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
mbed_official 3:7a284390b0ce 1 /* ----------------------------------------------------------------------
mbed_official 3:7a284390b0ce 2 * Copyright (C) 2010-2013 ARM Limited. All rights reserved.
mbed_official 3:7a284390b0ce 3 *
mbed_official 3:7a284390b0ce 4 * $Date: 17. January 2013
mbed_official 3:7a284390b0ce 5 * $Revision: V1.4.1
mbed_official 3:7a284390b0ce 6 *
mbed_official 3:7a284390b0ce 7 * Project: CMSIS DSP Library
mbed_official 3:7a284390b0ce 8 * Title: arm_scale_q31.c
mbed_official 3:7a284390b0ce 9 *
mbed_official 3:7a284390b0ce 10 * Description: Multiplies a Q31 vector by a scalar.
mbed_official 3:7a284390b0ce 11 *
emilmont 1:fdd22bb7aa52 12 * Target Processor: Cortex-M4/Cortex-M3/Cortex-M0
mbed_official 3:7a284390b0ce 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.
emilmont 1:fdd22bb7aa52 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
mbed_official 3:7a284390b0ce 43 /**
mbed_official 3:7a284390b0ce 44 * @ingroup groupMath
emilmont 1:fdd22bb7aa52 45 */
emilmont 1:fdd22bb7aa52 46
mbed_official 3:7a284390b0ce 47 /**
mbed_official 3:7a284390b0ce 48 * @addtogroup scale
mbed_official 3:7a284390b0ce 49 * @{
emilmont 1:fdd22bb7aa52 50 */
emilmont 1:fdd22bb7aa52 51
mbed_official 3:7a284390b0ce 52 /**
mbed_official 3:7a284390b0ce 53 * @brief Multiplies a Q31 vector by a scalar.
mbed_official 3:7a284390b0ce 54 * @param[in] *pSrc points to the input vector
mbed_official 3:7a284390b0ce 55 * @param[in] scaleFract fractional portion of the scale value
mbed_official 3:7a284390b0ce 56 * @param[in] shift number of bits to shift the result by
mbed_official 3:7a284390b0ce 57 * @param[out] *pDst points to the output vector
mbed_official 3:7a284390b0ce 58 * @param[in] blockSize number of samples in the vector
mbed_official 3:7a284390b0ce 59 * @return none.
mbed_official 3:7a284390b0ce 60 *
mbed_official 3:7a284390b0ce 61 * <b>Scaling and Overflow Behavior:</b>
mbed_official 3:7a284390b0ce 62 * \par
mbed_official 3:7a284390b0ce 63 * The input data <code>*pSrc</code> and <code>scaleFract</code> are in 1.31 format.
mbed_official 3:7a284390b0ce 64 * These are multiplied to yield a 2.62 intermediate result and this is shifted with saturation to 1.31 format.
emilmont 1:fdd22bb7aa52 65 */
emilmont 1:fdd22bb7aa52 66
emilmont 1:fdd22bb7aa52 67 void arm_scale_q31(
emilmont 1:fdd22bb7aa52 68 q31_t * pSrc,
emilmont 1:fdd22bb7aa52 69 q31_t scaleFract,
emilmont 1:fdd22bb7aa52 70 int8_t shift,
emilmont 1:fdd22bb7aa52 71 q31_t * pDst,
emilmont 1:fdd22bb7aa52 72 uint32_t blockSize)
emilmont 1:fdd22bb7aa52 73 {
emilmont 1:fdd22bb7aa52 74 int8_t kShift = shift + 1; /* Shift to apply after scaling */
emilmont 1:fdd22bb7aa52 75 int8_t sign = (kShift & 0x80);
emilmont 1:fdd22bb7aa52 76 uint32_t blkCnt; /* loop counter */
emilmont 1:fdd22bb7aa52 77 q31_t in, out;
emilmont 1:fdd22bb7aa52 78
mbed_official 3:7a284390b0ce 79 #ifndef ARM_MATH_CM0_FAMILY
emilmont 1:fdd22bb7aa52 80
emilmont 1:fdd22bb7aa52 81 /* Run the below code for Cortex-M4 and Cortex-M3 */
emilmont 1:fdd22bb7aa52 82
emilmont 1:fdd22bb7aa52 83 q31_t in1, in2, in3, in4; /* temporary input variables */
emilmont 1:fdd22bb7aa52 84 q31_t out1, out2, out3, out4; /* temporary output variabels */
emilmont 1:fdd22bb7aa52 85
emilmont 1:fdd22bb7aa52 86
emilmont 1:fdd22bb7aa52 87 /*loop Unrolling */
emilmont 1:fdd22bb7aa52 88 blkCnt = blockSize >> 2u;
emilmont 1:fdd22bb7aa52 89
emilmont 1:fdd22bb7aa52 90 if(sign == 0u)
emilmont 1:fdd22bb7aa52 91 {
mbed_official 3:7a284390b0ce 92 /* First part of the processing with loop unrolling. Compute 4 outputs at a time.
emilmont 1:fdd22bb7aa52 93 ** a second loop below computes the remaining 1 to 3 samples. */
emilmont 1:fdd22bb7aa52 94 while(blkCnt > 0u)
emilmont 1:fdd22bb7aa52 95 {
emilmont 1:fdd22bb7aa52 96 /* read four inputs from source */
emilmont 1:fdd22bb7aa52 97 in1 = *pSrc;
emilmont 1:fdd22bb7aa52 98 in2 = *(pSrc + 1);
emilmont 1:fdd22bb7aa52 99 in3 = *(pSrc + 2);
emilmont 1:fdd22bb7aa52 100 in4 = *(pSrc + 3);
emilmont 1:fdd22bb7aa52 101
emilmont 1:fdd22bb7aa52 102 /* multiply input with scaler value */
emilmont 1:fdd22bb7aa52 103 in1 = ((q63_t) in1 * scaleFract) >> 32;
emilmont 1:fdd22bb7aa52 104 in2 = ((q63_t) in2 * scaleFract) >> 32;
emilmont 1:fdd22bb7aa52 105 in3 = ((q63_t) in3 * scaleFract) >> 32;
emilmont 1:fdd22bb7aa52 106 in4 = ((q63_t) in4 * scaleFract) >> 32;
emilmont 1:fdd22bb7aa52 107
emilmont 1:fdd22bb7aa52 108 /* apply shifting */
emilmont 1:fdd22bb7aa52 109 out1 = in1 << kShift;
emilmont 1:fdd22bb7aa52 110 out2 = in2 << kShift;
emilmont 1:fdd22bb7aa52 111
emilmont 1:fdd22bb7aa52 112 /* saturate the results. */
emilmont 1:fdd22bb7aa52 113 if(in1 != (out1 >> kShift))
emilmont 1:fdd22bb7aa52 114 out1 = 0x7FFFFFFF ^ (in1 >> 31);
emilmont 1:fdd22bb7aa52 115
emilmont 1:fdd22bb7aa52 116 if(in2 != (out2 >> kShift))
emilmont 1:fdd22bb7aa52 117 out2 = 0x7FFFFFFF ^ (in2 >> 31);
emilmont 1:fdd22bb7aa52 118
emilmont 1:fdd22bb7aa52 119 out3 = in3 << kShift;
emilmont 1:fdd22bb7aa52 120 out4 = in4 << kShift;
emilmont 1:fdd22bb7aa52 121
emilmont 1:fdd22bb7aa52 122 *pDst = out1;
emilmont 1:fdd22bb7aa52 123 *(pDst + 1) = out2;
emilmont 1:fdd22bb7aa52 124
emilmont 1:fdd22bb7aa52 125 if(in3 != (out3 >> kShift))
emilmont 1:fdd22bb7aa52 126 out3 = 0x7FFFFFFF ^ (in3 >> 31);
emilmont 1:fdd22bb7aa52 127
emilmont 1:fdd22bb7aa52 128 if(in4 != (out4 >> kShift))
emilmont 1:fdd22bb7aa52 129 out4 = 0x7FFFFFFF ^ (in4 >> 31);
emilmont 1:fdd22bb7aa52 130
emilmont 1:fdd22bb7aa52 131 /* Store result destination */
emilmont 1:fdd22bb7aa52 132 *(pDst + 2) = out3;
emilmont 1:fdd22bb7aa52 133 *(pDst + 3) = out4;
emilmont 1:fdd22bb7aa52 134
emilmont 1:fdd22bb7aa52 135 /* Update pointers to process next sampels */
emilmont 1:fdd22bb7aa52 136 pSrc += 4u;
emilmont 1:fdd22bb7aa52 137 pDst += 4u;
emilmont 1:fdd22bb7aa52 138
emilmont 1:fdd22bb7aa52 139 /* Decrement the loop counter */
emilmont 1:fdd22bb7aa52 140 blkCnt--;
emilmont 1:fdd22bb7aa52 141 }
emilmont 1:fdd22bb7aa52 142
emilmont 1:fdd22bb7aa52 143 }
emilmont 1:fdd22bb7aa52 144 else
emilmont 1:fdd22bb7aa52 145 {
mbed_official 3:7a284390b0ce 146 /* First part of the processing with loop unrolling. Compute 4 outputs at a time.
emilmont 1:fdd22bb7aa52 147 ** a second loop below computes the remaining 1 to 3 samples. */
emilmont 1:fdd22bb7aa52 148 while(blkCnt > 0u)
emilmont 1:fdd22bb7aa52 149 {
emilmont 1:fdd22bb7aa52 150 /* read four inputs from source */
emilmont 1:fdd22bb7aa52 151 in1 = *pSrc;
emilmont 1:fdd22bb7aa52 152 in2 = *(pSrc + 1);
emilmont 1:fdd22bb7aa52 153 in3 = *(pSrc + 2);
emilmont 1:fdd22bb7aa52 154 in4 = *(pSrc + 3);
emilmont 1:fdd22bb7aa52 155
emilmont 1:fdd22bb7aa52 156 /* multiply input with scaler value */
emilmont 1:fdd22bb7aa52 157 in1 = ((q63_t) in1 * scaleFract) >> 32;
emilmont 1:fdd22bb7aa52 158 in2 = ((q63_t) in2 * scaleFract) >> 32;
emilmont 1:fdd22bb7aa52 159 in3 = ((q63_t) in3 * scaleFract) >> 32;
emilmont 1:fdd22bb7aa52 160 in4 = ((q63_t) in4 * scaleFract) >> 32;
emilmont 1:fdd22bb7aa52 161
emilmont 1:fdd22bb7aa52 162 /* apply shifting */
mbed_official 3:7a284390b0ce 163 out1 = in1 >> -kShift;
mbed_official 3:7a284390b0ce 164 out2 = in2 >> -kShift;
emilmont 1:fdd22bb7aa52 165
mbed_official 3:7a284390b0ce 166 out3 = in3 >> -kShift;
mbed_official 3:7a284390b0ce 167 out4 = in4 >> -kShift;
emilmont 1:fdd22bb7aa52 168
emilmont 1:fdd22bb7aa52 169 /* Store result destination */
emilmont 1:fdd22bb7aa52 170 *pDst = out1;
emilmont 1:fdd22bb7aa52 171 *(pDst + 1) = out2;
emilmont 1:fdd22bb7aa52 172
emilmont 1:fdd22bb7aa52 173 *(pDst + 2) = out3;
emilmont 1:fdd22bb7aa52 174 *(pDst + 3) = out4;
emilmont 1:fdd22bb7aa52 175
emilmont 1:fdd22bb7aa52 176 /* Update pointers to process next sampels */
emilmont 1:fdd22bb7aa52 177 pSrc += 4u;
emilmont 1:fdd22bb7aa52 178 pDst += 4u;
emilmont 1:fdd22bb7aa52 179
emilmont 1:fdd22bb7aa52 180 /* Decrement the loop counter */
emilmont 1:fdd22bb7aa52 181 blkCnt--;
emilmont 1:fdd22bb7aa52 182 }
emilmont 1:fdd22bb7aa52 183 }
mbed_official 3:7a284390b0ce 184 /* If the blockSize is not a multiple of 4, compute any remaining output samples here.
emilmont 1:fdd22bb7aa52 185 ** No loop unrolling is used. */
emilmont 1:fdd22bb7aa52 186 blkCnt = blockSize % 0x4u;
emilmont 1:fdd22bb7aa52 187
emilmont 1:fdd22bb7aa52 188 #else
emilmont 1:fdd22bb7aa52 189
emilmont 1:fdd22bb7aa52 190 /* Run the below code for Cortex-M0 */
emilmont 1:fdd22bb7aa52 191
emilmont 1:fdd22bb7aa52 192 /* Initialize blkCnt with number of samples */
emilmont 1:fdd22bb7aa52 193 blkCnt = blockSize;
emilmont 1:fdd22bb7aa52 194
mbed_official 3:7a284390b0ce 195 #endif /* #ifndef ARM_MATH_CM0_FAMILY */
emilmont 1:fdd22bb7aa52 196
mbed_official 3:7a284390b0ce 197 if(sign == 0)
emilmont 1:fdd22bb7aa52 198 {
mbed_official 3:7a284390b0ce 199 while(blkCnt > 0u)
mbed_official 3:7a284390b0ce 200 {
mbed_official 3:7a284390b0ce 201 /* C = A * scale */
mbed_official 3:7a284390b0ce 202 /* Scale the input and then store the result in the destination buffer. */
mbed_official 3:7a284390b0ce 203 in = *pSrc++;
mbed_official 3:7a284390b0ce 204 in = ((q63_t) in * scaleFract) >> 32;
mbed_official 3:7a284390b0ce 205
mbed_official 3:7a284390b0ce 206 out = in << kShift;
mbed_official 3:7a284390b0ce 207
mbed_official 3:7a284390b0ce 208 if(in != (out >> kShift))
mbed_official 3:7a284390b0ce 209 out = 0x7FFFFFFF ^ (in >> 31);
mbed_official 3:7a284390b0ce 210
mbed_official 3:7a284390b0ce 211 *pDst++ = out;
emilmont 1:fdd22bb7aa52 212
mbed_official 3:7a284390b0ce 213 /* Decrement the loop counter */
mbed_official 3:7a284390b0ce 214 blkCnt--;
mbed_official 3:7a284390b0ce 215 }
mbed_official 3:7a284390b0ce 216 }
mbed_official 3:7a284390b0ce 217 else
mbed_official 3:7a284390b0ce 218 {
mbed_official 3:7a284390b0ce 219 while(blkCnt > 0u)
mbed_official 3:7a284390b0ce 220 {
mbed_official 3:7a284390b0ce 221 /* C = A * scale */
mbed_official 3:7a284390b0ce 222 /* Scale the input and then store the result in the destination buffer. */
mbed_official 3:7a284390b0ce 223 in = *pSrc++;
mbed_official 3:7a284390b0ce 224 in = ((q63_t) in * scaleFract) >> 32;
emilmont 1:fdd22bb7aa52 225
mbed_official 3:7a284390b0ce 226 out = in >> -kShift;
mbed_official 3:7a284390b0ce 227
mbed_official 3:7a284390b0ce 228 *pDst++ = out;
emilmont 1:fdd22bb7aa52 229
mbed_official 3:7a284390b0ce 230 /* Decrement the loop counter */
mbed_official 3:7a284390b0ce 231 blkCnt--;
mbed_official 3:7a284390b0ce 232 }
mbed_official 3:7a284390b0ce 233
emilmont 1:fdd22bb7aa52 234 }
emilmont 1:fdd22bb7aa52 235 }
emilmont 1:fdd22bb7aa52 236
mbed_official 3:7a284390b0ce 237 /**
mbed_official 3:7a284390b0ce 238 * @} end of scale group
emilmont 1:fdd22bb7aa52 239 */