Aded CMSIS5 DSP and NN folder. Needs some work

Embed: (wiki syntax)

« Back to documentation index

Show/hide line numbers arm_abs_q7.c Source File

arm_abs_q7.c

00001 /* ----------------------------------------------------------------------
00002  * Project:      CMSIS DSP Library
00003  * Title:        arm_abs_q7.c
00004  * Description:  Q7 vector absolute value
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 BasicAbs
00037  * @{
00038  */
00039 
00040 /**
00041  * @brief Q7 vector absolute value.
00042  * @param[in]       *pSrc points to the input buffer
00043  * @param[out]      *pDst points to the output buffer
00044  * @param[in]       blockSize number of samples in each vector
00045  * @return none.
00046  *
00047  * \par Conditions for optimum performance
00048  *  Input and output buffers should be aligned by 32-bit
00049  *
00050  *
00051  * <b>Scaling and Overflow Behavior:</b>
00052  * \par
00053  * The function uses saturating arithmetic.
00054  * The Q7 value -1 (0x80) will be saturated to the maximum allowable positive value 0x7F.
00055  */
00056 
00057 void arm_abs_q7(
00058   q7_t * pSrc,
00059   q7_t * pDst,
00060   uint32_t blockSize)
00061 {
00062   uint32_t blkCnt;                               /* loop counter */
00063   q7_t in;                                       /* Input value1 */
00064 
00065 #if defined (ARM_MATH_DSP)
00066 
00067   /* Run the below code for Cortex-M4 and Cortex-M3 */
00068   q31_t in1, in2, in3, in4;                      /* temporary input variables */
00069   q31_t out1, out2, out3, out4;                  /* temporary output variables */
00070 
00071   /*loop Unrolling */
00072   blkCnt = blockSize >> 2U;
00073 
00074   /* First part of the processing with loop unrolling.  Compute 4 outputs at a time.
00075    ** a second loop below computes the remaining 1 to 3 samples. */
00076   while (blkCnt > 0U)
00077   {
00078     /* C = |A| */
00079     /* Read inputs */
00080     in1 = (q31_t) * pSrc;
00081     in2 = (q31_t) * (pSrc + 1);
00082     in3 = (q31_t) * (pSrc + 2);
00083 
00084     /* find absolute value */
00085     out1 = (in1 > 0) ? in1 : (q31_t)__QSUB8(0, in1);
00086 
00087     /* read input */
00088     in4 = (q31_t) * (pSrc + 3);
00089 
00090     /* find absolute value */
00091     out2 = (in2 > 0) ? in2 : (q31_t)__QSUB8(0, in2);
00092 
00093     /* store result to destination */
00094     *pDst = (q7_t) out1;
00095 
00096     /* find absolute value */
00097     out3 = (in3 > 0) ? in3 : (q31_t)__QSUB8(0, in3);
00098 
00099     /* find absolute value */
00100     out4 = (in4 > 0) ? in4 : (q31_t)__QSUB8(0, in4);
00101 
00102     /* store result to destination */
00103     *(pDst + 1) = (q7_t) out2;
00104 
00105     /* store result to destination */
00106     *(pDst + 2) = (q7_t) out3;
00107 
00108     /* store result to destination */
00109     *(pDst + 3) = (q7_t) out4;
00110 
00111     /* update pointers to process next samples */
00112     pSrc += 4U;
00113     pDst += 4U;
00114 
00115     /* Decrement the loop counter */
00116     blkCnt--;
00117   }
00118 
00119   /* If the blockSize is not a multiple of 4, compute any remaining output samples here.
00120    ** No loop unrolling is used. */
00121   blkCnt = blockSize % 0x4U;
00122 #else
00123 
00124   /* Run the below code for Cortex-M0 */
00125   blkCnt = blockSize;
00126 
00127 #endif /* #define ARM_MATH_CM0_FAMILY */
00128 
00129   while (blkCnt > 0U)
00130   {
00131     /* C = |A| */
00132     /* Read the input */
00133     in = *pSrc++;
00134 
00135     /* Store the Absolute result in the destination buffer */
00136     *pDst++ = (in > 0) ? in : ((in == (q7_t) 0x80) ? 0x7f : -in);
00137 
00138     /* Decrement the loop counter */
00139     blkCnt--;
00140   }
00141 }
00142 
00143 /**
00144  * @} end of BasicAbs group
00145  */
00146