Path: utzoo!censor!geac!torsqnt!news-server.csri.toronto.edu!cs.utexas.edu!wuarchive!udel!nigel.ee.udel.edu!mccalpin From: mccalpin@perelandra.cms.udel.edu (John D. McCalpin) Newsgroups: comp.benchmarks Subject: Re: SPEC vs. Dhrystone Message-ID: Date: 3 Jan 91 15:01:22 GMT References: <44342@mips.mips.COM> <15379@ogicse.ogi.edu> <44353@mips.mips.COM> <1685@marlin.NOSC.MIL> <15546@ogicse.ogi.edu> <44465@mips.mips.COM> Sender: usenet@ee.udel.edu Distribution: comp.benchmarks Organization: College of Marine Studies, U. Del. Lines: 30 Nntp-Posting-Host: perelandra.cms.udel.edu In-reply-to: mash@mips.COM's message of 3 Jan 91 06:18:59 GMT >>>>> On 3 Jan 91 06:18:59 GMT, mash@mips.COM (John Mashey) said: mash> 1C: Compiler gimmickry mash> For any important benchmark that is small, compilers will get tuned mash> in ways that are absolutely useless in real life. This has happened mash> at least with Whetstone, Dhrystone, and LINPACK. So what optimizations have been performed on the LINPACK 100x100 code that are "absolutely useless" in real life? I am honestly curious, since it seems to me that the vast majority of the time is spent in a pair of loops which repeatedly call SAXPY. SAXPY, in turn, is a short, vectorizable function with a number of parameter checks before the calculation. The only optimizations that I can think of that would be useful are inlining and constant propagation, and they are both generally helpful. Inlining can be useful in short vector codes, especially if combined with compile-time constant propagation, since that would allow most of the checks to be deleted. The checks are (potentially) useful in the BLAS library (though they are redundant when LINPACK calls the BLAS, since the parameters are already checked at least one before the BLAS calls). Nonetheless, optimizing them away by hand removes the check that should be there when the user calls the BLAS routines directly, as I sometimes do.... -- John D. McCalpin mccalpin@perelandra.cms.udel.edu Assistant Professor mccalpin@brahms.udel.edu College of Marine Studies, U. Del. J.MCCALPIN/OMNET