Sun, 17 Mar 2013 23:53:19 +0100 |
Thomas Wuerthinger |
Put exception handler entry back.
|
Sun, 17 Mar 2013 23:46:58 +0100 |
Thomas Wuerthinger |
Remove last usages of scratch register. Separate runtime calls from normal method calls in LIR.
|
Sun, 17 Mar 2013 22:57:47 +0100 |
Thomas Wuerthinger |
Explicitely allocate r10 for deoptimization operation.
|
Sun, 17 Mar 2013 21:27:09 +0100 |
Thomas Wuerthinger |
Enable convert deopts to guards optimization.
|
Sun, 17 Mar 2013 21:25:29 +0100 |
Thomas Wuerthinger |
Rename field.
|
Sun, 17 Mar 2013 21:23:14 +0100 |
Thomas Wuerthinger |
Do not schedule earliest for guard lowering.
|
Sun, 17 Mar 2013 21:20:39 +0100 |
Thomas Wuerthinger |
Support deoptimizing before the entry to a synchronized method.
|
Sun, 17 Mar 2013 21:14:35 +0100 |
Thomas Wuerthinger |
Remove PlaceHolderNode class definition.
|
Sun, 17 Mar 2013 19:49:46 +0100 |
Thomas Wuerthinger |
Add test case for deoptimizing out of a synchronized method. Made guard lowering phase use earliest schedule.
|
Sun, 17 Mar 2013 19:39:28 +0100 |
Thomas Wuerthinger |
Make scheduling phase customizable.
|
Sun, 17 Mar 2013 19:36:10 +0100 |
Thomas Wuerthinger |
Removed unused phase.
|
Sun, 17 Mar 2013 19:20:35 +0100 |
Thomas Wuerthinger |
Fix guard canonicalization.
|
Sun, 17 Mar 2013 17:36:29 +0100 |
Thomas Wuerthinger |
Remove unused code in LabelRef.
|
Sat, 16 Mar 2013 21:55:55 +0100 |
Christian Humer |
Build-graal needs to be versioned for non mx builds.
|
Sat, 16 Mar 2013 21:55:48 +0100 |
Christian Humer |
Build-graal needs to be versioned for non mx builds.
|
Sat, 16 Mar 2013 18:24:47 +0100 |
Christian Humer |
Added build-graal.xml .hgignore. Since it is generated there is no need to version it.
|
Sat, 16 Mar 2013 18:23:36 +0100 |
Christian Humer |
Removed build-graal.xml since it is generated it should not be versioned.
|
Sat, 16 Mar 2013 18:22:39 +0100 |
Christian Humer |
Fixed build-graal.xml generation was out of sync with build-graal.xml.
|
Sat, 16 Mar 2013 16:11:40 +0100 |
Christian Humer |
Fixed not specializing nodes were named Null instead of Default.
|
Sat, 16 Mar 2013 15:53:48 +0100 |
Christian Humer |
Cleanup.
|
Sat, 16 Mar 2013 15:53:04 +0100 |
Christian Humer |
Fixed minor bug for implicit parameters in sourcegen.
|
Sat, 16 Mar 2013 15:52:40 +0100 |
Christian Humer |
Fixed messages were sometimes not shown in sourcegen.
|
Sat, 16 Mar 2013 15:51:58 +0100 |
Christian Humer |
Fixed specialization id generation.
|
Fri, 15 Mar 2013 21:18:47 +0100 |
Christian Humer |
Merge.
|
Fri, 15 Mar 2013 21:18:33 +0100 |
Christian Humer |
Final fields of base node can be optionally passed to builtin specialization method. And a few fixes.
|
Fri, 15 Mar 2013 20:31:11 +0100 |
Doug Simon |
removed unused project dependency
|
Fri, 15 Mar 2013 20:08:10 +0100 |
Doug Simon |
Merge.
|
Fri, 15 Mar 2013 20:07:15 +0100 |
Doug Simon |
re-enabled frame omission
|
Fri, 15 Mar 2013 20:06:19 +0100 |
Doug Simon |
moved AMD64ConvertSnippets to (new) project com.oracle.graal.snippets.amd64
|
Fri, 15 Mar 2013 18:51:58 +0100 |
Doug Simon |
Merge.
|
Fri, 15 Mar 2013 18:50:39 +0100 |
Doug Simon |
fixed issue preventing frame omission
|
Fri, 15 Mar 2013 18:16:15 +0100 |
Doug Simon |
refactored HotSpotAMD64LIRGenerator into its own file
|
Fri, 15 Mar 2013 18:09:19 +0100 |
Doug Simon |
made build-graal.xml be generated by mx
|
Fri, 15 Mar 2013 17:52:55 +0100 |
Gilles Duboscq |
Merge
|
Fri, 15 Mar 2013 17:26:52 +0100 |
Gilles Duboscq |
Merge
|
Fri, 15 Mar 2013 17:26:22 +0100 |
Gilles Duboscq |
Remove orphaned condition in implicit null check detection
|
Fri, 15 Mar 2013 16:57:43 +0100 |
Gilles Duboscq |
Guards with different anchors should be commoned out in the lowering phase
|
Fri, 15 Mar 2013 13:36:24 +0100 |
Gilles Duboscq |
Make ValueAnchor canonicalization work in more cases
|
Fri, 15 Mar 2013 13:35:45 +0100 |
Gilles Duboscq |
Fix implicit null checks depending on themselves
|
Thu, 14 Mar 2013 21:36:28 +0100 |
Gilles Duboscq |
Implement implicit null check on HIR during guard lowering
|
Thu, 14 Mar 2013 14:02:23 +0100 |
Gilles Duboscq |
Remove dead implicit null checks code in LIRGenerator
|
Thu, 14 Mar 2013 13:43:57 +0100 |
Gilles Duboscq |
Cleanup branching and jumps in LIRGenerator: JumpOp/BranchOp do not need a state. In emitBranch, emitNullCheckBranch, emitCompareBranch, emitOverflowCheckBranch and emitIntegerTestBranch both targets are always provided.
|
Thu, 14 Mar 2013 13:25:28 +0100 |
Gilles Duboscq |
Remove emitDeoptimizeOnOverflow and emitGuardCheck.
|
Fri, 15 Mar 2013 15:38:55 +0100 |
Doug Simon |
Merge.
|
Fri, 15 Mar 2013 15:38:27 +0100 |
Doug Simon |
AMD64 HotSpot backend now models RBP as an incoming parameter that must be preserved until the end of the method. For non-leaf methods, the value is preserved in the special stack slot required by the HotSpot runtime for walking/inspecting frames of such methods.
|
Thu, 14 Mar 2013 18:03:00 +0100 |
Doug Simon |
fixed StackOverflowError in VirtualObject.toString()
|
Fri, 15 Mar 2013 02:18:46 +0100 |
Thomas Wuerthinger |
Disable deopt to guard conversion.
|
Fri, 15 Mar 2013 02:18:15 +0100 |
Thomas Wuerthinger |
Merge.
|
Thu, 14 Mar 2013 19:20:43 +0100 |
Thomas Wuerthinger |
Make OptimisticOptimizations enum public.
|
Thu, 14 Mar 2013 19:13:40 +0100 |
Thomas Wuerthinger |
Enable exception probabilities for operations by default.
|
Thu, 14 Mar 2013 19:00:42 +0100 |
Thomas Wuerthinger |
Make optimistic optimizations customizable.
|
Thu, 14 Mar 2013 14:15:11 +0100 |
Andreas Woess |
windows build fix: server+graal build needs TIERED; include graal in client build
|
Thu, 14 Mar 2013 14:35:53 +0100 |
Doug Simon |
moved logic for reserving a special area/slot in a frame (e.g., for use during deoptimization) out of FrameMap and into platform specific backend class
|
Thu, 14 Mar 2013 14:08:59 +0100 |
Doug Simon |
LIRGenerator is transmitted across backend passes instead of just the LIR so that backend-specific, per-compilation information can be attached to the backend-specific LIRGenerator object
|
Thu, 14 Mar 2013 13:13:59 +0100 |
Christian Humer |
Merge.
|
Thu, 14 Mar 2013 13:13:27 +0100 |
Christian Humer |
Fixed unsupported generic generation which could also take void as a type.
|
Thu, 14 Mar 2013 13:11:14 +0100 |
Christian Humer |
Fixed element ordering of methods which were inconsistent from javac to ecj.
|
Thu, 14 Mar 2013 13:10:29 +0100 |
Christian Humer |
Fixed behaviour of useSpecializations for Generic specializations.
|
Thu, 14 Mar 2013 12:19:01 +0100 |
Thomas Wuerthinger |
Fix convert deoptimize to guard phase.
|
Thu, 14 Mar 2013 01:14:23 +0100 |
Thomas Wuerthinger |
Fix import.
|