vlib.x: add a go like benchmark, that estimates automatically how many iterations are needed, to get a statistically significant result

[skeris]

For my task, I need a benchmark that measures memory consumption. I checked the default benchmark library and didn't find that feature. So I decided to write it myself.

As advised in the Telegram chat, "just rewrite the golang library to vlang".

I checked the golang benchmark code and decided it was strange and overly complicated due to its use as a command-line tool.

So I decided to create my own library.

Basically, I copied the memory measurement logic and the prediction of the number of iterations from golang to estimate the expected benchmark evaluation time. And from vlang, I got the logic of using the benchmark as a library, not as a command-line tool.

How to use it:

• Create a Benchmark using the constructor, provide a closure to benchmark, optionally provide the expected number of iterations or estimation time.
• Run the run_benchmark method.
• Wait until it ends.
• Use BenchmarkResult as you need - by default, it prints, but you can send it to a chat or to a database.

But I'm not sure if I'm using the gc_heap_usage fields correctly. I was confused by the boehm_gc documentation, it's too hard for me. So if anyone can help me with that - you are welcome

[Delta456]

There's already a benchmark library. How different is it from that library?

[JalonSolov]

The current benchmark library only measures speed, not memory.

[spytheman]

There are other differences too - the existing benchmark does not auto calibrate the number of iterations, till a given time passes, it relies on your code to run the loop and iterate manually.