Experiences with Fortran 90 compilers
Compilers for Fortran 90 are relatively new and many of them have problems. I have attempted to make SPARSEM, FSPAK90, REMLF90 and BLUPF90 running on many platforms. Although this effort has been successful, it was not painless. The table below shows my experiences with various F90 compilers.

Most of my software development was on the Sun, with the remaining work on Lahey and Absoft/Linux. My experience with the other compilers was limited to getting the programs compile and then run examples.
 
Compiler / version Problem Effects of contacting the customer support General impression
HP 1.0  Bugs with I/O (handled through third party) No response. Workaround done by me. OK after fixes
Lahey 3.5 Minor incompatibilities Help not obtained until a "flame" posted. Incompatibilities were explained as "true" standard. Workaround done by me. A bit idiosyncratic but solid. Good diagnostics.
Sun 1.0.1 I/O bug + transient problems Bug report filed through official channels acknowledged after a few months. Real technical support not obtained until after "crying wolf". Workarounds by me. Several easily avoidable bugs. Execution is slow. Memory usage up to 4 times higher due to unnecessary memory copying.
SGI 7.2 Compile error Compiler bug identified in a day and a workaround suggested OK
Absoft / Linux 5.0 Running problems System limitations and remedies explained in a matter of hours A solid workhorse
Absoft / Windows 5.0 Compile crash Compiler bug identified in a day and a workaround suggested Unix alike.
RS/6000 Makefile problems Xlf90 switches explained by netnews participants Solid.
Digital / OSF 5.1  Crash Problems acknowledged fast; suggested fixes did not solve problems. Workaround found by myself after a lengthy effort. Fast compilation and execution on Alpha.
 Go to previous page

May 6, 1998