C Code Optimization Using MATLAB Coder app for Hexagon DSP

Supported Hardware

Prerequisites

Launch the hardware setup and install the Qualcomm SDK. For more information, see Launch Hardware Setup.

Required Hardware

To run this example, you need the following hardware:

Generate Code using MATLAB Coder app

Open MATLAB Coder App and Select Source Files

3. Click Next to go to the Define Input Types step.

Define Input Types

The code generator determines the class, size, and complexity of all variables in the MATLAB files at code generation time, also known as compile time. Therefore, you must specify the properties of all entry-point function inputs. To specify input properties, you can:

3. Click Next to go to the Check for Run-Time Issues step.

Check for Run-Time Issues

The Check for Run-Time Issues step generates a MEX file from your entry-point functions, runs the MEX function, and reports any issues. A MEX function is generated code that can be called from within MATLAB. It is recommended to perform this step because it allows you to detect and fix run-time errors in the generated C code. By default, the MEX function includes memory integrity checks, which perform array bounds and dimension checking to detect violations of memory integrity in the code generated for MATLAB functions.

Upon clicking Next, the coder will check for any general runtime issues in the plain C code. This step is optional and can be skipped if the CRL is being used.

By default, the app collects line execution counts. These counts help you to see how well the test file ex_fir_hexagon_test.m exercised the function. To view line execution counts, click View MATLAB line execution counts. The app editor displays a color-coded bar to the left of the code. To extend the color highlighting over the code and to see line execution counts, place your cursor over the bar.

Generate C Code

In the Generate Code tab, configure this code generation settings:

After updating the code generation settings, click on Generate to initiate the code generation process.

Upon successful code generation, you will be able to view the generated code and the generated report.

Numerical Verification of Generated Code

To manually verify numerical accuracy, follow a workflow similar to the codegen command line. Click Run manually in MATLAB under VERIFY CODE to initiate the verification process. This action generates the <codegen function>_pil.mex file, which shares the same interface as ex_fir_hexagon_ml.m. You can then compare the outputs of these functions when provided with the same inputs.

After the above step, you can run the below code:

% numSamples = 120;
% sin1 = dsp.SineWave('Amplitude',1,'Frequency',1000,...
%                      'SampleRate',16000, 'SamplesPerFrame', numSamples,...
%                      'OutputDataType', 'single');
% sin2 = dsp.SineWave('Amplitude',4,'Frequency',5000,...
%                      'SampleRate',16000, 'SamplesPerFrame', numSamples,...
%                      'OutputDataType', 'single');
% 
% numSteps = 200;
% frameLength = sin1.SamplesPerFrame;
% yRef = zeros(frameLength,1,numSteps,'single');
% y = zeros(frameLength,1,numSteps,'single');
% 
% for k = 1:numSteps
%    x1k = sin1(); % generate 1KHz sine wave
%    x5k = sin2(); % generate 5KHz sine wave
%    n = randn(size(x1k), 'single')*sqrt(.05); % generate noise signal
%    u = x1k+x5k+n;
%    % Run with MATLAB code on host machine
%    yRef(:,:,k) = ex_fir_hexagon_ml(u);
%    % Run with generated code on target
%    y(:,:,k) = ex_fir_hexagon_ml_pil(u);
% end
% clear ex_fir_hexagon_ml;
% clear ex_fir_hexagon_ml_pil;

After the execution, you can compare the outputs of the PIL against the reference using any norm function.

% % norm comparison of y & yRef
% absoluteError = norm(y(:)-yRef(:),'inf');
% fprintf("Absolute error = %g \n",absoluteError);

Alternatively, you can use the verifyEqual function from the matlab.unittest.TestCase.forInteractiveUse test class. This function allows you to compare the outputs against the given absolute and relative tolerances, concluding the overall results.

% reltol = single(1e-5);
% abstol = single(1e-5);
% matlab.unittest.TestCase.forInteractiveUse.verifyEqual(y,yRef, ...
%     'RelTol',reltol, 'AbsTol', abstol);

Analyze Performance using Code Profile Analyzer

The MATLAB Coder App provides users with the ability to profile the generated code and offers various utilities for performance analysis. To perform profiling, follow this steps:

This enables the coder to generate <codegen function>_pil.mex file, which runs in MATLAB and interacts with the target to deploy and execute target-specific code. It's important to note that the verification process may take time, depending on the complexity of the test script.

Once the testing is completed, a green pop-up will appear at the bottom of the app, indicating the completion of the verification process.

After the verification process is complete, you can click on Stop PIL Verification. This action will enable a link in the Test Output window, allowing you to view the profiling information. By clicking on the link, you can access and analyze the profiling results for further performance evaluation.

Click the profiling report link to generate the report.

This will display various metrics of the profiled time, such as average execution time, etc.

Considering the maximum execution time of the ex_fir_hexagon_ml function, which is 17,072 cycles, repeating the code profiling with the Code Replacement Library set to None resulted in the same function taking 42,188 cycles. This indicates that enabling the CRL improved the cycle performance of ex_fir_hexagon_ml by approximately 50%.

Code Generation Summary

The summary of the coder app will have all the links to report, generated code locations and dependencies.

Note: For Qualcomm Hexagon QHL CRL, explicit alignment specification for buffers is necessary only when the input/output variables are directly used in the operators, functions and system-objects.

More About