arm_shift_q15.c
Go to the documentation of this file.
00001 /* ---------------------------------------------------------------------- 00002 * Copyright (C) 2011 ARM Limited. All rights reserved. 00003 * 00004 * $Date: 15. December 2011 00005 * $Revision: V2.0.0 00006 * 00007 * Project: Cortex-R 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-R4/R5 00013 * 00014 * Version 1.0.0 2011/03/08 00015 * Alpha release. 00016 * 00017 * Version 1.0.1 2011/09/30 00018 * Beta release. 00019 * 00020 * Version 2.0.0 2011/12/15 00021 * Final release. 00022 * 00023 * -------------------------------------------------------------------- */ 00024 #include "arm_math.h" 00025 00053 void arm_shift_q15( 00054 q15_t * pSrc, 00055 int8_t shiftBits, 00056 q15_t * pDst, 00057 uint32_t blockSize) 00058 { 00059 uint32_t blkCnt; /* loop counter */ 00060 uint8_t sign; /* Sign of shiftBits */ 00061 q15_t in1, in2, in3, in4; /* Temporary variables */ 00062 00063 00064 /*loop Unrolling */ 00065 blkCnt = blockSize >> 3u; 00066 00067 /* Getting the sign of shiftBits */ 00068 sign = (shiftBits & 0x80); 00069 00070 /* If the shift value is positive then do right shift else left shift */ 00071 if(sign == 0u) 00072 { 00073 /* First part of the processing with loop unrolling. Compute 8 outputs at a time. 00074 ** a second loop below computes the remaining 1 to 7 samples. */ 00075 while(blkCnt > 0u) 00076 { 00077 /* Read 2 inputs */ 00078 in1 = *pSrc; 00079 in2 = *(pSrc + 1); 00080 in3 = *(pSrc + 2); 00081 in4 = *(pSrc + 3); 00082 00083 /* C = A << shiftBits */ 00084 /* Shift the inputs and then store the results in the destination buffer. */ 00085 #ifndef ARM_MATH_BIG_ENDIAN 00086 00087 #ifdef CCS 00088 *__SIMD32(pDst)++ = __PKHBT(__SSATA((in1 << shiftBits), 0, 16), 00089 __SSATA((in2 << shiftBits), 0, 16), 16); 00090 *__SIMD32(pDst)++ = __PKHBT(__SSATA((in3 << shiftBits), 0, 16), 00091 __SSATA((in4 << shiftBits), 0, 16), 16); 00092 #else 00093 *__SIMD32(pDst)++ = __PKHBT(__SSAT((in1 << shiftBits), 16), 00094 __SSAT((in2 << shiftBits), 16), 16); 00095 *__SIMD32(pDst)++ = __PKHBT(__SSAT((in3 << shiftBits), 16), 00096 __SSAT((in4 << shiftBits), 16), 16); 00097 #endif // #ifdef CCS 00098 00099 #else 00100 00101 #ifdef CCS 00102 *__SIMD32(pDst)++ = __PKHBT(__SSATA((in2 << shiftBits), 0, 16), 00103 __SSATA((in1 << shiftBits), 0, 16), 16); 00104 *__SIMD32(pDst)++ = __PKHBT(__SSATA((in4 << shiftBits), 0, 16), 00105 __SSATA((in3 << shiftBits), 0, 16), 16); 00106 #else 00107 *__SIMD32(pDst)++ = __PKHBT(__SSAT((in2 << shiftBits), 16), 00108 __SSAT((in1 << shiftBits), 16), 16); 00109 *__SIMD32(pDst)++ = __PKHBT(__SSAT((in4 << shiftBits), 16), 00110 __SSAT((in3 << shiftBits), 16), 16); 00111 #endif // #ifdef CCS 00112 00113 #endif /* #ifndef ARM_MATH_BIG_ENDIAN */ 00114 in1 = *(pSrc + 4); 00115 in2 = *(pSrc + 5); 00116 in3 = *(pSrc + 6); 00117 in4 = *(pSrc + 7); 00118 00119 /* C = A << shiftBits */ 00120 /* Shift the inputs and then store the results in the destination buffer. */ 00121 #ifndef ARM_MATH_BIG_ENDIAN 00122 00123 #ifdef CCS 00124 *__SIMD32(pDst)++ = __PKHBT(__SSATA((in1 << shiftBits), 0, 16), 00125 __SSATA((in2 << shiftBits), 0, 16), 16); 00126 *__SIMD32(pDst)++ = __PKHBT(__SSATA((in3 << shiftBits), 0, 16), 00127 __SSATA((in4 << shiftBits), 0, 16), 16); 00128 #else 00129 *__SIMD32(pDst)++ = __PKHBT(__SSAT((in1 << shiftBits), 16), 00130 __SSAT((in2 << shiftBits), 16), 16); 00131 *__SIMD32(pDst)++ = __PKHBT(__SSAT((in3 << shiftBits), 16), 00132 __SSAT((in4 << shiftBits), 16), 16); 00133 #endif // #ifdef CCS 00134 00135 #else 00136 00137 #ifdef CCS 00138 *__SIMD32(pDst)++ = __PKHBT(__SSATA((in2 << shiftBits), 0, 16), 00139 __SSATA((in1 << shiftBits), 0, 16), 16); 00140 *__SIMD32(pDst)++ = __PKHBT(__SSATA((in4 << shiftBits), 0, 16), 00141 __SSATA((in3 << shiftBits), 0, 16), 16); 00142 #else 00143 *__SIMD32(pDst)++ = __PKHBT(__SSAT((in2 << shiftBits), 16), 00144 __SSAT((in1 << shiftBits), 16), 16); 00145 *__SIMD32(pDst)++ = __PKHBT(__SSAT((in4 << shiftBits), 16), 00146 __SSAT((in3 << shiftBits), 16), 16); 00147 #endif // #ifdef CCS 00148 00149 #endif /* #ifndef ARM_MATH_BIG_ENDIAN */ 00150 00151 /* Update source pointer to process next sampels */ 00152 pSrc += 8u; 00153 00154 /* Decrement the loop counter */ 00155 blkCnt--; 00156 } 00157 00158 /* If the blockSize is not a multiple of 8, compute any remaining output samples here. 00159 ** No loop unrolling is used. */ 00160 blkCnt = blockSize % 0x8u; 00161 00162 while(blkCnt > 0u) 00163 { 00164 /* C = A << shiftBits */ 00165 /* Shift and then store the results in the destination buffer. */ 00166 #ifdef CCS 00167 *pDst++ = __SSATA((*pSrc++ << shiftBits), 0, 16); 00168 #else 00169 *pDst++ = __SSAT((*pSrc++ << shiftBits), 16); 00170 #endif // #ifdef CCS 00171 /* Decrement the loop counter */ 00172 blkCnt--; 00173 } 00174 } 00175 else 00176 { 00177 shiftBits = -shiftBits; 00178 00179 /* First part of the processing with loop unrolling. Compute 8 outputs at a time. 00180 ** a second loop below computes the remaining 1 to 7 samples. */ 00181 while(blkCnt > 0u) 00182 { 00183 /* Read 2 inputs */ 00184 in1 = *pSrc; 00185 in2 = *(pSrc + 1); 00186 in3 = *(pSrc + 2); 00187 in4 = *(pSrc + 3); 00188 00189 /* C = A >> shiftBits */ 00190 /* Shift the inputs and then store the results in the destination buffer. */ 00191 00192 #ifndef ARM_MATH_BIG_ENDIAN 00193 00194 *__SIMD32(pDst)++ = __PKHBT((in1 >> shiftBits), 00195 (in2 >> shiftBits), 16); 00196 *__SIMD32(pDst)++ = __PKHBT((in3 >> shiftBits), 00197 (in4 >> shiftBits), 16); 00198 #else 00199 00200 *__SIMD32(pDst)++ = __PKHBT((in2 >> shiftBits), 00201 (in1 >> shiftBits), 16); 00202 *__SIMD32(pDst)++ = __PKHBT((in4 >> shiftBits), 00203 (in3 >> shiftBits), 16); 00204 00205 #endif /* #ifndef ARM_MATH_BIG_ENDIAN */ 00206 00207 in1 = *(pSrc + 4); 00208 in2 = *(pSrc + 5); 00209 in3 = *(pSrc + 6); 00210 in4 = *(pSrc + 7); 00211 00212 /* C = A >> shiftBits */ 00213 /* Shift the inputs and then store the results in the destination buffer. */ 00214 00215 #ifndef ARM_MATH_BIG_ENDIAN 00216 00217 *__SIMD32(pDst)++ = __PKHBT((in1 >> shiftBits), 00218 (in2 >> shiftBits), 16); 00219 *__SIMD32(pDst)++ = __PKHBT((in3 >> shiftBits), 00220 (in4 >> shiftBits), 16); 00221 #else 00222 00223 *__SIMD32(pDst)++ = __PKHBT((in2 >> shiftBits), 00224 (in1 >> shiftBits), 16); 00225 *__SIMD32(pDst)++ = __PKHBT((in4 >> shiftBits), 00226 (in3 >> shiftBits), 16); 00227 00228 #endif /* #ifndef ARM_MATH_BIG_ENDIAN */ 00229 00230 pSrc += 8u; 00231 00232 /* Decrement the loop counter */ 00233 blkCnt--; 00234 } 00235 00236 /* If the blockSize is not a multiple of 8, compute any remaining output samples here. 00237 ** No loop unrolling is used. */ 00238 blkCnt = blockSize % 0x8u; 00239 00240 while(blkCnt > 0u) 00241 { 00242 /* C = A >> shiftBits */ 00243 /* Shift the inputs and then store the results in the destination buffer. */ 00244 in1 = *pSrc++; 00245 00246 *pDst++ = (in1 >> shiftBits); 00247 00248 /* Decrement the loop counter */ 00249 blkCnt--; 00250 } 00251 } 00252 } 00253
