Freescale Semiconductor MSC8156EVM User guide

Freescale Semiconductor

Application Note

Document Number: AN4228

Rev. 0, 10/2010

MSC8156EVM Kernels Starting Guide

This document provides a starting guide to some commonly
used digital signal processing functions available for use
with the Freescale MSC8156EVM board. The example
projects are demonstrated in this guide. The objective of this
document is to help the users integrate various independent
projects using these kernels.

1 Introduction

The MSC8156EVM is supported by a collection of
commonly used digital signal processing kernels that
function with the SC3850 DSP core. The project described
in this document provides the kernel library consisting of C
and assembly callable kernel applications, as well as their
test harnesses. This tutorial guide demonstrates how to use
several of the most useful and representative kernel
examples such as FIR and IIR filters, FFT, Divide and
Matrix Inverse.

NOTE

Download the kernel software package from
the MSC8156EVM Tool Summary Page on

www.freescale.com.

Contents

1 Introduction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .1

2 What You Need to Run this Project . . . . . . . . . . . . . . . . .2

3 Test Procedures. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .3

4 Common Kernel Example Demonstration . . . . . . . . . . . .4

4.1 FIR_complex_16x16. . . . . . . . . . . . . . . . . . . . . . . . . . .5

4.2 Complex Radix-4 FFT/IFFT 16x16 . . . . . . . . . . . . . . . .6

4.3 Complex Radix-2 and Radix-4 FFT/ IFFT 16x16 . . . . .9

4.4 IIR . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .10

4.5 Division 16×16 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .12

4.6 Ln . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .13

4.7 Matrix Invers io n Co mp lex 2×2 . . . . . . . . . . . . . . . . . .14

4.8 Matrix Invers io n Co mp lex 4×4 . . . . . . . . . . . . . . . . . .15

© 2010 Freescale Semiconductor, Inc.

What You Need to Run this Project

2 What You Need to Run this Project

Running the DSP kernels requires the following devices:

• Personal computer (PC) with CodeWarrior for StarCore-Based DSP IDE for the MSC8156EVM
board connected to the PC

• MSC8156EVM board

The MSC8156EVM project includes the following kernels:

• FIR_complex_16×16

• Complex Radix-4 FFT/IFFT 16×16

• Complex Radix-4 and Radix-2 FFT/IFFT 16×16

• IIR

• Division

•Ln

• Matrix Inversion complex 2×2

• Matrix Inversion complex 4×4

Figure 1 shows the folder directory of all the kernel example projects.

Figure 1. Kernel Example Project Directory

MSC8156EVM Kernels Starting Guide, Rev. 0

2 Freescale Semiconduct or

3 Test Procedures

Use the following steps to prepare for and run the project:

1. Import the SC3850 DSP kernel library by dragging the .project file in

\fsl_sc3850_kernels\code\cw\sc3850_kernels to the CodeWarrior project window (Figure 2).

Figure 2. Importing the Project Files

2. Build the kernel by clicking on the build icon .

3. After building the kernel project, .elb files are created in the folder fsl_sc3850_kernels\lib.

Test Procedures

4. After the kernel is built, you can run one of the test cases in the \fsl_sc3850_kernels\test\
folder. Import the associated .project of the selected test case and build the project. After
building the test case, .eld files are created in the \fsl_sc3850_kernels\tests\<test_case>\cw
folder.

5. Load the project by clicking on the debug icon and selecting Debug Configurations.

6. Select the appropriate launch configuration, that is, assembly or C test (Figure 3), and click on the
Debug button. Note that not all test cases are available in both assembly (ASM) and C. Some test
cases only have one option.

Figure 3. Launch Configuration

7. Run the project by clicking on the run icon .

NOTE

See Section 4, Common Kernel Example Demonstration for details on how to run the DSP
kernel test cases.

MSC8156EVM Kernels Starting Guide, Rev. 0

Freescale Semiconduc tor 3

Common Kernel Example Demonstration

4 Common Kernel Example Demonstration

After the DSP kernel library is built, the u ser can run one of the kerne l test ca ses provided with the EVM.
This section provides detailed information for each kernel. For each kernel, the listing includes the
following:

• Location from which to import the file.

• Function

• ASM Prototype

• C Prototype

• Inputs

• Outputs

• Data alignment requirements (if applicable).

• Performance Measurement

The following notes apply for all kernels:

1. Import the kernel as described in Section 3, Test Procedures.

2. DPU is a defined function that enables cycle measurements

#ifdef DPU

#define INIT_CYCLE InitDPU()

#define GET_CYCLE ReadCountDPU()

#endif

3. The kernel is called twice in the example project. The first call brings the kernel to cache so we
can measure the performance of the second call more with warm cache.

4. The test results printed in the CodeWarrior console should show the cycles used to complete the
kernel process and check with the reference outputs.

MSC8156EVM Kernels Starting Guide, Rev. 0

4 Freescale Semiconduct or

4.1 FIR_complex_16x16

Location:

fsl_sc3850_kernels\tests\fir_complex_16x16\cw\test_sc3850_fir_complex_16x16

Function:

FIR filtering with 16-bit complex inputs and coefficients

ASM Prototype:

void sc3850_fir_complex_16x16_asm(Word32 x[], Word32 h[], Word16 y[], Word16 nr,
Word16 nh);

C Prototype:

Void sc3850_fir_complex_16x16_c(Word32 x[], Word32 h[], Word16 y[], Word16 nr, Word16
nh);

Inputs:

Word32 x[]: 32-bit complex inputs, 16 bits for real part and 16 bits for imaginary part
Word32 h[]: 32-bit complex coefficients, 16 bits for real and 16 bits for imaginary part
Word16 Nr: number of input data samples
Word16 Nh: number of elements in the filter

Common Kernel Example Demonstration

In the test source code, these inputs are defined as shown in Figure 4.

#define Nr 40
#define Nh 40

Word16 Inpu t[2 * (2 *Nr + Nh+ 2) ]= {

#include "../vectors/test_in _8 0. dat "

};

Word16 Coef f[2 * Nh ]={

#include "../vectors/coeff.d at "

};

Input data and coefficients are vectors stored
in .dat files

Test_in_80 has 244 entries
Coeff has 80 entries

Figure 4. Input Definitions

Output:

Word16 y[]:16-bit output. Interleaved real and imaginary part

In the test source code, the output is computed and stored as shown in Figure 5.

str ea m = fo pen( "../vectors/output_8 0.da t", "w+ " );

for

(i= 0 ;i<2*2*N r;i+ + )

{
fpr in t f(s tr eam,"%d,\n",( in t) Ou tp ut[i]);
}

fclose(strea m );

Figure 5. Output Definition

The output vector is stored to output_80.dat and compared with the reference output. If the
accuracy of the filter is verified, in the CodeWarrior console it displays:

No wrong results found

MSC8156EVM Kernels Starting Guide, Rev. 0

Freescale Semiconduc tor 5