Details | Last modification | View Log | RSS feed
| Rev | Author | Line No. | Line |
|---|---|---|---|
| 56 | mjames | 1 | /* ---------------------------------------------------------------------- |
| 2 | * Project: CMSIS DSP Library |
||
| 3 | * Title: arm_shift_q31.c |
||
| 4 | * Description: Shifts the elements of a Q31 vector by a specified number of bits |
||
| 5 | * |
||
| 6 | * $Date: 27. January 2017 |
||
| 7 | * $Revision: V.1.5.1 |
||
| 8 | * |
||
| 9 | * Target Processor: Cortex-M cores |
||
| 10 | * -------------------------------------------------------------------- */ |
||
| 11 | /* |
||
| 12 | * Copyright (C) 2010-2017 ARM Limited or its affiliates. All rights reserved. |
||
| 13 | * |
||
| 14 | * SPDX-License-Identifier: Apache-2.0 |
||
| 15 | * |
||
| 16 | * Licensed under the Apache License, Version 2.0 (the License); you may |
||
| 17 | * not use this file except in compliance with the License. |
||
| 18 | * You may obtain a copy of the License at |
||
| 19 | * |
||
| 20 | * www.apache.org/licenses/LICENSE-2.0 |
||
| 21 | * |
||
| 22 | * Unless required by applicable law or agreed to in writing, software |
||
| 23 | * distributed under the License is distributed on an AS IS BASIS, WITHOUT |
||
| 24 | * WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. |
||
| 25 | * See the License for the specific language governing permissions and |
||
| 26 | * limitations under the License. |
||
| 27 | */ |
||
| 28 | |||
| 29 | #include "arm_math.h" |
||
| 30 | |||
| 31 | /** |
||
| 32 | * @ingroup groupMath |
||
| 33 | */ |
||
| 34 | /** |
||
| 35 | * @defgroup shift Vector Shift |
||
| 36 | * |
||
| 37 | * Shifts the elements of a fixed-point vector by a specified number of bits. |
||
| 38 | * There are separate functions for Q7, Q15, and Q31 data types. |
||
| 39 | * The underlying algorithm used is: |
||
| 40 | * |
||
| 41 | * <pre> |
||
| 42 | * pDst[n] = pSrc[n] << shift, 0 <= n < blockSize. |
||
| 43 | * </pre> |
||
| 44 | * |
||
| 45 | * If <code>shift</code> is positive then the elements of the vector are shifted to the left. |
||
| 46 | * If <code>shift</code> is negative then the elements of the vector are shifted to the right. |
||
| 47 | * |
||
| 48 | * The functions support in-place computation allowing the source and destination |
||
| 49 | * pointers to reference the same memory buffer. |
||
| 50 | */ |
||
| 51 | |||
| 52 | /** |
||
| 53 | * @addtogroup shift |
||
| 54 | * @{ |
||
| 55 | */ |
||
| 56 | |||
| 57 | /** |
||
| 58 | * @brief Shifts the elements of a Q31 vector a specified number of bits. |
||
| 59 | * @param[in] *pSrc points to the input vector |
||
| 60 | * @param[in] shiftBits number of bits to shift. A positive value shifts left; a negative value shifts right. |
||
| 61 | * @param[out] *pDst points to the output vector |
||
| 62 | * @param[in] blockSize number of samples in the vector |
||
| 63 | * @return none. |
||
| 64 | * |
||
| 65 | * |
||
| 66 | * <b>Scaling and Overflow Behavior:</b> |
||
| 67 | * \par |
||
| 68 | * The function uses saturating arithmetic. |
||
| 69 | * Results outside of the allowable Q31 range [0x80000000 0x7FFFFFFF] will be saturated. |
||
| 70 | */ |
||
| 71 | |||
| 72 | void arm_shift_q31( |
||
| 73 | q31_t * pSrc, |
||
| 74 | int8_t shiftBits, |
||
| 75 | q31_t * pDst, |
||
| 76 | uint32_t blockSize) |
||
| 77 | { |
||
| 78 | uint32_t blkCnt; /* loop counter */ |
||
| 79 | uint8_t sign = (shiftBits & 0x80); /* Sign of shiftBits */ |
||
| 80 | |||
| 81 | #if defined (ARM_MATH_DSP) |
||
| 82 | |||
| 83 | q31_t in1, in2, in3, in4; /* Temporary input variables */ |
||
| 84 | q31_t out1, out2, out3, out4; /* Temporary output variables */ |
||
| 85 | |||
| 86 | /*loop Unrolling */ |
||
| 87 | blkCnt = blockSize >> 2U; |
||
| 88 | |||
| 89 | |||
| 90 | if (sign == 0U) |
||
| 91 | { |
||
| 92 | /* First part of the processing with loop unrolling. Compute 4 outputs at a time. |
||
| 93 | ** a second loop below computes the remaining 1 to 3 samples. */ |
||
| 94 | while (blkCnt > 0U) |
||
| 95 | { |
||
| 96 | /* C = A << shiftBits */ |
||
| 97 | /* Shift the input and then store the results in the destination buffer. */ |
||
| 98 | in1 = *pSrc; |
||
| 99 | in2 = *(pSrc + 1); |
||
| 100 | out1 = in1 << shiftBits; |
||
| 101 | in3 = *(pSrc + 2); |
||
| 102 | out2 = in2 << shiftBits; |
||
| 103 | in4 = *(pSrc + 3); |
||
| 104 | if (in1 != (out1 >> shiftBits)) |
||
| 105 | out1 = 0x7FFFFFFF ^ (in1 >> 31); |
||
| 106 | |||
| 107 | if (in2 != (out2 >> shiftBits)) |
||
| 108 | out2 = 0x7FFFFFFF ^ (in2 >> 31); |
||
| 109 | |||
| 110 | *pDst = out1; |
||
| 111 | out3 = in3 << shiftBits; |
||
| 112 | *(pDst + 1) = out2; |
||
| 113 | out4 = in4 << shiftBits; |
||
| 114 | |||
| 115 | if (in3 != (out3 >> shiftBits)) |
||
| 116 | out3 = 0x7FFFFFFF ^ (in3 >> 31); |
||
| 117 | |||
| 118 | if (in4 != (out4 >> shiftBits)) |
||
| 119 | out4 = 0x7FFFFFFF ^ (in4 >> 31); |
||
| 120 | |||
| 121 | *(pDst + 2) = out3; |
||
| 122 | *(pDst + 3) = out4; |
||
| 123 | |||
| 124 | /* Update destination pointer to process next sampels */ |
||
| 125 | pSrc += 4U; |
||
| 126 | pDst += 4U; |
||
| 127 | |||
| 128 | /* Decrement the loop counter */ |
||
| 129 | blkCnt--; |
||
| 130 | } |
||
| 131 | } |
||
| 132 | else |
||
| 133 | { |
||
| 134 | |||
| 135 | /* First part of the processing with loop unrolling. Compute 4 outputs at a time. |
||
| 136 | ** a second loop below computes the remaining 1 to 3 samples. */ |
||
| 137 | while (blkCnt > 0U) |
||
| 138 | { |
||
| 139 | /* C = A >> shiftBits */ |
||
| 140 | /* Shift the input and then store the results in the destination buffer. */ |
||
| 141 | in1 = *pSrc; |
||
| 142 | in2 = *(pSrc + 1); |
||
| 143 | in3 = *(pSrc + 2); |
||
| 144 | in4 = *(pSrc + 3); |
||
| 145 | |||
| 146 | *pDst = (in1 >> -shiftBits); |
||
| 147 | *(pDst + 1) = (in2 >> -shiftBits); |
||
| 148 | *(pDst + 2) = (in3 >> -shiftBits); |
||
| 149 | *(pDst + 3) = (in4 >> -shiftBits); |
||
| 150 | |||
| 151 | |||
| 152 | pSrc += 4U; |
||
| 153 | pDst += 4U; |
||
| 154 | |||
| 155 | blkCnt--; |
||
| 156 | } |
||
| 157 | |||
| 158 | } |
||
| 159 | |||
| 160 | /* If the blockSize is not a multiple of 4, compute any remaining output samples here. |
||
| 161 | ** No loop unrolling is used. */ |
||
| 162 | blkCnt = blockSize % 0x4U; |
||
| 163 | |||
| 164 | #else |
||
| 165 | |||
| 166 | /* Run the below code for Cortex-M0 */ |
||
| 167 | |||
| 168 | |||
| 169 | /* Initialize blkCnt with number of samples */ |
||
| 170 | blkCnt = blockSize; |
||
| 171 | |||
| 172 | #endif /* #if defined (ARM_MATH_DSP) */ |
||
| 173 | |||
| 174 | |||
| 175 | while (blkCnt > 0U) |
||
| 176 | { |
||
| 177 | /* C = A (>> or <<) shiftBits */ |
||
| 178 | /* Shift the input and then store the result in the destination buffer. */ |
||
| 179 | *pDst++ = (sign == 0U) ? clip_q63_to_q31((q63_t) * pSrc++ << shiftBits) : |
||
| 180 | (*pSrc++ >> -shiftBits); |
||
| 181 | |||
| 182 | /* Decrement the loop counter */ |
||
| 183 | blkCnt--; |
||
| 184 | } |
||
| 185 | |||
| 186 | |||
| 187 | } |
||
| 188 | |||
| 189 | /** |
||
| 190 | * @} end of shift group |
||
| 191 | */ |