From maintainers-request at octave dot org Mon Jan 24 09:23:47 2005 Subject: Re: Benchmark test of Octave on Windows. From: David Bateman To: Ole Jacob Hagen Cc: maintainers at octave dot org Date: Mon, 24 Jan 2005 16:24:59 +0100 Ole Jacob Hagen wrote: > > > Here is the timings on Octave-2.1.64 linked with ATLAS, but built with > gcc-3.2.3 on Cygwin. > > Octave Benchmark 2 > ================== > Number of times each test is run__________________________: 3 > > I. Matrix calculation > --------------------- > Creation, transp., deformation of a 1500x1500 matrix (sec): 1.039 > 800x800 normal distributed random matrix ^1000______ (sec): 0.264 You don't have the randn function from octave-forge installed for the MINGW build, which was much slower here > Sorting of 2,000,000 random values__________________ (sec): 1.204 > 700x700 cross-product matrix (b = a' * a)___________ (sec): 0.2917 > Linear regression over a 600x600 matrix (c = a \ b') (sec): 0.165 > ------------------------------------------------------ > Trimmed geom. mean (2 extremes eliminated): 0.4309 > > II. Matrix functions > -------------------- > FFT over 800,000 random values______________________ (sec): 0.857 > Eigenvalues of a 320x320 random matrix______________ (sec): 0.5477 > Determinant of a 650x650 random matrix______________ (sec): 0.198 > Cholesky decomposition of a 900x900 matrix__________ (sec): 0.18 > Inverse of a 400x400 random matrix__________________ (sec): 0.1207 > ------------------------------------------------------ > Trimmed geom. mean (2 extremes eliminated): 0.2692 > Spot-on the same, no surprise.. > III. Programmation > ------------------ > 750,000 Fibonacci numbers calculation (vector calc)_ (sec): 0.7607 This one was much slower. If I remember correctly this uses lots of memory, so you might need to run the script two or more time to get a reasonable time for this test > Creation of a 2250x2250 Hilbert matrix (matrix calc) (sec): 0.8133 > Grand common divisors of 70,000 pairs (recursion)___ (sec): 0.367 > Creation of a 220x220 Toeplitz matrix (loops)_______ (sec): 0.93 > Escoufier's method on a 37x37 matrix (mixed)________ (sec): 1.399 This aren't as bad as they could be for the MINGW built... All in all pretty good.. Cheers David -- David Bateman David dot Bateman at motorola dot com Motorola CRM +33 1 69 35 48 04 (Ph) Parc Les Algorithmes, Commune de St Aubin +33 1 69 35 77 01 (Fax) 91193 Gif-Sur-Yvette FRANCE The information contained in this communication has been classified as: [x] General Business Information [ ] Motorola Internal Use Only [ ] Motorola Confidential Proprietary