From mboxrd@z Thu Jan 1 00:00:00 1970 X-Spam-Checker-Version: SpamAssassin 3.4.4 (2020-01-24) on polar.synack.me X-Spam-Level: X-Spam-Status: No, score=-1.9 required=5.0 tests=BAYES_00 autolearn=ham autolearn_force=no version=3.4.4 X-Google-Language: ENGLISH,ASCII-7-bit X-Google-Thread: 1014db,304c86061dc69dba X-Google-Attributes: gid1014db,public X-Google-Thread: 109fba,304c86061dc69dba X-Google-Attributes: gid109fba,public X-Google-Thread: 1094ba,2746b4897cd9baa7 X-Google-Attributes: gid1094ba,public X-Google-Thread: 103376,5cb36983754f64da X-Google-Attributes: gid103376,public X-Google-ArrivalTime: 2004-02-12 17:32:30 PST Path: archiver1.google.com!news2.google.com!news.maxwell.syr.edu!peer1.stngva01.us.to.verio.net!news.verio.net!sea-read.news.verio.net.POSTED!not-for-mail Newsgroups: comp.lang.ada,comp.lang.c,comp.lang.c++,comp.lang.fortran Subject: Re: Ada performance (was No call for Ada ) References: <20040206174017.7E84F4C4114@lovelace.ada-france.org> <54759e7e.0402101819.95cec1d@posting.google.com> <88dc613b.0402121520.bf939f8@posting.google.com> Followup-To: comp.lang.fortran From: lindahl@pbm.com (Greg Lindahl) Organization: a guest of Shadow Island Games X-Newsreader: trn 4.0-test72 (19 April 1999) Originator: lindahl@xkey.com (Greg Lindahl) X-Original-NNTP-Posting-Host: h42-rt.sv.meer.net Message-ID: <402c2929$1@news.meer.net> X-Original-Trace: 12 Feb 2004 17:32:25 -0800, h42-rt.sv.meer.net Date: Fri, 13 Feb 2004 01:32:28 GMT NNTP-Posting-Host: 209.157.152.10 X-Complaints-To: abuse@verio.net X-Trace: sea-read.news.verio.net 1076635948 209.157.152.10 (Fri, 13 Feb 2004 01:32:28 GMT) NNTP-Posting-Date: Fri, 13 Feb 2004 01:32:28 GMT Xref: archiver1.google.com comp.lang.ada:5510 comp.lang.c:22198 comp.lang.c++:18956 comp.lang.fortran:4923 Date: 2004-02-13T01:32:28+00:00 List-Id: In article <88dc613b.0402121520.bf939f8@posting.google.com>, MSG wrote: >I replaced the dead Java group with comp.lang.fortran - maybe they'll >clue us in on the matrix multiplication performance issues (or at >least keep Adaists from feeding us strange explanations). Please don't cross-post a cross-posted flamewar into new newsgroups. If the Ada people are bothering you, stop bothering them. The Fortran view is that if you want a fast matrix multiply, you'll use the built-in language matmul(), or you'll call an appropriate BLAS 3 subroutine. This results in getting cache-blocked tweaked-assembly routines. Alternately, use a compiler which has built-in cache blocking, such as that generated by the KAP preprocessor, or the SGI IRIX or PathScale Opteron compilers. Followups to comp.lang.fortran. -- greg (disclaimer: I work for, but don't speak for, PathScale.)