arm_shift_q15.c

00001 /* ----------------------------------------------------------------------    
00002 * Copyright (C) 2010-2014 ARM Limited. All rights reserved.    
00003 *    
00004 * $Date:        19. March 2015
00005 * $Revision:    V.1.4.5
00006 *    
00007 * Project:      CMSIS DSP Library    
00008 * Title:        arm_shift_q15.c    
00009 *    
00010 * Description:  Shifts the elements of a Q15 vector by a specified number of bits.    
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 groupMath    
00045  */
00046 
00047 /**    
00048  * @addtogroup shift    
00049  * @{    
00050  */
00051 
00052 /**    
00053  * @brief  Shifts the elements of a Q15 vector a specified number of bits.    
00054  * @param[in]  *pSrc points to the input vector    
00055  * @param[in]  shiftBits number of bits to shift.  A positive value shifts left; a negative value shifts right.    
00056  * @param[out]  *pDst points to the output vector    
00057  * @param[in]  blockSize number of samples in the vector    
00058  * @return none.    
00059  *    
00060  * <b>Scaling and Overflow Behavior:</b>    
00061  * \par    
00062  * The function uses saturating arithmetic.    
00063  * Results outside of the allowable Q15 range [0x8000 0x7FFF] will be saturated.    
00064  */
00065 
00066 void arm_shift_q15(
00067   q15_t * pSrc,
00068   int8_t shiftBits,
00069   q15_t * pDst,
00070   uint32_t blockSize)
00071 {
00072   uint32_t blkCnt;                               /* loop counter */
00073   uint8_t sign;                                  /* Sign of shiftBits */
00074 
00075 #ifndef ARM_MATH_CM0_FAMILY
00076 
00077 /* Run the below code for Cortex-M4 and Cortex-M3 */
00078 
00079   q15_t in1, in2;                                /* Temporary variables */
00080 
00081 
00082   /*loop Unrolling */
00083   blkCnt = blockSize >> 2u;
00084 
00085   /* Getting the sign of shiftBits */
00086   sign = (shiftBits & 0x80);
00087 
00088   /* If the shift value is positive then do right shift else left shift */
00089   if(sign == 0u)
00090   {
00091     /* First part of the processing with loop unrolling.  Compute 4 outputs at a time.    
00092      ** a second loop below computes the remaining 1 to 3 samples. */
00093     while(blkCnt > 0u)
00094     {
00095       /* Read 2 inputs */
00096       in1 = *pSrc++;
00097       in2 = *pSrc++;
00098       /* C = A << shiftBits */
00099       /* Shift the inputs and then store the results in the destination buffer. */
00100 #ifndef  ARM_MATH_BIG_ENDIAN
00101 
00102       *__SIMD32(pDst)++ = __PKHBT(__SSAT((in1 << shiftBits), 16),
00103                                   __SSAT((in2 << shiftBits), 16), 16);
00104 
00105 #else
00106 
00107       *__SIMD32(pDst)++ = __PKHBT(__SSAT((in2 << shiftBits), 16),
00108                                   __SSAT((in1 << shiftBits), 16), 16);
00109 
00110 #endif /* #ifndef  ARM_MATH_BIG_ENDIAN    */
00111 
00112       in1 = *pSrc++;
00113       in2 = *pSrc++;
00114 
00115 #ifndef  ARM_MATH_BIG_ENDIAN
00116 
00117       *__SIMD32(pDst)++ = __PKHBT(__SSAT((in1 << shiftBits), 16),
00118                                   __SSAT((in2 << shiftBits), 16), 16);
00119 
00120 #else
00121 
00122       *__SIMD32(pDst)++ = __PKHBT(__SSAT((in2 << shiftBits), 16),
00123                                   __SSAT((in1 << shiftBits), 16), 16);
00124 
00125 #endif /* #ifndef  ARM_MATH_BIG_ENDIAN    */
00126 
00127       /* Decrement the loop counter */
00128       blkCnt--;
00129     }
00130 
00131     /* If the blockSize is not a multiple of 4, compute any remaining output samples here.    
00132      ** No loop unrolling is used. */
00133     blkCnt = blockSize % 0x4u;
00134 
00135     while(blkCnt > 0u)
00136     {
00137       /* C = A << shiftBits */
00138       /* Shift and then store the results in the destination buffer. */
00139       *pDst++ = __SSAT((*pSrc++ << shiftBits), 16);
00140 
00141       /* Decrement the loop counter */
00142       blkCnt--;
00143     }
00144   }
00145   else
00146   {
00147     /* First part of the processing with loop unrolling.  Compute 4 outputs at a time.    
00148      ** a second loop below computes the remaining 1 to 3 samples. */
00149     while(blkCnt > 0u)
00150     {
00151       /* Read 2 inputs */
00152       in1 = *pSrc++;
00153       in2 = *pSrc++;
00154 
00155       /* C = A >> shiftBits */
00156       /* Shift the inputs and then store the results in the destination buffer. */
00157 #ifndef  ARM_MATH_BIG_ENDIAN
00158 
00159       *__SIMD32(pDst)++ = __PKHBT((in1 >> -shiftBits),
00160                                   (in2 >> -shiftBits), 16);
00161 
00162 #else
00163 
00164       *__SIMD32(pDst)++ = __PKHBT((in2 >> -shiftBits),
00165                                   (in1 >> -shiftBits), 16);
00166 
00167 #endif /* #ifndef  ARM_MATH_BIG_ENDIAN    */
00168 
00169       in1 = *pSrc++;
00170       in2 = *pSrc++;
00171 
00172 #ifndef  ARM_MATH_BIG_ENDIAN
00173 
00174       *__SIMD32(pDst)++ = __PKHBT((in1 >> -shiftBits),
00175                                   (in2 >> -shiftBits), 16);
00176 
00177 #else
00178 
00179       *__SIMD32(pDst)++ = __PKHBT((in2 >> -shiftBits),
00180                                   (in1 >> -shiftBits), 16);
00181 
00182 #endif /* #ifndef  ARM_MATH_BIG_ENDIAN    */
00183 
00184       /* Decrement the loop counter */
00185       blkCnt--;
00186     }
00187 
00188     /* If the blockSize is not a multiple of 4, compute any remaining output samples here.    
00189      ** No loop unrolling is used. */
00190     blkCnt = blockSize % 0x4u;
00191 
00192     while(blkCnt > 0u)
00193     {
00194       /* C = A >> shiftBits */
00195       /* Shift the inputs and then store the results in the destination buffer. */
00196       *pDst++ = (*pSrc++ >> -shiftBits);
00197 
00198       /* Decrement the loop counter */
00199       blkCnt--;
00200     }
00201   }
00202 
00203 #else
00204 
00205   /* Run the below code for Cortex-M0 */
00206 
00207   /* Getting the sign of shiftBits */
00208   sign = (shiftBits & 0x80);
00209 
00210   /* If the shift value is positive then do right shift else left shift */
00211   if(sign == 0u)
00212   {
00213     /* Initialize blkCnt with number of samples */
00214     blkCnt = blockSize;
00215 
00216     while(blkCnt > 0u)
00217     {
00218       /* C = A << shiftBits */
00219       /* Shift and then store the results in the destination buffer. */
00220       *pDst++ = __SSAT(((q31_t) * pSrc++ << shiftBits), 16);
00221 
00222       /* Decrement the loop counter */
00223       blkCnt--;
00224     }
00225   }
00226   else
00227   {
00228     /* Initialize blkCnt with number of samples */
00229     blkCnt = blockSize;
00230 
00231     while(blkCnt > 0u)
00232     {
00233       /* C = A >> shiftBits */
00234       /* Shift the inputs and then store the results in the destination buffer. */
00235       *pDst++ = (*pSrc++ >> -shiftBits);
00236 
00237       /* Decrement the loop counter */
00238       blkCnt--;
00239     }
00240   }
00241 
00242 #endif /* #ifndef ARM_MATH_CM0_FAMILY */
00243 
00244 }
00245 
00246 /**    
00247  * @} end of shift group    
00248  */