arm_mat_sub_q31.c

00001 /* ----------------------------------------------------------------------
00002  * Project:      CMSIS DSP Library
00003  * Title:        arm_mat_sub_q31.c
00004  * Description:  Q31 matrix subtraction
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 groupMatrix
00033  */
00034 
00035 /**
00036  * @addtogroup MatrixSub
00037  * @{
00038  */
00039 
00040 /**
00041  * @brief Q31 matrix subtraction.
00042  * @param[in]       *pSrcA points to the first input matrix structure
00043  * @param[in]       *pSrcB points to the second input matrix structure
00044  * @param[out]      *pDst points to output matrix structure
00045  * @return          The function returns either
00046  * <code>ARM_MATH_SIZE_MISMATCH</code> or <code>ARM_MATH_SUCCESS</code> based on the outcome of size checking.
00047  *
00048  * <b>Scaling and Overflow Behavior:</b>
00049  * \par
00050  * The function uses saturating arithmetic.
00051  * Results outside of the allowable Q31 range [0x80000000 0x7FFFFFFF] will be saturated.
00052  */
00053 
00054 
00055 arm_status arm_mat_sub_q31(
00056   const arm_matrix_instance_q31 * pSrcA,
00057   const arm_matrix_instance_q31 * pSrcB,
00058   arm_matrix_instance_q31 * pDst)
00059 {
00060   q31_t *pIn1 = pSrcA->pData;                    /* input data matrix pointer A */
00061   q31_t *pIn2 = pSrcB->pData;                    /* input data matrix pointer B */
00062   q31_t *pOut = pDst->pData;                     /* output data matrix pointer */
00063   q31_t inA1, inB1;                              /* temporary variables */
00064 
00065 #if defined (ARM_MATH_DSP)
00066 
00067   q31_t inA2, inB2;                              /* temporary variables */
00068   q31_t out1, out2;                              /* temporary variables */
00069 
00070 #endif //      #if defined (ARM_MATH_DSP)
00071 
00072   uint32_t numSamples;                           /* total number of elements in the matrix  */
00073   uint32_t blkCnt;                               /* loop counters */
00074   arm_status status;                             /* status of matrix subtraction */
00075 
00076 
00077 #ifdef ARM_MATH_MATRIX_CHECK
00078   /* Check for matrix mismatch condition  */
00079   if ((pSrcA->numRows != pSrcB->numRows) ||
00080      (pSrcA->numCols != pSrcB->numCols) ||
00081      (pSrcA->numRows != pDst->numRows) || (pSrcA->numCols != pDst->numCols))
00082   {
00083     /* Set status as ARM_MATH_SIZE_MISMATCH */
00084     status = ARM_MATH_SIZE_MISMATCH;
00085   }
00086   else
00087 #endif
00088   {
00089     /* Total number of samples in the input matrix */
00090     numSamples = (uint32_t) pSrcA->numRows * pSrcA->numCols;
00091 
00092 #if defined (ARM_MATH_DSP)
00093 
00094     /* Run the below code for Cortex-M4 and Cortex-M3 */
00095 
00096     /* Loop Unrolling */
00097     blkCnt = numSamples >> 2U;
00098 
00099     /* First part of the processing with loop unrolling.  Compute 4 outputs at a time.
00100      ** a second loop below computes the remaining 1 to 3 samples. */
00101     while (blkCnt > 0U)
00102     {
00103       /* C(m,n) = A(m,n) - B(m,n) */
00104       /* Subtract, saturate and then store the results in the destination buffer. */
00105       /* Read values from source A */
00106       inA1 = pIn1[0];
00107 
00108       /* Read values from source B */
00109       inB1 = pIn2[0];
00110 
00111       /* Read values from source A */
00112       inA2 = pIn1[1];
00113 
00114       /* Subtract and saturate */
00115       out1 = __QSUB(inA1, inB1);
00116 
00117       /* Read values from source B */
00118       inB2 = pIn2[1];
00119 
00120       /* Read values from source A */
00121       inA1 = pIn1[2];
00122 
00123       /* Subtract and saturate */
00124       out2 = __QSUB(inA2, inB2);
00125 
00126       /* Read values from source B */
00127       inB1 = pIn2[2];
00128 
00129       /* Store result in destination */
00130       pOut[0] = out1;
00131       pOut[1] = out2;
00132 
00133       /* Read values from source A */
00134       inA2 = pIn1[3];
00135 
00136       /* Read values from source B */
00137       inB2 = pIn2[3];
00138 
00139       /* Subtract and saturate */
00140       out1 = __QSUB(inA1, inB1);
00141 
00142       /* Subtract and saturate */
00143       out2 = __QSUB(inA2, inB2);
00144 
00145       /* Store result in destination */
00146       pOut[2] = out1;
00147       pOut[3] = out2;
00148 
00149       /* update pointers to process next samples */
00150       pIn1 += 4U;
00151       pIn2 += 4U;
00152       pOut += 4U;
00153 
00154       /* Decrement the loop counter */
00155       blkCnt--;
00156     }
00157 
00158     /* If the numSamples is not a multiple of 4, compute any remaining output samples here.
00159      ** No loop unrolling is used. */
00160     blkCnt = numSamples % 0x4U;
00161 
00162 #else
00163 
00164     /* Run the below code for Cortex-M0 */
00165 
00166     /* Initialize blkCnt with number of samples */
00167     blkCnt = numSamples;
00168 
00169 #endif /* #if defined (ARM_MATH_DSP) */
00170 
00171     while (blkCnt > 0U)
00172     {
00173       /* C(m,n) = A(m,n) - B(m,n) */
00174       /* Subtract, saturate and then store the results in the destination buffer. */
00175       inA1 = *pIn1++;
00176       inB1 = *pIn2++;
00177 
00178       inA1 = __QSUB(inA1, inB1);
00179 
00180       *pOut++ = inA1;
00181 
00182       /* Decrement the loop counter */
00183       blkCnt--;
00184     }
00185 
00186     /* Set status as ARM_MATH_SUCCESS */
00187     status = ARM_MATH_SUCCESS;
00188   }
00189 
00190   /* Return to application */
00191   return (status);
00192 }
00193 
00194 /**
00195  * @} end of MatrixSub group
00196  */
00197