arm_scale_q31.c

00001 /* ----------------------------------------------------------------------
00002  * Project:      CMSIS DSP Library
00003  * Title:        arm_scale_q31.c
00004  * Description:  Multiplies a Q31 vector by a scalar
00005  *
00006  * $Date:        27. January 2017
00007  * $Revision:    V.1.5.1
00008  *
00009  * Target Processor: Cortex-M cores
00010  * -------------------------------------------------------------------- */
00011 /*
00012  * Copyright (C) 2010-2017 ARM Limited or its affiliates. All rights reserved.
00013  *
00014  * SPDX-License-Identifier: Apache-2.0
00015  *
00016  * Licensed under the Apache License, Version 2.0 (the License); you may
00017  * not use this file except in compliance with the License.
00018  * You may obtain a copy of the License at
00019  *
00020  * www.apache.org/licenses/LICENSE-2.0
00021  *
00022  * Unless required by applicable law or agreed to in writing, software
00023  * distributed under the License is distributed on an AS IS BASIS, WITHOUT
00024  * WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
00025  * See the License for the specific language governing permissions and
00026  * limitations under the License.
00027  */
00028 
00029 #include "arm_math.h"
00030 
00031 /**
00032  * @ingroup groupMath
00033  */
00034 
00035 /**
00036  * @addtogroup scale
00037  * @{
00038  */
00039 
00040 /**
00041  * @brief Multiplies a Q31 vector by a scalar.
00042  * @param[in]       *pSrc points to the input vector
00043  * @param[in]       scaleFract fractional portion of the scale value
00044  * @param[in]       shift number of bits to shift the result by
00045  * @param[out]      *pDst points to the output vector
00046  * @param[in]       blockSize number of samples in the vector
00047  * @return none.
00048  *
00049  * <b>Scaling and Overflow Behavior:</b>
00050  * \par
00051  * The input data <code>*pSrc</code> and <code>scaleFract</code> are in 1.31 format.
00052  * These are multiplied to yield a 2.62 intermediate result and this is shifted with saturation to 1.31 format.
00053  */
00054 
00055 void arm_scale_q31(
00056   q31_t * pSrc,
00057   q31_t scaleFract,
00058   int8_t shift,
00059   q31_t * pDst,
00060   uint32_t blockSize)
00061 {
00062   int8_t kShift = shift + 1;                     /* Shift to apply after scaling */
00063   int8_t sign = (kShift & 0x80);
00064   uint32_t blkCnt;                               /* loop counter */
00065   q31_t in, out;
00066 
00067 #if defined (ARM_MATH_DSP)
00068 
00069 /* Run the below code for Cortex-M4 and Cortex-M3 */
00070 
00071   q31_t in1, in2, in3, in4;                      /* temporary input variables */
00072   q31_t out1, out2, out3, out4;                  /* temporary output variabels */
00073 
00074 
00075   /*loop Unrolling */
00076   blkCnt = blockSize >> 2U;
00077 
00078   if (sign == 0U)
00079   {
00080     /* First part of the processing with loop unrolling.  Compute 4 outputs at a time.
00081      ** a second loop below computes the remaining 1 to 3 samples. */
00082     while (blkCnt > 0U)
00083     {
00084       /* read four inputs from source */
00085       in1 = *pSrc;
00086       in2 = *(pSrc + 1);
00087       in3 = *(pSrc + 2);
00088       in4 = *(pSrc + 3);
00089 
00090       /* multiply input with scaler value */
00091       in1 = ((q63_t) in1 * scaleFract) >> 32;
00092       in2 = ((q63_t) in2 * scaleFract) >> 32;
00093       in3 = ((q63_t) in3 * scaleFract) >> 32;
00094       in4 = ((q63_t) in4 * scaleFract) >> 32;
00095 
00096       /* apply shifting */
00097       out1 = in1 << kShift;
00098       out2 = in2 << kShift;
00099 
00100       /* saturate the results. */
00101       if (in1 != (out1 >> kShift))
00102         out1 = 0x7FFFFFFF ^ (in1 >> 31);
00103 
00104       if (in2 != (out2 >> kShift))
00105         out2 = 0x7FFFFFFF ^ (in2 >> 31);
00106 
00107       out3 = in3 << kShift;
00108       out4 = in4 << kShift;
00109 
00110       *pDst = out1;
00111       *(pDst + 1) = out2;
00112 
00113       if (in3 != (out3 >> kShift))
00114         out3 = 0x7FFFFFFF ^ (in3 >> 31);
00115 
00116       if (in4 != (out4 >> kShift))
00117         out4 = 0x7FFFFFFF ^ (in4 >> 31);
00118 
00119       /* Store result destination */
00120       *(pDst + 2) = out3;
00121       *(pDst + 3) = out4;
00122 
00123       /* Update pointers to process next sampels */
00124       pSrc += 4U;
00125       pDst += 4U;
00126 
00127       /* Decrement the loop counter */
00128       blkCnt--;
00129     }
00130 
00131   }
00132   else
00133   {
00134     /* First part of the processing with loop unrolling.  Compute 4 outputs at a time.
00135      ** a second loop below computes the remaining 1 to 3 samples. */
00136     while (blkCnt > 0U)
00137     {
00138       /* read four inputs from source */
00139       in1 = *pSrc;
00140       in2 = *(pSrc + 1);
00141       in3 = *(pSrc + 2);
00142       in4 = *(pSrc + 3);
00143 
00144       /* multiply input with scaler value */
00145       in1 = ((q63_t) in1 * scaleFract) >> 32;
00146       in2 = ((q63_t) in2 * scaleFract) >> 32;
00147       in3 = ((q63_t) in3 * scaleFract) >> 32;
00148       in4 = ((q63_t) in4 * scaleFract) >> 32;
00149 
00150       /* apply shifting */
00151       out1 = in1 >> -kShift;
00152       out2 = in2 >> -kShift;
00153 
00154       out3 = in3 >> -kShift;
00155       out4 = in4 >> -kShift;
00156 
00157       /* Store result destination */
00158       *pDst = out1;
00159       *(pDst + 1) = out2;
00160 
00161       *(pDst + 2) = out3;
00162       *(pDst + 3) = out4;
00163 
00164       /* Update pointers to process next sampels */
00165       pSrc += 4U;
00166       pDst += 4U;
00167 
00168       /* Decrement the loop counter */
00169       blkCnt--;
00170     }
00171   }
00172   /* If the blockSize is not a multiple of 4, compute any remaining output samples here.
00173    ** No loop unrolling is used. */
00174   blkCnt = blockSize % 0x4U;
00175 
00176 #else
00177 
00178   /* Run the below code for Cortex-M0 */
00179 
00180   /* Initialize blkCnt with number of samples */
00181   blkCnt = blockSize;
00182 
00183 #endif /* #if defined (ARM_MATH_DSP) */
00184 
00185   if (sign == 0)
00186   {
00187       while (blkCnt > 0U)
00188       {
00189         /* C = A * scale */
00190         /* Scale the input and then store the result in the destination buffer. */
00191         in = *pSrc++;
00192         in = ((q63_t) in * scaleFract) >> 32;
00193 
00194         out = in << kShift;
00195 
00196         if (in != (out >> kShift))
00197             out = 0x7FFFFFFF ^ (in >> 31);
00198 
00199         *pDst++ = out;
00200 
00201         /* Decrement the loop counter */
00202         blkCnt--;
00203       }
00204   }
00205   else
00206   {
00207       while (blkCnt > 0U)
00208       {
00209         /* C = A * scale */
00210         /* Scale the input and then store the result in the destination buffer. */
00211         in = *pSrc++;
00212         in = ((q63_t) in * scaleFract) >> 32;
00213 
00214         out = in >> -kShift;
00215 
00216         *pDst++ = out;
00217 
00218         /* Decrement the loop counter */
00219         blkCnt--;
00220       }
00221 
00222   }
00223 }
00224 
00225 /**
00226  * @} end of scale group
00227  */
00228