Path: utzoo!censor!geac!torsqnt!news-server.csri.toronto.edu!cs.utexas.edu!sdd.hp.com!usc!snorkelwacker.mit.edu!bloom-beacon!eru!hagbard!sunic!ugle.unit.no!nuug!ulrik!ulrik!oles From: oles@kelvin.uio.no (Ole Swang) Newsgroups: comp.benchmarks Subject: Re: Ole Swang's benchmark: Sum of Harmonic Series Message-ID: Date: 17 Dec 90 17:34:18 GMT References: <44125@mips.mips.COM> <44126@mips.mips.COM> Sender: news@ulrik.uio.no (USENET News System) Organization: University of Oslo, Norway Lines: 37 In-Reply-To: mark@mips.COM's message of 15 Dec 90 18:34:03 GMT >> >>It vectorizes fully on the vectorizing compilers I've tested it on >>(Cray and Convex). It has the advantage over the bc benchmark that >>it's the same code every time. >> >>Cray X/MP 216 0.29 >>MIPS RC6280 3.3 >>MIPS RC3230 Magnum 8.1 >>Convex C 120 8.7 <**** full vectorization ??? >>Sun SPARCstation II 10.0 >>DECstation 5000/200 10.5 >>...etc > >Since the Convex measurement is "surrounded" by little bitty workstations, >it might not have been running the fully vectorized object code. Maybe >someone can replicate the C-120 measurement above and see what the >generated code is doing. > >1/2 baked random idea: this pgm performs divide-and-accumulate, so perhaps >the machines that have a multiply-accumulate atomic instruction (Apollo >DN-10000, Intel i860, IBM RS-6000, etc.) might possibly excel. Or, >perhaps not. >-- > -- Mark Johnson > MIPS Computer Systems, 930 E. Arques M/S 2-02, Sunnyvale, CA 94086 > (408) 524-8308 mark@mips.com {or ...!decwrl!mips!mark} The compiler insists that the loop is fully vectorized - in scalar mode (compiler option -O1), the C120 uses 44 secs. The C120 isn't very fast... The quantum chemical program package Gaussian 88 (tm) runs about an order of magnitude faster on a Cray X/MP. This benchmark, however, runs about 25 times faster on the Cray. Perhaps the C120 is relatively weaker on divisions. -- ----------------------------------------------------------------------- Ole Swang assistant professor, Dept. of Chemistry, U .of Oslo -----------------------------------------------------------------------