Bug 151852 – [JSC] Extend the strength reduction of B3's BitAnd with booleans

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

[Benjamin Poulain]

[JSC] Extend the strength reduction of B3's BitAnd with booleans

[Benjamin Poulain]

Created attachment 266606 [details]
Patch

[Saam Barati]

Comment on attachment 266606 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=266606&action=review

r=me

> Source/JavaScriptCore/b3/B3ReduceStrength.cpp:578
> +                && m_value->child(0)->child(1)->asInt() & 1

I think you could also do: "m_value->child(0)->

[Saam Barati]

Comment on attachment 266606 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=266606&action=review

>> Source/JavaScriptCore/b3/B3ReduceStrength.cpp:578
>> +                && m_value->child(0)->child(1)->asInt() & 1
> 
> I think you could also do: "m_value->child(0)->

Oops I meant to delete that.

[WebKit Commit Bot]

Comment on attachment 266606 [details]
Patch

Clearing flags on attachment: 266606

Committed r193436: <http://trac.webkit.org/changeset/193436>

[WebKit Commit Bot]

All reviewed patches have been landed.  Closing bug.

[Filip Pizlo]

Comment on attachment 266606 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=266606&action=review

> Source/JavaScriptCore/b3/B3ReduceStrength.cpp:869
> +            // Turn this: Branch(BitAnd(bool, xyb1), then, else)
> +            // Into this: Branch(bool, then, else)
> +            if (branch->child(0)->opcode() == BitAnd
> +                && branch->child(0)->child(1)->hasInt()
> +                && branch->child(0)->child(1)->asInt() & 1
> +                && branch->child(0)->child(0)->returnsBool()) {
> +                branch->child(0) = branch->child(0)->child(0);
> +                m_changed = true;
> +            }
> +

You could have made this a rule that converts BitAnd(bool, xyb1) to bool, regardless of the user of the BitAnd.