EGCS performance ?

[Joe Buck]

>  I've taken the liberty of modifying some of the LFK kernels so they don't
> depend on compilers recognizing specific oddly coded loops or optimizing
> obsolete syntax. 

Many of those oddly coded loops are tuning for older Cray machines,
which do best if you can feed them lots of 64-element vector operations.
> This helps the commercial compilers as much as it does egcs,
> but makes the results more meaningful as a test of performance on reasonable
> source code.

However, many scientific programmers will be using Cray-tuned "dusty deck"
Fortran for a long time to come, based on the "if it isn't broke, don't
fix it" principle.  I know that the Stanford SUIF people put a lot of
effort into high-level loop transformations to completely rearrange these
kinds of nested loops.