Date: Thu, 30 Mar 2000 17:09:16 +1000 From: "Andrew Reilly" <areilly@nsw.bigpond.net.au> To: "Igor B. Bykhalo" <goshik@binep.ac.ru> Cc: "-questions@FreeBSD" <freebsd-questions@FreeBSD.ORG>, "-stable@FreeBSD" <freebsd-stable@FreeBSD.ORG> Subject: Re: Doing math - need advice Message-ID: <20000330170916.A5632@gurney.reilly.home> In-Reply-To: <01bf9a0e$ffb5fe00$0000e9c1@serv2.binep.ac.ru> References: <01bf9a0e$ffb5fe00$0000e9c1@serv2.binep.ac.ru>
next in thread | previous in thread | raw e-mail | index | archive | help
On Thu, Mar 30, 2000 at 09:12:59AM +0300, Igor B. Bykhalo wrote: > We are going to use mathematical package > (EPR spectra simulation). Program relies > on BLAS, LAPACK and ARPACK libraries - all > in Fortran 77. Program itself can be built > using gcc, egcs or pgcc. > > Questions: > > 1. What is the recommended way to build > f77 libraries to get maximum performance? > Should we use g77 from 4.0-Release, or > we can stay on 3-Stable and use f2c + pgcc > (from ports)? cd /usr/ports/math/blas; make install cd /usr/ports/math/lapack.... > 2. Does anyone know incarnations of the mentioned > libraries in C language? There's a clapack on netlib.org, so there's probably a c blas behind that somewhere. Dunno about ARPACK. > 3. Where to read some general advices on the optimization > of math perfomance (under FreeBSD)? :) There are a few generic pentium optimisation pages on the network. These may or may not be relevant to P6 based machines. Up to a certain level of performance, and for many code structures, memory system behaviour, and therefore data layout and access will dominate the application performance. You're therefore best served by getting things right, and studying locality of reference, rather than compiler optimisations. Compiler optimisations (and even assembly language tweaking) can come last, after some thorough profiling. (Yes, FreeBSD supports profiling.) -- Andrew To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-questions" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20000330170916.A5632>