Path: utzoo!utgpu!news-server.csri.toronto.edu!cs.utexas.edu!sdd.hp.com!zaphod.mps.ohio-state.edu!brutus.cs.uiuc.edu!apple!rutgers!deejay!gear!cadlab!staff
From: staff@cadlab.sublink.ORG (Alex Martelli)
Newsgroups: comp.lang.fortran
Subject: Re: the f2c program
Message-ID: <249@cadlab.sublink.ORG>
Date: 15 Aug 90 20:15:00 GMT
References: <1990Aug11.202542.9891@ux1.cso.uiuc.edu> <1990Aug12.005836.5356@zoo.toronto.edu> <937@gistdev.gist.com>
Organization: CAD.LAB, Bologna, Italia
Lines: 24

flint@gistdev.gist.com (Flint Pellett) writes:


>f2c seems to do a reasonable job of translating, if you don't want to
>have to work with/on the resulting C code, but only need to compile it.

Yes, that's its stated objective.
I'm still reeling from the shock of the performance of a test doing
2d fft on a 256 x 256 array of random complex numbers on a Sparcstation
1... Sun's Fortran (not the "sparcompiler", we don't have that yet)
took at least 18.2 seconds (and that with -dalign, top optimization,
and everything), while passing the code through f2c and Sun's C (with
-O3, etc) gave 13.8 seconds, and with -R on f2c and -fsingle on cc,
to do single precision floating arithmetic instead of C's double
precision default, 10.2 seconds.  Wow!  What does this say on the
quality of optimization of Sun's old f77...?  Anyway, I'm going to
measure f77 vs f2c+cc on many more workstations - I always did have a
nagging doubt about the quality of workstation Fortran compilers...

-- 
Alex Martelli - CAD.LAB s.p.a., v. Stalingrado 45, Bologna, Italia
Email: (work:) staff@cadlab.sublink.org, (home:) alex@am.sublink.org
Phone: (work:) ++39 (51) 371099, (home:) ++39 (51) 250434; 
Fax: ++39 (51) 366964 (work only; any time of day or night).