Skip to content

/github/workspace/src/TransformFunctions/kernels/plp_cfft_q32p_xpulpv2.c

Functions

Name
void plp_cfft_radix4by2_q32(int32_t * pSrc, uint32_t fftLen, const int32_t * pCoef, uint32_t nPE)
void plp_radix4_butterfly_q32(int32_t * pSrc, uint32_t fftLen, int32_t * pCoef, uint32_t twidCoefModifier, uint32_t nPE)
void plp_cfft_q32p_xpulpv2(void * args)
Parallel quantized 32 bit complex fast fourier transform for XPULPV2.

Defines

Name
MAX(x, y)
MIN(x, y)
multAcc_32x32_keep32_R(a, x, y)
multSub_32x32_keep32_R(a, x, y)
mult_32x32_keep32_R(a, x, y)

Functions Documentation

function plp_cfft_radix4by2_q32

void plp_cfft_radix4by2_q32(
    int32_t * pSrc,
    uint32_t fftLen,
    const int32_t * pCoef,
    uint32_t nPE
)

function plp_radix4_butterfly_q32

static void plp_radix4_butterfly_q32(
    int32_t * pSrc,
    uint32_t fftLen,
    int32_t * pCoef,
    uint32_t twidCoefModifier,
    uint32_t nPE
)

function plp_cfft_q32p_xpulpv2

void plp_cfft_q32p_xpulpv2(
    void * args
)

Parallel quantized 32 bit complex fast fourier transform for XPULPV2.

Parameters:

Macros Documentation

define MAX

#define MAX(
    x,
    y
)
(((x) > (y)) ? (x) : (y))

define MIN

#define MIN(
    x,
    y
)
(((x) < (y)) ? (x) : (y))

define multAcc_32x32_keep32_R

#define multAcc_32x32_keep32_R(
    a,
    x,
    y
)
    a = (int32_t) (((((int64_t) a) << 32) + ((int64_t) x * y) + 0x80000000LL ) >> 32)

define multSub_32x32_keep32_R

#define multSub_32x32_keep32_R(
    a,
    x,
    y
)
    a = (int32_t) (((((int64_t) a) << 32) - ((int64_t) x * y) + 0x80000000LL ) >> 32)

define mult_32x32_keep32_R

#define mult_32x32_keep32_R(
    a,
    x,
    y
)
    a = (int32_t) (((int64_t) x * y + 0x80000000LL ) >> 32)

Source code

/* =====================================================================
 * Project:      PULP DSP Library
 * Title:        plp_cfft_q32p_xpulpv2.c
 * Description:  32-bit fixed point Fast Fourier Transform on Compled Input Data
 *
 * $Date:        16. May 2022
 * $Revision:    V0
 *
 * Target Processor: PULP cores
 * ===================================================================== */
/*
 * Copyright (C) 2020 ETH Zurich and University of Bologna. All rights reserved.
 *
 * Author: Marco Bertuletti, ETH Zurich
 *
 * 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 "plp_math.h"
#define MAX(x, y) (((x) > (y)) ? (x) : (y))
#define MIN(x, y) (((x) < (y)) ? (x) : (y))

#define multAcc_32x32_keep32_R(a, x, y) \
    a = (int32_t) (((((int64_t) a) << 32) + ((int64_t) x * y) + 0x80000000LL ) >> 32)

#define multSub_32x32_keep32_R(a, x, y) \
    a = (int32_t) (((((int64_t) a) << 32) - ((int64_t) x * y) + 0x80000000LL ) >> 32)

#define mult_32x32_keep32_R(a, x, y) \
    a = (int32_t) (((int64_t) x * y + 0x80000000LL ) >> 32)


void plp_cfft_radix4by2_q32(int32_t *pSrc, 
                            uint32_t fftLen,
                            const int32_t *pCoef,
                            uint32_t nPE);

static void plp_radix4_butterfly_q32(int32_t *pSrc,
                                    uint32_t fftLen,
                                    int32_t *pCoef,
                                    uint32_t twidCoefModifier,
                                    uint32_t nPE);

void plp_cfft_q32p_xpulpv2(void *args)
{

    int core_id = hal_core_id();
    plp_cfft_instance_q32_parallel *a = (plp_cfft_instance_q32_parallel *) args;
    uint32_t L = a->S->fftLen;

    if (a->ifftFlag == 0) {
        switch (L) {
            case 16:
            case 64:
            case 256:
            case 1024:
            case 4096:
            plp_radix4_butterfly_q32(a->p1, L, (int32_t *)a->S->pTwiddle, 1, a->nPE);
            break;
            case 32:
            case 128:
            case 512:
            case 2048:
            plp_cfft_radix4by2_q32(a->p1, L, (int32_t *)a->S->pTwiddle, a->nPE);
            break;
        }
    }
    hal_team_barrier();

    if (a->bitReverseFlag)
        plp_bitreversal_32p_xpulpv2((uint32_t *)a->p1, a->S->bitRevLength, a->S->pBitRevTable, a->nPE);
}

void plp_cfft_radix4by2_q32(int32_t *pSrc, 
                            uint32_t fftLen,
                            const int32_t *pCoef,
                            uint32_t nPE)
{

    int core_id = hal_core_id();
    uint32_t i, l;
    uint32_t n2, nCores;
    int32_t xt, yt, cosVal, sinVal;
    int32_t p0, p1;

    n2 = fftLen >> 1U;
    if (n2 % nPE == 0) {
        nCores = n2/nPE;
    } else {
        nCores = n2/nPE + 1;
    }

    uint32_t core_offset = core_id*nCores;
    for (i = core_offset; i < MIN(n2,core_offset + nCores); i++)
    {
        cosVal = pCoef[2 * i];
        sinVal = pCoef[2 * i + 1];

        l = i + n2;

        xt =          (pSrc[2 * i] >> 2U) - (pSrc[2 * l] >> 2U);
        pSrc[2 * i] = (pSrc[2 * i] >> 2U) + (pSrc[2 * l] >> 2U);

        yt =              (pSrc[2 * i + 1] >> 2U) - (pSrc[2 * l + 1] >> 2U);
        pSrc[2 * i + 1] = (pSrc[2 * l + 1] >> 2U) + (pSrc[2 * i + 1] >> 2U);

        mult_32x32_keep32_R(p0, xt, cosVal);
        mult_32x32_keep32_R(p1, yt, cosVal);
        multAcc_32x32_keep32_R(p0, yt, sinVal);
        multSub_32x32_keep32_R(p1, xt, sinVal);

        pSrc[2 * l]     = p0 << 1;
        pSrc[2 * l + 1] = p1 << 1;
    }
    hal_team_barrier();

    if (nPE > 1) {
        if(core_id < nPE/2) {
            /* first col */
            plp_radix4_butterfly_q32 (pSrc, n2, (int32_t*)pCoef, 2U, nPE/2);
        } else {
            /* second col */
            plp_radix4_butterfly_q32 (pSrc + fftLen, n2, (int32_t*)pCoef, 2U, nPE - nPE/2);
        }
    } else {
        // first col
        plp_radix4_butterfly_q32 (pSrc, n2, (int32_t*)pCoef, 2U, nPE);
        // second col
        plp_radix4_butterfly_q32 (pSrc + fftLen, n2, (int32_t*)pCoef, 2U, nPE - nPE);
    }
    hal_team_barrier();

    n2 = fftLen >> 1U;
    for (i = core_offset; i < MIN((fftLen >> 1), core_offset + nCores); i++)
    {
        p0 = pSrc[4 * i + 0];
        p1 = pSrc[4 * i + 1];
        xt = pSrc[4 * i + 2];
        yt = pSrc[4 * i + 3];

        p0 <<= 1U;
        p1 <<= 1U;
        xt <<= 1U;
        yt <<= 1U;

        pSrc[4 * i + 0] = p0;
        pSrc[4 * i + 1] = p1;
        pSrc[4 * i + 2] = xt;
        pSrc[4 * i + 3] = yt;
    }
}

void plp_radix4_butterfly_q32(  int32_t *pSrc,
                                uint32_t fftLen,
                                int32_t *pCoef,
                                uint32_t twidCoefModifier,
                                uint32_t nPE) {

    int core_id = hal_core_id()%nPE;
    uint32_t n1, n2, ia1, ia2, ia3, i0, i1, i2, i3, j, k;
    int32_t t1, t2, r1, r2, s1, s2, co1, co2, co3, si1, si2, si3;

    int32_t xa, xb, xc, xd;
    int32_t ya, yb, yc, yd;
    int32_t xa_out, xb_out, xc_out, xd_out;
    int32_t ya_out, yb_out, yc_out, yd_out;

    int32_t *ptr1;

  /* Total process is divided into three stages */

  /* process first stage, middle stages, & last stage */


  /* start of first stage process */

  /*  Initializations for the first stage */
    n2 = fftLen;
    n1 = n2;
  /* n2 = fftLen/4 */
    n2 >>= 2U;

    ia1 = 0U;

    uint32_t step;
    if (n2 % nPE == 0) {
        step = n2/nPE;
    } else {
        step = n2/nPE + 1;
    }

    /*  Calculation of first stage */
    /* start of first stage process */
    for (i0 = core_id * step; i0 < MIN(core_id * step + step, n2); i0++)
    {
        /*  index calculation for the input as, */
        /*  pSrc[i0 + 0], pSrc[i0 + fftLen/4], pSrc[i0 + fftLen/2U], pSrc[i0 + 3fftLen/4] */
        i1 = i0 + n2;
        i2 = i1 + n2;
        i3 = i2 + n2;

        /*  Twiddle coefficients index modifier */
        ia1 = i0 * twidCoefModifier;

        /* input is in 1.31(q31) format and provide 4 guard bits for the input */
        /*  Butterfly implementation */
        /* xa + xc */
        r1 = (pSrc[(2U * i0)] >> 4U) + (pSrc[(2U * i2)] >> 4U);
        /* xa - xc */
        r2 = (pSrc[(2U * i0)] >> 4U) - (pSrc[(2U * i2)] >> 4U);

        /* xb + xd */
        t1 = (pSrc[(2U * i1)] >> 4U) + (pSrc[(2U * i3)] >> 4U);

        /* ya + yc */
        s1 = (pSrc[(2U * i0) + 1U] >> 4U) + (pSrc[(2U * i2) + 1U] >> 4U);
        /* ya - yc */
        s2 = (pSrc[(2U * i0) + 1U] >> 4U) - (pSrc[(2U * i2) + 1U] >> 4U);

        /* xa' = xa + xb + xc + xd */
        pSrc[2U * i0] = (r1 + t1);
        /* (xa + xc) - (xb + xd) */
        r1 = r1 - t1;
        /* yb + yd */
        t2 = (pSrc[(2U * i1) + 1U] >> 4U) + (pSrc[(2U * i3) + 1U] >> 4U);

        /* ya' = ya + yb + yc + yd */
        pSrc[(2U * i0) + 1U] = (s1 + t2);

        /* (ya + yc) - (yb + yd) */
        s1 = s1 - t2;

        /* yb - yd */
        t1 = (pSrc[(2U * i1) + 1U] >> 4U) - (pSrc[(2U * i3) + 1U] >> 4U);
        /* xb - xd */
        t2 = (pSrc[(2U * i1)] >> 4U) - (pSrc[(2U * i3)] >> 4U);

        /*  index calculation for the coefficients */
        ia2 = 2U * ia1;
        co2 = pCoef[(ia2 * 2U)];
        si2 = pCoef[(ia2 * 2U) + 1U];

        /* xc' = (xa-xb+xc-xd)co2 + (ya-yb+yc-yd)(si2) */
        pSrc[2U * i1] = (((int32_t) (((int64_t) r1 * co2) >> 32)) +
            ((int32_t) (((int64_t) s1 * si2) >> 32))) << 1U;

        /* yc' = (ya-yb+yc-yd)co2 - (xa-xb+xc-xd)(si2) */
        pSrc[(2U * i1) + 1U] = (((int32_t) (((int64_t) s1 * co2) >> 32)) -
            ((int32_t) (((int64_t) r1 * si2) >> 32))) << 1U;

        /* (xa - xc) + (yb - yd) */
        r1 = r2 + t1;
        /* (xa - xc) - (yb - yd) */
        r2 = r2 - t1;

        /* (ya - yc) - (xb - xd) */
        s1 = s2 - t2;
        /* (ya - yc) + (xb - xd) */
        s2 = s2 + t2;

        co1 = pCoef[(ia1 * 2U)];
        si1 = pCoef[(ia1 * 2U) + 1U];

        /* xb' = (xa+yb-xc-yd)co1 + (ya-xb-yc+xd)(si1) */
        pSrc[2U * i2] = (((int32_t) (((int64_t) r1 * co1) >> 32)) +
            ((int32_t) (((int64_t) s1 * si1) >> 32))) << 1U;

        /* yb' = (ya-xb-yc+xd)co1 - (xa+yb-xc-yd)(si1) */
        pSrc[(2U * i2) + 1U] = (((int32_t) (((int64_t) s1 * co1) >> 32)) -
            ((int32_t) (((int64_t) r1 * si1) >> 32))) << 1U;

        /*  index calculation for the coefficients */
        ia3 = 3U * ia1;
        co3 = pCoef[(ia3 * 2U)];
        si3 = pCoef[(ia3 * 2U) + 1U];

        /* xd' = (xa-yb-xc+yd)co3 + (ya+xb-yc-xd)(si3) */
        pSrc[2U * i3] = (((int32_t) (((int64_t) r2 * co3) >> 32)) +
            ((int32_t) (((int64_t) s2 * si3) >> 32))) << 1U;

        /* yd' = (ya+xb-yc-xd)co3 - (xa-yb-xc+yd)(si3) */
        pSrc[(2U * i3) + 1U] = (((int32_t) (((int64_t) s2 * co3) >> 32)) -
            ((int32_t) (((int64_t) r2 * si3) >> 32))) << 1U;

    }

    /* end of first stage process */
    /* data is in 5.27(q27) format */
    hal_team_barrier();

    /* start of Middle stages process */
    /* each stage in middle stages provides two down scaling of the input */
    twidCoefModifier <<= 2U;

    for (k = fftLen / 4U; k > 4U; k >>= 2U)
    {
    /*  Initializations for the first stage */
        n1 = n2;
        n2 >>= 2U;

        if (n2 % nPE == 0) {
            step = n2/nPE;
        } else {
            step = n2/nPE + 1;
        }

    /*  Calculation of first stage */
        for (j = core_id * step; j < MIN(core_id * step + step, n2); j++)
        {
            /*  Twiddle coefficients index modifier */
            ia1 = twidCoefModifier * j;
            /*  index calculation for the coefficients */
            ia2 = ia1 + ia1;
            ia3 = ia2 + ia1;
            co1 = pCoef[(ia1 * 2U)];
            si1 = pCoef[(ia1 * 2U) + 1U];
            co2 = pCoef[(ia2 * 2U)];
            si2 = pCoef[(ia2 * 2U) + 1U];
            co3 = pCoef[(ia3 * 2U)];
            si3 = pCoef[(ia3 * 2U) + 1U];

            for (i0 = j; i0 < fftLen; i0 += n1)
            {
                /*  index calculation for the input as, */
                /*  pSrc[i0 + 0], pSrc[i0 + fftLen/4], pSrc[i0 + fftLen/2U], pSrc[i0 + 3fftLen/4] */
                i1 = i0 + n2;
                i2 = i1 + n2;
                i3 = i2 + n2;

                /*  Butterfly implementation */
                /* xa + xc */
                r1 = pSrc[2U * i0] + pSrc[2U * i2];
                /* xa - xc */
                r2 = pSrc[2U * i0] - pSrc[2U * i2];

                /* ya + yc */
                s1 = pSrc[(2U * i0) + 1U] + pSrc[(2U * i2) + 1U];
                /* ya - yc */
                s2 = pSrc[(2U * i0) + 1U] - pSrc[(2U * i2) + 1U];

                /* xb + xd */
                t1 = pSrc[2U * i1] + pSrc[2U * i3];

                /* xa' = xa + xb + xc + xd */
                pSrc[2U * i0] = (r1 + t1) >> 2U;
                /* xa + xc -(xb + xd) */
                r1 = r1 - t1;

                /* yb + yd */
                t2 = pSrc[(2U * i1) + 1U] + pSrc[(2U * i3) + 1U];
                /* ya' = ya + yb + yc + yd */
                pSrc[(2U * i0) + 1U] = (s1 + t2) >> 2U;

                /* (ya + yc) - (yb + yd) */
                s1 = s1 - t2;

                /* (yb - yd) */
                t1 = pSrc[(2U * i1) + 1U] - pSrc[(2U * i3) + 1U];
                /* (xb - xd) */
                t2 = pSrc[2U * i1] - pSrc[2U * i3];

                /* xc' = (xa-xb+xc-xd)co2 + (ya-yb+yc-yd)(si2) */
                pSrc[2U * i1] = (((int32_t) (((int64_t) r1 * co2) >> 32)) +
                    ((int32_t) (((int64_t) s1 * si2) >> 32))) >> 1U;

                /* yc' = (ya-yb+yc-yd)co2 - (xa-xb+xc-xd)(si2) */
                pSrc[(2U * i1) + 1U] = (((int32_t) (((int64_t) s1 * co2) >> 32)) -
                    ((int32_t) (((int64_t) r1 * si2) >> 32))) >> 1U;

                /* (xa - xc) + (yb - yd) */
                r1 = r2 + t1;
                /* (xa - xc) - (yb - yd) */
                r2 = r2 - t1;

                /* (ya - yc) -  (xb - xd) */
                s1 = s2 - t2;
                /* (ya - yc) +  (xb - xd) */
                s2 = s2 + t2;

                /* xb' = (xa+yb-xc-yd)co1 + (ya-xb-yc+xd)(si1) */
                pSrc[2U * i2] = (((int32_t) (((int64_t) r1 * co1) >> 32)) +
                    ((int32_t) (((int64_t) s1 * si1) >> 32))) >> 1U;

                /* yb' = (ya-xb-yc+xd)co1 - (xa+yb-xc-yd)(si1) */
                pSrc[(2U * i2) + 1U] = (((int32_t) (((int64_t) s1 * co1) >> 32)) -
                    ((int32_t) (((int64_t) r1 * si1) >> 32))) >> 1U;

                /* xd' = (xa-yb-xc+yd)co3 + (ya+xb-yc-xd)(si3) */
                pSrc[2U * i3] = (((int32_t) (((int64_t) r2 * co3) >> 32)) +
                    ((int32_t) (((int64_t) s2 * si3) >> 32))) >> 1U;

                /* yd' = (ya+xb-yc-xd)co3 - (xa-yb-xc+yd)(si3) */
                pSrc[(2U * i3) + 1U] = (((int32_t) (((int64_t) s2 * co3) >> 32)) -
                    ((int32_t) (((int64_t) r2 * si3) >> 32))) >> 1U;
            }
        }
        twidCoefModifier <<= 2U;
        hal_team_barrier();
    }

    /* End of Middle stages process */

    /* data is in 11.21(q21) format for the 1024 point as there are 3 middle stages */
    /* data is in 9.23(q23) format for the 256 point as there are 2 middle stages */
    /* data is in 7.25(q25) format for the 64 point as there are 1 middle stage */
    /* data is in 5.27(q27) format for the 16 point as there are no middle stages */

    /* start of Last stage process */
    /*  Initializations for the last stage */
    n1 = n2;
    n2 >>= 2U;

    int steps;
    /* start of last stage process */
    steps = fftLen/n1;
    // printf("steps: %i at %i\n", steps, fftLen);
    if (steps % nPE == 0) {
        step = steps/nPE;
    } else {
        step = steps/nPE + 1;
    }

    /*  Calculations of last stage */
    for (i0 = core_id * step * n1; i0 < MIN((core_id * step + step) * n1, fftLen); i0 += n1)
    {
        i1 = i0 + n2;
        i2 = i1 + n2;
        i3 = i2 + n2;

        /* Read xa (real), ya(imag) input */
        xa = pSrc[2U * i0];
        ya = pSrc[2U * i0 + 1U];

        /* Read xb (real), yb(imag) input */
        xb = pSrc[2U * i1];
        yb = pSrc[2U * i1 + 1U];

        /* Read xc (real), yc(imag) input */
        xc = pSrc[2U * i2];
        yc = pSrc[2U * i2 + 1U];

        /* Read xc (real), yc(imag) input */
        xd = pSrc[2U * i3];
        yd = pSrc[2U * i3 + 1U];

        /* xa' = xa + xb + xc + xd */
        xa_out = xa + xb + xc + xd;

        /* ya' = ya + yb + yc + yd */
        ya_out = ya + yb + yc + yd;

        /* pointer updation for writing */
        ptr1 = ptr1 - 8U;

        /* writing xa' and ya' */
        pSrc[2U * i0] = xa_out;
        pSrc[2U * i0 + 1U] = ya_out;

        xc_out = (xa - xb + xc - xd);
        yc_out = (ya - yb + yc - yd);

        /* writing xc' and yc' */
        pSrc[2U * i1] = xc_out;
        pSrc[2U * i1 + 1U] = yc_out;

        xb_out = (xa + yb - xc - yd);
        yb_out = (ya - xb - yc + xd);

        /* writing xb' and yb' */
        pSrc[2U * i2] = xb_out;
        pSrc[2U * i2 + 1U] = yb_out;

        xd_out = (xa - yb - xc + yd);
        yd_out = (ya + xb - yc - xd);

        /* writing xd' and yd' */
        pSrc[2U * i3] = xd_out;
        pSrc[2U * i3 + 1U] = yd_out;


    }

  /* output is in 11.21(q21) format for the 1024 point */
  /* output is in 9.23(q23) format for the 256 point */
  /* output is in 7.25(q25) format for the 64 point */
  /* output is in 5.27(q27) format for the 16 point */

  /* End of last stage process */

}

Updated on 2023-03-01 at 16:16:33 +0000