Data = cell2mat(textscan( fid, '%d', 'delimiter', ';')); % convert to array from cell
N=10; % size of new array number elements/column
Data=reshape(Data,N,[]).'; % and recast to said size...
This presumes, of course, that mod(numel(Data),N) == 0 as you can't have "jagged" arrays in Matlab. If that isn't so you either must truncate to the multiple of N, fill with NaN or other value to make up the missing elements or use a cell array instead that can have a short row.
ADDENDUM
Again, on the presumption you know your data is commensurate in size, you can do as you first surmised on the initial read--
N=1024;
fmt=repmat('%f',1,N);
data=cell2mat(textscan(fid,fmt,inf));
This has the issue that will error out if there aren't enough data to fill the last row whereas the first will read all the data and give you a chance to fix up the result if desired. This form will return the data up to the point of the error, however, so if the intent is to read M records of N values, then it may be the solution desired.
Of course, again, you'll only get a reshaped output file if you then go ahead and write the data in the desired format to a new file (or overwrite the original which is potentially dangerous if you make a mistake on the way to the forum).
