Need help importing Fortran90 binary files into MATLAB

Hello everyone!

I recently started learning how to code in Fortan90 for an upcoming project. The plan is to perform simulations with a Fortran90 program (compiled with the Linux PGI compiler in case this is important), store the output of those simulations in 2D matrices and exporting them as binary files for later use in MATLAB.

For testing purposes I created the following subroutine to export the matrices:

subroutine save_data (x,y,z)

    integer, intent(in):: x,y,z
    integer::i
    real, dimension(x,y):: temp
    character(20):: filename = ''

      do i=1, z
        write(filename,'(a,i3.3,a)') 'output_matrix',i,'.dat'
        temp=output(1:x,1:y,i)
        open(unit=13, file=filename, action="write", status="replace", form="unformatted")
        write(13) temp
        close(unit=13)
      end do

end subroutine save_data

The problem now is, despite several days of trial & error I am still not able to import the "output_matrix***.dat"-files into MATLAB so that I can work with them. I did find several questions similar to mine on MATLABCentral (such as THIS or THIS) and tried to do what was suggested there, unfortunately with no luck. So far I was only able to achieve this:

clc
clear all
close all

fid = fopen('output_matrix001.dat');

h=fread(fid,[31,16]);

which resulted in h looking like this:

instead of this:

Could you please help me with this (for more advanced users supposedly not very hard to solve) issue? I attached the "output_matrix001.dat"-file as well as an m-file with what MATLAB should extract from the "output_matrix001.dat"-file.

Answer 1

[Walter Roberson]

The default for fread is uint8=>double . You need to specify a precision, probably '*single' or 'single=>double'

Answer 2

[dawaske]

Thank you for this very helpful comment! Now the numbers seem to be right (except for the one on the top left) independent of whether I use "*single' or 'single=>double' (what's the difference between those two options anyway?); however, there is still something wrong with the order:

Do you have any idea how to fix this?

EDIT

ok so playing around a little with fread I found that if I write

h=fread(fid,[16,31],'single=>double');

instead of

h=fread(fid,[31,16],'single=>double');

I get this:

which is almost what I it's supposed to look like except the matrix is transposed of course, the number in the top left is still wrong and the last column (or line in this case) is missing.

[Walter Roberson]

Remember that arrays fill down the columns first, so it is common to have to specify the number of columns first and rows second and then transpose the results.

Fortran binary files are permitted to have embedded structure. Fortran does not have any standards about the representation of binary files, except that the write and read needs to be consistent for any one compiler (on any given operating system). "Stream of bytes with no internal markers" is the portable but not guaranteed by the Fortran standards. Traditionally Fortran binary files were record oriented and so needed internal information about the record length (at least).

[James Tursa]

Note: The new Fortran way of writing streams of bytes is the ACCESS="STREAM" option in the OPEN statement. This should hopefully be more portable that simply using FORM="UNFORMATTED".