Bug 161639 – MapHash should do constant folding when it has a constant argument and its legal to hash that value

• Component: JavaScriptCore
• Status: RESOLVED FIXED

[Saam Barati]

....

[Saam Barati]

Created attachment 288544 [details]
WIP

I think this is the patch. Lets see what EWS thinks.

[Saam Barati]

Created attachment 288552 [details]
patch

[Saam Barati]

Created attachment 288554 [details]
patch

[Filip Pizlo]

Comment on attachment 288554 [details]
patch

View in context: https://bugs.webkit.org/attachment.cgi?id=288554&action=review

> Source/JavaScriptCore/dfg/DFGAbstractInterpreterInlines.h:987
> +                // We don't want this JSValue to devolve into a double so
> +                // we cast it to an int32_t. When we actually use this
> +                // value, we will use it as if it were unsigned.
> +                setConstant(node, JSValue(static_cast<int32_t>(*hash)));

Does MapHash ordinarily return an int32?  If so, then this comment would be less confusing as: "Although C++ code uses uint32_t for the hash, the closest type in DFG IR is Int32 and that's what MapHash returns.  So, we have to cast to int32_t to avoid large unsigned values becoming doubles."

Also, I would use jsNumber(int32_t) instead of JSValue().

[Saam Barati]

Created attachment 288594 [details]
patch for landing

[WebKit Commit Bot]

Comment on attachment 288594 [details]
patch for landing

Clearing flags on attachment: 288594

Committed r205819: <http://trac.webkit.org/changeset/205819>

[WebKit Commit Bot]

All reviewed patches have been landed.  Closing bug.