Aded CMSIS5 DSP and NN folder. Needs some work

Embed: (wiki syntax)

« Back to documentation index

Show/hide line numbers arm_q15_to_q31.c Source File

arm_q15_to_q31.c

00001 /* ----------------------------------------------------------------------
00002  * Project:      CMSIS DSP Library
00003  * Title:        arm_q15_to_q31.c
00004  * Description:  Converts the elements of the Q15 vector to Q31 vector
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 groupSupport
00033  */
00034 
00035 /**
00036  * @addtogroup q15_to_x
00037  * @{
00038  */
00039 
00040 /**
00041  * @brief Converts the elements of the Q15 vector to Q31 vector.
00042  * @param[in]       *pSrc points to the Q15 input vector
00043  * @param[out]      *pDst points to the Q31 output vector
00044  * @param[in]       blockSize length of the input vector
00045  * @return none.
00046  *
00047  * \par Description:
00048  *
00049  * The equation used for the conversion process is:
00050  *
00051  * <pre>
00052  *  pDst[n] = (q31_t) pSrc[n] << 16;   0 <= n < blockSize.
00053  * </pre>
00054  *
00055  */
00056 
00057 
00058 void arm_q15_to_q31(
00059   q15_t * pSrc,
00060   q31_t * pDst,
00061   uint32_t blockSize)
00062 {
00063   q15_t *pIn = pSrc;                             /* Src pointer */
00064   uint32_t blkCnt;                               /* loop counter */
00065 
00066 #if defined (ARM_MATH_DSP)
00067 
00068   /* Run the below code for Cortex-M4 and Cortex-M3 */
00069   q31_t in1, in2;
00070   q31_t out1, out2, out3, out4;
00071 
00072   /*loop Unrolling */
00073   blkCnt = blockSize >> 2U;
00074 
00075   /* First part of the processing with loop unrolling.  Compute 4 outputs at a time.
00076    ** a second loop below computes the remaining 1 to 3 samples. */
00077   while (blkCnt > 0U)
00078   {
00079     /* C = (q31_t)A << 16 */
00080     /* convert from q15 to q31 and then store the results in the destination buffer */
00081     in1 = *__SIMD32(pIn)++;
00082     in2 = *__SIMD32(pIn)++;
00083 
00084 #ifndef ARM_MATH_BIG_ENDIAN
00085 
00086     /* extract lower 16 bits to 32 bit result */
00087     out1 = in1 << 16U;
00088     /* extract upper 16 bits to 32 bit result */
00089     out2 = in1 & 0xFFFF0000;
00090     /* extract lower 16 bits to 32 bit result */
00091     out3 = in2 << 16U;
00092     /* extract upper 16 bits to 32 bit result */
00093     out4 = in2 & 0xFFFF0000;
00094 
00095 #else
00096 
00097     /* extract upper 16 bits to 32 bit result */
00098     out1 = in1 & 0xFFFF0000;
00099     /* extract lower 16 bits to 32 bit result */
00100     out2 = in1 << 16U;
00101     /* extract upper 16 bits to 32 bit result */
00102     out3 = in2 & 0xFFFF0000;
00103     /* extract lower 16 bits to 32 bit result */
00104     out4 = in2 << 16U;
00105 
00106 #endif //      #ifndef ARM_MATH_BIG_ENDIAN
00107 
00108     *pDst++ = out1;
00109     *pDst++ = out2;
00110     *pDst++ = out3;
00111     *pDst++ = out4;
00112 
00113     /* Decrement the loop counter */
00114     blkCnt--;
00115   }
00116 
00117   /* If the blockSize is not a multiple of 4, compute any remaining output samples here.
00118    ** No loop unrolling is used. */
00119   blkCnt = blockSize % 0x4U;
00120 
00121 #else
00122 
00123   /* Run the below code for Cortex-M0 */
00124 
00125   /* Loop over blockSize number of values */
00126   blkCnt = blockSize;
00127 
00128 #endif /* #if defined (ARM_MATH_DSP) */
00129 
00130   while (blkCnt > 0U)
00131   {
00132     /* C = (q31_t)A << 16 */
00133     /* convert from q15 to q31 and then store the results in the destination buffer */
00134     *pDst++ = (q31_t) * pIn++ << 16;
00135 
00136     /* Decrement the loop counter */
00137     blkCnt--;
00138   }
00139 
00140 }
00141 
00142 /**
00143  * @} end of q15_to_x group
00144  */
00145