• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
1 /* ----------------------------------------------------------------------
2  * Project:      CMSIS DSP Library
3  * Title:        arm_cmplx_conj_f16.c
4  * Description:  Floating-point complex conjugate
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/complex_math_functions_f16.h"
30 
31 #if defined(ARM_FLOAT16_SUPPORTED)
32 /**
33   @ingroup groupCmplxMath
34  */
35 
36 
37 /**
38   @addtogroup cmplx_conj
39   @{
40  */
41 
42 /**
43   @brief         Floating-point complex conjugate.
44   @param[in]     pSrc        points to the input vector
45   @param[out]    pDst        points to the output vector
46   @param[in]     numSamples  number of samples in each vector
47   @return        none
48  */
49 
50 #if defined(ARM_MATH_MVE_FLOAT16) && !defined(ARM_MATH_AUTOVECTORIZE)
51 
arm_cmplx_conj_f16(const float16_t * pSrc,float16_t * pDst,uint32_t numSamples)52 void arm_cmplx_conj_f16(
53     const float16_t * pSrc,
54     float16_t * pDst,
55     uint32_t numSamples)
56 {
57     static const float16_t cmplx_conj_sign[8] = { 1.0f, -1.0f, 1.0f, -1.0f, 1.0f, -1.0f, 1.0f, -1.0f };
58     uint32_t blockSize = numSamples * CMPLX_DIM;   /* loop counters */
59     uint32_t blkCnt;
60     f16x8_t vecSrc;
61     f16x8_t vecSign;
62 
63     /*
64      * load sign vector
65      */
66     vecSign = *(f16x8_t *) cmplx_conj_sign;
67 
68     /* Compute 4 real samples at a time */
69     blkCnt = blockSize >> 3U;
70 
71     while (blkCnt > 0U)
72     {
73         vecSrc = vld1q(pSrc);
74         vst1q(pDst,vmulq(vecSrc, vecSign));
75         /*
76          * Decrement the blkCnt loop counter
77          * Advance vector source and destination pointers
78          */
79         pSrc += 8;
80         pDst += 8;
81         blkCnt--;
82     }
83 
84      /* Tail */
85     blkCnt = (blockSize & 0x7) >> 1;
86 
87     while (blkCnt > 0U)
88     {
89       /* C[0] + jC[1] = A[0]+ j(-1)A[1] */
90 
91       /* Calculate Complex Conjugate and store result in destination buffer. */
92       *pDst++ =  *pSrc++;
93       *pDst++ = -(_Float16)*pSrc++;
94 
95       /* Decrement loop counter */
96       blkCnt--;
97     }
98 
99 }
100 
101 #else
arm_cmplx_conj_f16(const float16_t * pSrc,float16_t * pDst,uint32_t numSamples)102 void arm_cmplx_conj_f16(
103   const float16_t * pSrc,
104         float16_t * pDst,
105         uint32_t numSamples)
106 {
107         uint32_t blkCnt;                               /* Loop counter */
108 
109 #if defined (ARM_MATH_LOOPUNROLL) && !defined(ARM_MATH_AUTOVECTORIZE)
110 
111   /* Loop unrolling: Compute 4 outputs at a time */
112   blkCnt = numSamples >> 2U;
113 
114   while (blkCnt > 0U)
115   {
116     /* C[0] + jC[1] = A[0]+ j(-1)A[1] */
117 
118     /* Calculate Complex Conjugate and store result in destination buffer. */
119     *pDst++ =  *pSrc++;
120     *pDst++ = -(_Float16)*pSrc++;
121 
122     *pDst++ =  *pSrc++;
123     *pDst++ = -(_Float16)*pSrc++;
124 
125     *pDst++ =  *pSrc++;
126     *pDst++ = -(_Float16)*pSrc++;
127 
128     *pDst++ =  *pSrc++;
129     *pDst++ = -(_Float16)*pSrc++;
130 
131     /* Decrement loop counter */
132     blkCnt--;
133   }
134 
135   /* Loop unrolling: Compute remaining outputs */
136   blkCnt = numSamples % 0x4U;
137 
138 #else
139 
140   /* Initialize blkCnt with number of samples */
141   blkCnt = numSamples;
142 
143 #endif /* #if defined (ARM_MATH_LOOPUNROLL) */
144 
145   while (blkCnt > 0U)
146   {
147     /* C[0] + jC[1] = A[0]+ j(-1)A[1] */
148 
149     /* Calculate Complex Conjugate and store result in destination buffer. */
150     *pDst++ =  *pSrc++;
151     *pDst++ = -(_Float16)*pSrc++;
152 
153     /* Decrement loop counter */
154     blkCnt--;
155   }
156 
157 }
158 #endif /* defined(ARM_MATH_MVEF) && !defined(ARM_MATH_AUTOVECTORIZE) */
159 
160 /**
161   @} end of cmplx_conj group
162  */
163 #endif /* #if defined(ARM_FLOAT16_SUPPORTED) */
164