This is the mail archive of the
gsl-discuss@sources.redhat.com
mailing list for the GSL project.
RE: [off-topic] Intel's C/C++ compiler ...
- From: Brian Gough <bjg at network-theory dot co dot uk>
- To: Mikael Adlers <mikael at mathcore dot com>
- Cc: gsl-discuss at sources dot redhat dot com
- Date: Fri, 1 Feb 2002 23:04:36 +0000 (GMT)
- Subject: RE: [off-topic] Intel's C/C++ compiler ...
- References: <4DFD94A38EB3104E838576A77AEBB36E166ED2@hermes.mathcore.se>
Mikael Adlers writes:
> there can be an great difference between compilers, for example
> I got a speedup of 1.5 when switching from lcc to MS Visual C++
> (sparse QR routine for matlab), so I think the article may be valid.
> However, they could have stated what compiler flags they where using!
I think you are too kind to them. The article is completely bogus in
my opinion. I don't think they even tried do a good benchmark.
As you can read on their site the people behind the magazine are
actually a marketing company, and are owned by Andover(Slashdot) so
their lack of standards is not a surprise to me.
As for the compiler efficiency my experience is that the variation due
to different options settings on a given compiler is often far greater
than between compilers. So the most important thing to alwys know the
various compilers options well, in particular the -f and -m otions on
gcc, and to profile the code thoroughly with gprof to identify the
bottlenecks.
Brian