Summary: | DFG SSA should use PutLocal/KillLocal instead of SetLocal to communicate what is flushed to the stack and when | ||||||
---|---|---|---|---|---|---|---|
Product: | WebKit | Reporter: | Filip Pizlo <fpizlo> | ||||
Component: | JavaScriptCore | Assignee: | Filip Pizlo <fpizlo> | ||||
Status: | RESOLVED FIXED | ||||||
Severity: | Normal | CC: | barraclough, ggaren, mark.lam, mhahnenb, msaboff, oliver, sam | ||||
Priority: | P2 | ||||||
Version: | 528+ (Nightly build) | ||||||
Hardware: | All | ||||||
OS: | All | ||||||
Bug Depends on: | |||||||
Bug Blocks: | 137307, 137168 | ||||||
Attachments: |
|
Description
Filip Pizlo
2014-09-29 19:48:39 PDT
Probably the best way to solve this problem is to have DFG SSA use some name other than "SetLocal" for this node. That way the SSA equivalent of the SetLocal NodeType can have NodeMustGenerate as a default flag. We could call this new node PutLocal. Created attachment 238988 [details]
the patch
Comment on attachment 238988 [details] the patch View in context: https://bugs.webkit.org/attachment.cgi?id=238988&action=review r=me > Source/JavaScriptCore/ChangeLog:36 > + - PutLocal means that that a value is to be stored to the stack. It makes a flush available. > + - KillLocal means that the value stored to the stack is no longer available for the purposes > + of OSR (i.e. it no longer accurately corresponds to what that actual bytecode variable > + would have been, so you have to fall back on node availability). I wonder if we can get the flush logic into the names of these nodes. FlushLocal KillFlushLocal ? (In reply to comment #3) > (From update of attachment 238988 [details]) > View in context: https://bugs.webkit.org/attachment.cgi?id=238988&action=review > > r=me > > > Source/JavaScriptCore/ChangeLog:36 > > + - PutLocal means that that a value is to be stored to the stack. It makes a flush available. > > + - KillLocal means that the value stored to the stack is no longer available for the purposes > > + of OSR (i.e. it no longer accurately corresponds to what that actual bytecode variable > > + would have been, so you have to fall back on node availability). > > I wonder if we can get the flush logic into the names of these nodes. > > FlushLocal > KillFlushLocal > > ? FlushLocal is probably better than PutLocal KillFlushLocal is a mouthful! I'll land this patch as-is but I've filed: https://bugs.webkit.org/show_bug.cgi?id=137307 Landed in http://trac.webkit.org/changeset/174165 |