Bug 207849 – [Wasm] REGRESSION(r256665): Wasm->JS call IC needs to save memory size register

• Status: RESOLVED
• Resolution: FIXED
• Component: JavaScriptCore

[Tadeu Zagallo]

...

[Tadeu Zagallo]

Created attachment 390944 [details]
Patch

[Tadeu Zagallo]

<rdar://problem/59504154>

[Mark Lam]

Comment on attachment 390944 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=390944&action=review

> Source/JavaScriptCore/wasm/js/WebAssemblyFunction.cpp:165
> +    // Pessimistically save callee saves in BoundsChecking mode since the LLInt always bounds checks
> +    return Wasm::PinnedRegisterInfo::get().toSave(Wasm::MemoryMode::BoundsChecking);

When would we ever not want to save the sizeRegister?  Why not get rid of the memoryMode condition and always save the register?  From gripping the code, I don't see a case where toSave() is ever called with anything other than MemoryMode::BoundsChecking.

[Mark Lam]

(In reply to Mark Lam from comment #3)
> > Source/JavaScriptCore/wasm/js/WebAssemblyFunction.cpp:165
> > +    // Pessimistically save callee saves in BoundsChecking mode since the LLInt always bounds checks
> > +    return Wasm::PinnedRegisterInfo::get().toSave(Wasm::MemoryMode::BoundsChecking);
> 
> When would we ever not want to save the sizeRegister?  Why not get rid of
> the memoryMode condition and always save the register?  From gripping the
> code, I don't see a case where toSave() is ever called with anything other
> than MemoryMode::BoundsChecking.

/gripping/grepping/

[Mark Lam]

Comment on attachment 390944 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=390944&action=review

r=me

>> Source/JavaScriptCore/wasm/js/WebAssemblyFunction.cpp:165
>> +    return Wasm::PinnedRegisterInfo::get().toSave(Wasm::MemoryMode::BoundsChecking);
> 
> When would we ever not want to save the sizeRegister?  Why not get rid of the memoryMode condition and always save the register?  From gripping the code, I don't see a case where toSave() is ever called with anything other than MemoryMode::BoundsChecking.

I'm wrong, there's still one place where we may not pass MemoryMode::BoundsChecking.

[Tadeu Zagallo]

*** Bug 207843 has been marked as a duplicate of this bug. ***

[WebKit Commit Bot]

Comment on attachment 390944 [details]
Patch

Clearing flags on attachment: 390944

Committed r256766: <https://trac.webkit.org/changeset/256766>

[WebKit Commit Bot]

All reviewed patches have been landed.  Closing bug.