Next: Predefined Calculations, Previous: What Is A Benchmark, Up: Introduction [Contents][Index]
A complementary approach to the analysis on the final code is the extraction of some fragment of the code and to measure its execution performance separated of the whole process. These measurements are usually compared between different implementation algorithms of the fragment, reference implementations and/or baselines. This process is usually called micro-benchmark.
The main unit of MicroBenchmark framework is the test case (see Test Reference.) A test case encapsulates the desired process of measurement, including its preparation and finalization (see Execution Stages.)
The execution environment of a test case can be constrained (see Test Constraints.) Size constraints produce a combinatorial number of executions to be measured (see Test Dimensions.) The concrete values are provided to the test through the state object (see State Reference.)
A collection of test cases is called a suite (see Suite Reference.) Its execution is store the collected data into a report (see Report Reference.)