1 /* ----------------------------------------------------------------------
2 * Project: CMSIS DSP Library
3 * Title: arm_cmplx_mult_real_q15.c
4 * Description: Q15 complex by real multiplication
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.h"
30
31 /**
32 @ingroup groupCmplxMath
33 */
34
35 /**
36 @addtogroup CmplxByRealMult
37 @{
38 */
39
40 /**
41 @brief Q15 complex-by-real multiplication.
42 @param[in] pSrcCmplx points to complex input vector
43 @param[in] pSrcReal points to real input vector
44 @param[out] pCmplxDst points to complex output vector
45 @param[in] numSamples number of samples in each vector
46
47 @par Scaling and Overflow Behavior
48 The function uses saturating arithmetic.
49 Results outside of the allowable Q15 range [0x8000 0x7FFF] are saturated.
50 */
51 #if defined(ARM_MATH_MVEI) && !defined(ARM_MATH_AUTOVECTORIZE)
52
arm_cmplx_mult_real_q15(const q15_t * pSrcCmplx,const q15_t * pSrcReal,q15_t * pCmplxDst,uint32_t numSamples)53 void arm_cmplx_mult_real_q15(
54 const q15_t * pSrcCmplx,
55 const q15_t * pSrcReal,
56 q15_t * pCmplxDst,
57 uint32_t numSamples)
58 {
59 static const uint16_t stride_cmplx_x_real_16[8] = {
60 0, 0, 1, 1, 2, 2, 3, 3
61 };
62 q15x8_t rVec;
63 q15x8_t cmplxVec;
64 q15x8_t dstVec;
65 uint16x8_t strideVec;
66 uint32_t blockSizeC = numSamples * CMPLX_DIM; /* loop counters */
67 uint32_t blkCnt;
68 q15_t in;
69
70 /*
71 * stride vector for pairs of real generation
72 */
73 strideVec = vld1q(stride_cmplx_x_real_16);
74
75 blkCnt = blockSizeC >> 3;
76
77 while (blkCnt > 0U)
78 {
79 cmplxVec = vld1q(pSrcCmplx);
80 rVec = vldrhq_gather_shifted_offset_s16(pSrcReal, strideVec);
81 dstVec = vqdmulhq(cmplxVec, rVec);
82 vst1q(pCmplxDst, dstVec);
83
84 pSrcReal += 4;
85 pSrcCmplx += 8;
86 pCmplxDst += 8;
87 blkCnt --;
88 }
89
90 /* Tail */
91 blkCnt = (blockSizeC & 7) >> 1;
92 while (blkCnt > 0U)
93 {
94 /* C[2 * i ] = A[2 * i ] * B[i]. */
95 /* C[2 * i + 1] = A[2 * i + 1] * B[i]. */
96
97 in = *pSrcReal++;
98 /* store the result in the destination buffer. */
99 *pCmplxDst++ = (q15_t) __SSAT((((q31_t) *pSrcCmplx++ * in) >> 15), 16);
100 *pCmplxDst++ = (q15_t) __SSAT((((q31_t) *pSrcCmplx++ * in) >> 15), 16);
101
102 /* Decrement loop counter */
103 blkCnt--;
104 }
105 }
106 #else
arm_cmplx_mult_real_q15(const q15_t * pSrcCmplx,const q15_t * pSrcReal,q15_t * pCmplxDst,uint32_t numSamples)107 void arm_cmplx_mult_real_q15(
108 const q15_t * pSrcCmplx,
109 const q15_t * pSrcReal,
110 q15_t * pCmplxDst,
111 uint32_t numSamples)
112 {
113 uint32_t blkCnt; /* Loop counter */
114 q15_t in; /* Temporary variable */
115
116 #if defined (ARM_MATH_LOOPUNROLL)
117
118 #if defined (ARM_MATH_DSP)
119 q31_t inA1, inA2; /* Temporary variables to hold input data */
120 q31_t inB1; /* Temporary variables to hold input data */
121 q15_t out1, out2, out3, out4; /* Temporary variables to hold output data */
122 q31_t mul1, mul2, mul3, mul4; /* Temporary variables to hold intermediate data */
123 #endif
124
125 /* Loop unrolling: Compute 4 outputs at a time */
126 blkCnt = numSamples >> 2U;
127
128 while (blkCnt > 0U)
129 {
130 /* C[2 * i ] = A[2 * i ] * B[i]. */
131 /* C[2 * i + 1] = A[2 * i + 1] * B[i]. */
132
133 #if defined (ARM_MATH_DSP)
134 /* read 2 complex numbers both real and imaginary from complex input buffer */
135 inA1 = read_q15x2_ia (&pSrcCmplx);
136 inA2 = read_q15x2_ia (&pSrcCmplx);
137 /* read 2 real values at a time from real input buffer */
138 inB1 = read_q15x2_ia (&pSrcReal);
139
140 /* multiply complex number with real numbers */
141 #ifndef ARM_MATH_BIG_ENDIAN
142 mul1 = (q31_t) ((q15_t) (inA1) * (q15_t) (inB1));
143 mul2 = (q31_t) ((q15_t) (inA1 >> 16) * (q15_t) (inB1));
144 mul3 = (q31_t) ((q15_t) (inA2) * (q15_t) (inB1 >> 16));
145 mul4 = (q31_t) ((q15_t) (inA2 >> 16) * (q15_t) (inB1 >> 16));
146 #else
147 mul2 = (q31_t) ((q15_t) (inA1 >> 16) * (q15_t) (inB1 >> 16));
148 mul1 = (q31_t) ((q15_t) inA1 * (q15_t) (inB1 >> 16));
149 mul4 = (q31_t) ((q15_t) (inA2 >> 16) * (q15_t) inB1);
150 mul3 = (q31_t) ((q15_t) inA2 * (q15_t) inB1);
151 #endif /* #ifndef ARM_MATH_BIG_ENDIAN */
152
153 /* saturate the result */
154 out1 = (q15_t) __SSAT(mul1 >> 15U, 16);
155 out2 = (q15_t) __SSAT(mul2 >> 15U, 16);
156 out3 = (q15_t) __SSAT(mul3 >> 15U, 16);
157 out4 = (q15_t) __SSAT(mul4 >> 15U, 16);
158
159 /* pack real and imaginary outputs and store them to destination */
160 write_q15x2_ia (&pCmplxDst, __PKHBT(out1, out2, 16));
161 write_q15x2_ia (&pCmplxDst, __PKHBT(out3, out4, 16));
162
163 inA1 = read_q15x2_ia (&pSrcCmplx);
164 inA2 = read_q15x2_ia (&pSrcCmplx);
165 inB1 = read_q15x2_ia (&pSrcReal);
166
167 #ifndef ARM_MATH_BIG_ENDIAN
168 mul1 = (q31_t) ((q15_t) (inA1) * (q15_t) (inB1));
169 mul2 = (q31_t) ((q15_t) (inA1 >> 16) * (q15_t) (inB1));
170 mul3 = (q31_t) ((q15_t) (inA2) * (q15_t) (inB1 >> 16));
171 mul4 = (q31_t) ((q15_t) (inA2 >> 16) * (q15_t) (inB1 >> 16));
172 #else
173 mul2 = (q31_t) ((q15_t) (inA1 >> 16) * (q15_t) (inB1 >> 16));
174 mul1 = (q31_t) ((q15_t) inA1 * (q15_t) (inB1 >> 16));
175 mul4 = (q31_t) ((q15_t) (inA2 >> 16) * (q15_t) inB1);
176 mul3 = (q31_t) ((q15_t) inA2 * (q15_t) inB1);
177 #endif /* #ifndef ARM_MATH_BIG_ENDIAN */
178
179 out1 = (q15_t) __SSAT(mul1 >> 15U, 16);
180 out2 = (q15_t) __SSAT(mul2 >> 15U, 16);
181 out3 = (q15_t) __SSAT(mul3 >> 15U, 16);
182 out4 = (q15_t) __SSAT(mul4 >> 15U, 16);
183
184 write_q15x2_ia (&pCmplxDst, __PKHBT(out1, out2, 16));
185 write_q15x2_ia (&pCmplxDst, __PKHBT(out3, out4, 16));
186 #else
187 in = *pSrcReal++;
188 *pCmplxDst++ = (q15_t) __SSAT((((q31_t) *pSrcCmplx++ * in) >> 15), 16);
189 *pCmplxDst++ = (q15_t) __SSAT((((q31_t) *pSrcCmplx++ * in) >> 15), 16);
190
191 in = *pSrcReal++;
192 *pCmplxDst++ = (q15_t) __SSAT((((q31_t) *pSrcCmplx++ * in) >> 15), 16);
193 *pCmplxDst++ = (q15_t) __SSAT((((q31_t) *pSrcCmplx++ * in) >> 15), 16);
194
195 in = *pSrcReal++;
196 *pCmplxDst++ = (q15_t) __SSAT((((q31_t) *pSrcCmplx++ * in) >> 15), 16);
197 *pCmplxDst++ = (q15_t) __SSAT((((q31_t) *pSrcCmplx++ * in) >> 15), 16);
198
199 in = *pSrcReal++;
200 *pCmplxDst++ = (q15_t) __SSAT((((q31_t) *pSrcCmplx++ * in) >> 15), 16);
201 *pCmplxDst++ = (q15_t) __SSAT((((q31_t) *pSrcCmplx++ * in) >> 15), 16);
202 #endif
203
204 /* Decrement loop counter */
205 blkCnt--;
206 }
207
208 /* Loop unrolling: Compute remaining outputs */
209 blkCnt = numSamples % 0x4U;
210
211 #else
212
213 /* Initialize blkCnt with number of samples */
214 blkCnt = numSamples;
215
216 #endif /* #if defined (ARM_MATH_LOOPUNROLL) */
217
218 while (blkCnt > 0U)
219 {
220 /* C[2 * i ] = A[2 * i ] * B[i]. */
221 /* C[2 * i + 1] = A[2 * i + 1] * B[i]. */
222
223 in = *pSrcReal++;
224 /* store the result in the destination buffer. */
225 *pCmplxDst++ = (q15_t) __SSAT((((q31_t) *pSrcCmplx++ * in) >> 15), 16);
226 *pCmplxDst++ = (q15_t) __SSAT((((q31_t) *pSrcCmplx++ * in) >> 15), 16);
227
228 /* Decrement loop counter */
229 blkCnt--;
230 }
231
232 }
233 #endif /* defined(ARM_MATH_MVEI) */
234
235 /**
236 @} end of CmplxByRealMult group
237 */
238
