cranelift: add a new resumable_trapnz instruction;

[bnjbvr]

This is useful to have to allow resumable_trap to happen in loop
headers, for instance. This is the correct way to implement interrupt
checks in Spidermonkey, which are effectively resumable traps. Previous
implementation was using traps, which is wrong, since traps semantically
can't be resumed after.

[github-actions]

Subscribe to Label Action

cc @bnjbvr

This issue or pull request has been labeled: "cranelift", "cranelift:area:aarch64", "cranelift:meta"

Thus the following users have been cc'd because of the following labels:

• bnjbvr: cranelift

To subscribe or unsubscribe from this label, edit the .github/subscribe-to-label.json configuration file.

Learn more.

[bnjbvr]

Just for context: the reason why I added a new instruction, rather than creating manually the blocks and use the resumable_trap instruction is that when lowering wasm to clif, we can't mutate the CFG in place (we can't create new blocks). Allow mutation of the CFG at this place would be nice in the future, and could allow new things (of the same kind that the IR template proposal would allow), but this is out of scope for what appears to be really a bug fix here.

[bnjbvr]

Fuzz-target failure is being discussed in XAMPPRocky/remove_dir_all#19.

[cfallin]

LGTM, thanks!

[fitzgen]

https://github.com/bytecodealliance/wasmtime/blob/master/cranelift/codegen/src/regalloc/safepoint.rs#L54 should probably switch to something like

if pos.func.dfg[inst].is_resumable_trap() { ... }

that returns true for both resumable_trap and resumable_trapnz.

[fitzgen]

Thanks!

[bnjbvr]

@fitzgen Thanks! Added something to that effect. (It didn't seem worth to make it a flag visible from the meta crate, but let me know if you think it's better (maybe for consistency? probably the other way around would be better imo))

[fitzgen]

(It didn't seem worth to make it a flag visible from the meta crate, but let me know if you think it's better (maybe for consistency? probably the other way around would be better imo))

Not 100% sure what you're getting at here, but the changes in this PR look great to me!