[HDL Coder] Bug: Fixed point converter does not descend into System Objects

Can I somehow tell the fixed point converter to explicitly convert a certain function? This would temporarily fix the following bug.

The Bug I found can be reproduced with these files:

dut_tb.m

a = dut(1);

dut.m

function out = dut(in)

    persistent system;

    if isempty(system)
        system = dutSystem(1.1);
    end
    
    out = system.step(in);
end

dutSystem.m

classdef dutSystem < matlab.System
    
    properties (Nontunable)
        number
    end
    
    methods (Access = public)
        function obj = dutSystem(number)
            obj.number = number;
        end
    end
    
    methods (Access = protected)   
        function out = stepImpl(obj,in)
            out = buggy_function(in,obj.number);
        end
    end
end

buggy_function.m

function out = buggy_function(in1, in2)
    out = round(in1 + in2);
end

convert.m

function [] = convert()
    exArgs = {1};
    fc = coder.config('fixpt');
    fc.TestBenchName = 'dut_tb';
    hc = coder.config('hdl');
    codegen -float2fixed fc -args exArgs -config hc dut
end

Execute convert. The HDL coder will stop, because round is supposedly not supported by the HDL coder. But it is supported. The fixed-point converter just did not convert buggy_function.m

Answer 1

[Kiran Kintali]

Unfortunately the development team cannot reproduce the issue. Is it possible the attached reproduction steps are not sufficient? Can you share a zip file with the code and a snapshot of the error message when running the convert() function? Thanks

[Jan Siegmund]

Sorry, somehow, MATLAB 2020a in contrast to 2019b accepts double inputs in round.

But this does not mean, that the error is not reproduceable. The problematic extract from my dutSystem_fixpt.m:

...
function out = stepImpl(obj,in)
    fm = get_fimath();

    out = fi(buggy_function(in,obj.number), 0, 14, 15, fm);
end
...

It uses the standard buggy_function and not buggy_function_fixpt.

Use this buggy_function.m instead and you will see an error pop up.

function out = buggy_function(in1, in2)
    out = in1/(2^in2);
end