1 /* ----------------------------------------------------------------------
2 * Project: CMSIS DSP Library
3 * Title: arm_mat_trans_f16.c
4 * Description: Floating-point matrix transpose
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/matrix_functions_f16.h"
30
31 #if defined(ARM_FLOAT16_SUPPORTED)
32
33
34 /**
35 @ingroup groupMatrix
36 */
37
38 /**
39 @addtogroup MatrixTrans
40 @{
41 */
42
43 /**
44 @brief Floating-point matrix transpose.
45 @param[in] pSrc points to input matrix
46 @param[out] pDst points to output matrix
47 @return execution status
48 - \ref ARM_MATH_SUCCESS : Operation successful
49 - \ref ARM_MATH_SIZE_MISMATCH : Matrix size check failed
50 */
51 #if defined(ARM_MATH_MVE_FLOAT16) && !defined(ARM_MATH_AUTOVECTORIZE)
52
53 #include "arm_helium_utils.h"
54
arm_mat_trans_f16(const arm_matrix_instance_f16 * pSrc,arm_matrix_instance_f16 * pDst)55 arm_status arm_mat_trans_f16(
56 const arm_matrix_instance_f16 * pSrc,
57 arm_matrix_instance_f16 * pDst)
58 {
59 arm_status status; /* status of matrix transpose */
60
61 #ifdef ARM_MATH_MATRIX_CHECK
62
63 /* Check for matrix mismatch condition */
64 if ((pSrc->numRows != pDst->numCols) ||
65 (pSrc->numCols != pDst->numRows) )
66 {
67 /* Set status as ARM_MATH_SIZE_MISMATCH */
68 status = ARM_MATH_SIZE_MISMATCH;
69 }
70 else
71
72 #endif /* #ifdef ARM_MATH_MATRIX_CHECK */
73
74 {
75 if (pDst->numRows == pDst->numCols)
76 {
77 if (pDst->numCols == 1)
78 {
79 pDst->pData[0] = pSrc->pData[0];
80 return(ARM_MATH_SUCCESS);
81 }
82 if (pDst->numCols == 2)
83 return arm_mat_trans_16bit_2x2((uint16_t *)pSrc->pData, (uint16_t *)pDst->pData);
84 if (pDst->numCols == 3)
85 return arm_mat_trans_16bit_3x3_mve((uint16_t *)pSrc->pData, (uint16_t *)pDst->pData);
86 if (pDst->numCols == 4)
87 return arm_mat_trans_16bit_4x4_mve((uint16_t *)pSrc->pData, (uint16_t *)pDst->pData);
88 }
89
90 arm_mat_trans_16bit_generic(pSrc->numRows, pSrc->numCols, (uint16_t *)pSrc->pData, (uint16_t *)pDst->pData);
91 /* Set status as ARM_MATH_SUCCESS */
92 status = ARM_MATH_SUCCESS;
93 }
94
95 /* Return to application */
96 return (status);
97 }
98
99 #else
100
arm_mat_trans_f16(const arm_matrix_instance_f16 * pSrc,arm_matrix_instance_f16 * pDst)101 arm_status arm_mat_trans_f16(
102 const arm_matrix_instance_f16 * pSrc,
103 arm_matrix_instance_f16 * pDst)
104 {
105 float16_t *pIn = pSrc->pData; /* input data matrix pointer */
106 float16_t *pOut = pDst->pData; /* output data matrix pointer */
107 float16_t *px; /* Temporary output data matrix pointer */
108 uint16_t nRows = pSrc->numRows; /* number of rows */
109 uint16_t nCols = pSrc->numCols; /* number of columns */
110 uint32_t col, row = nRows, i = 0U; /* Loop counters */
111 arm_status status; /* status of matrix transpose */
112
113 #ifdef ARM_MATH_MATRIX_CHECK
114
115 /* Check for matrix mismatch condition */
116 if ((pSrc->numRows != pDst->numCols) ||
117 (pSrc->numCols != pDst->numRows) )
118 {
119 /* Set status as ARM_MATH_SIZE_MISMATCH */
120 status = ARM_MATH_SIZE_MISMATCH;
121 }
122 else
123
124 #endif /* #ifdef ARM_MATH_MATRIX_CHECK */
125
126 {
127 /* Matrix transpose by exchanging the rows with columns */
128 /* row loop */
129 do
130 {
131 /* Pointer px is set to starting address of column being processed */
132 px = pOut + i;
133
134 #if defined (ARM_MATH_LOOPUNROLL)
135
136 /* Loop unrolling: Compute 4 outputs at a time */
137 col = nCols >> 2U;
138
139 while (col > 0U) /* column loop */
140 {
141 /* Read and store input element in destination */
142 *px = *pIn++;
143 /* Update pointer px to point to next row of transposed matrix */
144 px += nRows;
145
146 *px = *pIn++;
147 px += nRows;
148
149 *px = *pIn++;
150 px += nRows;
151
152 *px = *pIn++;
153 px += nRows;
154
155 /* Decrement column loop counter */
156 col--;
157 }
158
159 /* Loop unrolling: Compute remaining outputs */
160 col = nCols % 0x4U;
161
162 #else
163
164 /* Initialize col with number of samples */
165 col = nCols;
166
167 #endif /* #if defined (ARM_MATH_LOOPUNROLL) */
168
169 while (col > 0U)
170 {
171 /* Read and store input element in destination */
172 *px = *pIn++;
173
174 /* Update pointer px to point to next row of transposed matrix */
175 px += nRows;
176
177 /* Decrement column loop counter */
178 col--;
179 }
180
181 i++;
182
183 /* Decrement row loop counter */
184 row--;
185
186 } while (row > 0U); /* row loop end */
187
188 /* Set status as ARM_MATH_SUCCESS */
189 status = ARM_MATH_SUCCESS;
190 }
191
192 /* Return to application */
193 return (status);
194 }
195 #endif /* defined(ARM_MATH_MVEF) && !defined(ARM_MATH_AUTOVECTORIZE) */
196
197 /**
198 * @} end of MatrixTrans group
199 */
200
201 #endif /* #if defined(ARM_FLOAT16_SUPPORTED) */
202
203