alexcrichton requested fitzgen for a review on PR #9190.
alexcrichton requested wasmtime-compiler-reviewers for a review on PR #9190.
alexcrichton opened PR #9190 from alexcrichton:refactor-call-info
to bytecodealliance:main
:
This commit is aimed at making it easier to extend
CallInfo
amongst the backends with more fields. Most backends already share almost all the same fields except for minor differences so this commit moves backends to all roughly the same track for handling this to enable having one definition ofCallInfo
that is shared everywhere.
fitzgen submitted PR review.
fitzgen created PR review comment:
I don't think this is anything we necessarily need to address right now, and some backends are already like this, but by not having the name in the
CallInfo
we are inflating the size of the instruction a bit. The downside is that now we would need direct vs indirect variants, if we kept everything inside the boxedCallInfo
.
fitzgen created PR review comment:
This seems unrelated?
fitzgen created PR review comment:
Nevermind, I see that this is factored out from the call implementation
bjorn3 submitted PR review.
bjorn3 created PR review comment:
Maybe this should accept CallInfo pre-boxed? That would save some copying around.
alexcrichton updated PR #9190.
alexcrichton updated PR #9190.
alexcrichton submitted PR review.
alexcrichton created PR review comment:
It's true yeah, I was assuming that the x64 style of moving indirect-vs-not into the
Inst
was ok to apply to all backends. If we end up undoing that though backends could also have something likestruct X64CallInfo { info: CallInfo, ... }
or similar
alexcrichton has enabled auto merge for PR #9190.
alexcrichton updated PR #9190.
alexcrichton has disabled auto merge for PR #9190.
alexcrichton commented on PR #9190:
Ok, as predicted size assertions got the best of me. I've refactored the other direction for backends to include the name/destination in the
CallInfo
uniformly across all backends.That was a large-ish change though so mind taking another look @fitzgen?
fitzgen commented on PR #9190:
I don't have time today, and I'm going backpacking next week, so while I don't mind taking a look when I get back you might want to find another reviewer :-p
alexcrichton updated PR #9190.
alexcrichton updated PR #9190.
alexcrichton commented on PR #9190:
@elliottt would you be ok reviewing this?
elliottt submitted PR review:
This all looks good to me!
alexcrichton updated PR #9190.
alexcrichton has enabled auto merge for PR #9190.
alexcrichton merged PR #9190.
Last updated: Dec 23 2024 at 12:05 UTC