arm_mat_scale_q15.c

00001 /* ----------------------------------------------------------------------    
00002 * Copyright (C) 2010-2014 ARM Limited. All rights reserved.    
00003 *    
00004 * $Date:        12. March 2014 
00005 * $Revision:    V1.4.3
00006 *    
00007 * Project:      CMSIS DSP Library    
00008 * Title:        arm_mat_scale_q15.c    
00009 *    
00010 * Description:  Multiplies a Q15 matrix by a scalar.    
00011 *    
00012 * Target Processor: Cortex-M4/Cortex-M3/Cortex-M0
00013 *  
00014 * Redistribution and use in source and binary forms, with or without 
00015 * modification, are permitted provided that the following conditions
00016 * are met:
00017 *   - Redistributions of source code must retain the above copyright
00018 *     notice, this list of conditions and the following disclaimer.
00019 *   - Redistributions in binary form must reproduce the above copyright
00020 *     notice, this list of conditions and the following disclaimer in
00021 *     the documentation and/or other materials provided with the 
00022 *     distribution.
00023 *   - Neither the name of ARM LIMITED nor the names of its contributors
00024 *     may be used to endorse or promote products derived from this
00025 *     software without specific prior written permission.
00026 *
00027 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
00028 * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
00029 * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
00030 * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE 
00031 * COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
00032 * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING,
00033 * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
00034 * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
00035 * CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
00036 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN
00037 * ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
00038 * POSSIBILITY OF SUCH DAMAGE.  
00039 * -------------------------------------------------------------------- */
00040 
00041 #include "arm_math.h"
00042 
00043 /**    
00044  * @ingroup groupMatrix    
00045  */
00046 
00047 /**    
00048  * @addtogroup MatrixScale    
00049  * @{    
00050  */
00051 
00052 /**    
00053  * @brief Q15 matrix scaling.    
00054  * @param[in]       *pSrc points to input matrix    
00055  * @param[in]       scaleFract fractional portion of the scale factor    
00056  * @param[in]       shift number of bits to shift the result by    
00057  * @param[out]      *pDst points to output matrix structure    
00058  * @return          The function returns either    
00059  * <code>ARM_MATH_SIZE_MISMATCH</code> or <code>ARM_MATH_SUCCESS</code> based on the outcome of size checking.    
00060  *    
00061  * @details    
00062  * <b>Scaling and Overflow Behavior:</b>    
00063  * \par    
00064  * The input data <code>*pSrc</code> and <code>scaleFract</code> are in 1.15 format.    
00065  * These are multiplied to yield a 2.30 intermediate result and this is shifted with saturation to 1.15 format.    
00066  */
00067 
00068 arm_status arm_mat_scale_q15(
00069   const arm_matrix_instance_q15 * pSrc,
00070   q15_t scaleFract,
00071   int32_t shift,
00072   arm_matrix_instance_q15 * pDst)
00073 {
00074   q15_t *pIn = pSrc->pData;                      /* input data matrix pointer */
00075   q15_t *pOut = pDst->pData;                     /* output data matrix pointer */
00076   uint32_t numSamples;                           /* total number of elements in the matrix */
00077   int32_t totShift = 15 - shift;                 /* total shift to apply after scaling */
00078   uint32_t blkCnt;                               /* loop counters */
00079   arm_status status;                             /* status of matrix scaling     */
00080 
00081 #ifndef ARM_MATH_CM0_FAMILY
00082 
00083   q15_t in1, in2, in3, in4;
00084   q31_t out1, out2, out3, out4;
00085   q31_t inA1, inA2;
00086 
00087 #endif //     #ifndef ARM_MATH_CM0_FAMILY
00088 
00089 #ifdef ARM_MATH_MATRIX_CHECK
00090   /* Check for matrix mismatch */
00091   if((pSrc->numRows != pDst->numRows) || (pSrc->numCols != pDst->numCols))
00092   {
00093     /* Set status as ARM_MATH_SIZE_MISMATCH */
00094     status = ARM_MATH_SIZE_MISMATCH;
00095   }
00096   else
00097 #endif //    #ifdef ARM_MATH_MATRIX_CHECK
00098   {
00099     /* Total number of samples in the input matrix */
00100     numSamples = (uint32_t) pSrc->numRows * pSrc->numCols;
00101 
00102 #ifndef ARM_MATH_CM0_FAMILY
00103 
00104     /* Run the below code for Cortex-M4 and Cortex-M3 */
00105     /* Loop Unrolling */
00106     blkCnt = numSamples >> 2;
00107 
00108     /* First part of the processing with loop unrolling.  Compute 4 outputs at a time.    
00109      ** a second loop below computes the remaining 1 to 3 samples. */
00110     while(blkCnt > 0u)
00111     {
00112       /* C(m,n) = A(m,n) * k */
00113       /* Scale, saturate and then store the results in the destination buffer. */
00114       /* Reading 2 inputs from memory */
00115       inA1 = _SIMD32_OFFSET(pIn);
00116       inA2 = _SIMD32_OFFSET(pIn + 2);
00117 
00118       /* C = A * scale */
00119       /* Scale the inputs and then store the 2 results in the destination buffer        
00120        * in single cycle by packing the outputs */
00121       out1 = (q31_t) ((q15_t) (inA1 >> 16) * scaleFract);
00122       out2 = (q31_t) ((q15_t) inA1 * scaleFract);
00123       out3 = (q31_t) ((q15_t) (inA2 >> 16) * scaleFract);
00124       out4 = (q31_t) ((q15_t) inA2 * scaleFract);
00125 
00126       out1 = out1 >> totShift;
00127       inA1 = _SIMD32_OFFSET(pIn + 4);
00128       out2 = out2 >> totShift;
00129       inA2 = _SIMD32_OFFSET(pIn + 6);
00130       out3 = out3 >> totShift;
00131       out4 = out4 >> totShift;
00132 
00133       in1 = (q15_t) (__SSAT(out1, 16));
00134       in2 = (q15_t) (__SSAT(out2, 16));
00135       in3 = (q15_t) (__SSAT(out3, 16));
00136       in4 = (q15_t) (__SSAT(out4, 16));
00137 
00138       _SIMD32_OFFSET(pOut) = __PKHBT(in2, in1, 16);
00139       _SIMD32_OFFSET(pOut + 2) = __PKHBT(in4, in3, 16);
00140 
00141       /* update pointers to process next sampels */
00142       pIn += 4u;
00143       pOut += 4u;
00144 
00145 
00146       /* Decrement the numSamples loop counter */
00147       blkCnt--;
00148     }
00149 
00150     /* If the numSamples is not a multiple of 4, compute any remaining output samples here.        
00151      ** No loop unrolling is used. */
00152     blkCnt = numSamples % 0x4u;
00153 
00154 #else
00155 
00156     /* Run the below code for Cortex-M0 */
00157 
00158     /* Initialize blkCnt with number of samples */
00159     blkCnt = numSamples;
00160 
00161 #endif /* #ifndef ARM_MATH_CM0_FAMILY */
00162 
00163     while(blkCnt > 0u)
00164     {
00165       /* C(m,n) = A(m,n) * k */
00166       /* Scale, saturate and then store the results in the destination buffer. */
00167       *pOut++ =
00168         (q15_t) (__SSAT(((q31_t) (*pIn++) * scaleFract) >> totShift, 16));
00169 
00170       /* Decrement the numSamples loop counter */
00171       blkCnt--;
00172     }
00173     /* Set status as ARM_MATH_SUCCESS */
00174     status = ARM_MATH_SUCCESS;
00175   }
00176 
00177   /* Return to application */
00178   return (status);
00179 }
00180 
00181 /**        
00182  * @} end of MatrixScale group        
00183  */