gfortran speed benchmarks

[Paul Thomas]

Steve,

>There are several people on c.l.f that are willing to
>tell you why the above statement may not be true.  Of
>course, the author of test_fpu.f90 has a different 
>point of view.
>  
>
Fine!  Where are their alternatives?  I had a limited objective to see 
if gfortran was in the same ballpark as commercial compilers.  test_fpu 
exposed an error in the dependency analysis, as I described on the list 
a while back, and I am just tonight putting  in some time on it.  Other 
tests in the suite have exposed specific defects in gfortran; to wit, 
dot_product and mod deficiencies.

If there are any other useful, free tests out there, I will gladly work 
on them; I extracted as much as I could from the LLNL kernel flops test.

Paul