Bug 202833 – GenerateAndAllocateRegisters can trivially elide self moves at end of liveness

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

[Keith Miller]

GenerateAndAllocateRegisters can trivially elide self moves at end of liveness

[Keith Miller]

Created attachment 380708 [details]
Patch

[Saam Barati]

Comment on attachment 380708 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=380708&action=review

> Source/JavaScriptCore/b3/air/AirAllocateRegistersAndStackAndGenerateCode.cpp:54
> +        if (!reg)
> +            return;

should we assert here that nothing in m_currentAllocation points to tmp?

[Keith Miller]

Comment on attachment 380708 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=380708&action=review

>> Source/JavaScriptCore/b3/air/AirAllocateRegistersAndStackAndGenerateCode.cpp:54
>> +            return;
> 
> should we assert here that nothing in m_currentAllocation points to tmp?

My intention was that would be handled in the lower loop. I suppose the lower loop could say a disallowed register is allocated to tmp. But that seems very unlikely and probably wouldn’t break anything anyway. Maybe I’m missing something though?

[Keith Miller]

Mac debug builder just seems to be down...

[Saam Barati]

Comment on attachment 380708 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=380708&action=review

>>> Source/JavaScriptCore/b3/air/AirAllocateRegistersAndStackAndGenerateCode.cpp:54
>>> +            return;
>> 
>> should we assert here that nothing in m_currentAllocation points to tmp?
> 
> My intention was that would be handled in the lower loop. I suppose the lower loop could say a disallowed register is allocated to tmp. But that seems very unlikely and probably wouldn’t break anything anyway. Maybe I’m missing something though?

nope, it's handled. Ignore me.

[WebKit Commit Bot]

Comment on attachment 380708 [details]
Patch

Clearing flags on attachment: 380708

Committed r250999: <https://trac.webkit.org/changeset/250999>

[WebKit Commit Bot]

All reviewed patches have been landed.  Closing bug.

[Radar WebKit Bug Importer]

<rdar://problem/56178257>

[Truitt Savell]

It looks like the changes in https://trac.webkit.org/changeset/250999/webkit

has caused 3 tests to time out on Mac Debug:
wasm/iframe-parent-postmessage.html
wasm/iframe-postmessage.html
wasm/window-postmessage.html

Not these tests are taking 25 seconds to pass, or they time out. before they were taking only a few seconds to pass. I was able to reproduce this locally by running the three tests in iterations.

[Keith Miller]

(In reply to Truitt Savell from comment #9)
> It looks like the changes in https://trac.webkit.org/changeset/250999/webkit
> 
> has caused 3 tests to time out on Mac Debug:
> wasm/iframe-parent-postmessage.html
> wasm/iframe-postmessage.html
> wasm/window-postmessage.html
> 
> Not these tests are taking 25 seconds to pass, or they time out. before they
> were taking only a few seconds to pass. I was able to reproduce this locally
> by running the three tests in iterations.

Opened https://bugs.webkit.org/show_bug.cgi?id=203050 with a fix.