Renesas / mbed-dsp-neon

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。


Diff: cmsis_dsp/TransformFunctions/arm_bitreversal.c

Revision:
1:fdd22bb7aa52
Child:
2:da51fb522205
--- /dev/null	Thu Jan 01 00:00:00 1970 +0000
+++ b/cmsis_dsp/TransformFunctions/arm_bitreversal.c	Wed Nov 28 12:30:09 2012 +0000
@@ -0,0 +1,222 @@
+/* ----------------------------------------------------------------------    
+* Copyright (C) 2010 ARM Limited. All rights reserved.    
+*    
+* $Date:        15. February 2012  
+* $Revision:     V1.1.0  
+*    
+* Project:         CMSIS DSP Library    
+* Title:        arm_bitreversal.c    
+*    
+* Description:    This file has common tables like Bitreverse, reciprocal etc which are used across different functions    
+*    
+* Target Processor: Cortex-M4/Cortex-M3/Cortex-M0
+*  
+* Version 1.1.0 2012/02/15 
+*    Updated with more optimizations, bug fixes and minor API changes.  
+*   
+* Version 1.0.10 2011/7/15  
+*    Initial Version  
+* -------------------------------------------------------------------- */
+
+#include "arm_math.h"
+#include "arm_common_tables.h"
+
+/*    
+ * @brief  In-place bit reversal function.   
+ * @param[in, out] *pSrc        points to the in-place buffer of floating-point data type.   
+ * @param[in]      fftSize      length of the FFT.   
+ * @param[in]      bitRevFactor bit reversal modifier that supports different size FFTs with the same bit reversal table.   
+ * @param[in]      *pBitRevTab  points to the bit reversal table.   
+ * @return none.   
+ */
+
+void arm_bitreversal_f32(
+  float32_t * pSrc,
+  uint16_t fftSize,
+  uint16_t bitRevFactor,
+  uint16_t * pBitRevTab)
+{
+  uint16_t fftLenBy2, fftLenBy2p1;
+  uint16_t i, j;
+  float32_t in;
+
+  /*  Initializations */
+  j = 0u;
+  fftLenBy2 = fftSize >> 1u;
+  fftLenBy2p1 = (fftSize >> 1u) + 1u;
+
+  /* Bit Reversal Implementation */
+  for (i = 0u; i <= (fftLenBy2 - 2u); i += 2u)
+  {
+    if(i < j)
+    {
+      /*  pSrc[i] <-> pSrc[j]; */
+      in = pSrc[2u * i];
+      pSrc[2u * i] = pSrc[2u * j];
+      pSrc[2u * j] = in;
+
+      /*  pSrc[i+1u] <-> pSrc[j+1u] */
+      in = pSrc[(2u * i) + 1u];
+      pSrc[(2u * i) + 1u] = pSrc[(2u * j) + 1u];
+      pSrc[(2u * j) + 1u] = in;
+
+      /*  pSrc[i+fftLenBy2p1] <-> pSrc[j+fftLenBy2p1] */
+      in = pSrc[2u * (i + fftLenBy2p1)];
+      pSrc[2u * (i + fftLenBy2p1)] = pSrc[2u * (j + fftLenBy2p1)];
+      pSrc[2u * (j + fftLenBy2p1)] = in;
+
+      /*  pSrc[i+fftLenBy2p1+1u] <-> pSrc[j+fftLenBy2p1+1u] */
+      in = pSrc[(2u * (i + fftLenBy2p1)) + 1u];
+      pSrc[(2u * (i + fftLenBy2p1)) + 1u] =
+        pSrc[(2u * (j + fftLenBy2p1)) + 1u];
+      pSrc[(2u * (j + fftLenBy2p1)) + 1u] = in;
+
+    }
+
+    /*  pSrc[i+1u] <-> pSrc[j+1u] */
+    in = pSrc[2u * (i + 1u)];
+    pSrc[2u * (i + 1u)] = pSrc[2u * (j + fftLenBy2)];
+    pSrc[2u * (j + fftLenBy2)] = in;
+
+    /*  pSrc[i+2u] <-> pSrc[j+2u] */
+    in = pSrc[(2u * (i + 1u)) + 1u];
+    pSrc[(2u * (i + 1u)) + 1u] = pSrc[(2u * (j + fftLenBy2)) + 1u];
+    pSrc[(2u * (j + fftLenBy2)) + 1u] = in;
+
+    /*  Reading the index for the bit reversal */
+    j = *pBitRevTab;
+
+    /*  Updating the bit reversal index depending on the fft length  */
+    pBitRevTab += bitRevFactor;
+  }
+}
+
+
+
+/*    
+ * @brief  In-place bit reversal function.   
+ * @param[in, out] *pSrc        points to the in-place buffer of Q31 data type.   
+ * @param[in]      fftLen       length of the FFT.   
+ * @param[in]      bitRevFactor bit reversal modifier that supports different size FFTs with the same bit reversal table   
+ * @param[in]      *pBitRevTab  points to bit reversal table.   
+ * @return none.   
+ */
+
+void arm_bitreversal_q31(
+  q31_t * pSrc,
+  uint32_t fftLen,
+  uint16_t bitRevFactor,
+  uint16_t * pBitRevTable)
+{
+  uint32_t fftLenBy2, fftLenBy2p1, i, j;
+  q31_t in;
+
+  /*  Initializations      */
+  j = 0u;
+  fftLenBy2 = fftLen / 2u;
+  fftLenBy2p1 = (fftLen / 2u) + 1u;
+
+  /* Bit Reversal Implementation */
+  for (i = 0u; i <= (fftLenBy2 - 2u); i += 2u)
+  {
+    if(i < j)
+    {
+      /*  pSrc[i] <-> pSrc[j]; */
+      in = pSrc[2u * i];
+      pSrc[2u * i] = pSrc[2u * j];
+      pSrc[2u * j] = in;
+
+      /*  pSrc[i+1u] <-> pSrc[j+1u] */
+      in = pSrc[(2u * i) + 1u];
+      pSrc[(2u * i) + 1u] = pSrc[(2u * j) + 1u];
+      pSrc[(2u * j) + 1u] = in;
+
+      /*  pSrc[i+fftLenBy2p1] <-> pSrc[j+fftLenBy2p1] */
+      in = pSrc[2u * (i + fftLenBy2p1)];
+      pSrc[2u * (i + fftLenBy2p1)] = pSrc[2u * (j + fftLenBy2p1)];
+      pSrc[2u * (j + fftLenBy2p1)] = in;
+
+      /*  pSrc[i+fftLenBy2p1+1u] <-> pSrc[j+fftLenBy2p1+1u] */
+      in = pSrc[(2u * (i + fftLenBy2p1)) + 1u];
+      pSrc[(2u * (i + fftLenBy2p1)) + 1u] =
+        pSrc[(2u * (j + fftLenBy2p1)) + 1u];
+      pSrc[(2u * (j + fftLenBy2p1)) + 1u] = in;
+
+    }
+
+    /*  pSrc[i+1u] <-> pSrc[j+1u] */
+    in = pSrc[2u * (i + 1u)];
+    pSrc[2u * (i + 1u)] = pSrc[2u * (j + fftLenBy2)];
+    pSrc[2u * (j + fftLenBy2)] = in;
+
+    /*  pSrc[i+2u] <-> pSrc[j+2u] */
+    in = pSrc[(2u * (i + 1u)) + 1u];
+    pSrc[(2u * (i + 1u)) + 1u] = pSrc[(2u * (j + fftLenBy2)) + 1u];
+    pSrc[(2u * (j + fftLenBy2)) + 1u] = in;
+
+    /*  Reading the index for the bit reversal */
+    j = *pBitRevTable;
+
+    /*  Updating the bit reversal index depending on the fft length */
+    pBitRevTable += bitRevFactor;
+  }
+}
+
+
+
+/*    
+   * @brief  In-place bit reversal function.   
+   * @param[in, out] *pSrc        points to the in-place buffer of Q15 data type.   
+   * @param[in]      fftLen       length of the FFT.   
+   * @param[in]      bitRevFactor bit reversal modifier that supports different size FFTs with the same bit reversal table   
+   * @param[in]      *pBitRevTab  points to bit reversal table.   
+   * @return none.   
+ */
+
+void arm_bitreversal_q15(
+  q15_t * pSrc16,
+  uint32_t fftLen,
+  uint16_t bitRevFactor,
+  uint16_t * pBitRevTab)
+{
+  q31_t *pSrc = (q31_t *) pSrc16;
+  q31_t in;
+  uint32_t fftLenBy2, fftLenBy2p1;
+  uint32_t i, j;
+
+  /*  Initializations */
+  j = 0u;
+  fftLenBy2 = fftLen / 2u;
+  fftLenBy2p1 = (fftLen / 2u) + 1u;
+
+  /* Bit Reversal Implementation */
+  for (i = 0u; i <= (fftLenBy2 - 2u); i += 2u)
+  {
+    if(i < j)
+    {
+      /*  pSrc[i] <-> pSrc[j]; */
+      /*  pSrc[i+1u] <-> pSrc[j+1u] */
+      in = pSrc[i];
+      pSrc[i] = pSrc[j];
+      pSrc[j] = in;
+
+      /*  pSrc[i + fftLenBy2p1] <-> pSrc[j + fftLenBy2p1];  */
+      /*  pSrc[i + fftLenBy2p1+1u] <-> pSrc[j + fftLenBy2p1+1u] */
+      in = pSrc[i + fftLenBy2p1];
+      pSrc[i + fftLenBy2p1] = pSrc[j + fftLenBy2p1];
+      pSrc[j + fftLenBy2p1] = in;
+    }
+
+    /*  pSrc[i+1u] <-> pSrc[j+fftLenBy2];         */
+    /*  pSrc[i+2] <-> pSrc[j+fftLenBy2+1u]  */
+    in = pSrc[i + 1u];
+    pSrc[i + 1u] = pSrc[j + fftLenBy2];
+    pSrc[j + fftLenBy2] = in;
+
+    /*  Reading the index for the bit reversal */
+    j = *pBitRevTab;
+
+    /*  Updating the bit reversal index depending on the fft length  */
+    pBitRevTab += bitRevFactor;
+  }
+}