Fork of mbed-dsp. CMSIS-DSP library of supporting NEON

Dependents:   mbed-os-example-cmsis_dsp_neon

Fork of mbed-dsp by mbed official

Information

Japanese version is available in lower part of this page.
このページの後半に日本語版が用意されています.

CMSIS-DSP of supporting NEON

What is this ?

A library for CMSIS-DSP of supporting NEON.
We supported the NEON to CMSIS-DSP Ver1.4.3(CMSIS V4.1) that ARM supplied, has achieved the processing speed improvement.
If you use the mbed-dsp library, you can use to replace this library.
CMSIS-DSP of supporting NEON is provied as a library.

Library Creation environment

CMSIS-DSP library of supporting NEON was created by the following environment.

  • Compiler
    ARMCC Version 5.03
  • Compile option switch[C Compiler]
   -DARM_MATH_MATRIX_CHECK -DARM_MATH_ROUNDING -O3 -Otime --cpu=Cortex-A9 --littleend --arm 
   --apcs=/interwork --no_unaligned_access --fpu=vfpv3_fp16 --fpmode=fast --apcs=/hardfp 
   --vectorize --asm
  • Compile option switch[Assembler]
   --cpreproc --cpu=Cortex-A9 --littleend --arm --apcs=/interwork --no_unaligned_access 
   --fpu=vfpv3_fp16 --fpmode=fast --apcs=/hardfp


Effects of NEON support

In the data which passes to each function, large size will be expected more effective than small size.
Also if the data is a multiple of 16, effect will be expected in every function in the CMSIS-DSP.


NEON対応CMSIS-DSP

概要

NEON対応したCMSIS-DSPのライブラリです。
ARM社提供のCMSIS-DSP Ver1.4.3(CMSIS V4.1)をターゲットにNEON対応を行ない、処理速度向上を実現しております。
mbed-dspライブラリを使用している場合は、本ライブラリに置き換えて使用することができます。
NEON対応したCMSIS-DSPはライブラリで提供します。

ライブラリ作成環境

NEON対応CMSIS-DSPライブラリは、以下の環境で作成しています。

  • コンパイラ
    ARMCC Version 5.03
  • コンパイルオプションスイッチ[C Compiler]
   -DARM_MATH_MATRIX_CHECK -DARM_MATH_ROUNDING -O3 -Otime --cpu=Cortex-A9 --littleend --arm 
   --apcs=/interwork --no_unaligned_access --fpu=vfpv3_fp16 --fpmode=fast --apcs=/hardfp 
   --vectorize --asm
  • コンパイルオプションスイッチ[Assembler]
   --cpreproc --cpu=Cortex-A9 --littleend --arm --apcs=/interwork --no_unaligned_access 
   --fpu=vfpv3_fp16 --fpmode=fast --apcs=/hardfp


NEON対応による効果について

CMSIS-DSP内の各関数へ渡すデータは、小さいサイズよりも大きいサイズの方が効果が見込めます。
また、16の倍数のデータであれば、CMSIS-DSP内のどの関数でも効果が見込めます。


Committer:
mbed_official
Date:
Mon Jun 23 09:30:09 2014 +0100
Revision:
4:9cee975aadce
Parent:
3:7a284390b0ce
Synchronized with git revision 6e7c7bcec41226f536474daae3c13d49e4c0e865

Full URL: https://github.com/mbedmicro/mbed/commit/6e7c7bcec41226f536474daae3c13d49e4c0e865/

Fix signed unsigned compare in dsp library

Who changed what in which revision?

UserRevisionLine numberNew contents of line
emilmont 1:fdd22bb7aa52 1 /* ----------------------------------------------------------------------
mbed_official 3:7a284390b0ce 2 * Copyright (C) 2010-2013 ARM Limited. All rights reserved.
emilmont 1:fdd22bb7aa52 3 *
mbed_official 3:7a284390b0ce 4 * $Date: 17. January 2013
mbed_official 3:7a284390b0ce 5 * $Revision: V1.4.1
emilmont 1:fdd22bb7aa52 6 *
emilmont 2:da51fb522205 7 * Project: CMSIS DSP Library
emilmont 2:da51fb522205 8 * Title: arm_mat_mult_q15.c
emilmont 1:fdd22bb7aa52 9 *
emilmont 2:da51fb522205 10 * Description: Q15 matrix multiplication.
emilmont 1:fdd22bb7aa52 11 *
emilmont 1:fdd22bb7aa52 12 * Target Processor: Cortex-M4/Cortex-M3/Cortex-M0
emilmont 1:fdd22bb7aa52 13 *
mbed_official 3:7a284390b0ce 14 * Redistribution and use in source and binary forms, with or without
mbed_official 3:7a284390b0ce 15 * modification, are permitted provided that the following conditions
mbed_official 3:7a284390b0ce 16 * are met:
mbed_official 3:7a284390b0ce 17 * - Redistributions of source code must retain the above copyright
mbed_official 3:7a284390b0ce 18 * notice, this list of conditions and the following disclaimer.
mbed_official 3:7a284390b0ce 19 * - Redistributions in binary form must reproduce the above copyright
mbed_official 3:7a284390b0ce 20 * notice, this list of conditions and the following disclaimer in
mbed_official 3:7a284390b0ce 21 * the documentation and/or other materials provided with the
mbed_official 3:7a284390b0ce 22 * distribution.
mbed_official 3:7a284390b0ce 23 * - Neither the name of ARM LIMITED nor the names of its contributors
mbed_official 3:7a284390b0ce 24 * may be used to endorse or promote products derived from this
mbed_official 3:7a284390b0ce 25 * software without specific prior written permission.
mbed_official 3:7a284390b0ce 26 *
mbed_official 3:7a284390b0ce 27 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
mbed_official 3:7a284390b0ce 28 * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
mbed_official 3:7a284390b0ce 29 * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
mbed_official 3:7a284390b0ce 30 * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE
mbed_official 3:7a284390b0ce 31 * COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
mbed_official 3:7a284390b0ce 32 * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING,
mbed_official 3:7a284390b0ce 33 * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
mbed_official 3:7a284390b0ce 34 * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
mbed_official 3:7a284390b0ce 35 * CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
mbed_official 3:7a284390b0ce 36 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN
mbed_official 3:7a284390b0ce 37 * ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
mbed_official 3:7a284390b0ce 38 * POSSIBILITY OF SUCH DAMAGE.
emilmont 1:fdd22bb7aa52 39 * -------------------------------------------------------------------- */
emilmont 1:fdd22bb7aa52 40
emilmont 1:fdd22bb7aa52 41 #include "arm_math.h"
emilmont 1:fdd22bb7aa52 42
emilmont 1:fdd22bb7aa52 43 /**
emilmont 1:fdd22bb7aa52 44 * @ingroup groupMatrix
emilmont 1:fdd22bb7aa52 45 */
emilmont 1:fdd22bb7aa52 46
emilmont 1:fdd22bb7aa52 47 /**
emilmont 1:fdd22bb7aa52 48 * @addtogroup MatrixMult
emilmont 1:fdd22bb7aa52 49 * @{
emilmont 1:fdd22bb7aa52 50 */
emilmont 1:fdd22bb7aa52 51
emilmont 1:fdd22bb7aa52 52
emilmont 1:fdd22bb7aa52 53 /**
emilmont 1:fdd22bb7aa52 54 * @brief Q15 matrix multiplication
emilmont 1:fdd22bb7aa52 55 * @param[in] *pSrcA points to the first input matrix structure
emilmont 1:fdd22bb7aa52 56 * @param[in] *pSrcB points to the second input matrix structure
emilmont 1:fdd22bb7aa52 57 * @param[out] *pDst points to output matrix structure
emilmont 2:da51fb522205 58 * @param[in] *pState points to the array for storing intermediate results
emilmont 2:da51fb522205 59 * @return The function returns either
emilmont 1:fdd22bb7aa52 60 * <code>ARM_MATH_SIZE_MISMATCH</code> or <code>ARM_MATH_SUCCESS</code> based on the outcome of size checking.
emilmont 1:fdd22bb7aa52 61 *
emilmont 1:fdd22bb7aa52 62 * @details
emilmont 1:fdd22bb7aa52 63 * <b>Scaling and Overflow Behavior:</b>
emilmont 1:fdd22bb7aa52 64 *
emilmont 1:fdd22bb7aa52 65 * \par
emilmont 1:fdd22bb7aa52 66 * The function is implemented using a 64-bit internal accumulator. The inputs to the
emilmont 1:fdd22bb7aa52 67 * multiplications are in 1.15 format and multiplications yield a 2.30 result.
emilmont 1:fdd22bb7aa52 68 * The 2.30 intermediate
emilmont 1:fdd22bb7aa52 69 * results are accumulated in a 64-bit accumulator in 34.30 format. This approach
emilmont 1:fdd22bb7aa52 70 * provides 33 guard bits and there is no risk of overflow. The 34.30 result is then
emilmont 1:fdd22bb7aa52 71 * truncated to 34.15 format by discarding the low 15 bits and then saturated to
emilmont 1:fdd22bb7aa52 72 * 1.15 format.
emilmont 1:fdd22bb7aa52 73 *
emilmont 1:fdd22bb7aa52 74 * \par
emilmont 1:fdd22bb7aa52 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.
emilmont 1:fdd22bb7aa52 76 *
emilmont 1:fdd22bb7aa52 77 */
emilmont 1:fdd22bb7aa52 78
emilmont 1:fdd22bb7aa52 79 arm_status arm_mat_mult_q15(
emilmont 1:fdd22bb7aa52 80 const arm_matrix_instance_q15 * pSrcA,
emilmont 1:fdd22bb7aa52 81 const arm_matrix_instance_q15 * pSrcB,
emilmont 1:fdd22bb7aa52 82 arm_matrix_instance_q15 * pDst,
mbed_official 3:7a284390b0ce 83 q15_t * pState CMSIS_UNUSED)
emilmont 1:fdd22bb7aa52 84 {
emilmont 1:fdd22bb7aa52 85 q63_t sum; /* accumulator */
emilmont 1:fdd22bb7aa52 86
mbed_official 3:7a284390b0ce 87 #ifndef ARM_MATH_CM0_FAMILY
emilmont 1:fdd22bb7aa52 88
emilmont 1:fdd22bb7aa52 89 /* Run the below code for Cortex-M4 and Cortex-M3 */
emilmont 1:fdd22bb7aa52 90
emilmont 1:fdd22bb7aa52 91 q15_t *pSrcBT = pState; /* input data matrix pointer for transpose */
emilmont 1:fdd22bb7aa52 92 q15_t *pInA = pSrcA->pData; /* input data matrix pointer A of Q15 type */
emilmont 1:fdd22bb7aa52 93 q15_t *pInB = pSrcB->pData; /* input data matrix pointer B of Q15 type */
emilmont 1:fdd22bb7aa52 94 q15_t *px; /* Temporary output data matrix pointer */
emilmont 1:fdd22bb7aa52 95 uint16_t numRowsA = pSrcA->numRows; /* number of rows of input matrix A */
emilmont 1:fdd22bb7aa52 96 uint16_t numColsB = pSrcB->numCols; /* number of columns of input matrix B */
emilmont 1:fdd22bb7aa52 97 uint16_t numColsA = pSrcA->numCols; /* number of columns of input matrix A */
emilmont 1:fdd22bb7aa52 98 uint16_t numRowsB = pSrcB->numRows; /* number of rows of input matrix A */
emilmont 1:fdd22bb7aa52 99 uint16_t col, i = 0u, row = numRowsB, colCnt; /* loop counters */
emilmont 1:fdd22bb7aa52 100 arm_status status; /* status of matrix multiplication */
emilmont 1:fdd22bb7aa52 101
emilmont 1:fdd22bb7aa52 102 #ifndef UNALIGNED_SUPPORT_DISABLE
emilmont 1:fdd22bb7aa52 103
emilmont 1:fdd22bb7aa52 104 q31_t in; /* Temporary variable to hold the input value */
emilmont 1:fdd22bb7aa52 105 q31_t pSourceA1, pSourceB1, pSourceA2, pSourceB2;
emilmont 1:fdd22bb7aa52 106
emilmont 1:fdd22bb7aa52 107 #else
emilmont 1:fdd22bb7aa52 108
emilmont 1:fdd22bb7aa52 109 q15_t in; /* Temporary variable to hold the input value */
emilmont 1:fdd22bb7aa52 110 q15_t inA1, inB1, inA2, inB2;
emilmont 1:fdd22bb7aa52 111
emilmont 2:da51fb522205 112 #endif /* #ifndef UNALIGNED_SUPPORT_DISABLE */
emilmont 1:fdd22bb7aa52 113
emilmont 1:fdd22bb7aa52 114 #ifdef ARM_MATH_MATRIX_CHECK
emilmont 1:fdd22bb7aa52 115 /* Check for matrix mismatch condition */
emilmont 1:fdd22bb7aa52 116 if((pSrcA->numCols != pSrcB->numRows) ||
emilmont 1:fdd22bb7aa52 117 (pSrcA->numRows != pDst->numRows) || (pSrcB->numCols != pDst->numCols))
emilmont 1:fdd22bb7aa52 118 {
emilmont 1:fdd22bb7aa52 119 /* Set status as ARM_MATH_SIZE_MISMATCH */
emilmont 1:fdd22bb7aa52 120 status = ARM_MATH_SIZE_MISMATCH;
emilmont 1:fdd22bb7aa52 121 }
emilmont 1:fdd22bb7aa52 122 else
emilmont 1:fdd22bb7aa52 123 #endif /* #ifdef ARM_MATH_MATRIX_CHECK */
emilmont 1:fdd22bb7aa52 124 {
emilmont 1:fdd22bb7aa52 125 /* Matrix transpose */
emilmont 1:fdd22bb7aa52 126 do
emilmont 1:fdd22bb7aa52 127 {
emilmont 1:fdd22bb7aa52 128 /* Apply loop unrolling and exchange the columns with row elements */
emilmont 1:fdd22bb7aa52 129 col = numColsB >> 2;
emilmont 1:fdd22bb7aa52 130
emilmont 1:fdd22bb7aa52 131 /* The pointer px is set to starting address of the column being processed */
emilmont 1:fdd22bb7aa52 132 px = pSrcBT + i;
emilmont 1:fdd22bb7aa52 133
emilmont 1:fdd22bb7aa52 134 /* First part of the processing with loop unrolling. Compute 4 outputs at a time.
emilmont 1:fdd22bb7aa52 135 ** a second loop below computes the remaining 1 to 3 samples. */
emilmont 1:fdd22bb7aa52 136 while(col > 0u)
emilmont 1:fdd22bb7aa52 137 {
emilmont 1:fdd22bb7aa52 138 #ifndef UNALIGNED_SUPPORT_DISABLE
emilmont 1:fdd22bb7aa52 139
emilmont 1:fdd22bb7aa52 140 /* Read two elements from the row */
emilmont 1:fdd22bb7aa52 141 in = *__SIMD32(pInB)++;
emilmont 1:fdd22bb7aa52 142
emilmont 1:fdd22bb7aa52 143 /* Unpack and store one element in the destination */
emilmont 1:fdd22bb7aa52 144 #ifndef ARM_MATH_BIG_ENDIAN
emilmont 1:fdd22bb7aa52 145
emilmont 1:fdd22bb7aa52 146 *px = (q15_t) in;
emilmont 1:fdd22bb7aa52 147
emilmont 1:fdd22bb7aa52 148 #else
emilmont 1:fdd22bb7aa52 149
emilmont 1:fdd22bb7aa52 150 *px = (q15_t) ((in & (q31_t) 0xffff0000) >> 16);
emilmont 1:fdd22bb7aa52 151
emilmont 1:fdd22bb7aa52 152 #endif /* #ifndef ARM_MATH_BIG_ENDIAN */
emilmont 1:fdd22bb7aa52 153
emilmont 1:fdd22bb7aa52 154 /* Update the pointer px to point to the next row of the transposed matrix */
emilmont 1:fdd22bb7aa52 155 px += numRowsB;
emilmont 1:fdd22bb7aa52 156
emilmont 1:fdd22bb7aa52 157 /* Unpack and store the second element in the destination */
emilmont 1:fdd22bb7aa52 158 #ifndef ARM_MATH_BIG_ENDIAN
emilmont 1:fdd22bb7aa52 159
emilmont 1:fdd22bb7aa52 160 *px = (q15_t) ((in & (q31_t) 0xffff0000) >> 16);
emilmont 1:fdd22bb7aa52 161
emilmont 1:fdd22bb7aa52 162 #else
emilmont 1:fdd22bb7aa52 163
emilmont 1:fdd22bb7aa52 164 *px = (q15_t) in;
emilmont 1:fdd22bb7aa52 165
emilmont 1:fdd22bb7aa52 166 #endif /* #ifndef ARM_MATH_BIG_ENDIAN */
emilmont 1:fdd22bb7aa52 167
emilmont 1:fdd22bb7aa52 168 /* Update the pointer px to point to the next row of the transposed matrix */
emilmont 1:fdd22bb7aa52 169 px += numRowsB;
emilmont 1:fdd22bb7aa52 170
emilmont 1:fdd22bb7aa52 171 /* Read two elements from the row */
emilmont 1:fdd22bb7aa52 172 in = *__SIMD32(pInB)++;
emilmont 1:fdd22bb7aa52 173
emilmont 1:fdd22bb7aa52 174 /* Unpack and store one element in the destination */
emilmont 1:fdd22bb7aa52 175 #ifndef ARM_MATH_BIG_ENDIAN
emilmont 1:fdd22bb7aa52 176
emilmont 1:fdd22bb7aa52 177 *px = (q15_t) in;
emilmont 1:fdd22bb7aa52 178
emilmont 1:fdd22bb7aa52 179 #else
emilmont 1:fdd22bb7aa52 180
emilmont 1:fdd22bb7aa52 181 *px = (q15_t) ((in & (q31_t) 0xffff0000) >> 16);
emilmont 1:fdd22bb7aa52 182
emilmont 1:fdd22bb7aa52 183 #endif /* #ifndef ARM_MATH_BIG_ENDIAN */
emilmont 1:fdd22bb7aa52 184
emilmont 1:fdd22bb7aa52 185 /* Update the pointer px to point to the next row of the transposed matrix */
emilmont 1:fdd22bb7aa52 186 px += numRowsB;
emilmont 1:fdd22bb7aa52 187
emilmont 1:fdd22bb7aa52 188 /* Unpack and store the second element in the destination */
emilmont 1:fdd22bb7aa52 189
emilmont 1:fdd22bb7aa52 190 #ifndef ARM_MATH_BIG_ENDIAN
emilmont 1:fdd22bb7aa52 191
emilmont 1:fdd22bb7aa52 192 *px = (q15_t) ((in & (q31_t) 0xffff0000) >> 16);
emilmont 1:fdd22bb7aa52 193
emilmont 1:fdd22bb7aa52 194 #else
emilmont 1:fdd22bb7aa52 195
emilmont 1:fdd22bb7aa52 196 *px = (q15_t) in;
emilmont 1:fdd22bb7aa52 197
emilmont 1:fdd22bb7aa52 198 #endif /* #ifndef ARM_MATH_BIG_ENDIAN */
emilmont 1:fdd22bb7aa52 199
emilmont 1:fdd22bb7aa52 200 /* Update the pointer px to point to the next row of the transposed matrix */
emilmont 1:fdd22bb7aa52 201 px += numRowsB;
emilmont 1:fdd22bb7aa52 202
emilmont 1:fdd22bb7aa52 203 #else
emilmont 1:fdd22bb7aa52 204
emilmont 1:fdd22bb7aa52 205 /* Read one element from the row */
emilmont 1:fdd22bb7aa52 206 in = *pInB++;
emilmont 1:fdd22bb7aa52 207
emilmont 1:fdd22bb7aa52 208 /* Store one element in the destination */
emilmont 1:fdd22bb7aa52 209 *px = in;
emilmont 1:fdd22bb7aa52 210
emilmont 1:fdd22bb7aa52 211 /* Update the pointer px to point to the next row of the transposed matrix */
emilmont 1:fdd22bb7aa52 212 px += numRowsB;
emilmont 1:fdd22bb7aa52 213
emilmont 1:fdd22bb7aa52 214 /* Read one element from the row */
emilmont 1:fdd22bb7aa52 215 in = *pInB++;
emilmont 1:fdd22bb7aa52 216
emilmont 1:fdd22bb7aa52 217 /* Store one element in the destination */
emilmont 1:fdd22bb7aa52 218 *px = in;
emilmont 1:fdd22bb7aa52 219
emilmont 1:fdd22bb7aa52 220 /* Update the pointer px to point to the next row of the transposed matrix */
emilmont 1:fdd22bb7aa52 221 px += numRowsB;
emilmont 1:fdd22bb7aa52 222
emilmont 1:fdd22bb7aa52 223 /* Read one element from the row */
emilmont 1:fdd22bb7aa52 224 in = *pInB++;
emilmont 1:fdd22bb7aa52 225
emilmont 1:fdd22bb7aa52 226 /* Store one element in the destination */
emilmont 1:fdd22bb7aa52 227 *px = in;
emilmont 1:fdd22bb7aa52 228
emilmont 1:fdd22bb7aa52 229 /* Update the pointer px to point to the next row of the transposed matrix */
emilmont 1:fdd22bb7aa52 230 px += numRowsB;
emilmont 1:fdd22bb7aa52 231
emilmont 1:fdd22bb7aa52 232 /* Read one element from the row */
emilmont 1:fdd22bb7aa52 233 in = *pInB++;
emilmont 1:fdd22bb7aa52 234
emilmont 1:fdd22bb7aa52 235 /* Store one element in the destination */
emilmont 1:fdd22bb7aa52 236 *px = in;
emilmont 1:fdd22bb7aa52 237
emilmont 1:fdd22bb7aa52 238 /* Update the pointer px to point to the next row of the transposed matrix */
emilmont 1:fdd22bb7aa52 239 px += numRowsB;
emilmont 1:fdd22bb7aa52 240
emilmont 2:da51fb522205 241 #endif /* #ifndef UNALIGNED_SUPPORT_DISABLE */
emilmont 1:fdd22bb7aa52 242
emilmont 1:fdd22bb7aa52 243 /* Decrement the column loop counter */
emilmont 1:fdd22bb7aa52 244 col--;
emilmont 1:fdd22bb7aa52 245 }
emilmont 1:fdd22bb7aa52 246
emilmont 1:fdd22bb7aa52 247 /* If the columns of pSrcB is not a multiple of 4, compute any remaining output samples here.
emilmont 1:fdd22bb7aa52 248 ** No loop unrolling is used. */
emilmont 1:fdd22bb7aa52 249 col = numColsB % 0x4u;
emilmont 1:fdd22bb7aa52 250
emilmont 1:fdd22bb7aa52 251 while(col > 0u)
emilmont 1:fdd22bb7aa52 252 {
emilmont 1:fdd22bb7aa52 253 /* Read and store the input element in the destination */
emilmont 1:fdd22bb7aa52 254 *px = *pInB++;
emilmont 1:fdd22bb7aa52 255
emilmont 1:fdd22bb7aa52 256 /* Update the pointer px to point to the next row of the transposed matrix */
emilmont 1:fdd22bb7aa52 257 px += numRowsB;
emilmont 1:fdd22bb7aa52 258
emilmont 1:fdd22bb7aa52 259 /* Decrement the column loop counter */
emilmont 1:fdd22bb7aa52 260 col--;
emilmont 1:fdd22bb7aa52 261 }
emilmont 1:fdd22bb7aa52 262
emilmont 1:fdd22bb7aa52 263 i++;
emilmont 1:fdd22bb7aa52 264
emilmont 1:fdd22bb7aa52 265 /* Decrement the row loop counter */
emilmont 1:fdd22bb7aa52 266 row--;
emilmont 1:fdd22bb7aa52 267
emilmont 1:fdd22bb7aa52 268 } while(row > 0u);
emilmont 1:fdd22bb7aa52 269
emilmont 1:fdd22bb7aa52 270 /* Reset the variables for the usage in the following multiplication process */
emilmont 1:fdd22bb7aa52 271 row = numRowsA;
emilmont 1:fdd22bb7aa52 272 i = 0u;
emilmont 1:fdd22bb7aa52 273 px = pDst->pData;
emilmont 1:fdd22bb7aa52 274
emilmont 1:fdd22bb7aa52 275 /* The following loop performs the dot-product of each row in pSrcA with each column in pSrcB */
emilmont 1:fdd22bb7aa52 276 /* row loop */
emilmont 1:fdd22bb7aa52 277 do
emilmont 1:fdd22bb7aa52 278 {
emilmont 1:fdd22bb7aa52 279 /* For every row wise process, the column loop counter is to be initiated */
emilmont 1:fdd22bb7aa52 280 col = numColsB;
emilmont 1:fdd22bb7aa52 281
emilmont 1:fdd22bb7aa52 282 /* For every row wise process, the pIn2 pointer is set
emilmont 1:fdd22bb7aa52 283 ** to the starting address of the transposed pSrcB data */
emilmont 1:fdd22bb7aa52 284 pInB = pSrcBT;
emilmont 1:fdd22bb7aa52 285
emilmont 1:fdd22bb7aa52 286 /* column loop */
emilmont 1:fdd22bb7aa52 287 do
emilmont 1:fdd22bb7aa52 288 {
emilmont 1:fdd22bb7aa52 289 /* Set the variable sum, that acts as accumulator, to zero */
emilmont 1:fdd22bb7aa52 290 sum = 0;
emilmont 1:fdd22bb7aa52 291
emilmont 1:fdd22bb7aa52 292 /* Apply loop unrolling and compute 2 MACs simultaneously. */
emilmont 1:fdd22bb7aa52 293 colCnt = numColsA >> 2;
emilmont 1:fdd22bb7aa52 294
emilmont 1:fdd22bb7aa52 295 /* Initiate the pointer pIn1 to point to the starting address of the column being processed */
emilmont 1:fdd22bb7aa52 296 pInA = pSrcA->pData + i;
emilmont 1:fdd22bb7aa52 297
emilmont 1:fdd22bb7aa52 298
emilmont 1:fdd22bb7aa52 299 /* matrix multiplication */
emilmont 1:fdd22bb7aa52 300 while(colCnt > 0u)
emilmont 1:fdd22bb7aa52 301 {
emilmont 1:fdd22bb7aa52 302 /* c(m,n) = a(1,1)*b(1,1) + a(1,2) * b(2,1) + .... + a(m,p)*b(p,n) */
emilmont 1:fdd22bb7aa52 303 #ifndef UNALIGNED_SUPPORT_DISABLE
emilmont 1:fdd22bb7aa52 304
emilmont 1:fdd22bb7aa52 305 /* read real and imag values from pSrcA and pSrcB buffer */
emilmont 1:fdd22bb7aa52 306 pSourceA1 = *__SIMD32(pInA)++;
emilmont 1:fdd22bb7aa52 307 pSourceB1 = *__SIMD32(pInB)++;
emilmont 1:fdd22bb7aa52 308
emilmont 1:fdd22bb7aa52 309 pSourceA2 = *__SIMD32(pInA)++;
emilmont 1:fdd22bb7aa52 310 pSourceB2 = *__SIMD32(pInB)++;
emilmont 1:fdd22bb7aa52 311
emilmont 1:fdd22bb7aa52 312 /* Multiply and Accumlates */
emilmont 1:fdd22bb7aa52 313 sum = __SMLALD(pSourceA1, pSourceB1, sum);
emilmont 1:fdd22bb7aa52 314 sum = __SMLALD(pSourceA2, pSourceB2, sum);
emilmont 1:fdd22bb7aa52 315
emilmont 1:fdd22bb7aa52 316 #else
emilmont 1:fdd22bb7aa52 317 /* read real and imag values from pSrcA and pSrcB buffer */
emilmont 1:fdd22bb7aa52 318 inA1 = *pInA++;
emilmont 1:fdd22bb7aa52 319 inB1 = *pInB++;
emilmont 1:fdd22bb7aa52 320 inA2 = *pInA++;
emilmont 1:fdd22bb7aa52 321 /* Multiply and Accumlates */
emilmont 1:fdd22bb7aa52 322 sum += inA1 * inB1;
emilmont 1:fdd22bb7aa52 323 inB2 = *pInB++;
emilmont 1:fdd22bb7aa52 324
emilmont 1:fdd22bb7aa52 325 inA1 = *pInA++;
emilmont 1:fdd22bb7aa52 326 inB1 = *pInB++;
emilmont 1:fdd22bb7aa52 327 /* Multiply and Accumlates */
emilmont 1:fdd22bb7aa52 328 sum += inA2 * inB2;
emilmont 1:fdd22bb7aa52 329 inA2 = *pInA++;
emilmont 1:fdd22bb7aa52 330 inB2 = *pInB++;
emilmont 1:fdd22bb7aa52 331
emilmont 1:fdd22bb7aa52 332 /* Multiply and Accumlates */
emilmont 1:fdd22bb7aa52 333 sum += inA1 * inB1;
emilmont 1:fdd22bb7aa52 334 sum += inA2 * inB2;
emilmont 1:fdd22bb7aa52 335
emilmont 2:da51fb522205 336 #endif /* #ifndef UNALIGNED_SUPPORT_DISABLE */
emilmont 1:fdd22bb7aa52 337
emilmont 1:fdd22bb7aa52 338 /* Decrement the loop counter */
emilmont 1:fdd22bb7aa52 339 colCnt--;
emilmont 1:fdd22bb7aa52 340 }
emilmont 1:fdd22bb7aa52 341
emilmont 1:fdd22bb7aa52 342 /* process remaining column samples */
emilmont 1:fdd22bb7aa52 343 colCnt = numColsA & 3u;
emilmont 1:fdd22bb7aa52 344
emilmont 1:fdd22bb7aa52 345 while(colCnt > 0u)
emilmont 1:fdd22bb7aa52 346 {
emilmont 1:fdd22bb7aa52 347 /* c(m,n) = a(1,1)*b(1,1) + a(1,2) * b(2,1) + .... + a(m,p)*b(p,n) */
emilmont 1:fdd22bb7aa52 348 sum += *pInA++ * *pInB++;
emilmont 1:fdd22bb7aa52 349
emilmont 1:fdd22bb7aa52 350 /* Decrement the loop counter */
emilmont 1:fdd22bb7aa52 351 colCnt--;
emilmont 1:fdd22bb7aa52 352 }
emilmont 1:fdd22bb7aa52 353
emilmont 1:fdd22bb7aa52 354 /* Saturate and store the result in the destination buffer */
emilmont 1:fdd22bb7aa52 355 *px = (q15_t) (__SSAT((sum >> 15), 16));
emilmont 1:fdd22bb7aa52 356 px++;
emilmont 1:fdd22bb7aa52 357
emilmont 1:fdd22bb7aa52 358 /* Decrement the column loop counter */
emilmont 1:fdd22bb7aa52 359 col--;
emilmont 1:fdd22bb7aa52 360
emilmont 1:fdd22bb7aa52 361 } while(col > 0u);
emilmont 1:fdd22bb7aa52 362
emilmont 1:fdd22bb7aa52 363 i = i + numColsA;
emilmont 1:fdd22bb7aa52 364
emilmont 1:fdd22bb7aa52 365 /* Decrement the row loop counter */
emilmont 1:fdd22bb7aa52 366 row--;
emilmont 1:fdd22bb7aa52 367
emilmont 1:fdd22bb7aa52 368 } while(row > 0u);
emilmont 1:fdd22bb7aa52 369
emilmont 1:fdd22bb7aa52 370 #else
emilmont 1:fdd22bb7aa52 371
emilmont 1:fdd22bb7aa52 372 /* Run the below code for Cortex-M0 */
emilmont 1:fdd22bb7aa52 373
emilmont 1:fdd22bb7aa52 374 q15_t *pIn1 = pSrcA->pData; /* input data matrix pointer A */
emilmont 1:fdd22bb7aa52 375 q15_t *pIn2 = pSrcB->pData; /* input data matrix pointer B */
emilmont 1:fdd22bb7aa52 376 q15_t *pInA = pSrcA->pData; /* input data matrix pointer A of Q15 type */
emilmont 1:fdd22bb7aa52 377 q15_t *pInB = pSrcB->pData; /* input data matrix pointer B of Q15 type */
emilmont 1:fdd22bb7aa52 378 q15_t *pOut = pDst->pData; /* output data matrix pointer */
emilmont 1:fdd22bb7aa52 379 q15_t *px; /* Temporary output data matrix pointer */
emilmont 1:fdd22bb7aa52 380 uint16_t numColsB = pSrcB->numCols; /* number of columns of input matrix B */
emilmont 1:fdd22bb7aa52 381 uint16_t numColsA = pSrcA->numCols; /* number of columns of input matrix A */
emilmont 1:fdd22bb7aa52 382 uint16_t numRowsA = pSrcA->numRows; /* number of rows of input matrix A */
emilmont 1:fdd22bb7aa52 383 uint16_t col, i = 0u, row = numRowsA, colCnt; /* loop counters */
emilmont 1:fdd22bb7aa52 384 arm_status status; /* status of matrix multiplication */
emilmont 1:fdd22bb7aa52 385
emilmont 1:fdd22bb7aa52 386 #ifdef ARM_MATH_MATRIX_CHECK
emilmont 1:fdd22bb7aa52 387
emilmont 1:fdd22bb7aa52 388 /* Check for matrix mismatch condition */
emilmont 1:fdd22bb7aa52 389 if((pSrcA->numCols != pSrcB->numRows) ||
emilmont 1:fdd22bb7aa52 390 (pSrcA->numRows != pDst->numRows) || (pSrcB->numCols != pDst->numCols))
emilmont 1:fdd22bb7aa52 391 {
emilmont 1:fdd22bb7aa52 392 /* Set status as ARM_MATH_SIZE_MISMATCH */
emilmont 1:fdd22bb7aa52 393 status = ARM_MATH_SIZE_MISMATCH;
emilmont 1:fdd22bb7aa52 394 }
emilmont 1:fdd22bb7aa52 395 else
emilmont 1:fdd22bb7aa52 396 #endif /* #ifdef ARM_MATH_MATRIX_CHECK */
emilmont 1:fdd22bb7aa52 397
emilmont 1:fdd22bb7aa52 398 {
emilmont 1:fdd22bb7aa52 399 /* The following loop performs the dot-product of each row in pSrcA with each column in pSrcB */
emilmont 1:fdd22bb7aa52 400 /* row loop */
emilmont 1:fdd22bb7aa52 401 do
emilmont 1:fdd22bb7aa52 402 {
emilmont 1:fdd22bb7aa52 403 /* Output pointer is set to starting address of the row being processed */
emilmont 1:fdd22bb7aa52 404 px = pOut + i;
emilmont 1:fdd22bb7aa52 405
emilmont 1:fdd22bb7aa52 406 /* For every row wise process, the column loop counter is to be initiated */
emilmont 1:fdd22bb7aa52 407 col = numColsB;
emilmont 1:fdd22bb7aa52 408
emilmont 1:fdd22bb7aa52 409 /* For every row wise process, the pIn2 pointer is set
emilmont 1:fdd22bb7aa52 410 ** to the starting address of the pSrcB data */
emilmont 1:fdd22bb7aa52 411 pIn2 = pSrcB->pData;
emilmont 1:fdd22bb7aa52 412
emilmont 1:fdd22bb7aa52 413 /* column loop */
emilmont 1:fdd22bb7aa52 414 do
emilmont 1:fdd22bb7aa52 415 {
emilmont 1:fdd22bb7aa52 416 /* Set the variable sum, that acts as accumulator, to zero */
emilmont 1:fdd22bb7aa52 417 sum = 0;
emilmont 1:fdd22bb7aa52 418
emilmont 1:fdd22bb7aa52 419 /* Initiate the pointer pIn1 to point to the starting address of pSrcA */
emilmont 1:fdd22bb7aa52 420 pIn1 = pInA;
emilmont 1:fdd22bb7aa52 421
emilmont 1:fdd22bb7aa52 422 /* Matrix A columns number of MAC operations are to be performed */
emilmont 1:fdd22bb7aa52 423 colCnt = numColsA;
emilmont 1:fdd22bb7aa52 424
emilmont 1:fdd22bb7aa52 425 /* matrix multiplication */
emilmont 1:fdd22bb7aa52 426 while(colCnt > 0u)
emilmont 1:fdd22bb7aa52 427 {
emilmont 1:fdd22bb7aa52 428 /* c(m,n) = a(1,1)*b(1,1) + a(1,2) * b(2,1) + .... + a(m,p)*b(p,n) */
emilmont 1:fdd22bb7aa52 429 /* Perform the multiply-accumulates */
emilmont 1:fdd22bb7aa52 430 sum += (q31_t) * pIn1++ * *pIn2;
emilmont 1:fdd22bb7aa52 431 pIn2 += numColsB;
emilmont 1:fdd22bb7aa52 432
emilmont 1:fdd22bb7aa52 433 /* Decrement the loop counter */
emilmont 1:fdd22bb7aa52 434 colCnt--;
emilmont 1:fdd22bb7aa52 435 }
emilmont 1:fdd22bb7aa52 436
emilmont 1:fdd22bb7aa52 437 /* Convert the result from 34.30 to 1.15 format and store the saturated value in destination buffer */
emilmont 1:fdd22bb7aa52 438 /* Saturate and store the result in the destination buffer */
emilmont 1:fdd22bb7aa52 439 *px++ = (q15_t) __SSAT((sum >> 15), 16);
emilmont 1:fdd22bb7aa52 440
emilmont 1:fdd22bb7aa52 441 /* Decrement the column loop counter */
emilmont 1:fdd22bb7aa52 442 col--;
emilmont 1:fdd22bb7aa52 443
emilmont 1:fdd22bb7aa52 444 /* Update the pointer pIn2 to point to the starting address of the next column */
emilmont 1:fdd22bb7aa52 445 pIn2 = pInB + (numColsB - col);
emilmont 1:fdd22bb7aa52 446
emilmont 1:fdd22bb7aa52 447 } while(col > 0u);
emilmont 1:fdd22bb7aa52 448
emilmont 1:fdd22bb7aa52 449 /* Update the pointer pSrcA to point to the starting address of the next row */
emilmont 1:fdd22bb7aa52 450 i = i + numColsB;
emilmont 1:fdd22bb7aa52 451 pInA = pInA + numColsA;
emilmont 1:fdd22bb7aa52 452
emilmont 1:fdd22bb7aa52 453 /* Decrement the row loop counter */
emilmont 1:fdd22bb7aa52 454 row--;
emilmont 1:fdd22bb7aa52 455
emilmont 1:fdd22bb7aa52 456 } while(row > 0u);
emilmont 1:fdd22bb7aa52 457
mbed_official 3:7a284390b0ce 458 #endif /* #ifndef ARM_MATH_CM0_FAMILY */
emilmont 1:fdd22bb7aa52 459 /* set status as ARM_MATH_SUCCESS */
emilmont 1:fdd22bb7aa52 460 status = ARM_MATH_SUCCESS;
emilmont 1:fdd22bb7aa52 461 }
emilmont 1:fdd22bb7aa52 462
emilmont 1:fdd22bb7aa52 463 /* Return to application */
emilmont 1:fdd22bb7aa52 464 return (status);
emilmont 1:fdd22bb7aa52 465 }
emilmont 1:fdd22bb7aa52 466
emilmont 1:fdd22bb7aa52 467 /**
emilmont 1:fdd22bb7aa52 468 * @} end of MatrixMult group
emilmont 1:fdd22bb7aa52 469 */