Jump to content

Talk:Dhrystone

Page contents not supported in other languages.
From Wikipedia, the free encyclopedia

MDIPS

[edit]
If MDIPS are used with the Dhrystone method, then why isn't it mentioned in this article? Couldn't we add a secion here and then we could "merge" the two articles. The MDIPS page is an almost empty stub. Andrew Powell 20:14, 9 March 2006 (UTC)[reply]

The factor for converting Dhrystone to DMIPS in the page is 1757, but on the reference is 1857. Which is the correct one ?

Outdated

[edit]

The article seems to be a little dated in its praise of Dhrystone for use as a benchmark. I added criticisms of the algorithm and reference to a statement made by the author himself that the benchmark was outmoded and should not be used. Starseeker shkm 03:48, 15 November 2007 (UTC)[reply]

Probably true, but right now I am interested in benchmarks for older computers, running older compilers, where it might not be so bad. But the Instructions_per_second#MIPS page seems to have Dhrystone results for computers too old, and too new. Was it really run on all those computers? Gah4 (talk) 09:09, 13 August 2016 (UTC)[reply]

Unfair criticism and comparison to Coremark

[edit]

The criticism and comparisom section take as a given that benchmarks are supposed to measure the performance of the hardware, as independent as possible from the compiler. The sensitivity tocompiler optimizations is presented purely as a weakness in the article. But when doing benchmarking to compare combined hardware / software systems or when comparing compilers it is a strength. The comparison pretends that all that matters is that both Dhrystone and Coremark can be obtained free of charge. But Dhrystone is free software, while Coremark is non-free software that may not be redistributed and can only be obtained by agreeing to EEMBC's terms. They forbid redistribution. The terms even require that you delete your own copies of old Coremark versions whenever EEMBC releases a new version. See Coremark license. 93.216.193.194 (talk) 08:25, 31 July 2015 (UTC)[reply]

Citation needed?

[edit]

Therefore, an optimizer can replace a string copy with a sequence of word moves without any loops, which will be much faster. This optimization consequently overstates system performance, sometimes by more than 30%[citation needed]. Which is the citation needed for? That optimizers do that, or that the performance change is 30%? Gah4 (talk) 09:07, 13 August 2016 (UTC)[reply]