Bug 79617

Summary: Implement support for op_negate and op_bitnot in the DFG JIT.
Product: WebKit Reporter: Gavin Barraclough <barraclough>
Component: JavaScriptCoreAssignee: Gavin Barraclough <barraclough>
Status: RESOLVED FIXED    
Severity: Normal CC: fpizlo, ggaren, sam
Priority: P2    
Version: 528+ (Nightly build)   
Hardware: Unspecified   
OS: Unspecified   
Attachments:
Description Flags
Fix
none
Remove op_bitnot
sam: review+
ArithNegate fpizlo: review+

Gavin Barraclough
Reported 2012-02-26 19:04:40 PST
To keep the IR simple, we can implement these as sub & xor operations.
Attachments
Fix (33.81 KB, patch)
2012-02-26 22:39 PST, Gavin Barraclough 		no flags
DetailsFormatted DiffDiff
Remove op_bitnot (25.17 KB, patch)
2012-02-27 10:18 PST, Gavin Barraclough 		sam: review+
DetailsFormatted DiffDiff
ArithNegate (22.43 KB, patch)
2012-02-27 15:22 PST, Gavin Barraclough 		fpizlo: review+
DetailsFormatted DiffDiff
Show Obsolete (1) View All Add attachment proposed patch, testcase, etc.
Gavin Barraclough
Comment 1 2012-02-26 22:39:28 PST
Created attachment 128955 [details] Fix
Gavin Barraclough
Comment 2 2012-02-26 23:25:05 PST
Gah, just finished testing, there is a bug here. To implement negate as a subtract, it is -0.0 - X, rather than 0 - X, d'oh, should have remembered that. Optimizing this would probably going to mean adding special cases for subtract-from-negative-zero throughout the code, so it's probably easier to just keep op_negate, and add support to the DFG JIT. In the case of op_bitnot, removing this was a nice simplification so I'll probably stick with that approach.
Gavin Barraclough
Comment 3 2012-02-27 10:18:27 PST
Created attachment 129064 [details] Remove op_bitnot
Sam Weinig
Comment 4 2012-02-27 10:22:40 PST
Comment on attachment 129064 [details] Remove op_bitnot View in context: https://bugs.webkit.org/attachment.cgi?id=129064&action=review > Source/JavaScriptCore/ChangeLog:8 > + Remove op_bitnop รข this is redundant, ~x === x^-1. Weird unicode character. > Source/JavaScriptCore/ChangeLog:10 > + Please put a comment about removing the canUse part of ResultType.
Gavin Barraclough
Comment 5 2012-02-27 10:26:49 PST
op_bitnot change landed in r109007.
Gavin Barraclough
Comment 6 2012-02-27 15:22:43 PST
Created attachment 129110 [details] ArithNegate
Gavin Barraclough
Comment 7 2012-02-27 16:32:01 PST
Fixed in r109038.
Note You need to log in before you can comment on or make changes to this bug.