In article <1993Jan20.124731.7725@hubcap.clemson.edu> owhadi@sirocco.rice.edu (Eric Owhadi) writes:
>
>Does anyone knows if the CMfortran optimisation is good enough to generate good data parallel code ? ie: Can one do better using a lower level programming model (and more sweat) (with the CMMD library and message passing routines).
Well, that is moot, since we have found CMF 2.0 beta2 to produce incorrect
code unless "-g" is included. In other words, forget it until 2.1 is
released sometime in March.
There is currently no way to compile vectorized code for nodes.