Mailing-List: contact cygwin-help AT sourceware DOT cygnus DOT com; run by ezmlm
List-Subscribe: <mailto:cygwin-subscribe AT sources DOT redhat DOT com>
List-Archive: <http://sources.redhat.com/ml/cygwin/>
List-Post: <mailto:cygwin AT sources DOT redhat DOT com>
List-Help: <mailto:cygwin-help AT sources DOT redhat DOT com>, <http://sources.redhat.com/ml/#faqs>
Sender: cygwin-owner AT sources DOT redhat DOT com
Delivered-To: mailing list cygwin AT sources DOT redhat DOT com
Message-ID: <001901c0fb1b$28503dd0$9865fea9@timayum4srqln4>
From: "Tim Prince" <tprince AT computer DOT org>
To: "Anthony B. Williams" <seaship AT pacbell DOT net>, <cygwin AT cygwin DOT com>
References: <000701c0faea$a1c557d0$6171a440 AT anthonyzls2oby>
Subject: Re: Intel Fortran 5.0 for Linux Beta
Date: Fri, 22 Jun 2001 05:59:20 -0700
MIME-Version: 1.0
Content-Type: text/plain;
	charset="iso-8859-1"
Content-Transfer-Encoding: 7bit
X-Priority: 3
X-MSMail-Priority: Normal
X-Mailer: Microsoft Outlook Express 5.50.4522.1200
X-MimeOLE: Produced By Microsoft MimeOLE V5.50.4522.1200

In case you don't insist on going totally Off Topic:

You don't make it clear what you are trying to do, but there's nothing
magic about the Intel compilers.  The Windows versions run on Windows
(and under cygwin) and make .exe's to run on Windows (and under cygwin).
I have built and run Windows MPI jobs using cygwin parallel make and
tcsh with MPI/Pro, icl, ifl, to emulate the original Unix environment.
The linux compilers run on red hat 6.2 and make executables to run on
red hat 6.2.

If g77 will do the job, and you want to try out the -msse2 option of
gcc-3.1 on cygwin, you can build and install the latest binutils, with
the base SECTION_ALIGNMENT parameter in bfd/coff-i386.c set to 4.  You
can install it separately from the cygwin binutils, e.g. in /usr/local,
in order not to break the cygwin g++.  In that case, you must add
the --with-as=<path to execute as> parameter when you configure gcc.
Note also the agreement among several of us that --with-sjlj-exceptions
is required for gcc.  Since gcc doesn't yet use any parallel sse, the
maximum speedup for g77 -msse2 over cygwin
g77 -march=pentiumpro -ffast-math is 70% (assuming that you use aligned
binutils with both), and the typical is none.  I don't know where you
could look for more than a 90% speed increase for parallel sse over
serial sse.  P4 runs well with -O2 and -O3.  Try setting -malign-loops=0
and -malign-jumps=0, or, look for a -mpentium4 option to appear.  You
will note that the method used by gcc-2.95 for INT() (cast float to int)
is far superior to the one used by gcc-3.x.

(OT) The current Intel linux compilers don't support red hat 7.1.  They
run on 6.x and make files which run on 6.x.  The C compiler is clearly
incompatible with 7.x; you might get lucky with ifc only or ifc with gcc
on 7.x, but why are you asking for free support so far beyond the
documented supported environment?  Intel were unprepared for the
interest in Fortran from people wanting to use it as intended.
Disclaimer:  obviously not speaking for my employer....

----- Original Message -----
From: "Anthony B. Williams" <seaship AT pacbell DOT net>
To: <cygwin AT cygwin DOT com>
Sent: Friday, June 22, 2001 12:12 AM
Subject: Intel Fortran 5.0 for Linux Beta


> Can Intel's fast Fortran compiler live happily within Cygwin? That is,
I
> would like to install Cygwin into Win2K Pro on a Pentium 4, download
> Fortran-for-Linux from Intel, substitute that for GNU f77 Fortran with
> appropriate changes to libraries for the link, and run at the 4X
speedup
> plausible under the SSE2 SIMD instruction set on the P4. Is there any
> obvious problem with this approach? The alternative is of course to
dual
> boot Red Hat Linux 7.1 on the same SCSI hard drive that has Win2K now,
and
> proceed as above from there, but there are compatibility issues with
my
> Win2K backup devices (CMSProducts USB ABS external 20 GB hard drive,
plus
> Roxio's GoBack 3.02). Parallel question has been submitted to Intel
Premier
> Support folk, but want both parties to examine for flaws from own
viewpoint.
> Thanks for your time, Tony.



--
Want to unsubscribe from this list?
Check out: http://cygwin.com/ml/#unsubscribe-simple