Benchmark execution times of runtime calls using Wasm execution.

[jimpo]

These benchmarks were useful in measuring the impact of #3011.

[bkchr]

We already have such benchmarks here: https://github.com/paritytech/substrate/blob/master/core/sr-api-macros/benches/bench.rs

[jimpo]

@bkchr Thanks, didn't know about those. I don't see any benchmarks for a full block import. Do you know if that exists? Also, I need to check that going through the runtime API doesn't add much overhead over the more targeted benchmark using the lower-level APIs, but I assume it wouldn't.

[bkchr]

The runtime api uses the same api internally as you are using. I'm not aware of any full block import benchmark.

[kianenigma]

I have a question: How close are these to a realistic execution?

(I am pretty sure I asked this in some technical chat but have no memory of what was the answer): How often do we, in reality, spawn a new executor?

About the code: we also directly benchmark executor in node/executor. If you really care about something within the executor, then it makes the most sense to actually do it from there imo.

[bkchr]

As he wanted to measure the impact of the runtime caching, it is more like a high level benchmark anyway.

[pepyakin]

As an aside, I wonder if test-runtime would be a better place than sr-api-macros for housing the benches?

[jimpo]

Closing for now until I have time to rework it.