I/O performance blues

[N.M. Maclaren]

On Nov 26 2013, Arjen Markus wrote:

>I have not studied the code in detail, but what struck is the use of
>BACKSPACE. Could that cause a performance penalty?

BACKSPACE is a disaster area, and always has been, but it is primarily
a cause of unreliability and only fairly rarely performance (though it
can happen).  However, 99% of the time, performance problems in formatted
I/O are due to the conversion.

It doesn't surprise me that gfortran is a lot slower than gcc, because
Fortran I/O is a lot more complicated than C I/O, and so needs a lot more
effort to keep the overheads under control.  I don't know that is where
the problem arises, but is where I would look first.


Regards,
Nick Maclaren.