Question

After running a SPICE simulation I get an .a2d file which I am parsing with Matlab. After parsing the file I get a cell array with the names of the variables
Ex: vars = {‘s0′;’s1′;’s2′;’a0′;’a1’}
and a data matrix with the signals transitions, where each row represents the data of each signal and columns represent time. All the data is binary, i.e., it’s only 0 or 1.

What I want to do is to create an algorithm that detects “words” based on the names stored in vars. For example, s0, s1, and s2 form a 3-bit word ‘s’; a0 and a1 a 2-bit word ‘a’.
Finally, what I need is to break the data matrix into one array for each word (converting from binary to decimal).

I am doing that by hand, but I want to know if there is any way to do it by a script.

Answer 1

Here’s a possible solution, putting the final values into the fields of a struct. I’ve used a regexp to extract the variable names and shifts as you didn’t say how long they might be.

% fake data
vars = {'sums4', 'sums2', 'sums1', 'a0', 'a5'};
data(1,:) = logical([0 0 1 0 0 0 1]);
data(2,:) = logical([0 1 1 1 1 0 0]);
data(3,:) = logical([1 0 1 1 0 0 1]);
data(4,:) = logical([0 0 1 0 1 1 1]);
data(5,:) = logical([0 1 1 1 0 0 0]);

output = struct();
for i = 1:length(vars) 
    matches = regexp(vars{i}, '(\D+)(\d+)', 'tokens');
    var = matches{1}{1};
    shiftsize = str2num(matches{1}{2});

    % initialise if not already present
    if (~isfield(output, var))
        output.(var) = zeros(1, size(data, 2));
    end

    % add apropriate shifted value to current values
    output.(var) = output.(var) + (bitshift(1, shiftsize) * data(i, :));
end

output ends up with fields named after the string parts of vars

output = 

    sums: [2 4 22 6 4 0 18]
       a: [0 32 33 32 1 1 1]