Changes

<font style="font-size:90%">From ARM White paper</font><br /> When first released, the Dhrystone benchmark fulfilled a useful function – at least it gavean alternative indicator to vendors’ literal MIPS ratings. However, almost twenty yearslater, there are undoubtedly better benchmarks available for measuring processorperformance. Code profiling demonstrates that Dhrystone is not a particularly good predictor of real-world performance, especially for complex embedded systems. The benchmark’ssusceptibility to being optimized, yielding large variations in results, is a considerableweakness if the intent is to use the numbers for comparative purposes. ARM’s ownmeasurements, presented here, demonstrate that each method of optimization canincrease the apparent DMIPS/MHz performance, and when combined can improve theDhrystone score by up to three times.ARM’s customers still occasionally ask for Dhrystone MIPS data, and so ARM willcontinue to provide these numbers. However, ARM does not recommend that Dhrystoneresults be used as part of any embedded processor evaluation exercise, due to its manynoted deficiencies. Where ARM provides a Dhrystone figure, it will be based on itsstandard tool suite under the conditions outlined above, and will therefore be 100%publicly and independently reproducible.