Fork of mbed-dsp. CMSIS-DSP library of supporting NEON
Dependents: mbed-os-example-cmsis_dsp_neon
Fork of mbed-dsp by
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。
cmsis_dsp/BasicMathFunctions/arm_scale_q31.c@1:fdd22bb7aa52, 2012-11-28 (annotated)
- Committer:
- emilmont
- Date:
- Wed Nov 28 12:30:09 2012 +0000
- Revision:
- 1:fdd22bb7aa52
- Child:
- 2:da51fb522205
DSP library code
Who changed what in which revision?
User | Revision | Line number | New contents of line |
---|---|---|---|
emilmont | 1:fdd22bb7aa52 | 1 | /* ---------------------------------------------------------------------- |
emilmont | 1:fdd22bb7aa52 | 2 | * Copyright (C) 2010 ARM Limited. All rights reserved. |
emilmont | 1:fdd22bb7aa52 | 3 | * |
emilmont | 1:fdd22bb7aa52 | 4 | * $Date: 15. May 2012 |
emilmont | 1:fdd22bb7aa52 | 5 | * $Revision: V1.1.0 |
emilmont | 1:fdd22bb7aa52 | 6 | * |
emilmont | 1:fdd22bb7aa52 | 7 | * Project: CMSIS DSP Library |
emilmont | 1:fdd22bb7aa52 | 8 | * Title: arm_scale_q31.c |
emilmont | 1:fdd22bb7aa52 | 9 | * |
emilmont | 1:fdd22bb7aa52 | 10 | * Description: Multiplies a Q31 vector by a scalar. |
emilmont | 1:fdd22bb7aa52 | 11 | * |
emilmont | 1:fdd22bb7aa52 | 12 | * Target Processor: Cortex-M4/Cortex-M3/Cortex-M0 |
emilmont | 1:fdd22bb7aa52 | 13 | * |
emilmont | 1:fdd22bb7aa52 | 14 | * Version 1.1.0 2012/02/15 |
emilmont | 1:fdd22bb7aa52 | 15 | * Updated with more optimizations, bug fixes and minor API changes. |
emilmont | 1:fdd22bb7aa52 | 16 | * |
emilmont | 1:fdd22bb7aa52 | 17 | * Version 1.0.10 2011/7/15 |
emilmont | 1:fdd22bb7aa52 | 18 | * Big Endian support added and Merged M0 and M3/M4 Source code. |
emilmont | 1:fdd22bb7aa52 | 19 | * |
emilmont | 1:fdd22bb7aa52 | 20 | * Version 1.0.3 2010/11/29 |
emilmont | 1:fdd22bb7aa52 | 21 | * Re-organized the CMSIS folders and updated documentation. |
emilmont | 1:fdd22bb7aa52 | 22 | * |
emilmont | 1:fdd22bb7aa52 | 23 | * Version 1.0.2 2010/11/11 |
emilmont | 1:fdd22bb7aa52 | 24 | * Documentation updated. |
emilmont | 1:fdd22bb7aa52 | 25 | * |
emilmont | 1:fdd22bb7aa52 | 26 | * Version 1.0.1 2010/10/05 |
emilmont | 1:fdd22bb7aa52 | 27 | * Production release and review comments incorporated. |
emilmont | 1:fdd22bb7aa52 | 28 | * |
emilmont | 1:fdd22bb7aa52 | 29 | * Version 1.0.0 2010/09/20 |
emilmont | 1:fdd22bb7aa52 | 30 | * Production release and review comments incorporated |
emilmont | 1:fdd22bb7aa52 | 31 | * |
emilmont | 1:fdd22bb7aa52 | 32 | * Version 0.0.7 2010/06/10 |
emilmont | 1:fdd22bb7aa52 | 33 | * Misra-C changes done |
emilmont | 1:fdd22bb7aa52 | 34 | * -------------------------------------------------------------------- */ |
emilmont | 1:fdd22bb7aa52 | 35 | |
emilmont | 1:fdd22bb7aa52 | 36 | #include "arm_math.h" |
emilmont | 1:fdd22bb7aa52 | 37 | |
emilmont | 1:fdd22bb7aa52 | 38 | /** |
emilmont | 1:fdd22bb7aa52 | 39 | * @ingroup groupMath |
emilmont | 1:fdd22bb7aa52 | 40 | */ |
emilmont | 1:fdd22bb7aa52 | 41 | |
emilmont | 1:fdd22bb7aa52 | 42 | /** |
emilmont | 1:fdd22bb7aa52 | 43 | * @addtogroup scale |
emilmont | 1:fdd22bb7aa52 | 44 | * @{ |
emilmont | 1:fdd22bb7aa52 | 45 | */ |
emilmont | 1:fdd22bb7aa52 | 46 | |
emilmont | 1:fdd22bb7aa52 | 47 | /** |
emilmont | 1:fdd22bb7aa52 | 48 | * @brief Multiplies a Q31 vector by a scalar. |
emilmont | 1:fdd22bb7aa52 | 49 | * @param[in] *pSrc points to the input vector |
emilmont | 1:fdd22bb7aa52 | 50 | * @param[in] scaleFract fractional portion of the scale value |
emilmont | 1:fdd22bb7aa52 | 51 | * @param[in] shift number of bits to shift the result by |
emilmont | 1:fdd22bb7aa52 | 52 | * @param[out] *pDst points to the output vector |
emilmont | 1:fdd22bb7aa52 | 53 | * @param[in] blockSize number of samples in the vector |
emilmont | 1:fdd22bb7aa52 | 54 | * @return none. |
emilmont | 1:fdd22bb7aa52 | 55 | * |
emilmont | 1:fdd22bb7aa52 | 56 | * <b>Scaling and Overflow Behavior:</b> |
emilmont | 1:fdd22bb7aa52 | 57 | * \par |
emilmont | 1:fdd22bb7aa52 | 58 | * The input data <code>*pSrc</code> and <code>scaleFract</code> are in 1.31 format. |
emilmont | 1:fdd22bb7aa52 | 59 | * These are multiplied to yield a 2.62 intermediate result and this is shifted with saturation to 1.31 format. |
emilmont | 1:fdd22bb7aa52 | 60 | */ |
emilmont | 1:fdd22bb7aa52 | 61 | |
emilmont | 1:fdd22bb7aa52 | 62 | void arm_scale_q31( |
emilmont | 1:fdd22bb7aa52 | 63 | q31_t * pSrc, |
emilmont | 1:fdd22bb7aa52 | 64 | q31_t scaleFract, |
emilmont | 1:fdd22bb7aa52 | 65 | int8_t shift, |
emilmont | 1:fdd22bb7aa52 | 66 | q31_t * pDst, |
emilmont | 1:fdd22bb7aa52 | 67 | uint32_t blockSize) |
emilmont | 1:fdd22bb7aa52 | 68 | { |
emilmont | 1:fdd22bb7aa52 | 69 | int8_t kShift = shift + 1; /* Shift to apply after scaling */ |
emilmont | 1:fdd22bb7aa52 | 70 | int8_t sign = (kShift & 0x80); |
emilmont | 1:fdd22bb7aa52 | 71 | uint32_t blkCnt; /* loop counter */ |
emilmont | 1:fdd22bb7aa52 | 72 | q31_t in, out; |
emilmont | 1:fdd22bb7aa52 | 73 | |
emilmont | 1:fdd22bb7aa52 | 74 | #ifndef ARM_MATH_CM0 |
emilmont | 1:fdd22bb7aa52 | 75 | |
emilmont | 1:fdd22bb7aa52 | 76 | /* Run the below code for Cortex-M4 and Cortex-M3 */ |
emilmont | 1:fdd22bb7aa52 | 77 | |
emilmont | 1:fdd22bb7aa52 | 78 | q31_t in1, in2, in3, in4; /* temporary input variables */ |
emilmont | 1:fdd22bb7aa52 | 79 | q31_t out1, out2, out3, out4; /* temporary output variabels */ |
emilmont | 1:fdd22bb7aa52 | 80 | |
emilmont | 1:fdd22bb7aa52 | 81 | |
emilmont | 1:fdd22bb7aa52 | 82 | /*loop Unrolling */ |
emilmont | 1:fdd22bb7aa52 | 83 | blkCnt = blockSize >> 2u; |
emilmont | 1:fdd22bb7aa52 | 84 | |
emilmont | 1:fdd22bb7aa52 | 85 | if(sign == 0u) |
emilmont | 1:fdd22bb7aa52 | 86 | { |
emilmont | 1:fdd22bb7aa52 | 87 | /* First part of the processing with loop unrolling. Compute 4 outputs at a time. |
emilmont | 1:fdd22bb7aa52 | 88 | ** a second loop below computes the remaining 1 to 3 samples. */ |
emilmont | 1:fdd22bb7aa52 | 89 | while(blkCnt > 0u) |
emilmont | 1:fdd22bb7aa52 | 90 | { |
emilmont | 1:fdd22bb7aa52 | 91 | /* read four inputs from source */ |
emilmont | 1:fdd22bb7aa52 | 92 | in1 = *pSrc; |
emilmont | 1:fdd22bb7aa52 | 93 | in2 = *(pSrc + 1); |
emilmont | 1:fdd22bb7aa52 | 94 | in3 = *(pSrc + 2); |
emilmont | 1:fdd22bb7aa52 | 95 | in4 = *(pSrc + 3); |
emilmont | 1:fdd22bb7aa52 | 96 | |
emilmont | 1:fdd22bb7aa52 | 97 | /* multiply input with scaler value */ |
emilmont | 1:fdd22bb7aa52 | 98 | in1 = ((q63_t) in1 * scaleFract) >> 32; |
emilmont | 1:fdd22bb7aa52 | 99 | in2 = ((q63_t) in2 * scaleFract) >> 32; |
emilmont | 1:fdd22bb7aa52 | 100 | in3 = ((q63_t) in3 * scaleFract) >> 32; |
emilmont | 1:fdd22bb7aa52 | 101 | in4 = ((q63_t) in4 * scaleFract) >> 32; |
emilmont | 1:fdd22bb7aa52 | 102 | |
emilmont | 1:fdd22bb7aa52 | 103 | /* apply shifting */ |
emilmont | 1:fdd22bb7aa52 | 104 | out1 = in1 << kShift; |
emilmont | 1:fdd22bb7aa52 | 105 | out2 = in2 << kShift; |
emilmont | 1:fdd22bb7aa52 | 106 | |
emilmont | 1:fdd22bb7aa52 | 107 | /* saturate the results. */ |
emilmont | 1:fdd22bb7aa52 | 108 | if(in1 != (out1 >> kShift)) |
emilmont | 1:fdd22bb7aa52 | 109 | out1 = 0x7FFFFFFF ^ (in1 >> 31); |
emilmont | 1:fdd22bb7aa52 | 110 | |
emilmont | 1:fdd22bb7aa52 | 111 | if(in2 != (out2 >> kShift)) |
emilmont | 1:fdd22bb7aa52 | 112 | out2 = 0x7FFFFFFF ^ (in2 >> 31); |
emilmont | 1:fdd22bb7aa52 | 113 | |
emilmont | 1:fdd22bb7aa52 | 114 | out3 = in3 << kShift; |
emilmont | 1:fdd22bb7aa52 | 115 | out4 = in4 << kShift; |
emilmont | 1:fdd22bb7aa52 | 116 | |
emilmont | 1:fdd22bb7aa52 | 117 | *pDst = out1; |
emilmont | 1:fdd22bb7aa52 | 118 | *(pDst + 1) = out2; |
emilmont | 1:fdd22bb7aa52 | 119 | |
emilmont | 1:fdd22bb7aa52 | 120 | if(in3 != (out3 >> kShift)) |
emilmont | 1:fdd22bb7aa52 | 121 | out3 = 0x7FFFFFFF ^ (in3 >> 31); |
emilmont | 1:fdd22bb7aa52 | 122 | |
emilmont | 1:fdd22bb7aa52 | 123 | if(in4 != (out4 >> kShift)) |
emilmont | 1:fdd22bb7aa52 | 124 | out4 = 0x7FFFFFFF ^ (in4 >> 31); |
emilmont | 1:fdd22bb7aa52 | 125 | |
emilmont | 1:fdd22bb7aa52 | 126 | /* Store result destination */ |
emilmont | 1:fdd22bb7aa52 | 127 | *(pDst + 2) = out3; |
emilmont | 1:fdd22bb7aa52 | 128 | *(pDst + 3) = out4; |
emilmont | 1:fdd22bb7aa52 | 129 | |
emilmont | 1:fdd22bb7aa52 | 130 | /* Update pointers to process next sampels */ |
emilmont | 1:fdd22bb7aa52 | 131 | pSrc += 4u; |
emilmont | 1:fdd22bb7aa52 | 132 | pDst += 4u; |
emilmont | 1:fdd22bb7aa52 | 133 | |
emilmont | 1:fdd22bb7aa52 | 134 | /* Decrement the loop counter */ |
emilmont | 1:fdd22bb7aa52 | 135 | blkCnt--; |
emilmont | 1:fdd22bb7aa52 | 136 | } |
emilmont | 1:fdd22bb7aa52 | 137 | |
emilmont | 1:fdd22bb7aa52 | 138 | } |
emilmont | 1:fdd22bb7aa52 | 139 | else |
emilmont | 1:fdd22bb7aa52 | 140 | { |
emilmont | 1:fdd22bb7aa52 | 141 | kShift = -kShift; |
emilmont | 1:fdd22bb7aa52 | 142 | |
emilmont | 1:fdd22bb7aa52 | 143 | /* First part of the processing with loop unrolling. Compute 4 outputs at a time. |
emilmont | 1:fdd22bb7aa52 | 144 | ** a second loop below computes the remaining 1 to 3 samples. */ |
emilmont | 1:fdd22bb7aa52 | 145 | while(blkCnt > 0u) |
emilmont | 1:fdd22bb7aa52 | 146 | { |
emilmont | 1:fdd22bb7aa52 | 147 | /* read four inputs from source */ |
emilmont | 1:fdd22bb7aa52 | 148 | in1 = *pSrc; |
emilmont | 1:fdd22bb7aa52 | 149 | in2 = *(pSrc + 1); |
emilmont | 1:fdd22bb7aa52 | 150 | in3 = *(pSrc + 2); |
emilmont | 1:fdd22bb7aa52 | 151 | in4 = *(pSrc + 3); |
emilmont | 1:fdd22bb7aa52 | 152 | |
emilmont | 1:fdd22bb7aa52 | 153 | /* multiply input with scaler value */ |
emilmont | 1:fdd22bb7aa52 | 154 | in1 = ((q63_t) in1 * scaleFract) >> 32; |
emilmont | 1:fdd22bb7aa52 | 155 | in2 = ((q63_t) in2 * scaleFract) >> 32; |
emilmont | 1:fdd22bb7aa52 | 156 | in3 = ((q63_t) in3 * scaleFract) >> 32; |
emilmont | 1:fdd22bb7aa52 | 157 | in4 = ((q63_t) in4 * scaleFract) >> 32; |
emilmont | 1:fdd22bb7aa52 | 158 | |
emilmont | 1:fdd22bb7aa52 | 159 | /* apply shifting */ |
emilmont | 1:fdd22bb7aa52 | 160 | out1 = in1 >> kShift; |
emilmont | 1:fdd22bb7aa52 | 161 | out2 = in2 >> kShift; |
emilmont | 1:fdd22bb7aa52 | 162 | |
emilmont | 1:fdd22bb7aa52 | 163 | out3 = in3 >> kShift; |
emilmont | 1:fdd22bb7aa52 | 164 | out4 = in4 >> kShift; |
emilmont | 1:fdd22bb7aa52 | 165 | |
emilmont | 1:fdd22bb7aa52 | 166 | /* Store result destination */ |
emilmont | 1:fdd22bb7aa52 | 167 | *pDst = out1; |
emilmont | 1:fdd22bb7aa52 | 168 | *(pDst + 1) = out2; |
emilmont | 1:fdd22bb7aa52 | 169 | |
emilmont | 1:fdd22bb7aa52 | 170 | *(pDst + 2) = out3; |
emilmont | 1:fdd22bb7aa52 | 171 | *(pDst + 3) = out4; |
emilmont | 1:fdd22bb7aa52 | 172 | |
emilmont | 1:fdd22bb7aa52 | 173 | /* Update pointers to process next sampels */ |
emilmont | 1:fdd22bb7aa52 | 174 | pSrc += 4u; |
emilmont | 1:fdd22bb7aa52 | 175 | pDst += 4u; |
emilmont | 1:fdd22bb7aa52 | 176 | |
emilmont | 1:fdd22bb7aa52 | 177 | /* Decrement the loop counter */ |
emilmont | 1:fdd22bb7aa52 | 178 | blkCnt--; |
emilmont | 1:fdd22bb7aa52 | 179 | } |
emilmont | 1:fdd22bb7aa52 | 180 | } |
emilmont | 1:fdd22bb7aa52 | 181 | /* If the blockSize is not a multiple of 4, compute any remaining output samples here. |
emilmont | 1:fdd22bb7aa52 | 182 | ** No loop unrolling is used. */ |
emilmont | 1:fdd22bb7aa52 | 183 | blkCnt = blockSize % 0x4u; |
emilmont | 1:fdd22bb7aa52 | 184 | |
emilmont | 1:fdd22bb7aa52 | 185 | #else |
emilmont | 1:fdd22bb7aa52 | 186 | |
emilmont | 1:fdd22bb7aa52 | 187 | /* Run the below code for Cortex-M0 */ |
emilmont | 1:fdd22bb7aa52 | 188 | if(sign != 0u) |
emilmont | 1:fdd22bb7aa52 | 189 | kShift = -kShift; |
emilmont | 1:fdd22bb7aa52 | 190 | |
emilmont | 1:fdd22bb7aa52 | 191 | /* Initialize blkCnt with number of samples */ |
emilmont | 1:fdd22bb7aa52 | 192 | blkCnt = blockSize; |
emilmont | 1:fdd22bb7aa52 | 193 | |
emilmont | 1:fdd22bb7aa52 | 194 | #endif /* #ifndef ARM_MATH_CM0 */ |
emilmont | 1:fdd22bb7aa52 | 195 | |
emilmont | 1:fdd22bb7aa52 | 196 | while(blkCnt > 0u) |
emilmont | 1:fdd22bb7aa52 | 197 | { |
emilmont | 1:fdd22bb7aa52 | 198 | /* C = A * scale */ |
emilmont | 1:fdd22bb7aa52 | 199 | /* Scale the input and then store the result in the destination buffer. */ |
emilmont | 1:fdd22bb7aa52 | 200 | in = *pSrc++; |
emilmont | 1:fdd22bb7aa52 | 201 | in = ((q63_t) in * scaleFract) >> 32; |
emilmont | 1:fdd22bb7aa52 | 202 | |
emilmont | 1:fdd22bb7aa52 | 203 | if(sign == 0) |
emilmont | 1:fdd22bb7aa52 | 204 | { |
emilmont | 1:fdd22bb7aa52 | 205 | out = in << kShift; |
emilmont | 1:fdd22bb7aa52 | 206 | if(in != (out >> kShift)) |
emilmont | 1:fdd22bb7aa52 | 207 | out = 0x7FFFFFFF ^ (in >> 31); |
emilmont | 1:fdd22bb7aa52 | 208 | } |
emilmont | 1:fdd22bb7aa52 | 209 | else |
emilmont | 1:fdd22bb7aa52 | 210 | { |
emilmont | 1:fdd22bb7aa52 | 211 | out = in >> kShift; |
emilmont | 1:fdd22bb7aa52 | 212 | } |
emilmont | 1:fdd22bb7aa52 | 213 | |
emilmont | 1:fdd22bb7aa52 | 214 | *pDst++ = out; |
emilmont | 1:fdd22bb7aa52 | 215 | |
emilmont | 1:fdd22bb7aa52 | 216 | /* Decrement the loop counter */ |
emilmont | 1:fdd22bb7aa52 | 217 | blkCnt--; |
emilmont | 1:fdd22bb7aa52 | 218 | } |
emilmont | 1:fdd22bb7aa52 | 219 | } |
emilmont | 1:fdd22bb7aa52 | 220 | |
emilmont | 1:fdd22bb7aa52 | 221 | /** |
emilmont | 1:fdd22bb7aa52 | 222 | * @} end of scale group |
emilmont | 1:fdd22bb7aa52 | 223 | */ |