1 /* ----------------------------------------------------------------------
2 * Project: CMSIS DSP Library
3 * Title: arm_q7_to_float.c
4 * Description: Converts the elements of the Q7 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 q7_to_x Convert 8-bit fixed point value
37 */
38
39 /**
40 @addtogroup q7_to_x
41 @{
42 */
43
44 /**
45 @brief Converts the elements of the Q7 vector to floating-point vector.
46 @param[in] pSrc points to the Q7 input vector
47 @param[out] pDst points to the floating-point output vector
48 @param[in] blockSize number of samples in each vector
49
50 @par Details
51 The equation used for the conversion process is:
52 <pre>
53 pDst[n] = (float32_t) pSrc[n] / 128; 0 <= n < blockSize.
54 </pre>
55 */
56 #if defined(ARM_MATH_MVEF) && !defined(ARM_MATH_AUTOVECTORIZE)
arm_q7_to_float(const q7_t * pSrc,float32_t * pDst,uint32_t blockSize)57 ARM_DSP_ATTRIBUTE void arm_q7_to_float(
58 const q7_t * pSrc,
59 float32_t * pDst,
60 uint32_t blockSize)
61 {
62 uint32_t blkCnt; /* loop counters */
63 q7x16_t vecDst;
64 q7_t const *pSrcVec;
65
66 pSrcVec = (q7_t const *) pSrc;
67 blkCnt = blockSize >> 2;
68 while (blkCnt > 0U)
69 {
70 /* C = (float32_t) A / 32768 */
71 /* convert from q7 to float and then store the results in the destination buffer */
72 vecDst = vldrbq_s32(pSrcVec);
73 pSrcVec += 4;
74 vstrwq(pDst, vcvtq_n_f32_s32((int32x4_t)vecDst, 7));
75 pDst += 4;
76 /*
77 * Decrement the blockSize loop counter
78 */
79 blkCnt--;
80 }
81
82 blkCnt = blockSize & 3;
83 while (blkCnt > 0U)
84 {
85 /* C = (float32_t) A / 128 */
86
87 /* Convert from q7 to float and store result in destination buffer */
88 *pDst++ = ((float32_t) * pSrcVec++ / 128.0f);
89
90 /* Decrement loop counter */
91 blkCnt--;
92 }
93 }
94 #else
95 #if defined(ARM_MATH_NEON)
arm_q7_to_float(const q7_t * pSrc,float32_t * pDst,uint32_t blockSize)96 ARM_DSP_ATTRIBUTE void arm_q7_to_float(
97 const q7_t * pSrc,
98 float32_t * pDst,
99 uint32_t blockSize)
100 {
101 const q7_t *pIn = pSrc; /* Src pointer */
102 uint32_t blkCnt; /* loop counter */
103
104 int8x16_t inV;
105 int16x8_t inVLO, inVHI;
106 int32x4_t inVLL, inVLH, inVHL, inVHH;
107 float32x4_t outV;
108
109 blkCnt = blockSize >> 4U;
110
111 /* Compute 16 outputs at a time.
112 ** a second loop below computes the remaining 1 to 15 samples. */
113 while (blkCnt > 0U)
114 {
115 /* C = (float32_t) A / 128 */
116 /* Convert from q7 to float and then store the results in the destination buffer */
117 inV = vld1q_s8(pIn);
118 pIn += 16;
119
120 inVLO = vmovl_s8(vget_low_s8(inV));
121 inVHI = vmovl_s8(vget_high_s8(inV));
122
123 inVLL = vmovl_s16(vget_low_s16(inVLO));
124 inVLH = vmovl_s16(vget_high_s16(inVLO));
125 inVHL = vmovl_s16(vget_low_s16(inVHI));
126 inVHH = vmovl_s16(vget_high_s16(inVHI));
127
128 outV = vcvtq_n_f32_s32(inVLL,7);
129 vst1q_f32(pDst, outV);
130 pDst += 4;
131
132 outV = vcvtq_n_f32_s32(inVLH,7);
133 vst1q_f32(pDst, outV);
134 pDst += 4;
135
136 outV = vcvtq_n_f32_s32(inVHL,7);
137 vst1q_f32(pDst, outV);
138 pDst += 4;
139
140 outV = vcvtq_n_f32_s32(inVHH,7);
141 vst1q_f32(pDst, outV);
142 pDst += 4;
143
144 /* Decrement the loop counter */
145 blkCnt--;
146 }
147
148 /* If the blockSize is not a multiple of 16, compute any remaining output samples here.
149 ** No loop unrolling is used. */
150 blkCnt = blockSize & 0xF;
151
152 while (blkCnt > 0U)
153 {
154 /* C = (float32_t) A / 128 */
155 /* Convert from q7 to float and then store the results in the destination buffer */
156 *pDst++ = ((float32_t) * pIn++ / 128.0f);
157
158 /* Decrement the loop counter */
159 blkCnt--;
160 }
161 }
162 #else
arm_q7_to_float(const q7_t * pSrc,float32_t * pDst,uint32_t blockSize)163 ARM_DSP_ATTRIBUTE void arm_q7_to_float(
164 const q7_t * pSrc,
165 float32_t * pDst,
166 uint32_t blockSize)
167 {
168 uint32_t blkCnt; /* Loop counter */
169 const q7_t *pIn = pSrc; /* Source pointer */
170
171 #if defined (ARM_MATH_LOOPUNROLL)
172
173 /* Loop unrolling: Compute 4 outputs at a time */
174 blkCnt = blockSize >> 2U;
175
176 while (blkCnt > 0U)
177 {
178 /* C = (float32_t) A / 128 */
179
180 /* Convert from q7 to float and store result in destination buffer */
181 *pDst++ = ((float32_t) * pIn++ / 128.0f);
182 *pDst++ = ((float32_t) * pIn++ / 128.0f);
183 *pDst++ = ((float32_t) * pIn++ / 128.0f);
184 *pDst++ = ((float32_t) * pIn++ / 128.0f);
185
186 /* Decrement loop counter */
187 blkCnt--;
188 }
189
190 /* Loop unrolling: Compute remaining outputs */
191 blkCnt = blockSize % 0x4U;
192
193 #else
194
195 /* Initialize blkCnt with number of samples */
196 blkCnt = blockSize;
197
198 #endif /* #if defined (ARM_MATH_LOOPUNROLL) */
199
200 while (blkCnt > 0U)
201 {
202 /* C = (float32_t) A / 128 */
203
204 /* Convert from q7 to float and store result in destination buffer */
205 *pDst++ = ((float32_t) * pIn++ / 128.0f);
206
207 /* Decrement loop counter */
208 blkCnt--;
209 }
210
211 }
212 #endif /* #if defined(ARM_MATH_NEON) */
213 #endif /* defined(ARM_MATH_MVEF) && !defined(ARM_MATH_AUTOVECTORIZE) */
214
215 /**
216 @} end of q7_to_x group
217 */
218
