1 /* ----------------------------------------------------------------------
2 * Project: CMSIS DSP Library
3 * Title: arm_float_to_q15.c
4 * Description: Converts the elements of the floating-point vector to Q15 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_f16.h"
30
31 #if defined(ARM_FLOAT16_SUPPORTED)
32
33
34 /**
35 @ingroup groupSupport
36 */
37
38 /**
39 @addtogroup float_to_x
40 @{
41 */
42
43 /**
44 @brief Converts the elements of the floating-point vector to f16 vector.
45 @param[in] pSrc points to the f32 input vector
46 @param[out] pDst points to the f16 output vector
47 @param[in] blockSize number of samples in each vector
48 @return none
49
50 */
51
52 #if defined(ARM_MATH_MVE_FLOAT16) && !defined(ARM_MATH_AUTOVECTORIZE) && defined(__CMSIS_GCC_H)
53 #pragma message "Scalar version of arm_float_to_f16 built. Helium version has build issues with gcc."
54 #endif
55
56 #if defined(ARM_MATH_MVE_FLOAT16) && !defined(ARM_MATH_AUTOVECTORIZE) && !defined(__CMSIS_GCC_H)
57
arm_float_to_f16(const float32_t * pSrc,float16_t * pDst,uint32_t blockSize)58 void arm_float_to_f16(
59 const float32_t * pSrc,
60 float16_t * pDst,
61 uint32_t blockSize)
62 {
63 int32_t blkCnt; /* loop counters */
64 float32x4x2_t tmp;
65 float16x8_t vecDst;
66 float32_t const *pSrcVec;
67
68
69 pSrcVec = (float32_t const *) pSrc;
70 blkCnt = blockSize >> 3;
71 while (blkCnt > 0)
72 {
73 /* convert from float32 to float16 and then store the results in the destination buffer */
74 tmp = vld2q(pSrcVec); pSrcVec += 8;
75 /* narrow / merge */
76 vecDst = vcvtbq_f16_f32(vecDst, tmp.val[0]);
77 vecDst = vcvttq_f16_f32(vecDst, tmp.val[1]);
78 vst1q(pDst, vecDst); pDst += 8;
79 /*
80 * Decrement the blockSize loop counter
81 */
82 blkCnt--;
83 }
84
85 /*
86 * tail
87 */
88 blkCnt = blockSize & 7;
89 if (blkCnt > 0)
90 {
91 mve_pred16_t p0 = vctp16q(blkCnt);
92 tmp = vld2q(pSrcVec);
93 vecDst = vcvtbq_f16_f32(vecDst, tmp.val[0]);
94 vecDst = vcvttq_f16_f32(vecDst, tmp.val[1]);
95 vstrhq_p(pDst, vecDst, p0);
96 }
97 }
98
99 #else
100
arm_float_to_f16(const float32_t * pSrc,float16_t * pDst,uint32_t blockSize)101 void arm_float_to_f16(
102 const float32_t * pSrc,
103 float16_t * pDst,
104 uint32_t blockSize)
105 {
106 const float32_t *pIn = pSrc; /* Src pointer */
107 uint32_t blkCnt; /* loop counter */
108
109 /*
110 * Loop over blockSize number of values
111 */
112 blkCnt = blockSize;
113
114 while (blkCnt > 0U)
115 {
116
117 *pDst++ = (float16_t) * pIn++;
118 /*
119 * Decrement the loop counter
120 */
121 blkCnt--;
122 }
123 }
124 #endif /* defined(ARM_MATH_MVEF) && !defined(ARM_MATH_AUTOVECTORIZE) */
125
126 /**
127 @} end of float_to_x group
128 */
129
130 #endif /* #if defined(ARM_FLOAT16_SUPPORTED) */
131
132
