Re: complaints about SPEC and a request for FORTRAN validation codes

moudgill@cs.cornell.EDU (Mayan Moudgill) writes:
>Are the FORTRAN programs in SPEC supposed to be Fortran77 compliant? ...


jwreilly@mipos2.intel.com (Jeffrey Reilly) writes:
> No, the SPEC benchmarks are not guaranteed to be Fortran77 compliant.


> SPEC's goal was to take existing programs/applications that people use and
> develop workloads for them to provide a common reference point for
> evaluation. ...


To amplify this, since some of it was even before Jeff's time at SPEC:


1) To survive thru the process, programs have to fit the "defacto
practice" enough to get most of the SPEC steering committee's machines to
run it and get the "same" (i.e., with the tolerances) answers. Hence,
programs that use really unusual features supported by one vendor never
make it.


2) At the very first Benchathon, held at MIPS a long time ago, it was
somewhat of shock to some people who'd carefully implemented stanadrds to
discover:
a) They couldn't compile some program.
b) Most everybody else could.
because there was some widely-used extension that everyone else had done.


Hence, the effect is that SPEC benchmarks tend to obey either dejure
standards *or* defacto ones .... and this is not at all bad.


-john mashey
UUCP: mash@sgi.com, DDD: 415-390-3090
USPS: Silicon Graphics 7U-005, 2011 N. Shoreline Blvd,
Mountain View, CA 94039-7311
--

Return to the comp.compilers page.
Search the comp.compilers archives again.