1 /* ----------------------------------------------------------------------
2  * Project:      CMSIS DSP Library
3  * Title:        arm_q15_to_q31.c
4  * Description:  Converts the elements of the Q15 vector to Q31 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   @addtogroup q15_to_x
37   @{
38  */
39 
40 /**
41   @brief         Converts the elements of the Q15 vector to Q31 vector.
42   @param[in]     pSrc       points to the Q15 input vector
43   @param[out]    pDst       points to the Q31 output vector
44   @param[in]     blockSize  number of samples in each vector
45 
46   @par           Details
47                    The equation used for the conversion process is:
48   <pre>
49       pDst[n] = (q31_t) pSrc[n] << 16;   0 <= n < blockSize.
50   </pre>
51  */
52 #if defined(ARM_MATH_MVEI) && !defined(ARM_MATH_AUTOVECTORIZE)
arm_q15_to_q31(const q15_t * pSrc,q31_t * pDst,uint32_t blockSize)53 ARM_DSP_ATTRIBUTE void arm_q15_to_q31(
54   const q15_t * pSrc,
55         q31_t * pDst,
56         uint32_t blockSize)
57 {
58 
59   uint32_t blkCnt;
60 
61   q31x4_t         vecDst;
62 
63   blkCnt = blockSize>> 2;
64   while (blkCnt > 0U)
65   {
66 
67         /* C = (q31_t)A << 16 */
68         /* convert from q15 to q31 and then store the results in the destination buffer */
69         /* load q15 + 32-bit widening */
70         vecDst = vldrhq_s32((q15_t const *) pSrc);
71         vecDst = vshlq_n(vecDst, 16);
72         vstrwq_s32(pDst, vecDst);
73 
74         /*
75          * Decrement the blockSize loop counter
76          * Advance vector source and destination pointers
77          */
78         pDst += 4;
79         pSrc += 4;
80         blkCnt --;
81     }
82 
83   blkCnt = blockSize & 3;
84   while (blkCnt > 0U)
85   {
86     /* C = (q31_t) A << 16 */
87 
88     /* Convert from q15 to q31 and store result in destination buffer */
89     *pDst++ = (q31_t) *pSrc++ << 16;
90 
91     /* Decrement loop counter */
92     blkCnt--;
93   }
94 }
95 #else
arm_q15_to_q31(const q15_t * pSrc,q31_t * pDst,uint32_t blockSize)96 ARM_DSP_ATTRIBUTE void arm_q15_to_q31(
97   const q15_t * pSrc,
98         q31_t * pDst,
99         uint32_t blockSize)
100 {
101         uint32_t blkCnt;                               /* Loop counter */
102   const q15_t *pIn = pSrc;                             /* Source pointer */
103 
104 #if defined (ARM_MATH_LOOPUNROLL)
105         q31_t in1, in2;
106         q31_t out1, out2, out3, out4;
107 #endif
108 
109 #if defined (ARM_MATH_LOOPUNROLL)
110 
111   /* Loop unrolling: Compute 4 outputs at a time */
112   blkCnt = blockSize >> 2U;
113 
114   while (blkCnt > 0U)
115   {
116     /* C = (q31_t)A << 16 */
117 
118     /* Convert from q15 to q31 and store result in destination buffer */
119     in1 = read_q15x2_ia (&pIn);
120     in2 = read_q15x2_ia (&pIn);
121 
122 #ifndef ARM_MATH_BIG_ENDIAN
123 
124     /* extract lower 16 bits to 32 bit result */
125     out1 = in1 << 16U;
126     /* extract upper 16 bits to 32 bit result */
127     out2 = in1 & 0xFFFF0000;
128     /* extract lower 16 bits to 32 bit result */
129     out3 = in2 << 16U;
130     /* extract upper 16 bits to 32 bit result */
131     out4 = in2 & 0xFFFF0000;
132 
133 #else
134 
135     /* extract upper 16 bits to 32 bit result */
136     out1 = in1 & 0xFFFF0000;
137     /* extract lower 16 bits to 32 bit result */
138     out2 = in1 << 16U;
139     /* extract upper 16 bits to 32 bit result */
140     out3 = in2 & 0xFFFF0000;
141     /* extract lower 16 bits to 32 bit result */
142     out4 = in2 << 16U;
143 
144 #endif /* #ifndef ARM_MATH_BIG_ENDIAN */
145 
146     *pDst++ = out1;
147     *pDst++ = out2;
148     *pDst++ = out3;
149     *pDst++ = out4;
150 
151     /* Decrement loop counter */
152     blkCnt--;
153   }
154 
155   /* Loop unrolling: Compute remaining outputs */
156   blkCnt = blockSize % 0x4U;
157 
158 #else
159 
160   /* Initialize blkCnt with number of samples */
161   blkCnt = blockSize;
162 
163 #endif /* #if defined (ARM_MATH_LOOPUNROLL) */
164 
165   while (blkCnt > 0U)
166   {
167     /* C = (q31_t) A << 16 */
168 
169     /* Convert from q15 to q31 and store result in destination buffer */
170     *pDst++ = (q31_t) *pIn++ << 16;
171 
172     /* Decrement loop counter */
173     blkCnt--;
174   }
175 
176 }
177 #endif /* defined(ARM_MATH_MVEI) */
178 
179 /**
180   @} end of q15_to_x group
181  */
182