Closed adlerjohn closed 1 year ago
What is our current strategy for pricing?
Will it have to be based upon real world benchmarking?
Run lots of benchmarks with as clean a setup as possible. That's really the only way.
@adlerjohn can we placeholder some initial values for now, than adjust them based upon real world benchmarks?
This would help just for implementation and testing before we get to benchmarking.
The implementation can just choose some sane values. There's no reason to pollute the specs with that.
@adlerjohn this could likely be closed, as we now have a gas schedule benchmarking framework that we are iterating on.
Don't we want them properly documented somewhere? If I wanted to know individual op costs (which I do when writing optimisers) where should I look?
We are working on a framework to autogenerate a yaml file of the costs based on bechmarks, and also generate a rust file which contains the instantiated gas costs.
As we expect the gas costs to change over time as we add more functionality like merkle trees or optimize the vm/client, I'm not sure if statically binding them to the specs is the best move right now.
Opcodes currently don't have gas costs. These need to be specified.