HDL Coder Error: HDL code generation does not support variable-size matrix type.

need help please!

I tried to use HDL CODER to convert my matlab code into verilog code,but failed in Workflow Advisor-HDL Code Generation.

This is partial code which shows error:

(H is a 16x16 complex matrix.)

SNR = [0:1:20];

ea=1;

es=ea*16;

sigma_n2 = es/snr;

snr = 10.^(SNR(1)/10);

complex_scalar = complex(sigma_n2/ea);

complex_eye = complex(eye(16));

temp = H*H' + complex_scalar*complex_eye;

temp2=H*H';

value = complex(zeros(16,16)); % 初始化为复数矩阵

tol = 1e-15;

MatrixB = complex(eye(16)); % 使用complex函数确保MatrixB是复数

Y = value;

array1 = complex(1:16);

for i=1:15

b=i;

% temp5=temp(b:16,b);

% temp4=real(temp5).^2 + imag(temp5).^2;

% abs_temp = sqrt(temp4);

temp_slice(1:16-b+1) = temp(b:16,b); % 获取切片

temp5(1:16-b+1,1) = temp_slice(1:16-b+1);

temp4(1:16-b+1,1) = real(temp_slice(1:16-b+1)).^2 + imag(temp_slice(1:16-b+1)).^2;

abs_temp(1:16-b+1,1) = sqrt(temp4(1:16-b+1,1));

[~, j] = max(abs_temp);

num1 = temp(i,:);

result=real(i);

temp(i,:) = temp(result,:);

temp(i,:) = num1;

% num1(1,i) = array1(1,i);

% array1(i) = array1(i);

% array1(1,i) = num1(1,i);

for j=i+1:16

%%%%%%%%%

% num1 = temp(j,i)/temp(i,i);

%%%%%%%%%

% coder.varsize('num1');

conj_temp=conj(temp(i,i));

numerator = temp(j,i)*conj_temp;

denominator =temp(i,i)*conj_temp;

result_real = real(numerator) / real(denominator);

result_imag = imag(numerator) / real(denominator);

temp6=0;

temp6=complex(temp6);

temp6=complex(result_real, result_imag);

%%%%%%%%%

temp(j,i) = temp6;

temp(j,i+1:16) = temp(j,i+1:16) - temp6*temp(i,i+1:16); //THIS shows errors

end

end

for i=1:16

Y(1,i) = MatrixB(array1(1),i);

for j=2:16

Y(j,i) = MatrixB(array1(j),i) - temp(j,1:j-1)*Y(1:j-1,i); //AND THIS ONE

end

%%%%%%%%%

% value(N,i) = Y(N,i)/temp(N,N);

%%%%%%%%%

conj_fenmu=conj(temp(16,16));

numerator=Y(16,i)*conj_fenmu;

denominator=temp(16,16)*conj_fenmu;

result_real = real(numerator) / real(denominator);

result_imag = imag(numerator) / real(denominator);

value(16,i)=complex(result_real, result_imag);

%%%%%%%%%

for j=16-1:-1:1

%%%%%%%%%

% value(j,i) = (Y(j,i) - temp(j,j+1:N)*value(j+1:N,i))/tempj,j);

%%%%%%%%%

fenzi=Y(j,i) - temp(j,j+1:16)*value(j+1:16,i);

conj_fenmu=conj(temp(j,j));

numerator =fenzi*conj_fenmu;

denominator =temp(j,j)*conj_fenmu;

result_real = real(numerator) / real(denominator);

result_imag = imag(numerator) / real(denominator);

value(j,i)=complex(result_real, result_imag);

end

end

The purpose of this code is to find the inverse matrix.Because inv() cant used in HDL CODER to calucate a complex matrix,so I have to do it by myself.I run it in HDL CODER,and in Workflow Advisor-HDL Code Generation,it shows several errors.

I dont know where these a0,b0,br,y,pre,pi matrix come from,i dont even define and use them.All of the matrices that I use have already been defined in terms of size.I have no idea why it shows error.Could someone help me please:(

I have uploaded my entire code

[SHUHENG]

can anyone help me😭

Answer 1

[Walter Roberson]

temp_slice(1:16-b+1) = temp(b:16,b); %

You do not declare temp_slice before use, so you are growing it within a loop, which requires dynamic memory allocation

[SHUHENG]

I declare it in the full code

temp_slice = complex(zeros(16,1));

and i used disp(size(temp_slice)) to show the size of it in loops,every loop it shows 16×1.