Contracts::put_code benchmarks

[expenses]

#3885 and #3819 note that put_code can take too long on large wasm files, and cause block production to fail. This PR adds benchmarks for this using compiled ink examples.

[expenses]

Here's the resulting time for erc20 by the way:

wasms/erc20.wasm        time:   [6.4462 ms 6.5215 ms 6.5997 ms]
                        thrpt:  [7.6973 MiB/s 7.7897 MiB/s 7.8806 MiB/s]

This seems pretty fast, so maybe the problems only occur under certain conditions?

[expenses]

@Stefie maybe you could run these benchmarks on your machine?

[gui1117]

this does benchmark put_code in native. I don't remember precisely but thoughts wasm was used always for block production. maybe in wasm the put_code performance are very different 🤔 ?

[expenses]

Oh, good point! What would be the best way to benchmark in wasm then?

[gui1117]

Oh, good point! What would be the best way to benchmark in wasm then?

I don't know @pepyakin ?

[bkchr]

@expenses you can setup a wasm blob like this one: https://github.com/paritytech/substrate/tree/master/primitives/runtime-interface/test-wasm

Expose a function benchmark_put_code or whatever. This function internally calls 1000x put_code. And you call this function from your benchmark. See here for on how to call it.

[Stefie]

@Stefie maybe you could run these benchmarks on your machine?

Sorry, I was out of office yesterday. If you tell me how, I'm happy to do that! I never ran any benchmarking tests for Substrate.

The times I mentioned in #3885 were with the latest Docker image from https://hub.docker.com/r/parity/substrate/tags , btw

[expenses]

@bkchr is this even possible, given that GenesisConfigs and TestExternalities only works under std?

[bkchr]

@expenses please look at the linked code, you create the storage etc from std and just call into wasm to execute the benchmark method.

[expenses]

Closing due to solutions having arised.