bnjbvr opened Issue #1617:
This will imply some work in
regalloc.rs
to tell where spilled ref-typed values might be, and some integration in the new backend.cc @julian-seward1 @cfallin
bnjbvr labeled Issue #1617:
This will imply some work in
regalloc.rs
to tell where spilled ref-typed values might be, and some integration in the new backend.cc @julian-seward1 @cfallin
bnjbvr labeled Issue #1617:
This will imply some work in
regalloc.rs
to tell where spilled ref-typed values might be, and some integration in the new backend.cc @julian-seward1 @cfallin
github-actions[bot] commented on Issue #1617:
Subscribe to Label Action
cc @bnjbvr
<details>
This issue or pull request has been labeled: "cranelift"Thus the following users have been cc'd because of the following labels:
- bnjbvr: cranelift
To subscribe or unsubscribe from this label, edit the <code>.github/subscribe-to-label.json</code> configuration file.
Learn more.
</details>
bjorn3 commented on Issue #1617:
This will imply some work in regalloc.rs to tell where spilled ref-typed values might be
Also for refs stored in registers, right?
bnjbvr commented on Issue #1617:
Also for refs stored in registers, right?
Eventually yes, but we don't require it in the short term at least in Spidermonkey, which scans the stack only at times where everything has been spilled onto the stack. (I wonder if that was accidentally working with the old regalloc, because it would inefficiently spill callee-preserved registers across a call...)
julian-seward1 commented on Issue #1617:
Sure; this will require tracking which
VirtualRange
s are ref-typed, or something like that. It might turn into a whole barrel of fun once we take into account copy coalescing, too.
bjorn3 commented on Issue #1617:
It would be nice to immediately do the more general version of tracking which
Value
is stored where. That is necessary to implement debuginfo for locals.
cfallin commented on Issue #1617:
Closing now, given that #1852 and subsequent bugfixes have landed; #1886 will track the wasmtime-specific work (which seems mainly to involve getting the right debug-info bits in place for the stack walk to work).
cfallin closed Issue #1617:
This will imply some work in
regalloc.rs
to tell where spilled ref-typed values might be, and some integration in the new backend.cc @julian-seward1 @cfallin
Last updated: Dec 23 2024 at 12:05 UTC