NEW 210694
VM should have a bottomCellForSetLocal that is understood to be a true bottom value by the DFG 
https://bugs.webkit.org/show_bug.cgi?id=210694
Summary VM should have a bottomCellForSetLocal that is understood to be a true bottom...
Keith Miller
Reported 2020-04-18 12:26:46 PDT
Right now we handle bottom values semantically needed for a SetLocal by vending a JSValue that roughly matches the predicted value for the non-bottom path. This comes up for op_iterator_open/next. Unfortunately, this still causes AI to be unhappy. If we had a constant cell that was understood by the DFG to be a bottom value (and presumably crashes whenever accessed), the DFG would generate better code for op_iterator_open/next.
Attachments
Add attachment proposed patch, testcase, etc.
Note You need to log in before you can comment on or make changes to this bug.