1 /*
2  * Copyright (C) 2010-2018 Arm Limited or its affiliates. All rights reserved.
3  *
4  * SPDX-License-Identifier: Apache-2.0
5  *
6  * Licensed under the Apache License, Version 2.0 (the License); you may
7  * not use this file except in compliance with the License.
8  * You may obtain a copy of the License at
9  *
10  * www.apache.org/licenses/LICENSE-2.0
11  *
12  * Unless required by applicable law or agreed to in writing, software
13  * distributed under the License is distributed on an AS IS BASIS, WITHOUT
14  * WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
15  * See the License for the specific language governing permissions and
16  * limitations under the License.
17  */
18 
19 /* ----------------------------------------------------------------------
20  * Project:      CMSIS NN Library
21  * Title:        arm_q7_to_q15_no_shift.c
22  * Description:  Converts the elements of the Q7 vector to Q15 vector without left-shift
23  *
24  * $Date:        17. January 2018
25  * $Revision:    V.1.0.0
26  *
27  * Target Processor:  Cortex-M cores
28  *
29  * -------------------------------------------------------------------- */
30 
31 #include "arm_nnsupportfunctions.h"
32 
33 /**
34  * @ingroup groupSupport
35  */
36 
37 /**
38  * @addtogroup nndata_convert
39  * @{
40  */
41 
42 /**
43  * @brief Converts the elements of the Q7 vector to Q15 vector without left-shift
44  * @param[in]       *pSrc points to the Q7 input vector
45  * @param[out]      *pDst points to the Q15 output vector
46  * @param[in]       blockSize length of the input vector
47  * @return none.
48  *
49  * \par Description:
50  *
51  * The equation used for the conversion process is:
52  *
53  * <pre>
54  * 	pDst[n] = (q15_t) pSrc[n];   0 <= n < blockSize.
55  * </pre>
56  *
57  */
58 
arm_q7_to_q15_no_shift(const q7_t * pSrc,q15_t * pDst,uint32_t blockSize)59 void arm_q7_to_q15_no_shift(const q7_t * pSrc, q15_t * pDst, uint32_t blockSize)
60 {
61     const q7_t *pIn = pSrc;     /* Src pointer */
62     uint32_t  blkCnt;           /* loop counter */
63 
64 #ifndef ARM_MATH_CM0_FAMILY
65     q31_t     in;
66     q31_t     in1, in2;
67     q31_t     out1, out2;
68 
69     /* Run the below code for Cortex-M4 and Cortex-M3 */
70 
71     /*loop Unrolling */
72     blkCnt = blockSize >> 2u;
73 
74     /* First part of the processing with loop unrolling.  Compute 4 outputs at a time.
75      ** a second loop below computes the remaining 1 to 3 samples. */
76     while (blkCnt > 0u)
77     {
78         /* C = (q15_t) A << 8 */
79         /* convert from q7 to q15 and then store the results in the destination buffer */
80         in = *__SIMD32(pIn)++;
81 
82         /* rotatate in by 8 and extend two q7_t values to q15_t values */
83         in1 = __SXTB16(__ROR(in, 8));
84 
85         /* extend remainig two q7_t values to q15_t values */
86         in2 = __SXTB16(in);
87 
88 #ifndef ARM_MATH_BIG_ENDIAN
89 
90         out2 = __PKHTB(in1, in2, 16);
91         out1 = __PKHBT(in2, in1, 16);
92 
93 #else
94 
95         out1 = __PKHTB(in1, in2, 16);
96         out2 = __PKHBT(in2, in1, 16);
97 
98 #endif
99 
100         *__SIMD32(pDst)++ = out1;
101         *__SIMD32(pDst)++ = out2;
102 
103         /* Decrement the loop counter */
104         blkCnt--;
105     }
106 
107     /* If the blockSize is not a multiple of 4, compute any remaining output samples here.
108      ** No loop unrolling is used. */
109     blkCnt = blockSize % 0x4u;
110 
111 #else
112 
113     /* Run the below code for Cortex-M0 */
114 
115     /* Loop over blockSize number of values */
116     blkCnt = blockSize;
117 
118 #endif                          /* #ifndef ARM_MATH_CM0_FAMILY */
119 
120     while (blkCnt > 0u)
121     {
122         /* C = (q15_t) A << 8 */
123         /* convert from q7 to q15 and then store the results in the destination buffer */
124         *pDst++ = (q15_t) * pIn++;
125 
126         /* Decrement the loop counter */
127         blkCnt--;
128     }
129 
130 }
131 
132 /**
133  * @} end of nndata_convert group
134  */
135