1  /* ----------------------------------------------------------------------
  2   * Project:      CMSIS DSP Library
  3   * Title:        arm_cmplx_mult_real_q31.c
  4   * Description:  Q31 complex by real multiplication
  5   *
  6   * $Date:        23 April 2021
  7   * $Revision:    V1.9.0
  8   *
  9   * Target Processor: Cortex-M and Cortex-A cores
 10   * -------------------------------------------------------------------- */
 11  /*
 12   * Copyright (C) 2010-2021 ARM Limited or its affiliates. All rights reserved.
 13   *
 14   * SPDX-License-Identifier: Apache-2.0
 15   *
 16   * Licensed under the Apache License, Version 2.0 (the License); you may
 17   * not use this file except in compliance with the License.
 18   * You may obtain a copy of the License at
 19   *
 20   * www.apache.org/licenses/LICENSE-2.0
 21   *
 22   * Unless required by applicable law or agreed to in writing, software
 23   * distributed under the License is distributed on an AS IS BASIS, WITHOUT
 24   * WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 25   * See the License for the specific language governing permissions and
 26   * limitations under the License.
 27   */
 28  
 29  #include "dsp/complex_math_functions.h"
 30  
 31  /**
 32    @ingroup groupCmplxMath
 33   */
 34  
 35  /**
 36    @addtogroup CmplxByRealMult
 37    @{
 38   */
 39  
 40  /**
 41    @brief         Q31 complex-by-real multiplication.
 42    @param[in]     pSrcCmplx   points to complex input vector
 43    @param[in]     pSrcReal    points to real input vector
 44    @param[out]    pCmplxDst   points to complex output vector
 45    @param[in]     numSamples  number of samples in each vector
 46    @return        none
 47  
 48    @par           Scaling and Overflow Behavior
 49                     The function uses saturating arithmetic.
 50                     Results outside of the allowable Q31 range[0x80000000 0x7FFFFFFF] are saturated.
 51   */
 52  
 53  #if defined(ARM_MATH_MVEI) && !defined(ARM_MATH_AUTOVECTORIZE)
 54  void arm_cmplx_mult_real_q31(
 55    const q31_t * pSrcCmplx,
 56    const q31_t * pSrcReal,
 57          q31_t * pCmplxDst,
 58          uint32_t numSamples)
 59  {
 60  
 61      static const uint32_t stride_cmplx_x_real_32[4] = {
 62          0, 0, 1, 1
 63      };
 64      q31x4_t rVec;
 65      q31x4_t cmplxVec;
 66      q31x4_t dstVec;
 67      uint32x4_t strideVec;
 68      uint32_t blockSizeC = numSamples * CMPLX_DIM;   /* loop counters */
 69      uint32_t blkCnt;
 70      q31_t in;
 71  
 72      /*
 73       * stride vector for pairs of real generation
 74       */
 75      strideVec = vld1q(stride_cmplx_x_real_32);
 76  
 77      /* Compute 4 complex outputs at a time */
 78      blkCnt = blockSizeC >> 2;
 79      while (blkCnt > 0U) 
 80      {
 81          cmplxVec = vld1q(pSrcCmplx);
 82          rVec = vldrwq_gather_shifted_offset_s32(pSrcReal, strideVec);
 83          dstVec = vqdmulhq(cmplxVec, rVec);
 84          vst1q(pCmplxDst, dstVec);
 85  
 86          pSrcReal += 2;
 87          pSrcCmplx += 4;
 88          pCmplxDst += 4;
 89          blkCnt --;
 90      }
 91  
 92      blkCnt = (blockSizeC & 3) >> 1; 
 93      while (blkCnt > 0U)
 94      {
 95        /* C[2 * i    ] = A[2 * i    ] * B[i]. */
 96        /* C[2 * i + 1] = A[2 * i + 1] * B[i]. */
 97    
 98        in = *pSrcReal++;
 99        /* store saturated result in 1.31 format to destination buffer */
100        *pCmplxDst++ = (__SSAT((q31_t) (((q63_t) *pSrcCmplx++ * in) >> 32), 31) << 1);
101        *pCmplxDst++ = (__SSAT((q31_t) (((q63_t) *pSrcCmplx++ * in) >> 32), 31) << 1);
102    
103        /* Decrement loop counter */
104        blkCnt--;
105      }
106  }
107  #else
108  void arm_cmplx_mult_real_q31(
109    const q31_t * pSrcCmplx,
110    const q31_t * pSrcReal,
111          q31_t * pCmplxDst,
112          uint32_t numSamples)
113  {
114          uint32_t blkCnt;                               /* Loop counter */
115          q31_t in;                                      /* Temporary variable */
116  
117  #if defined (ARM_MATH_LOOPUNROLL)
118  
119    /* Loop unrolling: Compute 4 outputs at a time */
120    blkCnt = numSamples >> 2U;
121  
122    while (blkCnt > 0U)
123    {
124      /* C[2 * i    ] = A[2 * i    ] * B[i]. */
125      /* C[2 * i + 1] = A[2 * i + 1] * B[i]. */
126  
127      in = *pSrcReal++;
128  #if defined (ARM_MATH_DSP)
129      /* store saturated result in 1.31 format to destination buffer */
130      *pCmplxDst++ = (__SSAT((q31_t) (((q63_t) *pSrcCmplx++ * in) >> 32), 31) << 1);
131      *pCmplxDst++ = (__SSAT((q31_t) (((q63_t) *pSrcCmplx++ * in) >> 32), 31) << 1);
132  #else
133      /* store result in destination buffer. */
134      *pCmplxDst++ = (q31_t) clip_q63_to_q31(((q63_t) *pSrcCmplx++ * in) >> 31);
135      *pCmplxDst++ = (q31_t) clip_q63_to_q31(((q63_t) *pSrcCmplx++ * in) >> 31);
136  #endif
137  
138      in = *pSrcReal++;
139  #if defined (ARM_MATH_DSP)
140      *pCmplxDst++ = (__SSAT((q31_t) (((q63_t) *pSrcCmplx++ * in) >> 32), 31) << 1);
141      *pCmplxDst++ = (__SSAT((q31_t) (((q63_t) *pSrcCmplx++ * in) >> 32), 31) << 1);
142  #else
143      *pCmplxDst++ = (q31_t) clip_q63_to_q31(((q63_t) *pSrcCmplx++ * in) >> 31);
144      *pCmplxDst++ = (q31_t) clip_q63_to_q31(((q63_t) *pSrcCmplx++ * in) >> 31);
145  #endif
146  
147      in = *pSrcReal++;
148  #if defined (ARM_MATH_DSP)
149      *pCmplxDst++ = (__SSAT((q31_t) (((q63_t) *pSrcCmplx++ * in) >> 32), 31) << 1);
150      *pCmplxDst++ = (__SSAT((q31_t) (((q63_t) *pSrcCmplx++ * in) >> 32), 31) << 1);
151  #else
152      *pCmplxDst++ = (q31_t) clip_q63_to_q31(((q63_t) *pSrcCmplx++ * in) >> 31);
153      *pCmplxDst++ = (q31_t) clip_q63_to_q31(((q63_t) *pSrcCmplx++ * in) >> 31);
154  #endif
155  
156      in = *pSrcReal++;
157  #if defined (ARM_MATH_DSP)
158      *pCmplxDst++ = (__SSAT((q31_t) (((q63_t) *pSrcCmplx++ * in) >> 32), 31) << 1);
159      *pCmplxDst++ = (__SSAT((q31_t) (((q63_t) *pSrcCmplx++ * in) >> 32), 31) << 1);
160  #else
161      *pCmplxDst++ = (q31_t) clip_q63_to_q31(((q63_t) *pSrcCmplx++ * in) >> 31);
162      *pCmplxDst++ = (q31_t) clip_q63_to_q31(((q63_t) *pSrcCmplx++ * in) >> 31);
163  #endif
164  
165      /* Decrement loop counter */
166      blkCnt--;
167    }
168  
169    /* Loop unrolling: Compute remaining outputs */
170    blkCnt = numSamples % 0x4U;
171  
172  #else
173  
174    /* Initialize blkCnt with number of samples */
175    blkCnt = numSamples;
176  
177  #endif /* #if defined (ARM_MATH_LOOPUNROLL) */
178  
179    while (blkCnt > 0U)
180    {
181      /* C[2 * i    ] = A[2 * i    ] * B[i]. */
182      /* C[2 * i + 1] = A[2 * i + 1] * B[i]. */
183  
184      in = *pSrcReal++;
185  #if defined (ARM_MATH_DSP)
186      /* store saturated result in 1.31 format to destination buffer */
187      *pCmplxDst++ = (__SSAT((q31_t) (((q63_t) *pSrcCmplx++ * in) >> 32), 31) << 1);
188      *pCmplxDst++ = (__SSAT((q31_t) (((q63_t) *pSrcCmplx++ * in) >> 32), 31) << 1);
189  #else
190      /* store result in destination buffer. */
191      *pCmplxDst++ = (q31_t) clip_q63_to_q31(((q63_t) *pSrcCmplx++ * in) >> 31);
192      *pCmplxDst++ = (q31_t) clip_q63_to_q31(((q63_t) *pSrcCmplx++ * in) >> 31);
193  #endif
194  
195      /* Decrement loop counter */
196      blkCnt--;
197    }
198  
199  }
200  #endif /* defined(ARM_MATH_MVEI) */
201  
202  /**
203    @} end of CmplxByRealMult group
204   */