HDL Coder Variable Size Matrix

Hello, I'm trying to do matrix multipication in fpga. For this, I would like to convert my .m function given below.

function y=sindy_product(x1,x2)%mxn . nxk

y=zeros(1,225);%mxk

iteration=1;

coder.varsize('A',[1 1000001]);

coder.varsize('B',[1 1000001]);

for i = 1:1000001:1000001*15%n,mxn

for j =1:15%k

A=x1(i:i+1000001-1);%n

B=x2(j:15:1000001*15);%k,nxk

C=0;

for k = 1:1000001

C=C+A(k)*B(k);

end

y(iteration)=C;

iteration=iteration+1;

end

end

end

Error I get:

sindy_product_fixpt:17Error'B' : Error: variable-size matrix type is not supported for HDL code generation.sindy_product_fixpt:20Error'A' : Error: variable-size matrix type is not supported for HDL code generation.sindy_product_fixpt:16Error'A' : Error: variable-size matrix type is not supported for HDL code generation.sindy_product_fixpt:20Error'B' : Error: variable-size matrix type is not supported for HDL code generation.sindy_product_fixpt:20Error'A' : Error: variable-size matrix type is not supported for HDL code generation.sindy_product_fixpt:17Error'B' : Error: variable-size matrix type is not supported for HDL code generation.sindy_product_fixpt:16Error'A' : Error: variable-size matrix type is not supported for HDL code generation.sindy_product_fixpt:16Error'varargin_1' : Error: variable-size matrix type is not supported for HDL code generation.sindy_product_fixpt:16Error'varargin_1' : Error: variable-size matrix type is not supported for HDL code generation.sindy_product_fixpt:17Error'varargin_1' : Error: variable-size matrix type is not supported for HDL code generation.sindy_product_fixpt:17Error'varargin_1' : Error: variable-size matrix type is not supported for HDL code generation.sindy_product_fixpt:16Error'varargin_1' : Error: variable-size matrix type is not supported for HDL code generation.sindy_product_fixpt:16Error'varargin_1' : Error: variable-size matrix type is not supported for HDL code generation.sindy_product_fixpt:17Error'varargin_1' : Error: variable-size matrix type is not supported for HDL code generation.sindy_product_fixpt:17Error'varargin_1' : Error: variable-size matrix type is not supported for HDL code generation.sindy_product_fixpt:16Error'var1' : Error: variable-size matrix type is not supported for HDL code generation.sindy_product_fixpt:16Error'A' : Error: variable-size matrix type is not supported for HDL code generation.sindy_product_fixpt:16Error'var1' : Error: variable-size matrix type is not supported for HDL code generation.sindy_product_fixpt:17Error'var1' : Error: variable-size matrix type is not supported for HDL code generation.sindy_product_fixpt:17Error'B' : Error: variable-size matrix type is not supported for HDL code generation.sindy_product_fixpt:17Error'var1' : Error: variable-size matrix type is not supported for HDL code generation.sindy_product_fixpt:0ErrorMATLAB HDL Coder failed in the code generation phase. See HDL Coder conformance report.

I use matricies' row vector form because hdl coder doesnt allow to use matrix form. I think my all the variables have constant size but i still get the variable-size matrix error.

Answer 1

[Ryan Baird]

coder.varsize is declaring a variable size matrix:

https://www.mathworks.com/help/coder/ref/coder.varsize.html

Initializing the matrix with the zeros function may solve your issue.

[Hau]

I have read the problem and read the proposal solution and i write this code but i still have the same problem. Can i have some ideas ?

function out = test(in, n, intrlv)
    % Define the maximum possible input length (adjust to your needs)
    max_len = 100;
    
    % Pre-allocate output array with a fixed size
    out = zeros(1, max_len);
    
    % Calculate the block size
    bl_n = n * intrlv;
    
    % Calculate the maximum number of blocks (fixed size)
    max_blocks = max_len / bl_n;
    
    % Loop through each block with fixed indexing
    for b = 1:max_blocks
        % Calculate the index range (fixed)
        idx_start = (b-1) * bl_n + 1;
        idx_end = b * bl_n;
        
        % Ensure indices are within the input length
        if idx_end <= length(in)
            % Copy the block to the output
            out(idx_start:idx_end) = in(idx_start:idx_end);
        end
    end
end

Answer 2

[Kiran Kintali]

Can you share your design and testbench and the project file used?

Thanks