1 /* ----------------------------------------------------------------------
2 * Project: CMSIS DSP Library
3 * Title: arm_mat_trans_f64.c
4 * Description: Floating-point matrix transpose
5 *
6 * $Date: 10 August 2022
7 * $Revision: V1.9.1
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/matrix_functions.h"
30
31 /**
32 @ingroup groupMatrix
33 */
34
35 /**
36 @defgroup MatrixTrans Matrix Transpose
37
38 Tranposes a matrix.
39
40 Transposing an <code>M x N</code> matrix flips it around the center diagonal and results in an <code>N x M</code> matrix.
41 \image html MatrixTranspose.gif "Transpose of a 3 x 3 matrix"
42 */
43
44 /**
45 @addtogroup MatrixTrans
46 @{
47 */
48
49 /**
50 @brief Floating-point matrix transpose.
51 @param[in] pSrc points to input matrix
52 @param[out] pDst points to output matrix
53 @return execution status
54 - \ref ARM_MATH_SUCCESS : Operation successful
55 - \ref ARM_MATH_SIZE_MISMATCH : Matrix size check failed
56 */
57 #if defined(ARM_MATH_NEON) && defined(__aarch64__)
58
arm_mat_trans_f64(const arm_matrix_instance_f64 * pSrc,arm_matrix_instance_f64 * pDst)59 ARM_DSP_ATTRIBUTE arm_status arm_mat_trans_f64(
60 const arm_matrix_instance_f64 * pSrc,
61 arm_matrix_instance_f64 * pDst)
62 {
63 float64_t *pIn = pSrc->pData; /* input data matrix pointer */
64 float64_t *pOut = pDst->pData; /* output data matrix pointer */
65 float64_t *px; /* Temporary output data matrix pointer */
66 uint16_t nRows = pSrc->numRows; /* number of rows */
67 uint16_t nColumns = pSrc->numCols; /* number of columns */
68
69 uint16_t blkCnt, rowCnt, i = 0U, row = nRows; /* loop counters */
70 arm_status status; /* status of matrix transpose */
71
72 #ifdef ARM_MATH_MATRIX_CHECK
73
74 /* Check for matrix mismatch condition */
75 if ((pSrc->numRows != pDst->numCols) || (pSrc->numCols != pDst->numRows))
76 {
77 /* Set status as ARM_MATH_SIZE_MISMATCH */
78 status = ARM_MATH_SIZE_MISMATCH;
79 }
80 else
81 #endif /* #ifdef ARM_MATH_MATRIX_CHECK */
82
83 {
84 /* Matrix transpose by exchanging the rows with columns */
85 /* Row loop */
86 rowCnt = row >> 1;
87 while (rowCnt > 0U)
88 {
89 float64_t *row0,*row1;
90 float64x2x4_t raV;
91
92 blkCnt = nColumns >> 2;
93
94 /* The pointer px is set to starting address of the column being processed */
95 px = pOut + i;
96
97 /* Compute 4 outputs at a time.
98 ** a second loop below computes the remaining 1 to 3 samples. */
99 while (blkCnt > 0U) /* Column loop */
100 {
101 row0 = pIn;
102 row1 = pIn+nColumns;
103 pIn+=4;
104 raV = vld4q_lane_f64(row0, raV, 0);
105 raV = vld4q_lane_f64(row1, raV, 1);
106
107 vst1q_f64(px,raV.val[0]);
108 px += nRows;
109
110 vst1q_f64(px,raV.val[1]);
111 px += nRows;
112
113 vst1q_f64(px,raV.val[2]);
114 px += nRows;
115
116 vst1q_f64(px,raV.val[3]);
117 px += nRows;
118
119 /* Decrement the column loop counter */
120 blkCnt--;
121 }
122
123 /* Perform matrix transpose for last 3 samples here. */
124 blkCnt = nColumns % 0x4U;
125
126 while (blkCnt > 0U)
127 {
128 /* Read and store the input element in the destination */
129 *px++ = *pIn;
130 *px++ = *(pIn + 1 * nColumns);
131
132 px += (nRows - 2);
133 pIn++;
134
135 /* Decrement the column loop counter */
136 blkCnt--;
137 }
138
139 i += 2;
140 pIn += 1 * nColumns;
141
142 /* Decrement the row loop counter */
143 rowCnt--;
144
145 } /* Row loop end */
146
147 rowCnt = row & 1;
148 while (rowCnt > 0U)
149 {
150 blkCnt = nColumns ;
151 /* The pointer px is set to starting address of the column being processed */
152 px = pOut + i;
153
154 while (blkCnt > 0U)
155 {
156 /* Read and store the input element in the destination */
157 *px = *pIn++;
158
159 /* Update the pointer px to point to the next row of the transposed matrix */
160 px += nRows;
161
162 /* Decrement the column loop counter */
163 blkCnt--;
164 }
165 i++;
166 rowCnt -- ;
167 }
168
169 /* Set status as ARM_MATH_SUCCESS */
170 status = ARM_MATH_SUCCESS;
171 }
172
173 /* Return to application */
174 return (status);
175 }
176 #else
arm_mat_trans_f64(const arm_matrix_instance_f64 * pSrc,arm_matrix_instance_f64 * pDst)177 ARM_DSP_ATTRIBUTE arm_status arm_mat_trans_f64(
178 const arm_matrix_instance_f64 * pSrc,
179 arm_matrix_instance_f64 * pDst)
180 {
181 float64_t *pIn = pSrc->pData; /* input data matrix pointer */
182 float64_t *pOut = pDst->pData; /* output data matrix pointer */
183 float64_t *px; /* Temporary output data matrix pointer */
184 uint16_t nRows = pSrc->numRows; /* number of rows */
185 uint16_t nCols = pSrc->numCols; /* number of columns */
186 uint64_t col, row = nRows, i = 0U; /* Loop counters */
187 arm_status status; /* status of matrix transpose */
188
189 #ifdef ARM_MATH_MATRIX_CHECK
190
191 /* Check for matrix mismatch condition */
192 if ((pSrc->numRows != pDst->numCols) ||
193 (pSrc->numCols != pDst->numRows) )
194 {
195 /* Set status as ARM_MATH_SIZE_MISMATCH */
196 status = ARM_MATH_SIZE_MISMATCH;
197 }
198 else
199
200 #endif /* #ifdef ARM_MATH_MATRIX_CHECK */
201
202 {
203 /* Matrix transpose by exchanging the rows with columns */
204 /* row loop */
205 do
206 {
207 /* Pointer px is set to starting address of column being processed */
208 px = pOut + i;
209
210 #if defined (ARM_MATH_LOOPUNROLL)
211
212 /* Loop unrolling: Compute 4 outputs at a time */
213 col = nCols >> 2U;
214
215 while (col > 0U) /* column loop */
216 {
217 /* Read and store input element in destination */
218 *px = *pIn++;
219 /* Update pointer px to point to next row of transposed matrix */
220 px += nRows;
221
222 *px = *pIn++;
223 px += nRows;
224
225 *px = *pIn++;
226 px += nRows;
227
228 *px = *pIn++;
229 px += nRows;
230
231 /* Decrement column loop counter */
232 col--;
233 }
234
235 /* Loop unrolling: Compute remaining outputs */
236 col = nCols % 0x4U;
237
238 #else
239
240 /* Initialize col with number of samples */
241 col = nCols;
242
243 #endif /* #if defined (ARM_MATH_LOOPUNROLL) */
244
245 while (col > 0U)
246 {
247 /* Read and store input element in destination */
248 *px = *pIn++;
249
250 /* Update pointer px to point to next row of transposed matrix */
251 px += nRows;
252
253 /* Decrement column loop counter */
254 col--;
255 }
256
257 i++;
258
259 /* Decrement row loop counter */
260 row--;
261
262 } while (row > 0U); /* row loop end */
263
264 /* Set status as ARM_MATH_SUCCESS */
265 status = ARM_MATH_SUCCESS;
266 }
267
268 /* Return to application */
269 return (status);
270 }
271 #endif
272 /**
273 * @} end of MatrixTrans group
274 */
275
