1  /* ----------------------------------------------------------------------
  2   * Project:      CMSIS DSP Library
  3   * Title:        arm_scale_q15.c
  4   * Description:  Multiplies a Q15 vector by a scalar
  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/basic_math_functions.h"
 30  
 31  /**
 32    @ingroup groupMath
 33   */
 34  
 35  /**
 36    @addtogroup BasicScale
 37    @{
 38   */
 39  
 40  /**
 41    @brief         Multiplies a Q15 vector by a scalar.
 42    @param[in]     pSrc       points to the input vector
 43    @param[in]     scaleFract fractional portion of the scale value
 44    @param[in]     shift      number of bits to shift the result by
 45    @param[out]    pDst       points to the output vector
 46    @param[in]     blockSize  number of samples in each vector
 47    @return        none
 48  
 49    @par           Scaling and Overflow Behavior
 50                     The input data <code>*pSrc</code> and <code>scaleFract</code> are in 1.15 format.
 51                     These are multiplied to yield a 2.30 intermediate result and this is shifted with saturation to 1.15 format.
 52   */
 53  
 54  #if defined(ARM_MATH_MVEI) && !defined(ARM_MATH_AUTOVECTORIZE)
 55  
 56  #include "arm_helium_utils.h"
 57  
 58  void arm_scale_q15(
 59      const q15_t * pSrc,
 60      q15_t   scaleFract,
 61      int8_t  shift,
 62      q15_t * pDst,
 63      uint32_t blockSize)
 64  {
 65      uint32_t  blkCnt;           /* loop counters */
 66      q15x8_t vecSrc;
 67      q15x8_t vecDst;
 68  
 69  
 70      /* Compute 8 outputs at a time */
 71      blkCnt = blockSize >> 3;
 72  
 73      while (blkCnt > 0U)
 74      {
 75          /*
 76           * C = A * scale
 77           * Scale the input and then store the result in the destination buffer.
 78           */
 79          vecSrc = vld1q(pSrc);
 80          vecDst = vmulhq(vecSrc, vdupq_n_s16(scaleFract));
 81          vecDst = vqshlq_r(vecDst, shift + 1);
 82          vst1q(pDst, vecDst);
 83          /*
 84           * Decrement the blockSize loop counter
 85           */
 86          blkCnt--;
 87          /*
 88           * advance vector source and destination pointers
 89           */
 90          pSrc += 8;
 91          pDst += 8;
 92      }
 93      /*
 94       * tail
 95       */
 96      blkCnt = blockSize & 7;
 97      if (blkCnt > 0U)
 98      {
 99          mve_pred16_t p0 = vctp16q(blkCnt);;
100          vecSrc = vld1q(pSrc);
101          vecDst = vmulhq(vecSrc, vdupq_n_s16(scaleFract));
102          vecDst = vqshlq_r(vecDst, shift + 1);
103          vstrhq_p(pDst, vecDst, p0);
104      }
105  
106  }
107  
108  
109  #else
110  void arm_scale_q15(
111    const q15_t *pSrc,
112          q15_t scaleFract,
113          int8_t shift,
114          q15_t *pDst,
115          uint32_t blockSize)
116  {
117          uint32_t blkCnt;                               /* Loop counter */
118          int8_t kShift = 15 - shift;                    /* Shift to apply after scaling */
119  
120  #if defined (ARM_MATH_LOOPUNROLL)
121  #if defined (ARM_MATH_DSP)
122    q31_t inA1, inA2;
123    q31_t out1, out2, out3, out4;                  /* Temporary output variables */
124    q15_t in1, in2, in3, in4;                      /* Temporary input variables */
125  #endif
126  #endif
127  
128  #if defined (ARM_MATH_LOOPUNROLL)
129  
130    /* Loop unrolling: Compute 4 outputs at a time */
131    blkCnt = blockSize >> 2U;
132  
133    while (blkCnt > 0U)
134    {
135      /* C = A * scale */
136  
137  #if defined (ARM_MATH_DSP)
138      /* read 2 times 2 samples at a time from source */
139      inA1 = read_q15x2_ia (&pSrc);
140      inA2 = read_q15x2_ia (&pSrc);
141  
142      /* Scale inputs and store result in temporary variables
143       * in single cycle by packing the outputs */
144      out1 = (q31_t) ((q15_t) (inA1 >> 16) * scaleFract);
145      out2 = (q31_t) ((q15_t) (inA1      ) * scaleFract);
146      out3 = (q31_t) ((q15_t) (inA2 >> 16) * scaleFract);
147      out4 = (q31_t) ((q15_t) (inA2      ) * scaleFract);
148  
149      /* apply shifting */
150      out1 = out1 >> kShift;
151      out2 = out2 >> kShift;
152      out3 = out3 >> kShift;
153      out4 = out4 >> kShift;
154  
155      /* saturate the output */
156      in1 = (q15_t) (__SSAT(out1, 16));
157      in2 = (q15_t) (__SSAT(out2, 16));
158      in3 = (q15_t) (__SSAT(out3, 16));
159      in4 = (q15_t) (__SSAT(out4, 16));
160  
161      /* store result to destination */
162      write_q15x2_ia (&pDst, __PKHBT(in2, in1, 16));
163      write_q15x2_ia (&pDst, __PKHBT(in4, in3, 16));
164  #else
165      *pDst++ = (q15_t) (__SSAT(((q31_t) *pSrc++ * scaleFract) >> kShift, 16));
166      *pDst++ = (q15_t) (__SSAT(((q31_t) *pSrc++ * scaleFract) >> kShift, 16));
167      *pDst++ = (q15_t) (__SSAT(((q31_t) *pSrc++ * scaleFract) >> kShift, 16));
168      *pDst++ = (q15_t) (__SSAT(((q31_t) *pSrc++ * scaleFract) >> kShift, 16));
169  #endif
170  
171      /* Decrement loop counter */
172      blkCnt--;
173    }
174  
175    /* Loop unrolling: Compute remaining outputs */
176    blkCnt = blockSize % 0x4U;
177  
178  #else
179  
180    /* Initialize blkCnt with number of samples */
181    blkCnt = blockSize;
182  
183  #endif /* #if defined (ARM_MATH_LOOPUNROLL) */
184  
185    while (blkCnt > 0U)
186    {
187      /* C = A * scale */
188  
189      /* Scale input and store result in destination buffer. */
190      *pDst++ = (q15_t) (__SSAT(((q31_t) *pSrc++ * scaleFract) >> kShift, 16));
191  
192      /* Decrement loop counter */
193      blkCnt--;
194    }
195  
196  }
197  #endif /* defined(ARM_MATH_MVEI) */
198  
199  /**
200    @} end of BasicScale group
201   */