Arm Mbed and Pelion Device Management support forum

CMSIS DSP support forthcoming?

First off, great work on the IDE. I’ve been using mbed since '11, and I love the environment and the improvements. Keep up the good work!

Any progress on CMSIS DSP support within mbed? I have a K66F that I bought specifically for audio processing, and I’ve run into nothing but trouble trying to get the DSP libraries to work.

  • I can’t find a way to simply link with CMSIS/DSP/Lib/ARM/arm_cortexM4lf_math.lib from the CMSIS pack, which would be easiest.

  • Compiling features/unsupported/dsp/cmsis_dsp/TransformFunctions with ARMC6, I get the dreaded compile issues with arm_bitreversal2.S.

  • If I switch to GCC_ARM, I get a whole host of other issues, such as, among others, “int32_t” and “NULL” undefined, problems with ::Serial not being able to use ::Stream, etc.

So, any updates on this would be most welcome.

Thanks!

Aaron

Hello Aaron,

I can’t find a way to simply link with CMSIS/DSP/Lib/ARM/arm_cortexM4lf_math.lib from the CMSIS pack, which would be easiest.

You can do it with Mbed CLI as follows:

cd path_to_your_mbed_cli_project
  • Build your project:
mbed compile -t GCC_ARM -m K66F --source . --source relative_path_from_this_folder_to_the_folder_containing_the_libarm_cortexM4l_math.a_file

I tested it with the following example main.cpp:

/* ----------------------------------------------------------------------
* Copyright (C) 2010-2012 ARM Limited. All rights reserved.
*
* $Date:         17. January 2013
* $Revision:     V1.4.0
*
* Project:       CMSIS DSP Library
* Title:         arm_dotproduct_example_f32.c
*
* Description:   Example code computing dot product of two vectors.
*
* Target Processor: Cortex-M4/Cortex-M3
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* are met:
*   - Redistributions of source code must retain the above copyright
*     notice, this list of conditions and the following disclaimer.
*   - Redistributions in binary form must reproduce the above copyright
*     notice, this list of conditions and the following disclaimer in
*     the documentation and/or other materials provided with the
*     distribution.
*   - Neither the name of ARM LIMITED nor the names of its contributors
*     may be used to endorse or promote products derived from this
*     software without specific prior written permission.
*
* THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
* "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
* LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
* FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE
* COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
* INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING,
* BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
* LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
* CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
* LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN
* ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
* POSSIBILITY OF SUCH DAMAGE.
 * -------------------------------------------------------------------- */

/**
 * @ingroup groupExamples
 */

/**
 * @defgroup DotproductExample Dot Product Example
 *
 * \par Description:
 * \par
 * Demonstrates the use of the Multiply and Add functions to perform the dot product.
 * The dot product of two vectors is obtained by multiplying corresponding elements
 * and summing the products.

 * \par Algorithm:
 * \par
 * The two input vectors \c A and \c B with length \c n, are multiplied element-by-element
 * and then added to obtain dot product.
 * \par
 * This is denoted by the following equation:
 *         <pre>  dotProduct = A[0] * B[0] + A[1] * B[1] + ... + A[n-1] * B[n-1]</pre>
 *
 * \par Block Diagram:
 * \par
 * \image html dotProduct.gif
 *
 * \par Variables Description:
 * \par
 * \li \c srcA_buf_f32 points to first input vector
 * \li \c srcB_buf_f32 points to second input vector
 * \li \c testOutput   stores dot product of the two input vectors.
 *
 * \par CMSIS DSP Software Library Functions Used:
 * \par
 * - arm_mult_f32()
 * - arm_add_f32()
 *
 * <b> Refer  </b>
 * \link arm_dotproduct_example_f32.c \endlink
 *
 */


/** \example arm_dotproduct_example_f32.c
  */

#include <math.h>
#include "arm_math.h"

#include "mbed.h"


/* ----------------------------------------------------------------------
* Defines each of the tests performed
* ------------------------------------------------------------------- */
#define MAX_BLOCKSIZE     32
#define DELTA           (0.000001f)

/* ----------------------------------------------------------------------
* Test input data for Floating point Dot Product example for 32-blockSize
* Generated by the MATLAB randn() function
* ------------------------------------------------------------------- */
/* ----------------------------------------------------------------------
** Test input data of srcA for blockSize 32
** ------------------------------------------------------------------- */
float32_t srcA_buf_f32[MAX_BLOCKSIZE] =
{
  -0.4325648115282207,  -1.6655843782380970,  0.1253323064748307,
   0.2876764203585489,  -1.1464713506814637,  1.1909154656429988,
   1.1891642016521031,  -0.0376332765933176,  0.3272923614086541,
   0.1746391428209245,  -0.1867085776814394,  0.7257905482933027,
  -0.5883165430141887,   2.1831858181971011, -0.1363958830865957,
   0.1139313135208096,   1.0667682113591888,  0.0592814605236053,
  -0.0956484054836690,  -0.8323494636500225,  0.2944108163926404,
  -1.3361818579378040,   0.7143245518189522,  1.6235620644462707,
  -0.6917757017022868,   0.8579966728282626,  1.2540014216025324,
  -1.5937295764474768,  -1.4409644319010200,  0.5711476236581780,
  -0.3998855777153632,   0.6899973754643451
};

/* ----------------------------------------------------------------------
** Test input data of srcB for blockSize 32
** ------------------------------------------------------------------- */
float32_t srcB_buf_f32[MAX_BLOCKSIZE] =
{
   1.7491401329284098,  0.1325982188803279,   0.3252281811989881,
  -0.7938091410349637,  0.3149236145048914,  -0.5272704888029532,
   0.9322666565031119,  1.1646643544607362,  -2.0456694357357357,
  -0.6443728590041911,  1.7410657940825480,   0.4867684246821860,
   1.0488288293660140,  1.4885752747099299,   1.2705014969484090,
  -1.8561241921210170,  2.1343209047321410,   1.4358467535865909,
  -0.9173023332875400, -1.1060770780029008,   0.8105708062681296,
   0.6985430696369063, -0.4015827425012831,   1.2687512030669628,
  -0.7836083053674872,  0.2132664971465569,   0.7878984786088954,
   0.8966819356782295, -0.1869172943544062,   1.0131816724341454,
   0.2484350696132857,  0.0596083377937976
};

/* Reference dot product output */
float32_t  refDotProdOut = 5.9273644806352142;

/* ----------------------------------------------------------------------
* Declare Global variables
* ------------------------------------------------------------------- */
float32_t multOutput[MAX_BLOCKSIZE];  /* Intermediate output */
float32_t testOutput;  /* Final ouput */

arm_status status;   /* Status of the example */

int main(void)
{
  uint32_t i;       /* Loop counter */
  float32_t diff;     /* Difference between reference and test outputs */

  /* Multiplication of two input buffers */
  arm_mult_f32(srcA_buf_f32, srcB_buf_f32, multOutput, MAX_BLOCKSIZE);

  /* Accumulate the multiplication output values to
     get the dot product of the two inputs */
  for(i=0; i< MAX_BLOCKSIZE; i++)
  {
    arm_add_f32(&testOutput, &multOutput[i], &testOutput, 1);
  }

  /* absolute value of difference between ref and test */
  diff = fabsf(refDotProdOut - testOutput);

  /* Comparison of dot product value with reference */
  status = (diff > DELTA) ? ARM_MATH_TEST_FAILURE : ARM_MATH_SUCCESS;

  if (status != ARM_MATH_SUCCESS)
  {
    printf("FAILURE\n");
    while (1);                             /* main function does not return */
  }
  else
  {
    printf("SUCCESS\n");
    printf("-----------------------\r\n");
    printf("multOutput = \r\n");
    for (int i = 0; i < MAX_BLOCKSIZE; i++) {
        printf("%f\r\n", multOutput[i]);
    }
    printf("-----------------------\r\n");
    printf("testOutput = %f\r\n", testOutput);
    printf("-----------------------\r\n");
    while (1);                             /* main function does not return */
  }
}

Result:

SUCCESS
-----------------------
multOutput = 
-0.756616
-0.220854
0.040762
-0.228360
-0.361051
-0.627935
1.108618
-0.043830
-0.669532
-0.112533
-0.325072
0.353292
-0.617043
3.249836
-0.173291
-0.211471
2.276826
0.085119
0.087739
0.920643
0.238641
-0.933381
-0.286860
2.059896
0.542081
0.182982
0.988026
-1.429069
0.269341
0.578676
-0.099346
0.041130
-----------------------
testOutput = 5.927364
-----------------------

Best regards, Zoltan

Thanks so much, Zoltan! I was able to get the mbed CLI installed and the project compiled / linked, but now I’m getting an odd blink code. I’m guessing the chip doesn’t like the code, which is something I’ll have to figure out. I wonder if it’s related to linking with libarm_cortexM4l_math.a vs libarm_cortexM4lf_math.a, which is what I use in MCUXpresso, and the same code seems to work (before I upgraded the NXP SDK, but that’s a different matter).

From a philosophical standpoint, I’m familiar with compiling from the command line, but it seems like if we’re using the mbed Studio IDE, then we should be able to do the linking from within that. I like the IDE; it works well, aside from this one issue. Oh well.

Thanks again!

Aaron

Hello Aaron,

If you’d like to use the Mbed Studio IDE then try this:

  • Create a new project and replace the content of main.cpp with the example above.
  • Open the Libraries tab (bottom-right) and click on the [+] button
  • Type (or copy&paste) https://os.mbed.com/teams/mbed-official/code/mbed-dsp to the “URL” edit box and type mbed-dsp to the “Library Name” edit box.
  • Click on [Next] and select default for the branch and click on [Finish].
  • Open the file mbed-dsp/cmsis_dsp/TransformFunctions/arm_bitreversal2.S for editing and replace lines from 43 trough 63 with
	#define CODESECT AREA     ||.text||, CODE, READONLY, ALIGN=2
	#define LABEL
  • Built the project

Best regards, Zoltan

1 Like

Thanks again, Zoltan! (Note to self: read the whole message, next time, before trying various options - it’s for your own good. :wink: ) I searched the mbed code library for an official version of the DSP code, but must’ve missed the one you mentioned; whatever I tried didn’t work: it wouldn’t even load in the library loader. Your link did.

However, after that, I ran into SEGFAULT violations in stage_rfft_f32: it was accessing S->pTwiddleRFFT, which wasn’t being set in the official version of arm_rfft_fast_init_f32. It is set correctly in the “unsupported” branch of the DSP code, so I copied that version of the file into the official branch and everything seems to work correctly, now. So, for anyone using arm_rfft_fast_f32 in their code, be forewarned.

Thanks so much, Zoltan! Have a gold star! :star:

1 Like