Enable inlining by default

[alexcrichton]

This commit switches the default configuration of Wasmtime to enable cross-module inlining by default. This means that components with multiple modules will, by default, inline small functions across these modules. Additionally modules which use the GC proposal will have inlining performed by default. Intra-module inlining without GC, however, will continue to not happen.

This required two minor updates in tests. One is to disable inlining when the exact call stack is expected and another is to disable inlining when the stack-switching proposal is enabled because it currently falls over in Cranelift on the stack_switch instruction.

[alexcrichton]

cc @fitzgen and @cfallin, this is initially a draft since I wanted to have somewhere to ask the question of "does anyone know a reason to not do this?" and figured that a minimal PR is good enough. So, to follow-through with that, do y'all know of a reason to not enable this by default? My main motivation would be optimizations like #13194 where for WASIp3 we'll want to have that on-by-default.

[cfallin]

do y'all know of a reason to not enable this by default?

At an algorithmic level we'd want to make sure our heuristics are robust enough that we don't have blowups (in compilation time or in size of binary); but I guess we've been fuzzing this for a long time now and haven't seen any such issues. That would have been my only concern, otherwise 👍

[fitzgen]

Yeah I feel fine as far as our heuristics and fuzzing go. Ofc the heuristics need tweaks to be Good but that can happen at any time as separate work.

The one thing I would be tentative about here is the impact on compile times. I remember seeing some cases where they were quite a bit slower with inlining (although I never systematically investigated this) and I'd want to make sure that (a) compilation times for single core modules is not affected and (b) that we are okay with whatever slow down we actually see for compiling components.

[github-actions]

Label Messager: wasmtime:config

It looks like you are changing Wasmtime's configuration options. Make sure to
complete this check list:

• If you added a new Config method, you wrote extensive documentation for
  it.

  Details

  Our documentation should be of the following form:

  Short, simple summary sentence.
  
  More details. These details can be multiple paragraphs. There should be
  information about not just the method, but its parameters and results as
  well.
  
  Is this method fallible? If so, when can it return an error?
  
  Can this method panic? If so, when does it panic?
  
  # Example
  
  Optional example here.
  

• If you added a new Config method, or modified an existing one, you
  ensured that this configuration is exercised by the fuzz targets.

  Details

  For example, if you expose a new strategy for allocating the next instance
  slot inside the pooling allocator, you should ensure that at least one of our
  fuzz targets exercises that new strategy.

  Often, all that is required of you is to ensure that there is a knob for this
  configuration option in wasmtime_fuzzing::Config (or one
  of its nested structs).

  Rarely, this may require authoring a new fuzz target to specifically test this
  configuration. See our docs on fuzzing for more details.

• If you are enabling a configuration option by default, make sure that it
  has been fuzzed for at least two weeks before turning it on by default.

---

Details

To modify this label's message, edit the .github/label-messager/wasmtime-config.md file.

To add new label messages or remove existing label messages, edit the
.github/label-messager.json configuration file.

Learn more.

[alexcrichton]

Discussion at the Cranelift meeting today concluded:

1. One avenue is that it should be confirmed that the quite-different inlining compilation path, when no inlining actually happens, should be double-checked to have little-to-no overehead.
2. Another avenue is that we should strive to, by default, not tamper with wasm backtraces. While it's possible to use frame tables built for debugging to recover inlining decisions after-the-fact that's not currently implemented. Given that we should probably change the defaults to only allowing inlining of intrinsics, for example, and have a dedicated mode for "inline some wasm things, but beware backtraces will be weird"

[alexcrichton]

I've morphed this PR into #13254 and #13250, so I'll close this in favor of those.