);
if (status != ARM_MATH_SUCCESS)
{
return(status);
}
S->fftLenRFFT = 64U;
S->pTwiddleRFFT = (float32_t *) twiddleCoef_rfft_64;
return ARM_MATH_SUCCESS;
}
#endif
#if !defined(ARM_DSP_CONFIG_TABLES) || defined(ARM_ALL_FFT_TABLES) || (defined(ARM_TABLE_TWIDDLECOEF_F32_64) && defined(ARM_TABLE_BITREVIDX_FLT_64) && defined(ARM_TABLE_TWIDDLECOEF_RFFT_F32_128))
/**
@private
@brief Initialization function for the 128pt floating-point real FFT.
@param[in,out] S points to an arm_rfft_fast_instance_f32 structure
@return execution status
- \ref ARM_MATH_SUCCESS : Operation successful
- \ref ARM_MATH_ARGUMENT_ERROR : an error is detected
*/
static arm_status arm_rfft_128_fast_init_f32( arm_rfft_fast_instance_f32 * S ) {
arm_status status;
if( !S ) return ARM_MATH_ARGUMENT_ERROR;
status=arm_cfft_init_f32(&(S->Sint),64);
if (status != ARM_MATH_SUCCESS)
{
return(status);
}
S->fftLenRFFT = 128;
S->pTwiddleRFFT = (float32_t *) twiddleCoef_rfft_128;
return ARM_MATH_SUCCESS;
}
#endif
#if !defined(ARM_DSP_CONFIG_TABLES) || defined(ARM_ALL_FFT_TABLES) || (defined(ARM_TABLE_TWIDDLECOEF_F32_128) && defined(ARM_TABLE_BITREVIDX_FLT_128) && defined(ARM_TABLE_TWIDDLECOEF_RFFT_F32_256))
/**
@private
@brief Initialization function for the 256pt floating-point real FFT.
@param[in,out] S points to an arm_rfft_fast_instance_f32 structure
@return execution status
- \ref ARM_MATH_SUCCESS : Operation successful
- \ref ARM_MATH_ARGUMENT_ERROR : an error is detected
*/
static arm_status arm_rfft_256_fast_init_f32( arm_rfft_fast_instance_f32 * S ) {
arm_status status;
if( !S ) return ARM_MATH_ARGUMENT_ERROR;
status=arm_cfft_init_f32(&(S->Sint),128);
if (status != ARM_MATH_SUCCESS)
{
return(status);
}
S->fftLenRFFT = 256U;
S->pTwiddleRFFT = (float32_t *) twiddleCoef_rfft_256;
return ARM_MATH_SUCCESS;
}
#endif
#if !defined(ARM_DSP_CONFIG_TABLES) || defined(ARM_ALL_FFT_TABLES) || (defined(ARM_TABLE_TWIDDLECOEF_F32_256) && defined(ARM_TABLE_BITREVIDX_FLT_256) && defined(ARM_TABLE_TWIDDLECOEF_RFFT_F32_512))
/**
@private
@brief Initialization function for the 512pt floating-point real FFT.
@param[in,out] S points to an arm_rfft_fast_instance_f32 structure
@return execution status
- \ref ARM_MATH_SUCCESS : Operation successful
- \ref ARM_MATH_ARGUMENT_ERROR : an error is detected
*/
static arm_status arm_rfft_512_fast_init_f32( arm_rfft_fast_instance_f32 * S ) {
arm_status status;
if( !S ) return ARM_MATH_ARGUMENT_ERROR;
status=arm_cfft_init_f32(&(S->Sint),256);
if (status != ARM_MATH_SUCCESS)
{
return(status);
}
S->fftLenRFFT = 512U;
S->pTwiddleRFFT = (float32_t *) twiddleCoef_rfft_512;
return ARM_MATH_SUCCESS;
}
#endif
#if !defined(ARM_DSP_CONFIG_TABLES) || defined(ARM_ALL_FFT_TABLES) || (defined(ARM_TABLE_TWIDDLECOEF_F32_512) && defined(ARM_TABLE_BITREVIDX_FLT_512) && defined(ARM_TABLE_TWIDDLECOEF_RFFT_F32_1024))
/**
@private
@brief Initialization function for the 1024pt floating-point real FFT.
@param[in,out] S points to an arm_rfft_fast_instance_f32 structure
@return execution status
- \ref ARM_MATH_SUCCESS : Operation successful
- \ref ARM_MATH_ARGUMENT_ERROR : an error is detected
*/
static arm_status arm_rfft_1024_fast_init_f32( arm_rfft_fast_instance_f32 * S ) {
arm_status status;
if( !S ) return ARM_MATH_ARGUMENT_ERROR;
status=arm_cfft_init_f32(&(S->Sint),512);
if (status != ARM_MATH_SUCCESS)
{
return(status);
}
S->fftLenRFFT = 1024U;
S->pTwiddleRFFT = (float32_t *) twiddleCoef_rfft_1024;
return ARM_MATH_SUCCESS;
}
#endif
#if !defined(ARM_DSP_CONFIG_TABLES) || defined(ARM_ALL_FFT_TABLES) || (defined(ARM_TABLE_TWIDDLECOEF_F32_1024) && defined(ARM_TABLE_BITREVIDX_FLT_1024) && defined(ARM_TABLE_TWIDDLECOEF_RFFT_F32_2048))
/**
@private
@brief Initialization function for the 2048pt floating-point real FFT.
@param[in,out] S points to an arm_rfft_fast_instance_f32 structure
@return execution status
- \ref ARM_MATH_SUCCESS : Operation successful
- \ref ARM_MATH_ARGUMENT_ERROR : an error is detected
*/
static arm_status arm_rfft_2048_fast_init_f32( arm_rfft_fast_instance_f32 * S ) {
arm_status status;
if( !S ) return ARM_MATH_ARGUMENT_ERROR;
status=arm_cfft_init_f32(&(S->Sint),1024);
if (status != ARM_MATH_SUCCESS)
{
return(status);
}
S->fftLenRFFT = 2048U;
S->pTwiddleRFFT = (float32_t *) twiddleCoef_rfft_2048;
return ARM_MATH_SUCCESS;
}
#endif
#if !defined(ARM_DSP_CONFIG_TABLES) || defined(ARM_ALL_FFT_TABLES) || (defined(ARM_TABLE_TWIDDLECOEF_F32_2048) && defined(ARM_TABLE_BITREVIDX_FLT_2048) && defined(ARM_TABLE_TWIDDLECOEF_RFFT_F32_4096))
/**
@private
* @brief Initialization function for the 4096pt floating-point real FFT.
* @param[in,out] S points to an arm_rfft_fast_instance_f32 structure
@return execution status
- \ref ARM_MATH_SUCCESS : Operation successful
- \ref ARM_MATH_ARGUMENT_ERROR : an error is detected
*/
static arm_status arm_rfft_4096_fast_init_f32( arm_rfft_fast_instance_f32 * S ) {
arm_status status;
if( !S ) return ARM_MATH_ARGUMENT_ERROR;
status=arm_cfft_init_f32(&(S->Sint),2048);
if (status != ARM_MATH_SUCCESS)
{
return(status);
}
S->fftLenRFFT = 4096U;
S->pTwiddleRFFT = (float32_t *) twiddleCoef_rfft_4096;
return ARM_MATH_SUCCESS;
}
#endif
/**
@brief Initialization function for the floating-point real FFT.
@param[in,out] S points to an arm_rfft_fast_instance_f32 structure
@param[in] fftLen length of the Real Sequence
@return execution status
- \ref ARM_MATH_SUCCESS : Operation successful
- \ref ARM_MATH_ARGUMENT_ERROR : <code>fftLen</code> is not a supported length
@par Description
The parameter <code>fftLen</code> specifies the length of RFFT/CIFFT process.
Supported FFT Lengths are 32, 64, 128, 256, 512, 1024, 2048, 4096.
@par
This Function also initializes Twiddle factor table pointer and Bit reversal table pointer.
*/
arm_status arm_rfft_fast_init_f32(
arm_rfft_fast_instance_f32 * S,
uint16_t fftLen)
{
typedef arm_status(*fft_init_ptr)( arm_rfft_fast_instance_f32 *);
fft_init_ptr fptr = 0x0;
switch (fftLen)
{
#if !defined(ARM_DSP_CONFIG_TABLES) || defined(ARM_ALL_FFT_TABLES) || (defined(ARM_TABLE_TWIDDLECOEF_F32_2048) && defined(ARM_TABLE_BITREVIDX_FLT_2048) && defined(ARM_TABLE_TWIDDLECOEF_RFFT_F32_4096))
case 4096U:
fptr = arm_rfft_4096_fast_init_f32;
break;
#endif
#if !defined(ARM_DSP_CONFIG_TABLES) || defined(ARM_ALL_FFT_TABLES) || (defined(ARM_TABLE_TWIDDLECOEF_F32_1024) && defined(ARM_TABLE_BITREVIDX_FLT_1024) && defined(ARM_TABLE_TWIDDLECOEF_RFFT_F32_2048))
case 2048U:
fptr = arm_rfft_2048_fast_init_f32;
break;
#endif
#if !defined(ARM_DSP_CONFIG_TABLES) || defined(ARM_ALL_FFT_TABLES) || (defined(ARM_TABLE_TWIDDLECOEF_F32_512) && defined(ARM_TABLE_BITREVIDX_FLT_512) && defined(ARM_TABLE_TWIDDLECOEF_RFFT_F32_1024))
case 1024U:
fptr = arm_rfft_1024_fast_init_f32;
break;
#endif
#if !defined(ARM_DSP_CONFIG_TABLES) || defined(ARM_ALL_FFT_TABLES) || (defined(ARM_TABLE_TWIDDLECOEF_F32_256) && defined(ARM_TABLE_BITREVIDX_FLT_256) && defined(ARM_TABLE_TWIDDLECOEF_RFFT_F32_512))
case 512U:
fptr = arm_rfft_512_fast_init_f32;
break;
#endif
#if !defined(ARM_DSP_CONFIG_TABLES) || defined(ARM_ALL_FFT_TABLES) || (defined(ARM_TABLE_TWIDDLECOEF_F32_128) && defined(ARM_TABLE_BITREVIDX_FLT_128) && defined(ARM_TABLE_TWIDDLECOEF_RFFT_F32_256))
case 256U:
fptr = arm_rfft_256_fast_init_f32;
break;
#endif
#if !defined(ARM_DSP_CONFIG_TABLES) || defined(ARM_ALL_FFT_TABLES) || (defined(ARM_TABLE_TWIDDLECOEF_F32_64) && defined(ARM_TABLE_BITREVIDX_FLT_64) && defined(ARM_TABLE_TWIDDLECOEF_RFFT_F32_128))
case 128U:
fptr = arm_rfft_128_fast_init_f32;
break;
#endif
#if !defined(ARM_DSP_CONFIG_TABLES) || defined(ARM_ALL_FFT_TABLES) || (defined(ARM_TABLE_TWIDDLECOEF_F32_32) && defined(ARM_TABLE_BITREVIDX_FLT_32) && defined(ARM_TABLE_TWIDDLECOEF_RFFT_F32_64))
case 64U:
fptr = arm_rfft_64_fast_init_f32;
break;
#endif
#if !defined(ARM_DSP_CONFIG_TABLES) || defined(ARM_ALL_FFT_TABLES) || (defined(ARM_TABLE_TWIDDLECOEF_F32_16) && defined(ARM_TABLE_BITREVIDX_FLT_16) && defined(ARM_TABLE_TWIDDLECOEF_RFFT_F32_32))
case 32U:
fptr = arm_rfft_32_fast_init_f32;
break;
#endif
default:
break;
}
if( ! fptr ) return ARM_MATH_ARGUMENT_ERROR;
return fptr( S );
}
/**
@} end of RealFFT group
*/