Compile pricing for testnet

[rachel-bousfield]

This PR adds a basic 14 million gas charge to compile a Stylus program. In the future we'll price compilation based on the wasm, which will cut costs considerably for reasonable programs. In the interest of preventing DoS, however, we'll charge 14 million as a conservative back-stop.

Note that this PR includes all the plumbing necessary to introduce a complex pricing policy.

payForCompilation(burner burn.Burner, info *wasmPricingInfo) {
    return burner.Burn(11000000)
}

The above function doesn't do anything right now, but could use the info parameter to charge different amounts based on various factors like wasm size, memory footprint, number of functions, etc.

This PR makes additional changes, including

• Diagramming the go stack when crossing into Rust
• Adding additional constraints on the shape of wasms
• An extra safety check on the ability to form Machine's during native / jit execution
• Not using rayon in the JIT prover (not needed anyways due to block validation being embarrassingly parallel)

[codecov]

Codecov Report

Merging #113 (7a07fd4) into stylus (6a8b0bf) will decrease coverage by 0.03%.
Report is 1 commits behind head on stylus.
The diff coverage is 30.31%.

@@            Coverage Diff             @@
##           stylus     #113      +/-   ##
==========================================
- Coverage   56.20%   56.17%   -0.03%     
==========================================
  Files         272      272              
  Lines       41274    41397     +123     
==========================================
+ Hits        23197    23255      +58     
- Misses      15707    15787      +80     
+ Partials     2370     2355      -15     

[rauljordan]

No comments. Nice annotations, explainers, and good cleanup

[tsahee]

part of my onchain compilation PR was realising that we don't really need a machine for these things.. a module is enough.
See from_user_binary in machine.rs:
https://github.com/OffchainLabs/stylus/pull/95/files#diff-dd6bc6f8903cfa5ee9704487f1a234c34957f007b05cc7e8ae47e482f8ead0e3

(I can remove new_user_stub when I merge my PR later)

[rachel-bousfield]

That makes sense, we just need to make sure that whatever we do, we revert native execution if arbitrator fails (rn that includes creating a machine, so let’s keep this for now but remove in your PR)

[tsahee]

during parsing?

[rachel-bousfield]

Right now those are the same thing, though your PR will change that. Def feel free to remove parse too :)

[tsahee]

generally: some of the changes here will not be necessary in onchian-compilation world..
After these changes: stylus_compile is only called on compilation and not on callUserWasm. Also - styls_compile must compute the arbitrator machine to get the module hash, so it could just as well also compute compilation cost.

So after that stylus_parse_wasm will be removed and stylus_compile will return WasmPricingInfo.

[rachel-bousfield]

That’s works too once we have that, though do note you’ll need to bring a few things over the boundary in order to compute and assign gas deductions (like the memory model params).

[tsahee]

Not critical, but I don't really like this..
msg is basically only used to print error message, right?
Can't we just do something like

func extraData(data []byte, status UserStatus, debug bool) string

And use that when printing error messages?

[rachel-bousfield]

I don’t love it either, but I’m trying to avoid extra variables that are only used in the debug case (we drop data depending on the kind of failure, so you have to move the .into argument to before the call and give it a new name)