提交 | 用户 | age
|
bfc108
|
1 |
/* ----------------------------------------------------------------------
|
Q |
2 |
* Copyright (C) 2010-2014 ARM Limited. All rights reserved.
|
|
3 |
*
|
|
4 |
* $Date: 19. March 2015
|
|
5 |
* $Revision: V.1.4.5
|
|
6 |
*
|
|
7 |
* Project: CMSIS DSP Library
|
|
8 |
* Title: arm_mat_mult_q15.c
|
|
9 |
*
|
|
10 |
* Description: Q15 matrix multiplication.
|
|
11 |
*
|
|
12 |
* Target Processor: Cortex-M4/Cortex-M3/Cortex-M0
|
|
13 |
*
|
|
14 |
* Redistribution and use in source and binary forms, with or without
|
|
15 |
* modification, are permitted provided that the following conditions
|
|
16 |
* are met:
|
|
17 |
* - Redistributions of source code must retain the above copyright
|
|
18 |
* notice, this list of conditions and the following disclaimer.
|
|
19 |
* - Redistributions in binary form must reproduce the above copyright
|
|
20 |
* notice, this list of conditions and the following disclaimer in
|
|
21 |
* the documentation and/or other materials provided with the
|
|
22 |
* distribution.
|
|
23 |
* - Neither the name of ARM LIMITED nor the names of its contributors
|
|
24 |
* may be used to endorse or promote products derived from this
|
|
25 |
* software without specific prior written permission.
|
|
26 |
*
|
|
27 |
* THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
|
|
28 |
* "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
|
|
29 |
* LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
|
|
30 |
* FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE
|
|
31 |
* COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
|
|
32 |
* INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING,
|
|
33 |
* BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
|
|
34 |
* LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
|
|
35 |
* CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
|
|
36 |
* LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN
|
|
37 |
* ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
|
|
38 |
* POSSIBILITY OF SUCH DAMAGE.
|
|
39 |
* -------------------------------------------------------------------- */
|
|
40 |
|
|
41 |
#include "arm_math.h"
|
|
42 |
|
|
43 |
/**
|
|
44 |
* @ingroup groupMatrix
|
|
45 |
*/
|
|
46 |
|
|
47 |
/**
|
|
48 |
* @addtogroup MatrixMult
|
|
49 |
* @{
|
|
50 |
*/
|
|
51 |
|
|
52 |
|
|
53 |
/**
|
|
54 |
* @brief Q15 matrix multiplication
|
|
55 |
* @param[in] *pSrcA points to the first input matrix structure
|
|
56 |
* @param[in] *pSrcB points to the second input matrix structure
|
|
57 |
* @param[out] *pDst points to output matrix structure
|
|
58 |
* @param[in] *pState points to the array for storing intermediate results (Unused)
|
|
59 |
* @return The function returns either
|
|
60 |
* <code>ARM_MATH_SIZE_MISMATCH</code> or <code>ARM_MATH_SUCCESS</code> based on the outcome of size checking.
|
|
61 |
*
|
|
62 |
* @details
|
|
63 |
* <b>Scaling and Overflow Behavior:</b>
|
|
64 |
*
|
|
65 |
* \par
|
|
66 |
* The function is implemented using a 64-bit internal accumulator. The inputs to the
|
|
67 |
* multiplications are in 1.15 format and multiplications yield a 2.30 result.
|
|
68 |
* The 2.30 intermediate
|
|
69 |
* results are accumulated in a 64-bit accumulator in 34.30 format. This approach
|
|
70 |
* provides 33 guard bits and there is no risk of overflow. The 34.30 result is then
|
|
71 |
* truncated to 34.15 format by discarding the low 15 bits and then saturated to
|
|
72 |
* 1.15 format.
|
|
73 |
*
|
|
74 |
* \par
|
|
75 |
* Refer to <code>arm_mat_mult_fast_q15()</code> for a faster but less precise version of this function for Cortex-M3 and Cortex-M4.
|
|
76 |
*
|
|
77 |
*/
|
|
78 |
|
|
79 |
arm_status arm_mat_mult_q15(
|
|
80 |
const arm_matrix_instance_q15 * pSrcA,
|
|
81 |
const arm_matrix_instance_q15 * pSrcB,
|
|
82 |
arm_matrix_instance_q15 * pDst,
|
|
83 |
q15_t * pState CMSIS_UNUSED)
|
|
84 |
{
|
|
85 |
q63_t sum; /* accumulator */
|
|
86 |
|
|
87 |
#ifndef ARM_MATH_CM0_FAMILY
|
|
88 |
|
|
89 |
/* Run the below code for Cortex-M4 and Cortex-M3 */
|
|
90 |
|
|
91 |
q15_t *pSrcBT = pState; /* input data matrix pointer for transpose */
|
|
92 |
q15_t *pInA = pSrcA->pData; /* input data matrix pointer A of Q15 type */
|
|
93 |
q15_t *pInB = pSrcB->pData; /* input data matrix pointer B of Q15 type */
|
|
94 |
q15_t *px; /* Temporary output data matrix pointer */
|
|
95 |
uint16_t numRowsA = pSrcA->numRows; /* number of rows of input matrix A */
|
|
96 |
uint16_t numColsB = pSrcB->numCols; /* number of columns of input matrix B */
|
|
97 |
uint16_t numColsA = pSrcA->numCols; /* number of columns of input matrix A */
|
|
98 |
uint16_t numRowsB = pSrcB->numRows; /* number of rows of input matrix A */
|
|
99 |
uint16_t col, i = 0u, row = numRowsB, colCnt; /* loop counters */
|
|
100 |
arm_status status; /* status of matrix multiplication */
|
|
101 |
|
|
102 |
#ifndef UNALIGNED_SUPPORT_DISABLE
|
|
103 |
|
|
104 |
q31_t in; /* Temporary variable to hold the input value */
|
|
105 |
q31_t pSourceA1, pSourceB1, pSourceA2, pSourceB2;
|
|
106 |
|
|
107 |
#else
|
|
108 |
|
|
109 |
q15_t in; /* Temporary variable to hold the input value */
|
|
110 |
q15_t inA1, inB1, inA2, inB2;
|
|
111 |
|
|
112 |
#endif /* #ifndef UNALIGNED_SUPPORT_DISABLE */
|
|
113 |
|
|
114 |
#ifdef ARM_MATH_MATRIX_CHECK
|
|
115 |
/* Check for matrix mismatch condition */
|
|
116 |
if((pSrcA->numCols != pSrcB->numRows) ||
|
|
117 |
(pSrcA->numRows != pDst->numRows) || (pSrcB->numCols != pDst->numCols))
|
|
118 |
{
|
|
119 |
/* Set status as ARM_MATH_SIZE_MISMATCH */
|
|
120 |
status = ARM_MATH_SIZE_MISMATCH;
|
|
121 |
}
|
|
122 |
else
|
|
123 |
#endif /* #ifdef ARM_MATH_MATRIX_CHECK */
|
|
124 |
{
|
|
125 |
/* Matrix transpose */
|
|
126 |
do
|
|
127 |
{
|
|
128 |
/* Apply loop unrolling and exchange the columns with row elements */
|
|
129 |
col = numColsB >> 2;
|
|
130 |
|
|
131 |
/* The pointer px is set to starting address of the column being processed */
|
|
132 |
px = pSrcBT + i;
|
|
133 |
|
|
134 |
/* First part of the processing with loop unrolling. Compute 4 outputs at a time.
|
|
135 |
** a second loop below computes the remaining 1 to 3 samples. */
|
|
136 |
while(col > 0u)
|
|
137 |
{
|
|
138 |
#ifndef UNALIGNED_SUPPORT_DISABLE
|
|
139 |
|
|
140 |
/* Read two elements from the row */
|
|
141 |
in = *__SIMD32(pInB)++;
|
|
142 |
|
|
143 |
/* Unpack and store one element in the destination */
|
|
144 |
#ifndef ARM_MATH_BIG_ENDIAN
|
|
145 |
|
|
146 |
*px = (q15_t) in;
|
|
147 |
|
|
148 |
#else
|
|
149 |
|
|
150 |
*px = (q15_t) ((in & (q31_t) 0xffff0000) >> 16);
|
|
151 |
|
|
152 |
#endif /* #ifndef ARM_MATH_BIG_ENDIAN */
|
|
153 |
|
|
154 |
/* Update the pointer px to point to the next row of the transposed matrix */
|
|
155 |
px += numRowsB;
|
|
156 |
|
|
157 |
/* Unpack and store the second element in the destination */
|
|
158 |
#ifndef ARM_MATH_BIG_ENDIAN
|
|
159 |
|
|
160 |
*px = (q15_t) ((in & (q31_t) 0xffff0000) >> 16);
|
|
161 |
|
|
162 |
#else
|
|
163 |
|
|
164 |
*px = (q15_t) in;
|
|
165 |
|
|
166 |
#endif /* #ifndef ARM_MATH_BIG_ENDIAN */
|
|
167 |
|
|
168 |
/* Update the pointer px to point to the next row of the transposed matrix */
|
|
169 |
px += numRowsB;
|
|
170 |
|
|
171 |
/* Read two elements from the row */
|
|
172 |
in = *__SIMD32(pInB)++;
|
|
173 |
|
|
174 |
/* Unpack and store one element in the destination */
|
|
175 |
#ifndef ARM_MATH_BIG_ENDIAN
|
|
176 |
|
|
177 |
*px = (q15_t) in;
|
|
178 |
|
|
179 |
#else
|
|
180 |
|
|
181 |
*px = (q15_t) ((in & (q31_t) 0xffff0000) >> 16);
|
|
182 |
|
|
183 |
#endif /* #ifndef ARM_MATH_BIG_ENDIAN */
|
|
184 |
|
|
185 |
/* Update the pointer px to point to the next row of the transposed matrix */
|
|
186 |
px += numRowsB;
|
|
187 |
|
|
188 |
/* Unpack and store the second element in the destination */
|
|
189 |
|
|
190 |
#ifndef ARM_MATH_BIG_ENDIAN
|
|
191 |
|
|
192 |
*px = (q15_t) ((in & (q31_t) 0xffff0000) >> 16);
|
|
193 |
|
|
194 |
#else
|
|
195 |
|
|
196 |
*px = (q15_t) in;
|
|
197 |
|
|
198 |
#endif /* #ifndef ARM_MATH_BIG_ENDIAN */
|
|
199 |
|
|
200 |
/* Update the pointer px to point to the next row of the transposed matrix */
|
|
201 |
px += numRowsB;
|
|
202 |
|
|
203 |
#else
|
|
204 |
|
|
205 |
/* Read one element from the row */
|
|
206 |
in = *pInB++;
|
|
207 |
|
|
208 |
/* Store one element in the destination */
|
|
209 |
*px = in;
|
|
210 |
|
|
211 |
/* Update the pointer px to point to the next row of the transposed matrix */
|
|
212 |
px += numRowsB;
|
|
213 |
|
|
214 |
/* Read one element from the row */
|
|
215 |
in = *pInB++;
|
|
216 |
|
|
217 |
/* Store one element in the destination */
|
|
218 |
*px = in;
|
|
219 |
|
|
220 |
/* Update the pointer px to point to the next row of the transposed matrix */
|
|
221 |
px += numRowsB;
|
|
222 |
|
|
223 |
/* Read one element from the row */
|
|
224 |
in = *pInB++;
|
|
225 |
|
|
226 |
/* Store one element in the destination */
|
|
227 |
*px = in;
|
|
228 |
|
|
229 |
/* Update the pointer px to point to the next row of the transposed matrix */
|
|
230 |
px += numRowsB;
|
|
231 |
|
|
232 |
/* Read one element from the row */
|
|
233 |
in = *pInB++;
|
|
234 |
|
|
235 |
/* Store one element in the destination */
|
|
236 |
*px = in;
|
|
237 |
|
|
238 |
/* Update the pointer px to point to the next row of the transposed matrix */
|
|
239 |
px += numRowsB;
|
|
240 |
|
|
241 |
#endif /* #ifndef UNALIGNED_SUPPORT_DISABLE */
|
|
242 |
|
|
243 |
/* Decrement the column loop counter */
|
|
244 |
col--;
|
|
245 |
}
|
|
246 |
|
|
247 |
/* If the columns of pSrcB is not a multiple of 4, compute any remaining output samples here.
|
|
248 |
** No loop unrolling is used. */
|
|
249 |
col = numColsB % 0x4u;
|
|
250 |
|
|
251 |
while(col > 0u)
|
|
252 |
{
|
|
253 |
/* Read and store the input element in the destination */
|
|
254 |
*px = *pInB++;
|
|
255 |
|
|
256 |
/* Update the pointer px to point to the next row of the transposed matrix */
|
|
257 |
px += numRowsB;
|
|
258 |
|
|
259 |
/* Decrement the column loop counter */
|
|
260 |
col--;
|
|
261 |
}
|
|
262 |
|
|
263 |
i++;
|
|
264 |
|
|
265 |
/* Decrement the row loop counter */
|
|
266 |
row--;
|
|
267 |
|
|
268 |
} while(row > 0u);
|
|
269 |
|
|
270 |
/* Reset the variables for the usage in the following multiplication process */
|
|
271 |
row = numRowsA;
|
|
272 |
i = 0u;
|
|
273 |
px = pDst->pData;
|
|
274 |
|
|
275 |
/* The following loop performs the dot-product of each row in pSrcA with each column in pSrcB */
|
|
276 |
/* row loop */
|
|
277 |
do
|
|
278 |
{
|
|
279 |
/* For every row wise process, the column loop counter is to be initiated */
|
|
280 |
col = numColsB;
|
|
281 |
|
|
282 |
/* For every row wise process, the pIn2 pointer is set
|
|
283 |
** to the starting address of the transposed pSrcB data */
|
|
284 |
pInB = pSrcBT;
|
|
285 |
|
|
286 |
/* column loop */
|
|
287 |
do
|
|
288 |
{
|
|
289 |
/* Set the variable sum, that acts as accumulator, to zero */
|
|
290 |
sum = 0;
|
|
291 |
|
|
292 |
/* Apply loop unrolling and compute 2 MACs simultaneously. */
|
|
293 |
colCnt = numColsA >> 2;
|
|
294 |
|
|
295 |
/* Initiate the pointer pIn1 to point to the starting address of the column being processed */
|
|
296 |
pInA = pSrcA->pData + i;
|
|
297 |
|
|
298 |
|
|
299 |
/* matrix multiplication */
|
|
300 |
while(colCnt > 0u)
|
|
301 |
{
|
|
302 |
/* c(m,n) = a(1,1)*b(1,1) + a(1,2) * b(2,1) + .... + a(m,p)*b(p,n) */
|
|
303 |
#ifndef UNALIGNED_SUPPORT_DISABLE
|
|
304 |
|
|
305 |
/* read real and imag values from pSrcA and pSrcB buffer */
|
|
306 |
pSourceA1 = *__SIMD32(pInA)++;
|
|
307 |
pSourceB1 = *__SIMD32(pInB)++;
|
|
308 |
|
|
309 |
pSourceA2 = *__SIMD32(pInA)++;
|
|
310 |
pSourceB2 = *__SIMD32(pInB)++;
|
|
311 |
|
|
312 |
/* Multiply and Accumlates */
|
|
313 |
sum = __SMLALD(pSourceA1, pSourceB1, sum);
|
|
314 |
sum = __SMLALD(pSourceA2, pSourceB2, sum);
|
|
315 |
|
|
316 |
#else
|
|
317 |
/* read real and imag values from pSrcA and pSrcB buffer */
|
|
318 |
inA1 = *pInA++;
|
|
319 |
inB1 = *pInB++;
|
|
320 |
inA2 = *pInA++;
|
|
321 |
/* Multiply and Accumlates */
|
|
322 |
sum += inA1 * inB1;
|
|
323 |
inB2 = *pInB++;
|
|
324 |
|
|
325 |
inA1 = *pInA++;
|
|
326 |
inB1 = *pInB++;
|
|
327 |
/* Multiply and Accumlates */
|
|
328 |
sum += inA2 * inB2;
|
|
329 |
inA2 = *pInA++;
|
|
330 |
inB2 = *pInB++;
|
|
331 |
|
|
332 |
/* Multiply and Accumlates */
|
|
333 |
sum += inA1 * inB1;
|
|
334 |
sum += inA2 * inB2;
|
|
335 |
|
|
336 |
#endif /* #ifndef UNALIGNED_SUPPORT_DISABLE */
|
|
337 |
|
|
338 |
/* Decrement the loop counter */
|
|
339 |
colCnt--;
|
|
340 |
}
|
|
341 |
|
|
342 |
/* process remaining column samples */
|
|
343 |
colCnt = numColsA & 3u;
|
|
344 |
|
|
345 |
while(colCnt > 0u)
|
|
346 |
{
|
|
347 |
/* c(m,n) = a(1,1)*b(1,1) + a(1,2) * b(2,1) + .... + a(m,p)*b(p,n) */
|
|
348 |
sum += *pInA++ * *pInB++;
|
|
349 |
|
|
350 |
/* Decrement the loop counter */
|
|
351 |
colCnt--;
|
|
352 |
}
|
|
353 |
|
|
354 |
/* Saturate and store the result in the destination buffer */
|
|
355 |
*px = (q15_t) (__SSAT((sum >> 15), 16));
|
|
356 |
px++;
|
|
357 |
|
|
358 |
/* Decrement the column loop counter */
|
|
359 |
col--;
|
|
360 |
|
|
361 |
} while(col > 0u);
|
|
362 |
|
|
363 |
i = i + numColsA;
|
|
364 |
|
|
365 |
/* Decrement the row loop counter */
|
|
366 |
row--;
|
|
367 |
|
|
368 |
} while(row > 0u);
|
|
369 |
|
|
370 |
#else
|
|
371 |
|
|
372 |
/* Run the below code for Cortex-M0 */
|
|
373 |
|
|
374 |
q15_t *pIn1 = pSrcA->pData; /* input data matrix pointer A */
|
|
375 |
q15_t *pIn2 = pSrcB->pData; /* input data matrix pointer B */
|
|
376 |
q15_t *pInA = pSrcA->pData; /* input data matrix pointer A of Q15 type */
|
|
377 |
q15_t *pInB = pSrcB->pData; /* input data matrix pointer B of Q15 type */
|
|
378 |
q15_t *pOut = pDst->pData; /* output data matrix pointer */
|
|
379 |
q15_t *px; /* Temporary output data matrix pointer */
|
|
380 |
uint16_t numColsB = pSrcB->numCols; /* number of columns of input matrix B */
|
|
381 |
uint16_t numColsA = pSrcA->numCols; /* number of columns of input matrix A */
|
|
382 |
uint16_t numRowsA = pSrcA->numRows; /* number of rows of input matrix A */
|
|
383 |
uint16_t col, i = 0u, row = numRowsA, colCnt; /* loop counters */
|
|
384 |
arm_status status; /* status of matrix multiplication */
|
|
385 |
|
|
386 |
#ifdef ARM_MATH_MATRIX_CHECK
|
|
387 |
|
|
388 |
/* Check for matrix mismatch condition */
|
|
389 |
if((pSrcA->numCols != pSrcB->numRows) ||
|
|
390 |
(pSrcA->numRows != pDst->numRows) || (pSrcB->numCols != pDst->numCols))
|
|
391 |
{
|
|
392 |
/* Set status as ARM_MATH_SIZE_MISMATCH */
|
|
393 |
status = ARM_MATH_SIZE_MISMATCH;
|
|
394 |
}
|
|
395 |
else
|
|
396 |
#endif /* #ifdef ARM_MATH_MATRIX_CHECK */
|
|
397 |
|
|
398 |
{
|
|
399 |
/* The following loop performs the dot-product of each row in pSrcA with each column in pSrcB */
|
|
400 |
/* row loop */
|
|
401 |
do
|
|
402 |
{
|
|
403 |
/* Output pointer is set to starting address of the row being processed */
|
|
404 |
px = pOut + i;
|
|
405 |
|
|
406 |
/* For every row wise process, the column loop counter is to be initiated */
|
|
407 |
col = numColsB;
|
|
408 |
|
|
409 |
/* For every row wise process, the pIn2 pointer is set
|
|
410 |
** to the starting address of the pSrcB data */
|
|
411 |
pIn2 = pSrcB->pData;
|
|
412 |
|
|
413 |
/* column loop */
|
|
414 |
do
|
|
415 |
{
|
|
416 |
/* Set the variable sum, that acts as accumulator, to zero */
|
|
417 |
sum = 0;
|
|
418 |
|
|
419 |
/* Initiate the pointer pIn1 to point to the starting address of pSrcA */
|
|
420 |
pIn1 = pInA;
|
|
421 |
|
|
422 |
/* Matrix A columns number of MAC operations are to be performed */
|
|
423 |
colCnt = numColsA;
|
|
424 |
|
|
425 |
/* matrix multiplication */
|
|
426 |
while(colCnt > 0u)
|
|
427 |
{
|
|
428 |
/* c(m,n) = a(1,1)*b(1,1) + a(1,2) * b(2,1) + .... + a(m,p)*b(p,n) */
|
|
429 |
/* Perform the multiply-accumulates */
|
|
430 |
sum += (q31_t) * pIn1++ * *pIn2;
|
|
431 |
pIn2 += numColsB;
|
|
432 |
|
|
433 |
/* Decrement the loop counter */
|
|
434 |
colCnt--;
|
|
435 |
}
|
|
436 |
|
|
437 |
/* Convert the result from 34.30 to 1.15 format and store the saturated value in destination buffer */
|
|
438 |
/* Saturate and store the result in the destination buffer */
|
|
439 |
*px++ = (q15_t) __SSAT((sum >> 15), 16);
|
|
440 |
|
|
441 |
/* Decrement the column loop counter */
|
|
442 |
col--;
|
|
443 |
|
|
444 |
/* Update the pointer pIn2 to point to the starting address of the next column */
|
|
445 |
pIn2 = pInB + (numColsB - col);
|
|
446 |
|
|
447 |
} while(col > 0u);
|
|
448 |
|
|
449 |
/* Update the pointer pSrcA to point to the starting address of the next row */
|
|
450 |
i = i + numColsB;
|
|
451 |
pInA = pInA + numColsA;
|
|
452 |
|
|
453 |
/* Decrement the row loop counter */
|
|
454 |
row--;
|
|
455 |
|
|
456 |
} while(row > 0u);
|
|
457 |
|
|
458 |
#endif /* #ifndef ARM_MATH_CM0_FAMILY */
|
|
459 |
/* set status as ARM_MATH_SUCCESS */
|
|
460 |
status = ARM_MATH_SUCCESS;
|
|
461 |
}
|
|
462 |
|
|
463 |
/* Return to application */
|
|
464 |
return (status);
|
|
465 |
}
|
|
466 |
|
|
467 |
/**
|
|
468 |
* @} end of MatrixMult group
|
|
469 |
*/
|