f77 Optimizer (Summ

[Nikolaos Fotis]

Until now, there were two responses to my previous posting. Many thanks to
the responders, for clarifying some aspects of f77.  Here are (slightly
edited) their responses. I'll try again with SunOS 4.0 when we get it, but
I doubt if we'll have in the near future a machine with more than 4 MBytes
(Sigh..)

______
>from mcvax!Sun.COM!khb (Keith Bierman - SPD Languages Marketing -- MTS)

v1.2 of f77 (which is the current one) has a very large number of
improvements over early releases. Since you are still running 3.5, I
assume you are running f77 1.05 ... which is a serious mistake. 

If you want good performance from Sun compilers you should go with the
latest release of sunOS. I came onboard in September and have
experienced NO OS related bugs (but then my job is to nail down
compiler problems...). 

For 4.x it is imperative that you configure your kernels (not take the
GENERIC default) and you really should have 8Mb of memory.

______
>from mcvax!oconnor!porsche!miker (Mike Raffety)

1.  In order to compile a decent-sized FORTRAN program in any sort of
optimized fashion, you needs LOTS of compute power (we have several
programs that aren't terribly large, and they compile for 20-30 minutes on
a Sun-3/260 and use up to 30 MB of swap and up to 20 MB virtual image.
There's NO WAY you're going to compile that on a 3/50 with 4 MB with any
sort of -O flag (you're probably getting some sort of error message from
iropt, the global optimizer, such as "leaves out of order", which
indicates it's working file is corrupt).

Mine comment:
Sorry, but I forgot to say that the -O3 flag worked. The problem appears
when I try to combine -O with any other flags, like -mc68020, etc.
Finally, the whole story is somewhat mysterious, but that's the reason
we have computers:
We're here to make their work, not ours :-)

Nikolaos  Fotis		e-mail: mcvax!ariadne!theseas!nfotis
			or (outside Eunet) try nfotis at theseas.UU.NET
Home:	16 Esperidon St., Halandri, Athens
	GREECE, 152 32