mbed official / mbed-dsp

CMSIS DSP library

Dependents:   performance_timer Surfboard_ gps2rtty Capstone ... more

Legacy Warning

This is an mbed 2 library. To learn more about mbed OS 5, visit the docs.

cmsis_dsp/MatrixFunctions/arm_mat_cmplx_mult_q15.c@5:3762170b6d4d, 2015-11-20 (annotated)

Committer:
mbed_official
Date:
Fri Nov 20 08:45:18 2015 +0000
Revision:
5:3762170b6d4d
Synchronized with git revision 2eb940b9a73af188d3004a2575fdfbb05febe62b



Full URL: https://github.com/mbedmicro/mbed/commit/2eb940b9a73af188d3004a2575fdfbb05febe62b/



Added option to build rpc library. closes #1426

Who changed what in which revision?

UserRevisionLine numberNew contents of line
mbed_official 5:3762170b6d4d 1 /* ----------------------------------------------------------------------
mbed_official 5:3762170b6d4d 2 * Copyright (C) 2010-2014 ARM Limited. All rights reserved.
mbed_official 5:3762170b6d4d 3 *
mbed_official 5:3762170b6d4d 4 * $Date: 19. March 2015
mbed_official 5:3762170b6d4d 5 * $Revision: V.1.4.5
mbed_official 5:3762170b6d4d 6 *
mbed_official 5:3762170b6d4d 7 * Project: CMSIS DSP Library
mbed_official 5:3762170b6d4d 8 * Title: arm_cmplx_mat_mult_q15.c
mbed_official 5:3762170b6d4d 9 *
mbed_official 5:3762170b6d4d 10 * Description: Q15 complex matrix multiplication.
mbed_official 5:3762170b6d4d 11 *
mbed_official 5:3762170b6d4d 12 * Target Processor: Cortex-M4/Cortex-M3/Cortex-M0
mbed_official 5:3762170b6d4d 13 *
mbed_official 5:3762170b6d4d 14 * Redistribution and use in source and binary forms, with or without
mbed_official 5:3762170b6d4d 15 * modification, are permitted provided that the following conditions
mbed_official 5:3762170b6d4d 16 * are met:
mbed_official 5:3762170b6d4d 17 * - Redistributions of source code must retain the above copyright
mbed_official 5:3762170b6d4d 18 * notice, this list of conditions and the following disclaimer.
mbed_official 5:3762170b6d4d 19 * - Redistributions in binary form must reproduce the above copyright
mbed_official 5:3762170b6d4d 20 * notice, this list of conditions and the following disclaimer in
mbed_official 5:3762170b6d4d 21 * the documentation and/or other materials provided with the
mbed_official 5:3762170b6d4d 22 * distribution.
mbed_official 5:3762170b6d4d 23 * - Neither the name of ARM LIMITED nor the names of its contributors
mbed_official 5:3762170b6d4d 24 * may be used to endorse or promote products derived from this
mbed_official 5:3762170b6d4d 25 * software without specific prior written permission.
mbed_official 5:3762170b6d4d 26 *
mbed_official 5:3762170b6d4d 27 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
mbed_official 5:3762170b6d4d 28 * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
mbed_official 5:3762170b6d4d 29 * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
mbed_official 5:3762170b6d4d 30 * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE
mbed_official 5:3762170b6d4d 31 * COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
mbed_official 5:3762170b6d4d 32 * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING,
mbed_official 5:3762170b6d4d 33 * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
mbed_official 5:3762170b6d4d 34 * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
mbed_official 5:3762170b6d4d 35 * CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
mbed_official 5:3762170b6d4d 36 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN
mbed_official 5:3762170b6d4d 37 * ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
mbed_official 5:3762170b6d4d 38 * POSSIBILITY OF SUCH DAMAGE.
mbed_official 5:3762170b6d4d 39 * -------------------------------------------------------------------- */
mbed_official 5:3762170b6d4d 40 #include "arm_math.h"
mbed_official 5:3762170b6d4d 41
mbed_official 5:3762170b6d4d 42 /**
mbed_official 5:3762170b6d4d 43 * @ingroup groupMatrix
mbed_official 5:3762170b6d4d 44 */
mbed_official 5:3762170b6d4d 45
mbed_official 5:3762170b6d4d 46 /**
mbed_official 5:3762170b6d4d 47 * @addtogroup CmplxMatrixMult
mbed_official 5:3762170b6d4d 48 * @{
mbed_official 5:3762170b6d4d 49 */
mbed_official 5:3762170b6d4d 50
mbed_official 5:3762170b6d4d 51
mbed_official 5:3762170b6d4d 52 /**
mbed_official 5:3762170b6d4d 53 * @brief Q15 Complex matrix multiplication
mbed_official 5:3762170b6d4d 54 * @param[in] *pSrcA points to the first input complex matrix structure
mbed_official 5:3762170b6d4d 55 * @param[in] *pSrcB points to the second input complex matrix structure
mbed_official 5:3762170b6d4d 56 * @param[out] *pDst points to output complex matrix structure
mbed_official 5:3762170b6d4d 57 * @param[in] *pScratch points to the array for storing intermediate results
mbed_official 5:3762170b6d4d 58 * @return The function returns either
mbed_official 5:3762170b6d4d 59 * <code>ARM_MATH_SIZE_MISMATCH</code> or <code>ARM_MATH_SUCCESS</code> based on the outcome of size checking.
mbed_official 5:3762170b6d4d 60 *
mbed_official 5:3762170b6d4d 61 * \par Conditions for optimum performance
mbed_official 5:3762170b6d4d 62 * Input, output and state buffers should be aligned by 32-bit
mbed_official 5:3762170b6d4d 63 *
mbed_official 5:3762170b6d4d 64 * \par Restrictions
mbed_official 5:3762170b6d4d 65 * If the silicon does not support unaligned memory access enable the macro UNALIGNED_SUPPORT_DISABLE
mbed_official 5:3762170b6d4d 66 * In this case input, output, scratch buffers should be aligned by 32-bit
mbed_official 5:3762170b6d4d 67 *
mbed_official 5:3762170b6d4d 68 * @details
mbed_official 5:3762170b6d4d 69 * <b>Scaling and Overflow Behavior:</b>
mbed_official 5:3762170b6d4d 70 *
mbed_official 5:3762170b6d4d 71 * \par
mbed_official 5:3762170b6d4d 72 * The function is implemented using a 64-bit internal accumulator. The inputs to the
mbed_official 5:3762170b6d4d 73 * multiplications are in 1.15 format and multiplications yield a 2.30 result.
mbed_official 5:3762170b6d4d 74 * The 2.30 intermediate
mbed_official 5:3762170b6d4d 75 * results are accumulated in a 64-bit accumulator in 34.30 format. This approach
mbed_official 5:3762170b6d4d 76 * provides 33 guard bits and there is no risk of overflow. The 34.30 result is then
mbed_official 5:3762170b6d4d 77 * truncated to 34.15 format by discarding the low 15 bits and then saturated to
mbed_official 5:3762170b6d4d 78 * 1.15 format.
mbed_official 5:3762170b6d4d 79 *
mbed_official 5:3762170b6d4d 80 * \par
mbed_official 5:3762170b6d4d 81 * Refer to <code>arm_mat_mult_fast_q15()</code> for a faster but less precise version of this function.
mbed_official 5:3762170b6d4d 82 *
mbed_official 5:3762170b6d4d 83 */
mbed_official 5:3762170b6d4d 84
mbed_official 5:3762170b6d4d 85
mbed_official 5:3762170b6d4d 86
mbed_official 5:3762170b6d4d 87
mbed_official 5:3762170b6d4d 88 arm_status arm_mat_cmplx_mult_q15(
mbed_official 5:3762170b6d4d 89 const arm_matrix_instance_q15 * pSrcA,
mbed_official 5:3762170b6d4d 90 const arm_matrix_instance_q15 * pSrcB,
mbed_official 5:3762170b6d4d 91 arm_matrix_instance_q15 * pDst,
mbed_official 5:3762170b6d4d 92 q15_t * pScratch)
mbed_official 5:3762170b6d4d 93 {
mbed_official 5:3762170b6d4d 94 /* accumulator */
mbed_official 5:3762170b6d4d 95 q15_t *pSrcBT = pScratch; /* input data matrix pointer for transpose */
mbed_official 5:3762170b6d4d 96 q15_t *pInA = pSrcA->pData; /* input data matrix pointer A of Q15 type */
mbed_official 5:3762170b6d4d 97 q15_t *pInB = pSrcB->pData; /* input data matrix pointer B of Q15 type */
mbed_official 5:3762170b6d4d 98 q15_t *px; /* Temporary output data matrix pointer */
mbed_official 5:3762170b6d4d 99 uint16_t numRowsA = pSrcA->numRows; /* number of rows of input matrix A */
mbed_official 5:3762170b6d4d 100 uint16_t numColsB = pSrcB->numCols; /* number of columns of input matrix B */
mbed_official 5:3762170b6d4d 101 uint16_t numColsA = pSrcA->numCols; /* number of columns of input matrix A */
mbed_official 5:3762170b6d4d 102 uint16_t numRowsB = pSrcB->numRows; /* number of rows of input matrix A */
mbed_official 5:3762170b6d4d 103 uint16_t col, i = 0u, row = numRowsB, colCnt; /* loop counters */
mbed_official 5:3762170b6d4d 104 arm_status status; /* status of matrix multiplication */
mbed_official 5:3762170b6d4d 105 q63_t sumReal, sumImag;
mbed_official 5:3762170b6d4d 106
mbed_official 5:3762170b6d4d 107 #ifdef UNALIGNED_SUPPORT_DISABLE
mbed_official 5:3762170b6d4d 108 q15_t in; /* Temporary variable to hold the input value */
mbed_official 5:3762170b6d4d 109 q15_t a, b, c, d;
mbed_official 5:3762170b6d4d 110 #else
mbed_official 5:3762170b6d4d 111 q31_t in; /* Temporary variable to hold the input value */
mbed_official 5:3762170b6d4d 112 q31_t prod1, prod2;
mbed_official 5:3762170b6d4d 113 q31_t pSourceA, pSourceB;
mbed_official 5:3762170b6d4d 114 #endif
mbed_official 5:3762170b6d4d 115
mbed_official 5:3762170b6d4d 116 #ifdef ARM_MATH_MATRIX_CHECK
mbed_official 5:3762170b6d4d 117 /* Check for matrix mismatch condition */
mbed_official 5:3762170b6d4d 118 if((pSrcA->numCols != pSrcB->numRows) ||
mbed_official 5:3762170b6d4d 119 (pSrcA->numRows != pDst->numRows) || (pSrcB->numCols != pDst->numCols))
mbed_official 5:3762170b6d4d 120 {
mbed_official 5:3762170b6d4d 121 /* Set status as ARM_MATH_SIZE_MISMATCH */
mbed_official 5:3762170b6d4d 122 status = ARM_MATH_SIZE_MISMATCH;
mbed_official 5:3762170b6d4d 123 }
mbed_official 5:3762170b6d4d 124 else
mbed_official 5:3762170b6d4d 125 #endif
mbed_official 5:3762170b6d4d 126 {
mbed_official 5:3762170b6d4d 127 /* Matrix transpose */
mbed_official 5:3762170b6d4d 128 do
mbed_official 5:3762170b6d4d 129 {
mbed_official 5:3762170b6d4d 130 /* Apply loop unrolling and exchange the columns with row elements */
mbed_official 5:3762170b6d4d 131 col = numColsB >> 2;
mbed_official 5:3762170b6d4d 132
mbed_official 5:3762170b6d4d 133 /* The pointer px is set to starting address of the column being processed */
mbed_official 5:3762170b6d4d 134 px = pSrcBT + i;
mbed_official 5:3762170b6d4d 135
mbed_official 5:3762170b6d4d 136 /* First part of the processing with loop unrolling. Compute 4 outputs at a time.
mbed_official 5:3762170b6d4d 137 ** a second loop below computes the remaining 1 to 3 samples. */
mbed_official 5:3762170b6d4d 138 while(col > 0u)
mbed_official 5:3762170b6d4d 139 {
mbed_official 5:3762170b6d4d 140 #ifdef UNALIGNED_SUPPORT_DISABLE
mbed_official 5:3762170b6d4d 141 /* Read two elements from the row */
mbed_official 5:3762170b6d4d 142 in = *pInB++;
mbed_official 5:3762170b6d4d 143 *px = in;
mbed_official 5:3762170b6d4d 144 in = *pInB++;
mbed_official 5:3762170b6d4d 145 px[1] = in;
mbed_official 5:3762170b6d4d 146
mbed_official 5:3762170b6d4d 147 /* Update the pointer px to point to the next row of the transposed matrix */
mbed_official 5:3762170b6d4d 148 px += numRowsB * 2;
mbed_official 5:3762170b6d4d 149
mbed_official 5:3762170b6d4d 150 /* Read two elements from the row */
mbed_official 5:3762170b6d4d 151 in = *pInB++;
mbed_official 5:3762170b6d4d 152 *px = in;
mbed_official 5:3762170b6d4d 153 in = *pInB++;
mbed_official 5:3762170b6d4d 154 px[1] = in;
mbed_official 5:3762170b6d4d 155
mbed_official 5:3762170b6d4d 156 /* Update the pointer px to point to the next row of the transposed matrix */
mbed_official 5:3762170b6d4d 157 px += numRowsB * 2;
mbed_official 5:3762170b6d4d 158
mbed_official 5:3762170b6d4d 159 /* Read two elements from the row */
mbed_official 5:3762170b6d4d 160 in = *pInB++;
mbed_official 5:3762170b6d4d 161 *px = in;
mbed_official 5:3762170b6d4d 162 in = *pInB++;
mbed_official 5:3762170b6d4d 163 px[1] = in;
mbed_official 5:3762170b6d4d 164
mbed_official 5:3762170b6d4d 165 /* Update the pointer px to point to the next row of the transposed matrix */
mbed_official 5:3762170b6d4d 166 px += numRowsB * 2;
mbed_official 5:3762170b6d4d 167
mbed_official 5:3762170b6d4d 168 /* Read two elements from the row */
mbed_official 5:3762170b6d4d 169 in = *pInB++;
mbed_official 5:3762170b6d4d 170 *px = in;
mbed_official 5:3762170b6d4d 171 in = *pInB++;
mbed_official 5:3762170b6d4d 172 px[1] = in;
mbed_official 5:3762170b6d4d 173
mbed_official 5:3762170b6d4d 174 /* Update the pointer px to point to the next row of the transposed matrix */
mbed_official 5:3762170b6d4d 175 px += numRowsB * 2;
mbed_official 5:3762170b6d4d 176
mbed_official 5:3762170b6d4d 177 /* Decrement the column loop counter */
mbed_official 5:3762170b6d4d 178 col--;
mbed_official 5:3762170b6d4d 179 }
mbed_official 5:3762170b6d4d 180
mbed_official 5:3762170b6d4d 181 /* If the columns of pSrcB is not a multiple of 4, compute any remaining output samples here.
mbed_official 5:3762170b6d4d 182 ** No loop unrolling is used. */
mbed_official 5:3762170b6d4d 183 col = numColsB % 0x4u;
mbed_official 5:3762170b6d4d 184
mbed_official 5:3762170b6d4d 185 while(col > 0u)
mbed_official 5:3762170b6d4d 186 {
mbed_official 5:3762170b6d4d 187 /* Read two elements from the row */
mbed_official 5:3762170b6d4d 188 in = *pInB++;
mbed_official 5:3762170b6d4d 189 *px = in;
mbed_official 5:3762170b6d4d 190 in = *pInB++;
mbed_official 5:3762170b6d4d 191 px[1] = in;
mbed_official 5:3762170b6d4d 192 #else
mbed_official 5:3762170b6d4d 193
mbed_official 5:3762170b6d4d 194 /* Read two elements from the row */
mbed_official 5:3762170b6d4d 195 in = *__SIMD32(pInB)++;
mbed_official 5:3762170b6d4d 196
mbed_official 5:3762170b6d4d 197 *__SIMD32(px) = in;
mbed_official 5:3762170b6d4d 198
mbed_official 5:3762170b6d4d 199 /* Update the pointer px to point to the next row of the transposed matrix */
mbed_official 5:3762170b6d4d 200 px += numRowsB * 2;
mbed_official 5:3762170b6d4d 201
mbed_official 5:3762170b6d4d 202
mbed_official 5:3762170b6d4d 203 /* Read two elements from the row */
mbed_official 5:3762170b6d4d 204 in = *__SIMD32(pInB)++;
mbed_official 5:3762170b6d4d 205
mbed_official 5:3762170b6d4d 206 *__SIMD32(px) = in;
mbed_official 5:3762170b6d4d 207
mbed_official 5:3762170b6d4d 208 /* Update the pointer px to point to the next row of the transposed matrix */
mbed_official 5:3762170b6d4d 209 px += numRowsB * 2;
mbed_official 5:3762170b6d4d 210
mbed_official 5:3762170b6d4d 211 /* Read two elements from the row */
mbed_official 5:3762170b6d4d 212 in = *__SIMD32(pInB)++;
mbed_official 5:3762170b6d4d 213
mbed_official 5:3762170b6d4d 214 *__SIMD32(px) = in;
mbed_official 5:3762170b6d4d 215
mbed_official 5:3762170b6d4d 216 /* Update the pointer px to point to the next row of the transposed matrix */
mbed_official 5:3762170b6d4d 217 px += numRowsB * 2;
mbed_official 5:3762170b6d4d 218
mbed_official 5:3762170b6d4d 219 /* Read two elements from the row */
mbed_official 5:3762170b6d4d 220 in = *__SIMD32(pInB)++;
mbed_official 5:3762170b6d4d 221
mbed_official 5:3762170b6d4d 222 *__SIMD32(px) = in;
mbed_official 5:3762170b6d4d 223
mbed_official 5:3762170b6d4d 224 /* Update the pointer px to point to the next row of the transposed matrix */
mbed_official 5:3762170b6d4d 225 px += numRowsB * 2;
mbed_official 5:3762170b6d4d 226
mbed_official 5:3762170b6d4d 227 /* Decrement the column loop counter */
mbed_official 5:3762170b6d4d 228 col--;
mbed_official 5:3762170b6d4d 229 }
mbed_official 5:3762170b6d4d 230
mbed_official 5:3762170b6d4d 231 /* If the columns of pSrcB is not a multiple of 4, compute any remaining output samples here.
mbed_official 5:3762170b6d4d 232 ** No loop unrolling is used. */
mbed_official 5:3762170b6d4d 233 col = numColsB % 0x4u;
mbed_official 5:3762170b6d4d 234
mbed_official 5:3762170b6d4d 235 while(col > 0u)
mbed_official 5:3762170b6d4d 236 {
mbed_official 5:3762170b6d4d 237 /* Read two elements from the row */
mbed_official 5:3762170b6d4d 238 in = *__SIMD32(pInB)++;
mbed_official 5:3762170b6d4d 239
mbed_official 5:3762170b6d4d 240 *__SIMD32(px) = in;
mbed_official 5:3762170b6d4d 241 #endif
mbed_official 5:3762170b6d4d 242
mbed_official 5:3762170b6d4d 243 /* Update the pointer px to point to the next row of the transposed matrix */
mbed_official 5:3762170b6d4d 244 px += numRowsB * 2;
mbed_official 5:3762170b6d4d 245
mbed_official 5:3762170b6d4d 246 /* Decrement the column loop counter */
mbed_official 5:3762170b6d4d 247 col--;
mbed_official 5:3762170b6d4d 248 }
mbed_official 5:3762170b6d4d 249
mbed_official 5:3762170b6d4d 250 i = i + 2u;
mbed_official 5:3762170b6d4d 251
mbed_official 5:3762170b6d4d 252 /* Decrement the row loop counter */
mbed_official 5:3762170b6d4d 253 row--;
mbed_official 5:3762170b6d4d 254
mbed_official 5:3762170b6d4d 255 } while(row > 0u);
mbed_official 5:3762170b6d4d 256
mbed_official 5:3762170b6d4d 257 /* Reset the variables for the usage in the following multiplication process */
mbed_official 5:3762170b6d4d 258 row = numRowsA;
mbed_official 5:3762170b6d4d 259 i = 0u;
mbed_official 5:3762170b6d4d 260 px = pDst->pData;
mbed_official 5:3762170b6d4d 261
mbed_official 5:3762170b6d4d 262 /* The following loop performs the dot-product of each row in pSrcA with each column in pSrcB */
mbed_official 5:3762170b6d4d 263 /* row loop */
mbed_official 5:3762170b6d4d 264 do
mbed_official 5:3762170b6d4d 265 {
mbed_official 5:3762170b6d4d 266 /* For every row wise process, the column loop counter is to be initiated */
mbed_official 5:3762170b6d4d 267 col = numColsB;
mbed_official 5:3762170b6d4d 268
mbed_official 5:3762170b6d4d 269 /* For every row wise process, the pIn2 pointer is set
mbed_official 5:3762170b6d4d 270 ** to the starting address of the transposed pSrcB data */
mbed_official 5:3762170b6d4d 271 pInB = pSrcBT;
mbed_official 5:3762170b6d4d 272
mbed_official 5:3762170b6d4d 273 /* column loop */
mbed_official 5:3762170b6d4d 274 do
mbed_official 5:3762170b6d4d 275 {
mbed_official 5:3762170b6d4d 276 /* Set the variable sum, that acts as accumulator, to zero */
mbed_official 5:3762170b6d4d 277 sumReal = 0;
mbed_official 5:3762170b6d4d 278 sumImag = 0;
mbed_official 5:3762170b6d4d 279
mbed_official 5:3762170b6d4d 280 /* Apply loop unrolling and compute 2 MACs simultaneously. */
mbed_official 5:3762170b6d4d 281 colCnt = numColsA >> 1;
mbed_official 5:3762170b6d4d 282
mbed_official 5:3762170b6d4d 283 /* Initiate the pointer pIn1 to point to the starting address of the column being processed */
mbed_official 5:3762170b6d4d 284 pInA = pSrcA->pData + i * 2;
mbed_official 5:3762170b6d4d 285
mbed_official 5:3762170b6d4d 286
mbed_official 5:3762170b6d4d 287 /* matrix multiplication */
mbed_official 5:3762170b6d4d 288 while(colCnt > 0u)
mbed_official 5:3762170b6d4d 289 {
mbed_official 5:3762170b6d4d 290 /* c(m,n) = a(1,1)*b(1,1) + a(1,2) * b(2,1) + .... + a(m,p)*b(p,n) */
mbed_official 5:3762170b6d4d 291
mbed_official 5:3762170b6d4d 292 #ifdef UNALIGNED_SUPPORT_DISABLE
mbed_official 5:3762170b6d4d 293
mbed_official 5:3762170b6d4d 294 /* read real and imag values from pSrcA buffer */
mbed_official 5:3762170b6d4d 295 a = *pInA;
mbed_official 5:3762170b6d4d 296 b = *(pInA + 1u);
mbed_official 5:3762170b6d4d 297 /* read real and imag values from pSrcB buffer */
mbed_official 5:3762170b6d4d 298 c = *pInB;
mbed_official 5:3762170b6d4d 299 d = *(pInB + 1u);
mbed_official 5:3762170b6d4d 300
mbed_official 5:3762170b6d4d 301 /* Multiply and Accumlates */
mbed_official 5:3762170b6d4d 302 sumReal += (q31_t) a *c;
mbed_official 5:3762170b6d4d 303 sumImag += (q31_t) a *d;
mbed_official 5:3762170b6d4d 304 sumReal -= (q31_t) b *d;
mbed_official 5:3762170b6d4d 305 sumImag += (q31_t) b *c;
mbed_official 5:3762170b6d4d 306
mbed_official 5:3762170b6d4d 307 /* read next real and imag values from pSrcA buffer */
mbed_official 5:3762170b6d4d 308 a = *(pInA + 2u);
mbed_official 5:3762170b6d4d 309 b = *(pInA + 3u);
mbed_official 5:3762170b6d4d 310 /* read next real and imag values from pSrcB buffer */
mbed_official 5:3762170b6d4d 311 c = *(pInB + 2u);
mbed_official 5:3762170b6d4d 312 d = *(pInB + 3u);
mbed_official 5:3762170b6d4d 313
mbed_official 5:3762170b6d4d 314 /* update pointer */
mbed_official 5:3762170b6d4d 315 pInA += 4u;
mbed_official 5:3762170b6d4d 316
mbed_official 5:3762170b6d4d 317 /* Multiply and Accumlates */
mbed_official 5:3762170b6d4d 318 sumReal += (q31_t) a *c;
mbed_official 5:3762170b6d4d 319 sumImag += (q31_t) a *d;
mbed_official 5:3762170b6d4d 320 sumReal -= (q31_t) b *d;
mbed_official 5:3762170b6d4d 321 sumImag += (q31_t) b *c;
mbed_official 5:3762170b6d4d 322 /* update pointer */
mbed_official 5:3762170b6d4d 323 pInB += 4u;
mbed_official 5:3762170b6d4d 324 #else
mbed_official 5:3762170b6d4d 325 /* read real and imag values from pSrcA and pSrcB buffer */
mbed_official 5:3762170b6d4d 326 pSourceA = *__SIMD32(pInA)++;
mbed_official 5:3762170b6d4d 327 pSourceB = *__SIMD32(pInB)++;
mbed_official 5:3762170b6d4d 328
mbed_official 5:3762170b6d4d 329 /* Multiply and Accumlates */
mbed_official 5:3762170b6d4d 330 #ifdef ARM_MATH_BIG_ENDIAN
mbed_official 5:3762170b6d4d 331 prod1 = -__SMUSD(pSourceA, pSourceB);
mbed_official 5:3762170b6d4d 332 #else
mbed_official 5:3762170b6d4d 333 prod1 = __SMUSD(pSourceA, pSourceB);
mbed_official 5:3762170b6d4d 334 #endif
mbed_official 5:3762170b6d4d 335 prod2 = __SMUADX(pSourceA, pSourceB);
mbed_official 5:3762170b6d4d 336 sumReal += (q63_t) prod1;
mbed_official 5:3762170b6d4d 337 sumImag += (q63_t) prod2;
mbed_official 5:3762170b6d4d 338
mbed_official 5:3762170b6d4d 339 /* read real and imag values from pSrcA and pSrcB buffer */
mbed_official 5:3762170b6d4d 340 pSourceA = *__SIMD32(pInA)++;
mbed_official 5:3762170b6d4d 341 pSourceB = *__SIMD32(pInB)++;
mbed_official 5:3762170b6d4d 342
mbed_official 5:3762170b6d4d 343 /* Multiply and Accumlates */
mbed_official 5:3762170b6d4d 344 #ifdef ARM_MATH_BIG_ENDIAN
mbed_official 5:3762170b6d4d 345 prod1 = -__SMUSD(pSourceA, pSourceB);
mbed_official 5:3762170b6d4d 346 #else
mbed_official 5:3762170b6d4d 347 prod1 = __SMUSD(pSourceA, pSourceB);
mbed_official 5:3762170b6d4d 348 #endif
mbed_official 5:3762170b6d4d 349 prod2 = __SMUADX(pSourceA, pSourceB);
mbed_official 5:3762170b6d4d 350 sumReal += (q63_t) prod1;
mbed_official 5:3762170b6d4d 351 sumImag += (q63_t) prod2;
mbed_official 5:3762170b6d4d 352
mbed_official 5:3762170b6d4d 353 #endif /* #ifdef UNALIGNED_SUPPORT_DISABLE */
mbed_official 5:3762170b6d4d 354
mbed_official 5:3762170b6d4d 355 /* Decrement the loop counter */
mbed_official 5:3762170b6d4d 356 colCnt--;
mbed_official 5:3762170b6d4d 357 }
mbed_official 5:3762170b6d4d 358
mbed_official 5:3762170b6d4d 359 /* process odd column samples */
mbed_official 5:3762170b6d4d 360 if((numColsA & 0x1u) > 0u)
mbed_official 5:3762170b6d4d 361 {
mbed_official 5:3762170b6d4d 362 /* c(m,n) = a(1,1)*b(1,1) + a(1,2) * b(2,1) + .... + a(m,p)*b(p,n) */
mbed_official 5:3762170b6d4d 363
mbed_official 5:3762170b6d4d 364 #ifdef UNALIGNED_SUPPORT_DISABLE
mbed_official 5:3762170b6d4d 365
mbed_official 5:3762170b6d4d 366 /* read real and imag values from pSrcA and pSrcB buffer */
mbed_official 5:3762170b6d4d 367 a = *pInA++;
mbed_official 5:3762170b6d4d 368 b = *pInA++;
mbed_official 5:3762170b6d4d 369 c = *pInB++;
mbed_official 5:3762170b6d4d 370 d = *pInB++;
mbed_official 5:3762170b6d4d 371
mbed_official 5:3762170b6d4d 372 /* Multiply and Accumlates */
mbed_official 5:3762170b6d4d 373 sumReal += (q31_t) a *c;
mbed_official 5:3762170b6d4d 374 sumImag += (q31_t) a *d;
mbed_official 5:3762170b6d4d 375 sumReal -= (q31_t) b *d;
mbed_official 5:3762170b6d4d 376 sumImag += (q31_t) b *c;
mbed_official 5:3762170b6d4d 377
mbed_official 5:3762170b6d4d 378 #else
mbed_official 5:3762170b6d4d 379 /* read real and imag values from pSrcA and pSrcB buffer */
mbed_official 5:3762170b6d4d 380 pSourceA = *__SIMD32(pInA)++;
mbed_official 5:3762170b6d4d 381 pSourceB = *__SIMD32(pInB)++;
mbed_official 5:3762170b6d4d 382
mbed_official 5:3762170b6d4d 383 /* Multiply and Accumlates */
mbed_official 5:3762170b6d4d 384 #ifdef ARM_MATH_BIG_ENDIAN
mbed_official 5:3762170b6d4d 385 prod1 = -__SMUSD(pSourceA, pSourceB);
mbed_official 5:3762170b6d4d 386 #else
mbed_official 5:3762170b6d4d 387 prod1 = __SMUSD(pSourceA, pSourceB);
mbed_official 5:3762170b6d4d 388 #endif
mbed_official 5:3762170b6d4d 389 prod2 = __SMUADX(pSourceA, pSourceB);
mbed_official 5:3762170b6d4d 390 sumReal += (q63_t) prod1;
mbed_official 5:3762170b6d4d 391 sumImag += (q63_t) prod2;
mbed_official 5:3762170b6d4d 392
mbed_official 5:3762170b6d4d 393 #endif /* #ifdef UNALIGNED_SUPPORT_DISABLE */
mbed_official 5:3762170b6d4d 394
mbed_official 5:3762170b6d4d 395 }
mbed_official 5:3762170b6d4d 396
mbed_official 5:3762170b6d4d 397 /* Saturate and store the result in the destination buffer */
mbed_official 5:3762170b6d4d 398
mbed_official 5:3762170b6d4d 399 *px++ = (q15_t) (__SSAT(sumReal >> 15, 16));
mbed_official 5:3762170b6d4d 400 *px++ = (q15_t) (__SSAT(sumImag >> 15, 16));
mbed_official 5:3762170b6d4d 401
mbed_official 5:3762170b6d4d 402 /* Decrement the column loop counter */
mbed_official 5:3762170b6d4d 403 col--;
mbed_official 5:3762170b6d4d 404
mbed_official 5:3762170b6d4d 405 } while(col > 0u);
mbed_official 5:3762170b6d4d 406
mbed_official 5:3762170b6d4d 407 i = i + numColsA;
mbed_official 5:3762170b6d4d 408
mbed_official 5:3762170b6d4d 409 /* Decrement the row loop counter */
mbed_official 5:3762170b6d4d 410 row--;
mbed_official 5:3762170b6d4d 411
mbed_official 5:3762170b6d4d 412 } while(row > 0u);
mbed_official 5:3762170b6d4d 413
mbed_official 5:3762170b6d4d 414 /* set status as ARM_MATH_SUCCESS */
mbed_official 5:3762170b6d4d 415 status = ARM_MATH_SUCCESS;
mbed_official 5:3762170b6d4d 416 }
mbed_official 5:3762170b6d4d 417
mbed_official 5:3762170b6d4d 418 /* Return to application */
mbed_official 5:3762170b6d4d 419 return (status);
mbed_official 5:3762170b6d4d 420 }
mbed_official 5:3762170b6d4d 421
mbed_official 5:3762170b6d4d 422 /**
mbed_official 5:3762170b6d4d 423 * @} end of MatrixMult group
mbed_official 5:3762170b6d4d 424 */