Error: A cast between fixpt and floating point type is not supported

I am trying to convert my matlab code to HDL verilog code by HDL Coder. Fixed-Point Conversion shows no Error,but at the time HDL code Generation it show error that

" ErrorA cast between fixpt and floating point type is not supported, at Function 'cameraman_dwt_code1_fixpt' (#1.874.875), line 21, column 12."

My Fixed-Point Arithmatic code is-
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%                                                                          %
%           Generated by MATLAB 9.2 and Fixed-Point Designer 5.4           %
%                                                                          %
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%#codegen
function xavg12 = cameraman_dwt_code1_fixpt(xr1)
fm = get_fimath();

[fmo_1,fmo_2,fmo_3,fmo_4] = dwt2(xr1,'db3');
xavg = fi(fmo_1, 0, 14, 5, fm);
xhr = fi(fmo_2, 1, 14, 43, fm);
xvr = fi(fmo_3, 1, 14, 43, fm);
xdig = fi(fmo_4, 0, 14, 83, fm);

x1=fi(xavg(1,1), 0, 14, 5, fm);        %#This line creates error at xavg(1,1) section(above mentiond on error- line21,coloumn 12)
xavg12=fi(x1, 0, 14, 5, fm);
end

function fm = get_fimath()
	fm = fimath('RoundingMethod', 'Floor',...
	     'OverflowAction', 'Wrap',...
	     'ProductMode','FullPrecision',...
	     'MaxProductWordLength', 128,...
	     'SumMode','FullPrecision',...
	     'MaxSumWordLength', 128);
end

####################### My function is->

function xavg12 = cameraman_dwt_code1(xr1)
[xavg,xhr,xvr,xdig] = dwt2(xr1,'db3');
x1=xavg(1,1);
xavg12=x1;

####################### My TestBench code is->

x=158;
xavg11=cameraman_dwt_code1(x);
disp(xavg11);

############# It is actually for an Image processing code but I want to test it only for a single pixel value.

Overall code output is 1.2392 for pixel or input value 158 which is acurate but still I can not solved the Fixed-Point Arithmatic code.

Please Help me.Thank you

[Walter Roberson]

That is odd, column 12 appears to be the indexing.

Hmmm, if dwt2 is generating floating point then after indexing xavg(1,1) would be floating point, and perhaps the error is about casting between that floating point and fixed point?

[Sandip Paul]

Hi Walter,Thanks for giving answer on my problem.Please give me any idea or trick to solve this problem.I had already try a lot of way but still not solve it yet.

[Walter Roberson]

I have not used the Fixed Point toolbox, only read a little about it.

At the moment I am concerned that possibly dwt2() is not being transformed to fixed point.

I wonder if it would help to switch to an integer wavelet? https://www.mathworks.com/help/wavelet/ug/lifting-method-for-constructing-wavelets.html#bs9j3kj

[Sandip Paul]

Ok.Thank you very much.

Answer 1

[Kiran Kintali]

Hi Sandip,

The issue here is not related to fixed-point conversion; HDL Coder currently does not support dwt2 function out of the box for HDL code generation.

You need to rewrite dwt2_hdl(pixel_in) function to accomplish the algorithm on hardware.

Thanks

Answer 2

[Kiran Kintali]

Hi Sandip,

Since dwt2 is not supported out of the box, please consider using core MATLAB to HDL features and implement dwt2_fpga.m using first principles. There is lot of material on the web related to this topic. https://tinyurl.com/s8flm5j

The previously mentioned links show you how to write good MATLAB suitable for HDL code generation.

Thanks

Answer 3

[Sandip Paul]

Hi Kiran, very much thank you for your suggestion. Is there any other way to make a HDL representation of dwt2 from matlab code?