Thu, 21 Mar 2013 15:55:00 +0100 |
Doug Simon |
minor punctuation fix
|
Thu, 21 Mar 2013 15:38:42 +0100 |
Doug Simon |
moved tests for HotSpot specific method substitutions into HotSpotMethodSubstitutionsTest
|
Thu, 21 Mar 2013 15:36:31 +0100 |
Doug Simon |
temporary file is removed when unittests fail
|
Thu, 21 Mar 2013 14:38:35 +0100 |
Doug Simon |
rename: GraalIntrinsics -> GraalMethodSubstitutions
|
Thu, 21 Mar 2013 14:35:06 +0100 |
Doug Simon |
rename: originalMethod -> originalMember
|
Thu, 21 Mar 2013 13:45:27 +0100 |
Doug Simon |
small documentation fix
|
Thu, 21 Mar 2013 18:04:03 +0100 |
Christian Humer |
Merge.
|
Thu, 21 Mar 2013 18:03:32 +0100 |
Christian Humer |
Added annotation processor for verifing @ClassSubstitution and @MethodSubstitution annotation at compile time.
|
Thu, 21 Mar 2013 13:20:32 +0100 |
Christian Humer |
Merge.
|
Thu, 21 Mar 2013 13:13:07 +0100 |
Christian Humer |
Merge.
|
Thu, 21 Mar 2013 13:12:36 +0100 |
Christian Humer |
Fixed varargs attribute was not copied in ExectableElements.
|
Thu, 21 Mar 2013 13:12:04 +0100 |
Christian Humer |
Cleanup.
|
Thu, 21 Mar 2013 16:25:26 +0100 |
Roland Schatz |
Generalize FloatingReadPhase to support arbitrary floatable access nodes.
|
Thu, 21 Mar 2013 12:30:44 +0100 |
Doug Simon |
snippet graphs are now stored with a Snippet.class key value in the compiler storage of the snippet method
|
Thu, 21 Mar 2013 12:23:37 +0100 |
Doug Simon |
method substitutions are now stored with a MethodSubstitution.class key value in the compiler storage of the original (i.e., substituted) method
|
Thu, 21 Mar 2013 12:14:09 +0100 |
Doug Simon |
created com.oracle.graal.api.replacements project and moved ClassSubstitution and MethodSubstitution there
|
Thu, 21 Mar 2013 11:59:40 +0100 |
Doug Simon |
refactored MethodSubstitution and MacroSubstitution into top level classes
|
Thu, 21 Mar 2013 11:09:19 +0100 |
Doug Simon |
Merge.
|
Thu, 21 Mar 2013 09:44:47 +0100 |
Doug Simon |
Merge.
|
Wed, 20 Mar 2013 22:30:33 +0100 |
Doug Simon |
rename projects: *snippets* -> *replacements* (GRAAL-178)
|
Wed, 20 Mar 2013 22:23:14 +0100 |
Doug Simon |
rename packages: *snippets* -> *replacements*
|
Wed, 20 Mar 2013 21:41:37 +0100 |
Doug Simon |
rename: SnippetIntrinsificationVerificationPhase -> NodeIntrinsificationVerificationPhase
|
Wed, 20 Mar 2013 21:18:59 +0100 |
Doug Simon |
rename: SnippetsInterface -> Snippets
|
Wed, 20 Mar 2013 20:34:26 +0100 |
Doug Simon |
removed unnecessary interface
|
Wed, 20 Mar 2013 20:23:40 +0100 |
Doug Simon |
comment fix
|
Wed, 20 Mar 2013 20:23:23 +0100 |
Doug Simon |
renamings as part of a move away from the overloading of the "snippet" term for various code replacement/substitution mechanisms towards it being solely for the concept of IR lowering
|
Wed, 20 Mar 2013 18:16:45 +0100 |
Doug Simon |
rename: SnippetIntrinsificationPhase -> NodeIntrinsificationPhase
|
Wed, 20 Mar 2013 18:11:59 +0100 |
Doug Simon |
rename: IntrinsificationTest -> MethodSubstitutionTest
|
Wed, 20 Mar 2013 16:15:35 +0100 |
Doug Simon |
extra test case for System.identityHashCode()
|
Mon, 18 Mar 2013 19:18:42 +0100 |
Bernhard Urban |
unittest: call wrapper to avoid long command lines
|
Wed, 20 Mar 2013 18:38:55 +0100 |
Gilles Duboscq |
IGV should not crash when comparing graphs with references to deleted nodes
|
Wed, 20 Mar 2013 18:38:28 +0100 |
Gilles Duboscq |
Refactor GuardLoweringPhase: use the same iterator code for implict null checks and guard lowering
|
Wed, 20 Mar 2013 16:27:07 +0100 |
Gilles Duboscq |
Add asFixedNode to FloatingAccessNode so that GuardLoweringPhase supports any type of FloatingAccessNode
|
Wed, 20 Mar 2013 08:22:35 -0700 |
Christian Wimmer |
Remove flag and optimization for object method arguments in the register allocator. Needs more investigation of GC impact on HotSpot
|
Wed, 20 Mar 2013 14:01:43 +0100 |
Lukas Stadler |
Merge
|
Mon, 11 Mar 2013 18:41:33 +0100 |
Lukas Stadler |
added RegisterFinalizer.virtualize
|
Wed, 20 Mar 2013 13:41:44 +0100 |
Lukas Stadler |
DynamicCounterNode for debugging/profiling purposes
|
Wed, 20 Mar 2013 13:41:13 +0100 |
Lukas Stadler |
small change VirtualInstanceNode constructor
|
Wed, 20 Mar 2013 11:34:31 +0100 |
Lukas Stadler |
simplify VirtualInstanceNode
|
Wed, 20 Mar 2013 11:33:00 +0100 |
Lukas Stadler |
change naming in Access interface
|
Wed, 20 Mar 2013 11:32:41 +0100 |
Lukas Stadler |
don't use Access interface in ReadNode.canonicalizeRead
|
Wed, 20 Mar 2013 11:30:17 +0100 |
Lukas Stadler |
allow multiple identities in MemoryCheckpoint interface
|
Wed, 20 Mar 2013 12:00:18 +0100 |
Gilles Duboscq |
If IgnoreUnrecognizedVMOptions is set then the Graal checks for unsupported options is just a warning instead of an error
|
Wed, 20 Mar 2013 11:59:25 +0100 |
Gilles Duboscq |
Disable scaladacpo specs because it sometimes returns non-zero while there is no visible failure
|
Tue, 19 Mar 2013 14:55:58 +0100 |
Gilles Duboscq |
Fix version parsing in mx
|
Mon, 18 Mar 2013 14:45:36 +0100 |
Gilles Duboscq |
Use implicit null checks for all null checks
|
Tue, 19 Mar 2013 17:03:13 +0100 |
Lukas Stadler |
clean up unused true/false constants in ConditionalEliminationPhase
|
Wed, 20 Mar 2013 00:06:09 +0100 |
Thomas Wuerthinger |
Speical case ExceptionObjectNode in deopt to guard conversion.
|
Tue, 19 Mar 2013 23:08:03 +0100 |
Thomas Wuerthinger |
Bug fix and additional assertions in deoptimize to guard phase.
|
Tue, 19 Mar 2013 20:48:53 +0100 |
Thomas Wuerthinger |
Merge.
|
Tue, 19 Mar 2013 20:48:45 +0100 |
Thomas Wuerthinger |
Introduce a ControlSinkNode base class.
|
Tue, 19 Mar 2013 20:48:28 +0100 |
Thomas Wuerthinger |
Fixes a corner case when killing the control flowing into a deoptimize node.
|
Tue, 19 Mar 2013 18:05:49 +0100 |
Thomas Wuerthinger |
Merge.
|
Tue, 19 Mar 2013 18:05:35 +0100 |
Thomas Wuerthinger |
Killing cfg path to deoptimization.
|
Tue, 19 Mar 2013 10:00:43 -0700 |
Christian Wimmer |
InvokeWithException cannot be the lastLocationAccess for a FloatingReadNode. Since it is both the invoke and a control flow split, the scheduler cannot schedule anything immediately the invoke. It can only schedule in the normal or exceptional successor - and we have to tell the scheduler here which side it needs to choose by putting in the location identity on both successors.
|
Tue, 19 Mar 2013 17:01:39 +0100 |
Roland Schatz |
Merge.
|