lib/main/CMSIS/DSP/Source/BasicMathFunctions/arm_scale_q31.c

   1 /* ----------------------------------------------------------------------
   2  * Project:      CMSIS DSP Library
   3  * Title:        arm_scale_q31.c
   4  * Description:  Multiplies a Q31 vector by a scalar
   5  *
   6  * $Date:        27. January 2017
   7  * $Revision:    V.1.5.1
   8  *
   9  * Target Processor: Cortex-M cores
  10  * -------------------------------------------------------------------- */
  11 /*
  12  * Copyright (C) 2010-2017 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 "arm_math.h"
  30
  31 /**
  32  * @ingroup groupMath
  33  */
  34
  35 /**
  36  * @addtogroup scale
  37  * @{
  38  */
  39
  40 /**
  41  * @brief Multiplies a Q31 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 the vector
  47  * @return none.
  48  *
  49  * <b>Scaling and Overflow Behavior:</b>
  50  * \par
  51  * The input data <code>*pSrc</code> and <code>scaleFract</code> are in 1.31 format.
  52  * These are multiplied to yield a 2.62 intermediate result and this is shifted with saturation to 1.31 format.
  53  */
  54
  55 void arm_scale_q31(
  56   q31_t * pSrc,
  57   q31_t scaleFract,
  58   int8_t shift,
  59   q31_t * pDst,
  60   uint32_t blockSize)
  61 {
  62   int8_t kShift = shift + 1;                     /* Shift to apply after scaling */
  63   int8_t sign = (kShift & 0x80);
  64   uint32_t blkCnt;                               /* loop counter */
  65   q31_t in, out;
  66
  67 #if defined (ARM_MATH_DSP)
  68
  69 /* Run the below code for Cortex-M4 and Cortex-M3 */
  70
  71   q31_t in1, in2, in3, in4;                      /* temporary input variables */
  72   q31_t out1, out2, out3, out4;                  /* temporary output variabels */
  73
  74
  75   /*loop Unrolling */
  76   blkCnt = blockSize >> 2U;
  77
  78   if (sign == 0U)
  79   {
  80     /* First part of the processing with loop unrolling.  Compute 4 outputs at a time.
  81      ** a second loop below computes the remaining 1 to 3 samples. */
  82     while (blkCnt > 0U)
  83     {
  84       /* read four inputs from source */
  85       in1 = *pSrc;
  86       in2 = *(pSrc + 1);
  87       in3 = *(pSrc + 2);
  88       in4 = *(pSrc + 3);
  89
  90       /* multiply input with scaler value */
  91       in1 = ((q63_t) in1 * scaleFract) >> 32;
  92       in2 = ((q63_t) in2 * scaleFract) >> 32;
  93       in3 = ((q63_t) in3 * scaleFract) >> 32;
  94       in4 = ((q63_t) in4 * scaleFract) >> 32;
  95
  96       /* apply shifting */
  97       out1 = in1 << kShift;
  98       out2 = in2 << kShift;
  99
 100       /* saturate the results. */
 101       if (in1 != (out1 >> kShift))
 102         out1 = 0x7FFFFFFF ^ (in1 >> 31);
 103
 104       if (in2 != (out2 >> kShift))
 105         out2 = 0x7FFFFFFF ^ (in2 >> 31);
 106
 107       out3 = in3 << kShift;
 108       out4 = in4 << kShift;
 109
 110       *pDst = out1;
 111       *(pDst + 1) = out2;
 112
 113       if (in3 != (out3 >> kShift))
 114         out3 = 0x7FFFFFFF ^ (in3 >> 31);
 115
 116       if (in4 != (out4 >> kShift))
 117         out4 = 0x7FFFFFFF ^ (in4 >> 31);
 118
 119       /* Store result destination */
 120       *(pDst + 2) = out3;
 121       *(pDst + 3) = out4;
 122
 123       /* Update pointers to process next sampels */
 124       pSrc += 4U;
 125       pDst += 4U;
 126
 127       /* Decrement the loop counter */
 128       blkCnt--;
 129     }
 130
 131   }
 132   else
 133   {
 134     /* First part of the processing with loop unrolling.  Compute 4 outputs at a time.
 135      ** a second loop below computes the remaining 1 to 3 samples. */
 136     while (blkCnt > 0U)
 137     {
 138       /* read four inputs from source */
 139       in1 = *pSrc;
 140       in2 = *(pSrc + 1);
 141       in3 = *(pSrc + 2);
 142       in4 = *(pSrc + 3);
 143
 144       /* multiply input with scaler value */
 145       in1 = ((q63_t) in1 * scaleFract) >> 32;
 146       in2 = ((q63_t) in2 * scaleFract) >> 32;
 147       in3 = ((q63_t) in3 * scaleFract) >> 32;
 148       in4 = ((q63_t) in4 * scaleFract) >> 32;
 149
 150       /* apply shifting */
 151       out1 = in1 >> -kShift;
 152       out2 = in2 >> -kShift;
 153
 154       out3 = in3 >> -kShift;
 155       out4 = in4 >> -kShift;
 156
 157       /* Store result destination */
 158       *pDst = out1;
 159       *(pDst + 1) = out2;
 160
 161       *(pDst + 2) = out3;
 162       *(pDst + 3) = out4;
 163
 164       /* Update pointers to process next sampels */
 165       pSrc += 4U;
 166       pDst += 4U;
 167
 168       /* Decrement the loop counter */
 169       blkCnt--;
 170     }
 171   }
 172   /* If the blockSize is not a multiple of 4, compute any remaining output samples here.
 173    ** No loop unrolling is used. */
 174   blkCnt = blockSize % 0x4U;
 175
 176 #else
 177
 178   /* Run the below code for Cortex-M0 */
 179
 180   /* Initialize blkCnt with number of samples */
 181   blkCnt = blockSize;
 182
 183 #endif /* #if defined (ARM_MATH_DSP) */
 184
 185   if (sign == 0)
 186   {
 187           while (blkCnt > 0U)
 188           {
 189                 /* C = A * scale */
 190                 /* Scale the input and then store the result in the destination buffer. */
 191                 in = *pSrc++;
 192                 in = ((q63_t) in * scaleFract) >> 32;
 193
 194                 out = in << kShift;
 195
 196                 if (in != (out >> kShift))
 197                         out = 0x7FFFFFFF ^ (in >> 31);
 198
 199                 *pDst++ = out;
 200
 201                 /* Decrement the loop counter */
 202                 blkCnt--;
 203           }
 204   }
 205   else
 206   {
 207           while (blkCnt > 0U)
 208           {
 209                 /* C = A * scale */
 210                 /* Scale the input and then store the result in the destination buffer. */
 211                 in = *pSrc++;
 212                 in = ((q63_t) in * scaleFract) >> 32;
 213
 214                 out = in >> -kShift;
 215
 216                 *pDst++ = out;
 217
 218                 /* Decrement the loop counter */
 219                 blkCnt--;
 220           }
 221
 222   }
 223 }
 224
 225 /**
 226  * @} end of scale group
 227  */