#8 speed issues

[klaus]

Well, it is not really a bug, but may be of interest.

We run/develop a scientific software package since quite
a few years. We tested and run it an various 32/64 Linux
machines. Our standard compiler is the intel compiler
(ifort).

When I compile with G95, the runtime performace is
significantly lower ... sometimes by a factor of ten.
(in the number crunching routines)

Is there anything, I can do with options? Do the intel
guys mess around with the floating point operations to
spead it up?
Btw. Does G95 avoid multiplication, when a factor is 0.0?
Can the slow down be related to huge arrays and memory
read/write
things?

Interestingly we encountered a similar performance
breakdown with a pgf compiler on a 64 bit machine.

I would appreciate a short note, which might help us.