Multi-core benchmarks

[tillkruss]

It would be quite useful, if it's not too complex, to benchmark code on several processes simultaneously and collect results from each process, Relay is a good example.

Some code might scale quite well with each core creating higher throughput, other code might be bottlenecked and adding more processes/workers won't do anything.

[dantleech]

Probably this could be done via. executor config #[Executor('microtime', ['processes' => 4]) where the iterations will be distributed over the processes.

[sgmihai]

Has this been implemented yet ? Or are we supposed to multiply the score with the number of cores to get an approximate maximum performance number? (not including multi threading?)

[dantleech]

No it hasn't. Currently benchmarks will run in 1 core (i.e. there is only 1 process running). What would your use case be?