arm_q15_to_float.c
1 /* ---------------------------------------------------------------------- 2 * Project: CMSIS DSP Library 3 * Title: arm_q15_to_float.c 4 * Description: Converts the elements of the Q15 vector to floating-point vector 5 * 6 * $Date: 23 April 2021 7 * $Revision: V1.9.0 8 * 9 * Target Processor: Cortex-M and Cortex-A cores 10 * -------------------------------------------------------------------- */ 11 /* 12 * Copyright (C) 2010-2021 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 "dsp/support_functions.h" 30 31 /** 32 @ingroup groupSupport 33 */ 34 35 /** 36 * @defgroup q15_to_x Convert 16-bit fixed point value 37 */ 38 39 /** 40 @addtogroup q15_to_x 41 @{ 42 */ 43 44 /** 45 @brief Converts the elements of the Q15 vector to floating-point vector. 46 @param[in] pSrc points to the Q15 input vector 47 @param[out] pDst points to the floating-point output vector 48 @param[in] blockSize number of samples in each vector 49 @return none 50 51 @par Details 52 The equation used for the conversion process is: 53 <pre> 54 pDst[n] = (float32_t) pSrc[n] / 32768; 0 <= n < blockSize. 55 </pre> 56 */ 57 58 #if defined(ARM_MATH_MVEF) && !defined(ARM_MATH_AUTOVECTORIZE) 59 void arm_q15_to_float( 60 const q15_t * pSrc, 61 float32_t * pDst, 62 uint32_t blockSize) 63 { 64 uint32_t blkCnt; 65 66 q15x8_t vecDst; 67 q15_t const *pSrcVec; 68 69 pSrcVec = (q15_t const *) pSrc; 70 blkCnt = blockSize >> 2; 71 while (blkCnt > 0U) 72 { 73 /* C = (float32_t) A / 32768 */ 74 /* convert from q15 to float and then store the results in the destination buffer */ 75 vecDst = vldrhq_s32(pSrcVec); 76 pSrcVec += 4; 77 vstrwq(pDst, vcvtq_n_f32_s32((int32x4_t)vecDst, 15)); 78 pDst += 4; 79 /* 80 * Decrement the blockSize loop counter 81 */ 82 blkCnt--; 83 } 84 85 blkCnt = blockSize & 3; 86 while (blkCnt > 0U) 87 { 88 /* C = (float32_t) A / 32768 */ 89 90 /* Convert from q15 to float and store result in destination buffer */ 91 *pDst++ = ((float32_t) *pSrcVec++ / 32768.0f); 92 93 /* Decrement loop counter */ 94 blkCnt--; 95 } 96 } 97 #else 98 #if defined(ARM_MATH_NEON_EXPERIMENTAL) 99 void arm_q15_to_float( 100 const q15_t * pSrc, 101 float32_t * pDst, 102 uint32_t blockSize) 103 { 104 const q15_t *pIn = pSrc; /* Src pointer */ 105 uint32_t blkCnt; /* loop counter */ 106 107 int16x8_t inV; 108 int32x4_t inV0, inV1; 109 float32x4_t outV; 110 111 blkCnt = blockSize >> 3U; 112 113 /* Compute 8 outputs at a time. 114 ** a second loop below computes the remaining 1 to 7 samples. */ 115 while (blkCnt > 0U) 116 { 117 /* C = (float32_t) A / 32768 */ 118 /* convert from q15 to float and then store the results in the destination buffer */ 119 inV = vld1q_s16(pIn); 120 pIn += 8; 121 122 inV0 = vmovl_s16(vget_low_s16(inV)); 123 inV1 = vmovl_s16(vget_high_s16(inV)); 124 125 outV = vcvtq_n_f32_s32(inV0,15); 126 vst1q_f32(pDst, outV); 127 pDst += 4; 128 129 outV = vcvtq_n_f32_s32(inV1,15); 130 vst1q_f32(pDst, outV); 131 pDst += 4; 132 133 /* Decrement the loop counter */ 134 blkCnt--; 135 } 136 137 /* If the blockSize is not a multiple of 8, compute any remaining output samples here. 138 ** No loop unrolling is used. */ 139 blkCnt = blockSize & 7; 140 141 142 while (blkCnt > 0U) 143 { 144 /* C = (float32_t) A / 32768 */ 145 /* convert from q15 to float and then store the results in the destination buffer */ 146 *pDst++ = ((float32_t) * pIn++ / 32768.0f); 147 148 /* Decrement the loop counter */ 149 blkCnt--; 150 } 151 } 152 #else 153 void arm_q15_to_float( 154 const q15_t * pSrc, 155 float32_t * pDst, 156 uint32_t blockSize) 157 { 158 uint32_t blkCnt; /* Loop counter */ 159 const q15_t *pIn = pSrc; /* Source pointer */ 160 161 #if defined (ARM_MATH_LOOPUNROLL) 162 163 /* Loop unrolling: Compute 4 outputs at a time */ 164 blkCnt = blockSize >> 2U; 165 166 while (blkCnt > 0U) 167 { 168 /* C = (float32_t) A / 32768 */ 169 170 /* Convert from q15 to float and store result in destination buffer */ 171 *pDst++ = ((float32_t) * pIn++ / 32768.0f); 172 *pDst++ = ((float32_t) * pIn++ / 32768.0f); 173 *pDst++ = ((float32_t) * pIn++ / 32768.0f); 174 *pDst++ = ((float32_t) * pIn++ / 32768.0f); 175 176 /* Decrement loop counter */ 177 blkCnt--; 178 } 179 180 /* Loop unrolling: Compute remaining outputs */ 181 blkCnt = blockSize % 0x4U; 182 183 #else 184 185 /* Initialize blkCnt with number of samples */ 186 blkCnt = blockSize; 187 188 #endif /* #if defined (ARM_MATH_LOOPUNROLL) */ 189 190 while (blkCnt > 0U) 191 { 192 /* C = (float32_t) A / 32768 */ 193 194 /* Convert from q15 to float and store result in destination buffer */ 195 *pDst++ = ((float32_t) *pIn++ / 32768.0f); 196 197 /* Decrement loop counter */ 198 blkCnt--; 199 } 200 201 } 202 #endif /* #if defined(ARM_MATH_NEON) */ 203 #endif /* defined(ARM_MATH_MVEF) && !defined(ARM_MATH_AUTOVECTORIZE) */ 204 205 /** 206 @} end of q15_to_x group 207 */