CANpack code generation Problem for MCU without floating-point math capabilities

roundf has been part of the C standard library since C99. However it expects floating point input and returns floating point output, which is a problem if you are using a system that does not have floating point hardware.

On a system that does not have floating point hardware, you have two choices:

you can carefully write your code to only operate on the int* and uint* classes; or
you can convert your code to use fixed-point arithmetic with the Fixed-Point Toolbox. The fixed point toolbox can emulate many floating-point operations in software, including potentially to greater precision than double precision floating point. However, transcendental functions such as trig functions or log generally have to be rewritten; search for CORDIC for those.

1 个评论
显示 -1更早的评论隐藏 -1更早的评论

zhichao li 2023-9-17

thanks!

use fixed-point arithmetic is a good solution,I will test!

now I just delete the roundf, Refer to older versions of code generation(15b)

请先登录，再进行评论。

Answer 2

zhichao li 2023-9-17

0
链接

此回答的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/2021892-canpack-code-generation-problem-for-mcu-without-floating-point-math-capabilities#answer_1311982

在 MATLAB Online 中打开

I solved it by modify the function in can_helpers.tlc

%function roundRealValue(inVar, outVar, dataType)
    %openfile buffer
    %switch dataType
        %case "real32_T"
            %% %<outVar> = roundf(%<inVar>);    //before
            %<outVar> = %<inVar>;
            %break
        %case "real_T"
        %case "real64_T"
            %% %<outVar> = round(%<inVar>);     //before
            %<outVar> = %<inVar>;               
            %break
    %endswitch
    %closefile buffer
    %return buffer
%endfunction