Bug 172944 – OMG tier up checks should be a patchpoint

• Status: RESOLVED
• Resolution: FIXED
• Component: New Bugs

[Keith Miller]

OMG tier up checks should be a patchpoint

[Keith Miller]

Created attachment 312039 [details]
Patch

[Saam Barati]

Comment on attachment 312039 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=312039&action=review

> Source/JavaScriptCore/wasm/WasmB3IRGenerator.cpp:913
> +            MacroAssembler::Call call = jit.nearCall();
> +
> +            ScratchRegisterAllocator::restoreRegistersFromStackForCall(jit, registersToSpill, RegisterSet(), numberOfStackBytesUsedForRegisterPreservation, extraPaddingBytes);

This code looks wrong. You're not restoring SP after the call. Are tests passing? I'm also looking at our other wasm call code, and wondering how any of it works since none of it restores SP.

> Source/JavaScriptCore/wasm/WasmThunks.cpp:120
> +    jit.call(GPRInfo::argumentGPR2);

ditto

[Saam Barati]

Comment on attachment 312039 [details]
Patch

Ignore me, C CC says SP is callee save.

[WebKit Commit Bot]

Comment on attachment 312039 [details]
Patch

Clearing flags on attachment: 312039

Committed r217812: <http://trac.webkit.org/changeset/217812>

[WebKit Commit Bot]

All reviewed patches have been landed.  Closing bug.

[Ryan Haddad]

Reverted r217812 for reason:

This change caused test failures on arm64.

Committed r217833: <http://trac.webkit.org/changeset/217833>

[Ryan Haddad]

(In reply to Ryan Haddad from comment #6)
> Reverted r217812 for reason:
> 
> This change caused test failures on arm64.
> 
> Committed r217833: <http://trac.webkit.org/changeset/217833>
<rdar://problem/32590017>

[Keith Miller]

Created attachment 312121 [details]
Patch for landing

[WebKit Commit Bot]

Comment on attachment 312121 [details]
Patch for landing

Clearing flags on attachment: 312121

Committed r217861: <http://trac.webkit.org/changeset/217861>

[WebKit Commit Bot]

All reviewed patches have been landed.  Closing bug.