Benchmarking is a general approach to compare companies, workprocesses, products and computers
Traditionally computer benchmarks have been used to compare hardware (computers; CPU, Ram, etc), however benchmarks can also be used to evaluate software as to algorithm, operating system, compilers, programming language, etc. Last not least benchmarks can be used to verify hardware and software integerity in cybersecurity.
A good benchmark can defined by its documentation; essentially any algothm can be used for benchmarking as long as the algorithm being used and the software coding can be understood - thus made public. Ideally also needed inputs and outputs are documented as well as compiler and other relevant settings; needed ram, used operatings system. Also a good benchmark will provide a single number to allow to compare different programming languages, compilers, hardware. Special care should be taken in order to allow to compare multicore cpus/hardware. Therefore this single number value should not be an execution time (in units of ns, ms, sec) but rather execution time per cpu. An even better measure is using MFLOPS or MIPS either per CPU or for the complete system.