Add a benchmark of the hang-on-test-failure code path

[saethlin]

This is the code pattern that produces the performance problem in #2273

I figured out what I was stuck on in #2315 (comment). For a while I was just doing let x: &[u8] = &[0u8; 4096]; but that doesn't produce the runtime inside Stack::item_popped that I was looking for, I think because this allocation is never deallocated. But with Vec, I get the profile I'm looking for.

[RalfJung]

😂 the test runs OOM on 32bit machines.^^

[saethlin]

I'm pretty sure this will pass with size 4096 after #2315 due to the ~3x memory reduction that PR provides. But I also think this benchmark makes a poorer case for the implementation details of that PR with the smaller size.

The irony is cruel.

Also I'm pretty sure the fact that the test fails means that cargo miri test with a failing test will currently OOM any 32-bit system. But slowly.

[RalfJung]

We can also just block this on landing #2315. Or increase it back to 4096 in that PR.

[saethlin]

I think that making a benchmark slower as part of an optimization PR would be very confusing if someone were to use the benchmarks to evaluate the impact of PRs later.

So for now I think just sharing this up in a PR makes me feel less dishonest about using a new benchmark to justify an implementation choice, and we can land it after #2315, with the larger size.

[RalfJung]

Can you rebase and increase the size to 4096 again? Does that work now? :D

[RalfJung]

@bors r+

[bors]

📌 Commit 7d9f04f has been approved by RalfJung

It is now in the queue for this repository.

[bors]

⌛ Testing commit 7d9f04f with merge a45d6ef...

[bors]

☀️ Test successful - checks-actions
Approved by: RalfJung
Pushing a45d6ef to master...

[RalfJung]

I just noticed we already have bench-cargo-miri/backtraces/src/main.rs. But that seems to be something else?

[saethlin]

Yes. That benchmark exists because I noticed that I'm pretty sure the Debug impl for BacktraceFrame or whatever I'm formatting there has an oddly low hit rate in the Stack cache.