wasmtime / PR #8383 Avoid copying the frame for tail call... · git-wasmtime

Stream: git-wasmtime

Topic: wasmtime / PR #8383 Avoid copying the frame for tail call...

Wasmtime GitHub notifications bot (Apr 16 2024 at 17:25):

elliottt opened PR #8383 from elliottt:trevor/tail-calls-riscv64 to bytecodealliance:main:

To mirror the implementation in the x64 backend, switch to eagerly reserving enough space in the incoming argument area for any tail call present in the function being compiled.

NOTE: this doesn't make the change to enable callee-save registers with the tail calling convention on aarch64, but that will be a relatively small change following this PR.

Wasmtime GitHub notifications bot (Apr 16 2024 at 17:44):

github-actions[bot] commented on PR #8383:

Subscribe to Label Action

cc @cfallin, @fitzgen

<details>
This issue or pull request has been labeled: "cranelift", "isle"

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

cfallin: isle

fitzgen: isle

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

Learn more.
</details>

Wasmtime GitHub notifications bot (Apr 16 2024 at 20:43):

elliottt updated PR #8383.

Wasmtime GitHub notifications bot (Apr 16 2024 at 21:04):

elliottt updated PR #8383.

Wasmtime GitHub notifications bot (Apr 16 2024 at 21:08):

elliottt edited PR #8383:

To mirror the implementation in the x64 backend, switch to eagerly reserving enough space in the incoming argument area for any tail call present in the function being compiled.

The diff here is large due to the change of representation for incoming arguments in the riscv64 specific AMode type. Now that we reference incoming arguments with a negative offset from the beginning of the incoming argument area, all references to incoming arguments changed in the riscv64 filetests.

NOTE: this doesn't make the change to enable callee-save registers with the tail calling convention on aarch64, but that will be a relatively small change following this PR.

Wasmtime GitHub notifications bot (Apr 16 2024 at 21:09):

elliottt edited PR #8383:

To mirror the implementation in the x64 backend, switch to eagerly reserving enough space in the incoming argument area for any tail call present in the function being compiled.

The diff here is large due to the change of representation for incoming arguments in the riscv64 specific AMode type. Now that we reference incoming arguments with a negative offset from the beginning of the incoming argument area, all references to incoming arguments changed in the riscv64 filetests. These would have ended up being no-op changes in the disassembly, though I took this opportunity to make them SP relative now, to move closer to frame pointers not being strictly required for tail calls.

NOTE: this doesn't make the change to enable callee-save registers with the tail calling convention on aarch64, but that will be a relatively small change following this PR.

Wasmtime GitHub notifications bot (Apr 16 2024 at 21:09):

elliottt has marked PR #8383 as ready for review.

Wasmtime GitHub notifications bot (Apr 16 2024 at 21:09):

elliottt requested fitzgen for a review on PR #8383.

Wasmtime GitHub notifications bot (Apr 16 2024 at 21:09):

elliottt requested wasmtime-compiler-reviewers for a review on PR #8383.

Wasmtime GitHub notifications bot (Apr 16 2024 at 21:09):

elliottt requested jameysharp for a review on PR #8383.

Wasmtime GitHub notifications bot (Apr 16 2024 at 21:09):

elliottt requested afonso360 for a review on PR #8383.

Wasmtime GitHub notifications bot (Apr 16 2024 at 21:10):

elliottt edited PR #8383:

To mirror the implementation in the x64 backend, switch to eagerly reserving enough space in the incoming argument area for any tail call present in the function being compiled.

The diff here is large due to the change of representation for incoming arguments in the riscv64 specific AMode type. Now that we reference incoming arguments with a negative offset from the beginning of the incoming argument area, all references to incoming arguments changed in the riscv64 filetests. These would have ended up being no-op changes in the disassembly, though I took this opportunity to make them SP relative now, to move closer to frame pointers not being strictly required for tail calls.

As this is the last backend that was waiting to be migrated to the approach of resizing the incoming argument area in the function prologue, I've also deleted the shared code in machinst/abi.rs that's no longer needed.

NOTE: this doesn't make the change to enable callee-save registers with the tail calling convention on aarch64, but that will be a relatively small change following this PR.

Wasmtime GitHub notifications bot (Apr 16 2024 at 21:15):

elliottt submitted PR review.

Wasmtime GitHub notifications bot (Apr 16 2024 at 21:15):

elliottt created PR review comment:

Maybe this would be better written as fp(incoming_arg_area - {offset})?

Wasmtime GitHub notifications bot (Apr 16 2024 at 21:15):

elliottt edited PR review comment.

Wasmtime GitHub notifications bot (Apr 16 2024 at 22:21):

afonso360 submitted PR review:

:wave: Hey,

I don't think I can do a very good review here, I don't remember a lot about how our frames are set up and haven't been following the rest of the tail calls work very closely. So I'd prefer if someone else could review this.

Wasmtime GitHub notifications bot (Apr 16 2024 at 22:21):

afonso360 submitted PR review:

:wave: Hey,

I don't think I can do a very good review here, I don't remember a lot about how our frames are set up and haven't been following the rest of the tail calls work very closely. So I'd prefer if someone else could review this.

Wasmtime GitHub notifications bot (Apr 16 2024 at 22:21):

afonso360 created PR review comment:

Is this island check now done in some common code, or do we no longer need it? I remember having some issues with missing islands when we enabled compressed jumps since with the compressed extensions we have a very short jump range.

Wasmtime GitHub notifications bot (Apr 16 2024 at 23:20):

elliottt submitted PR review.

Wasmtime GitHub notifications bot (Apr 16 2024 at 23:20):

elliottt created PR review comment:

Thanks for catching this, that check isn't being done in any generic code. I'll add this back in!

Wasmtime GitHub notifications bot (Apr 16 2024 at 23:21):

jameysharp submitted PR review.

Wasmtime GitHub notifications bot (Apr 16 2024 at 23:21):

jameysharp submitted PR review.

Wasmtime GitHub notifications bot (Apr 16 2024 at 23:21):

jameysharp created PR review comment:

I'd also take "-{}(incoming_arg)" to signal that the offset is subtracted from the base. I don't think it's worth bikeshedding this very much but I think clarifying that it's a subtraction would help with readability.

Wasmtime GitHub notifications bot (Apr 17 2024 at 00:03):

jameysharp submitted PR review.

Wasmtime GitHub notifications bot (Apr 17 2024 at 00:03):

jameysharp created PR review comment:

On further inspection, there is generic code in VCode::emit that emits islands based on the backend's Inst::worst_case_size() estimate. Both gen_epilogue and these ReturnCall/ReturnCallInd pseudo-instructions need to fit within that worst-case size. These pseudo-instructions might be slightly longer than gen_epilogue right now but they're almost the same, so I think they don't need an explicit island check.

Wasmtime GitHub notifications bot (Apr 17 2024 at 00:18):

elliottt created PR review comment:

I did some more refactoring, and the resulting code generated for a return_call or return_call_indirect should now be pretty close to the same size as gen_epilogue. Lots of chatting with @jameysharp has me convinced that we shouldn't need to check for island emission at this point anymore, and I've removed the manual resets at the call-sites for emit_return_call_common_sequence to ensure that we trigger an assert if this isn't true.

Wasmtime GitHub notifications bot (Apr 17 2024 at 00:18):

elliottt submitted PR review.

Wasmtime GitHub notifications bot (Apr 17 2024 at 00:19):

elliottt updated PR #8383.

Wasmtime GitHub notifications bot (Apr 17 2024 at 00:27):

elliottt updated PR #8383.

Wasmtime GitHub notifications bot (Apr 17 2024 at 00:29):

elliottt updated PR #8383.

Wasmtime GitHub notifications bot (Apr 18 2024 at 16:40):

elliottt updated PR #8383.

Wasmtime GitHub notifications bot (Apr 18 2024 at 21:20):

elliottt updated PR #8383.

Wasmtime GitHub notifications bot (Apr 18 2024 at 21:28):

jameysharp submitted PR review:

I'm convinced this is all correct. Ship it!

Wasmtime GitHub notifications bot (Apr 18 2024 at 22:18):

elliottt merged PR #8383.

Last updated: Dec 23 2024 at 13:07 UTC