Learn in the textbookPerformance Evaluation
DefinitionA benchmark (ベンチマーク) is a method that actually runs a workload fixed as the standard on the system being evaluated and uses measured values such as processing time to evaluate and compare performance.
ExplanationBecause every system is measured with the same workload under the same conditions, results can be compared fairly even across different models and configurations. The workload is chosen to match what the system will be used for. It is also called a benchmark test (ベンチマークテスト).
AnalogyIt is like a test where everyone solves the same questions within the same time limit and ability is compared by score.
Memory hookThe word originally meant a mark cut by surveyors as a reference for height, and the sense of "a reference for comparison" carried over to performance evaluation.