pjw@usna.navy.mil (Peter J. Welcher) (12/06/89)
Pardon me if this has already been covered.
A user has been benchmarking f77 with various fpa options on a Sun 3/280.
He took a simple REAL benchmark, changed some types to COMPLEX and DOUBLE
COMPLEX types , and noted a slowdown by a factor of about 90. The obvious
assumption is that the COMPLEX library code doesn't use the FPA.
I haven't looked deeply into this (i.e. I scanned TFM but didn't note
anything relevant). (I'm not that interested in Fortran.)
For the f77 hackers:
Am I missing something ?
Should this occur ?
Is this a known bug ?
Is there anything I can do about it ?