Binary file discussion with John and all

Hi John and all,

Thank you very muchh for all the help.I have tried to proceed further.

Now, I have extracted the data from [b]my[/b] finite element software/program (Software A) and am now in process of writing the binary file needed for another finite element software (Software B).

I have got the information from the software B from its documentation about the variables used,etc.

I'm in the process of running an example and validated that the data extracted from software A is correct.

I now have written the binary file as follows;

  character*40 file_name   
  integer nelems  
  
  integer(KIND=8) :: NEL8,NEL4,NEL2,NUMMAT8,NUMMAT4
  integer(KIND=8) :: NUMMAT2,NUMNP,NDIM
  REAL(KIND=8):: X(NDIM,nelems)
  INTEGER(KIND=8)::IX8(9,nelems)

  

  

 
  file_name = 'my_d3pfile.d3plot' 
  
  open (unit=21, file=file_name, form='UNFORMATTED', iostat=iostat) 
  
    !write (*,*) ' File ',trim(file_name),' Opened for writing: iostat=', iostat 

  !Below, the CONTROL data is written to 
  
  write(21) NEL8 !Number of 8 noded brick elements
  write(21) NEL4 !Number of 4 noded quadrilateral elements

  write(21) NEL2  !Number of 2 noded line elements
  write(21) NUMMAT8 !Total number of materials used by 8 noded brick elements
  write(21) NUMMAT4 !Total number of materials used by 4 noded quadrilateral elements
  write(21) NUMMAT2 !Total number of materials used by 2 noded line el;ements
  write(21) NUMNP !Total number of nodal points
  write(21) NDIM  !Total number of dimensions (1,2 or 3)
  
  !Below the geometry data (i.e. the mesh is written to the d3plot file opened above)
  
  write (21) ((X(i,j),i=1,3),j=1,NUMNP)    !Nodal coordinates
  write(21) ((IX8(i,j),i=1,9),j=1,nelems)  !Element connectivity
  
  close(unit=21)
  

 

  end[/code]

Any suggestions?Please help-any suggestion on the next step?

Please help

To address some comments on the sample code you have listed:

Did this subroutine you have listed compile with FTN95 ? I would expect this is not the case.

Comments based on the line number of the listed code copied to NOTEPAD:

Lines 1 & 2 : You appear to be mixing free format and fixed format continue. I’d recommend free format and replace as: subroutine SK09WritingD3PlotFile (NEL8,NEL4,NEL2,NUMMAT8,NUMMAT4, & NUMMAT2,NUMNP,NDIM,nelems,X,IX8)

Line 4 : This is C syntax for a C include file; you need a fortran compatible version of “sc03.h” and use the statement: include 'sc03.ins' ! sc03.ins is (hopefully)converted from sc03.h

Lines 9-12 : (KIND=8 ) is not a valid value of KIND for FTN95. The correct value is probably (KIND=2), although I suggest you use REAL*8 and make the code more portable.

Line 21 : You appear to have ignored or not understood the comments about incompatible binary file formats. This OPEN statement will generate a binary file that can only easily be read by another program that is compiled using FTN95. Is this the case ? ( I suspect not as lines 9-12 look like Intel or Lahey syntax)

Line 39 : The 3 should be changed to NDIM as: write (21) ((X(i,j),i=1,NDIM),j=1,NUMNP)

To answer the second post: I'm not sure about these addresses. Are they byte addresses ? Does the documentation you have been reading discuss the file format ? Reporting byte addresses suggests that possibly the reading program can accomodate both Salford and Intel/Lahey binary file formats and will use their non-standard ACCESS='BINARY', which is similar to Salfords's ACCESS='TRANSPARENT'.

I'd again suggest you should read the documentation for the compiler on the OPEN statement, especially related to unformatted read/write and also the documentation of software B.

John

Thansk again

The file description indicates that the files are 'word addressable fixed length binary files'. This means that the files are fixed length random access files. Fixed length binary files do not have any record structure, as in the <header> and <trailer> as all records are the same length.

The form of the open is : Integer*4 rec_len, iostat Open (unit=21, file='filename', access='DIRECT', RECL=rec_len, iostat=iostat)

The write statement is of the form: write (21,rec=1) numnp, ndim, nel8, nel4, nel2, nummat

To write the node coordinates, you need a write (21,rec=2) ((X(i,j),i=1,ndim),j=1,NUMNP) !Nodal coordinates { note this will probably write multiple records 2, 3, 4... You need to understand this, although to make the record number easier to calculate, the file structure may not permit this }

To create this file you need more guidance on :

1. the value of rec_len, and
2. The contents of the first record (rec=1) as I suspect it's layout is very important. Your reference to 'NDIM has disk address 15' and 'NEL8 has disk address 23' refers to the word address of these variables in the first record. The length description also refers to 4 or 8 byte words. Which one you use needs to be defined. This is important.

There is also a lot of description of the file name conventions and the parameter FEMLEN which appears to describe the maximum file length. (Lahey was limited to 2gb file sizes). My apologies but I didn't read that too carefully ! I do find this description a bit unfriendly. What is 'software B'? If you are defining only node locations and connectivity, I'd expect you will use only 1 file. Multiple files will be for analysis results.

The reading program needs to know the record length and what length unit is being used. From the info listed, it could be described by bytes, 4-byte words or 8-byte words. This needs to be clarified; probably in the first record. This is important. Again I'm puzzled by the ambiguity of the description ! They appear to be trying to make it more difficult than it should be !!

In summary these are fixed length record random access files. You need to write them using the expected layout. I think you need to find a fortran book about this. Good luck

John