🚀 go-pugleaf

In article <1jhqkhINNno3@bigboote.WPI.EDU>, you write:
|> I have to write a Fortran program which writes and reads tremendous amount
|> of data to to and from hard-disc.I wrote it in C using binary files and
|> low level I/O.It worked quite well.I tried to use unformatted files in
|> Fortran but they're at least 4 times slower.I used following open, write
|> and read statements
|>     open(8,File='Filename',FORM='unformatted')
|>     write(8)(A(J),J=1,N) 
|>     read(8)(A(J),J=1,N)   ( A is a double )
|> I also tried direct-access unformatted files but i couldn't figure out
|> how to specify RECL most efficiently.
|> I'm working in UNIX.
|> I will greatly appreciate any help that will make I/O in the Fortran 
|> program faster.
|> Please respond to ustundag@blakey.WPI.EDU
|> thanks
|>   
 The first thing to observe is generally the array dimension, because it's
always faster to execute :
      dimension A(N)
      ...
      write(8)A

The second is to look about buffering on you computer : if you have big file
perhaps you can find a way to write and read without buffering.
If direct access is not necessary, forget it : it's always lower that sequential one.
Good luck !

michele.morin@der.edf.fr