Xref: news2.mv.net comp.os.msdos.djgpp:5903 From: Thomas Demmer Newsgroups: comp.os.msdos.djgpp Subject: Re: Interesting Timings Date: Fri, 12 Jul 1996 09:33:27 +0100 Organization: Lehrstuhl fuer Stroemungsmechanik Lines: 55 Message-ID: <31E60DD7.446B@LSTM.Ruhr-UNI-Bochum.De> References: <31E3CD8F DOT 94A AT anolis DOT bnr DOT usu DOT edu> NNTP-Posting-Host: bvb.lstm.ruhr-uni-bochum.de Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit To: djgpp AT delorie DOT com DJ-Gateway: from newsgroup comp.os.msdos.djgpp James W. Haefner wrote: > > Here are some runtimes for a genetic programming > problem written in C++ executed with Borland 4.5 > (in a Win3.1 DOS box) and djgpp-v2 in Win and straight > DOS using CWSDPMI... Note the overhead associated with > Win3.1 > > Pop 200, Gen 2, Seed 12345: > Borland: realtime=371.74 (sec) > Gnu in win: realtime=148.8 > Gnu in dos: realtime=97.0 > > Pop 800, Gen 1, Seed 12345: > Borland: realtime=940.54 > Gnu in dos: realtime=241.43 This will hopefully shut up everyone complaining about stupid things like .EXE-sizes. If not, here's another thing: Delaunay triangulation of 976 nodes, giving 2932 edges and 1830 elements: IBM RISC/6000, 380, 3AT with 99.3 SPECint92 and 187.2 SPECfp92: realtime 12s 486/100, no name realtime 22s The price for the IBM was around $30000, for the NoName about $2000. Running this on a P166 might be interesting. To be fair, the triangulation is a process of the order n*n, so on larger meshes the balance gets better for the IBM, but you get the idea. If I just had had DJGPP when I wrote my diploma-thesis... Who needs BC, anyway? -- Ciao Tom ************************************************************* * Thomas Demmer * * Lehrstuhl fuer Stroemungsmechanik * * Ruhr-Uni-Bochum * * Universitaetsstr. 150 * * D-44780 Bochum * * Tel: +49 234 700 6434 * * Fax: +49 234 709 4162 * * http://www.lstm.ruhr-uni-bochum.de/~demmer * *************************************************************