Tue, 13 Nov 2012 14:41:19 +0100 |
Gilles Duboscq |
Factor out the precondition checking code in ArrayCopySnippets
|
Fri, 30 Nov 2012 12:08:38 +0100 |
Lukas Stadler |
simplified TLAB allocation (don't use + and -)
|
Fri, 30 Nov 2012 11:13:36 +0100 |
Gilles Duboscq |
Merge
|
Fri, 30 Nov 2012 11:11:13 +0100 |
Gilles Duboscq |
Do not print code installation failures from Java.
|
Fri, 30 Nov 2012 10:40:07 +0100 |
Christian Haeubl |
changed System.err output to TTY to avoid that DaCapo benchmarks fail digest validation
|
Fri, 30 Nov 2012 08:30:22 +0100 |
Christian Haeubl |
Merge.
|
Thu, 29 Nov 2012 17:19:13 +0100 |
Christian Haeubl |
minor fix
|
Thu, 29 Nov 2012 16:20:42 +0100 |
Christian Haeubl |
Merge.
|
Thu, 29 Nov 2012 15:44:49 +0100 |
Christian Haeubl |
fixed some issues that precluded compilation of Graal compiler methods
|
Thu, 29 Nov 2012 11:00:32 +0100 |
Christian Haeubl |
print a message when optimistic optimizations are disabled
|
Thu, 29 Nov 2012 17:43:48 -0800 |
Christian Wimmer |
Merge
|
Thu, 29 Nov 2012 17:43:09 -0800 |
Christian Wimmer |
Make constructors of Constant private to ensure proper encapuslation of the type-overloaded primitive field; reduce to one constructor that sets all fields.
|
Thu, 29 Nov 2012 16:53:44 -0800 |
Christian Wimmer |
Remove the ResolvedJavaType.Representation for static fields. It is not necessary to emit the constant holder for static fields in the GraphBuilder. Only the VM-specific lowering needs to create the constants, and it can access the VM-specific metadata objects directly.
|
Thu, 29 Nov 2012 15:05:39 -0800 |
Christian Wimmer |
Eliminate bogus conversion that Constant.asLong() performed; simplify asXxx methods so that they do not have control flow.
|
Thu, 29 Nov 2012 22:53:21 +0100 |
Doug Simon |
Merge.
|
Thu, 29 Nov 2012 22:53:03 +0100 |
Doug Simon |
better encoding of dependencies during dependency recording by Graal
|
Thu, 29 Nov 2012 17:57:03 +0100 |
Doug Simon |
removed Graal specific initialization of the ciObjectFactory
|
Thu, 29 Nov 2012 12:17:02 -0800 |
Christian Wimmer |
Cleanup of Kind class: remove use of Unsafe class
|
Thu, 29 Nov 2012 11:27:23 -0800 |
Christian Wimmer |
Cleanup of Kind class: remove isXxx methods
|
Thu, 29 Nov 2012 10:10:03 -0800 |
Christian Wimmer |
Replace ResolvedJavaType.isAssignableTo with isAssignableFrom to be consistent with java.lang.Class
|
Thu, 29 Nov 2012 09:13:59 -0800 |
Christian Wimmer |
Remove ResolvedJavaType.isClass()
|
Thu, 29 Nov 2012 17:13:59 +0100 |
Doug Simon |
Merge.
|
Thu, 29 Nov 2012 17:13:13 +0100 |
Doug Simon |
removed remaining use of the C++ "compiler interface" (i.e. ci) by Graal for installing dependencies
|
Thu, 29 Nov 2012 10:18:32 +0100 |
Doug Simon |
removed unnecessary and incorrect initialization of ciEnv._dependencies
|
Thu, 29 Nov 2012 15:13:19 +0100 |
Gilles Duboscq |
For code installation failure, start the counter so that we don't get an error message if an installation fails shortly after the VM has started
|
Thu, 29 Nov 2012 13:31:48 +0100 |
Gilles Duboscq |
Canoncialize UnsafeLoad/Store
|
Thu, 29 Nov 2012 13:24:08 +0100 |
Gilles Duboscq |
Common out some parts of UnsafeLoad/Store in UnsafeAccess
|
Wed, 28 Nov 2012 20:39:43 +0100 |
Doug Simon |
added metering of code installation failure rate to detect excessive failure caused by overly optimistic assumptions
|
Wed, 28 Nov 2012 19:18:39 +0100 |
Doug Simon |
guard against nmethod allocation failure in implementation of PrintMachineCodeToFile option
|
Wed, 28 Nov 2012 18:48:17 +0100 |
Doug Simon |
made HotSpotResolvedObjectType.findUniqueConcreteSubtype() more conservative for array types to fix issue with frequent invalidation of Graal compiled methods
|
Wed, 28 Nov 2012 18:45:54 +0100 |
Doug Simon |
added toString() implementation to Assumption classes
|
Wed, 28 Nov 2012 14:28:53 +0100 |
Gilles Duboscq |
Merge
|
Wed, 28 Nov 2012 14:07:25 +0100 |
Gilles Duboscq |
HotSpotResolvedJavaType is now the HotSpot implementation of ResolvedJavaType,
|
Wed, 28 Nov 2012 12:18:34 +0100 |
Gilles Duboscq |
Remove unsafe mirror() method from HotSpotTypeUnresolved. Use a HotSpotMirrorHolder interface for HotSpotResolvedJavaType and HotSpotTypePrimitive which always hold valid mirrors.
|
Wed, 28 Nov 2012 14:01:17 +0100 |
Christian Haeubl |
readded IntrinsifyArrayCopyPhase that was removed accidently
|
Wed, 28 Nov 2012 09:01:39 +0100 |
Christian Haeubl |
Merge.
|
Wed, 28 Nov 2012 09:01:15 +0100 |
Christian Haeubl |
disabled control-flow-based inlining
|
Tue, 27 Nov 2012 13:14:30 +0100 |
Christian Haeubl |
re-applied changeset d9930fec7553
|
Tue, 27 Nov 2012 13:02:43 +0100 |
Christian Haeubl |
Merge.
|
Tue, 27 Nov 2012 12:16:47 +0100 |
Christian Haeubl |
Merge.
|
Tue, 27 Nov 2012 12:15:18 +0100 |
Christian Haeubl |
fixes after merge
|
Tue, 27 Nov 2012 12:12:02 +0100 |
Christian Haeubl |
Merge.
|
Tue, 27 Nov 2012 10:49:12 +0100 |
Christian Haeubl |
added some more assertions to ensure that metadata constant are not emitted inline
|
Tue, 27 Nov 2012 10:41:42 +0100 |
Christian Haeubl |
bugfix to prevent metadata constants from being emitted inline
|
Mon, 26 Nov 2012 14:32:48 +0100 |
Christian Haeubl |
fix to avoid emitting a range switch for non-integer keys
|
Mon, 26 Nov 2012 11:02:50 +0100 |
Christian Haeubl |
fixes after merge
|
Mon, 26 Nov 2012 10:57:02 +0100 |
Christian Haeubl |
Merge.
|
Mon, 26 Nov 2012 10:51:39 +0100 |
Christian Haeubl |
minor changes
|
Fri, 23 Nov 2012 14:08:03 +0100 |
Christian Haeubl |
removed high-level interpreter code from HotSpot
|
Fri, 23 Nov 2012 12:06:31 +0100 |
Christian Haeubl |
fixes after merge
|
Fri, 23 Nov 2012 11:50:27 +0100 |
Christian Haeubl |
Merge.
|
Fri, 23 Nov 2012 11:40:17 +0100 |
Christian Haeubl |
fixed logging of inlining decisions
|
Thu, 22 Nov 2012 13:30:14 +0100 |
Christian Haeubl |
first part of refactoring the InliningPhase
|
Thu, 15 Nov 2012 15:10:41 +0100 |
Christian Haeubl |
merged inlining and intrinsification phases
|
Thu, 15 Nov 2012 11:40:50 +0100 |
Christian Haeubl |
increased CompileThreshold to increase reliability of profiling information
|
Tue, 27 Nov 2012 22:28:19 +0100 |
Doug Simon |
replaced MetaUtil.isJavaLangObject() with type.isClass(Object.class)
|
Tue, 27 Nov 2012 22:26:18 +0100 |
Doug Simon |
rename: ResolvedJavaType.isArrayClass() -> ResolvedJavaType.isArray()
|
Tue, 27 Nov 2012 22:24:39 +0100 |
Doug Simon |
removed ResolvedJavaType.toJava() and introduced ResolvedJavaType.isPrimitive()
|
Tue, 27 Nov 2012 20:36:46 +0100 |
Doug Simon |
Merge.
|
Tue, 27 Nov 2012 20:34:15 +0100 |
Doug Simon |
rename: findFieldWithOffset -> findInstanceFieldWithOffset
|
Tue, 27 Nov 2012 20:32:52 +0100 |
Doug Simon |
added test for ResolvedJavaType.findFieldWithOffset
|
Tue, 27 Nov 2012 18:33:13 +0100 |
Doug Simon |
Merge.
|
Tue, 27 Nov 2012 18:30:20 +0100 |
Doug Simon |
removed usages of ResolvedJavaType.toJava() from bytecode interpreter
|
Tue, 27 Nov 2012 18:03:49 +0100 |
Doug Simon |
removed more usages of ResolvedJavaType.toJava()
|
Tue, 27 Nov 2012 17:22:19 +0100 |
Doug Simon |
removed a usage of ResolvedJavaType.toJava()
|
Tue, 27 Nov 2012 18:25:02 +0100 |
Gilles Duboscq |
Make frame omission optional
|
Tue, 27 Nov 2012 17:33:52 +0100 |
Lukas Stadler |
Merge
|
Tue, 27 Nov 2012 17:32:40 +0100 |
Lukas Stadler |
virtualization of unsafeload and unsafestore
|
Tue, 27 Nov 2012 17:29:42 +0100 |
Lukas Stadler |
remove createSignature from CompilerToVM interface
|
Tue, 27 Nov 2012 13:44:16 +0100 |
Lukas Stadler |
remove MonitorEnter interface from DimensionsNode
|
Tue, 27 Nov 2012 13:43:04 +0100 |
Lukas Stadler |
VM support for deferred reads and writes: ScopeDesc, DebugInfo, DebugInfoRecorder
|
Tue, 27 Nov 2012 16:37:02 +0100 |
Doug Simon |
bug fix for HotSpotReslvedJavaType.findUniqueConcreteSubtype() on arrays of interfaces
|
Tue, 27 Nov 2012 16:09:05 +0100 |
Doug Simon |
added unit tests for JavaType, JavaMethod and ResolvedJavaMethod
|
Tue, 27 Nov 2012 15:10:50 +0100 |
Doug Simon |
improved documentation for ResolvedJavaType.findUniqueConcreteSubtype()
|
Tue, 27 Nov 2012 13:03:08 +0100 |
Gilles Duboscq |
Merge
|
Tue, 27 Nov 2012 11:29:06 +0100 |
Gilles Duboscq |
Avoid call to MetaUtil.format in GraphBuilder if it is not necessary
|
Tue, 27 Nov 2012 11:08:02 +0100 |
Gilles Duboscq |
HotSpotSignature.getParameterType should not try eager resolving.
|
Tue, 27 Nov 2012 12:10:41 +0100 |
Lukas Stadler |
Merge
|
Tue, 27 Nov 2012 11:51:13 +0100 |
Lukas Stadler |
PEA: remove shouldRevirtualize logic
|
Tue, 27 Nov 2012 11:21:48 +0100 |
Doug Simon |
added unit tests for ResolvedJavaType
|
Tue, 27 Nov 2012 11:18:34 +0100 |
Doug Simon |
added more #ifdefs for GRAAL specific code
|
Mon, 26 Nov 2012 19:30:54 -0800 |
Christian Wimmer |
Simplify how stamp is preserved after lowering of allocation nodes
|
Mon, 26 Nov 2012 18:58:28 -0800 |
Christian Wimmer |
More flexible handling of stamp for word type: Define a singleton Stamp instance for words that is then re-written to the target-specific primitive word stamp by the WordTypeRewriterPhase. This allows nodes to have a word stamp without any dependency on global or static state.
|
Mon, 26 Nov 2012 18:25:44 -0800 |
Christian Wimmer |
Allow intrinsification without a GraalCodeCacheProvider (workaround until TODO is resolved)
|
Mon, 26 Nov 2012 18:24:08 -0800 |
Christian Wimmer |
close() method for dump handlers should not throw IOException
|
Mon, 26 Nov 2012 18:17:14 -0800 |
Christian Wimmer |
Better implementation of UnsafeCastNode
|
Mon, 26 Nov 2012 18:03:47 -0800 |
Christian Wimmer |
Fix getInstanceFields for arryas and interfaces
|
Mon, 26 Nov 2012 18:21:22 +0100 |
Doug Simon |
Merge.
|
Mon, 26 Nov 2012 18:20:34 +0100 |
Doug Simon |
added tests for instanceof with array types
|
Mon, 26 Nov 2012 16:51:43 +0100 |
Doug Simon |
CompilerToVM.lookupType() now fails with an exception if eagerResolve is true and resolution fails
|
Mon, 26 Nov 2012 16:18:56 +0100 |
Doug Simon |
fixed bug in creation of the name for a HotSpotTypeUnresolved
|
Mon, 26 Nov 2012 16:08:10 +0100 |
Gilles Duboscq |
Insert less placeholder nodes in snippet code
|
Mon, 26 Nov 2012 13:20:00 +0100 |
Lukas Stadler |
only perform PEA iteration if there are analyzable nodes
|
Mon, 26 Nov 2012 13:19:32 +0100 |
Lukas Stadler |
Merge
|
Mon, 26 Nov 2012 12:03:46 +0100 |
Lukas Stadler |
Merge
|
Fri, 23 Nov 2012 15:01:37 +0100 |
Lukas Stadler |
let VirtualObjectNodes react to materialization
|
Mon, 26 Nov 2012 11:22:26 +0100 |
Doug Simon |
fixed implementation of HotSpotResolvedJavaType.findUniqueConcreteSubtype() for array types
|
Mon, 26 Nov 2012 11:21:18 +0100 |
Doug Simon |
cleanly handle a failure to install Graal compiled code due to failed dependency (re)checking
|
Fri, 23 Nov 2012 16:03:39 +0100 |
Gilles Duboscq |
Put code that is used only in ASSERT mode behind #ifdef ASSERT
|
Fri, 23 Nov 2012 12:18:01 +0100 |
Gilles Duboscq |
Rename Invoke.isMegamorphic to isPolymorphic since it's what it is actually used for.
|
Fri, 23 Nov 2012 15:30:48 +0100 |
Doug Simon |
Merge.
|
Fri, 23 Nov 2012 15:30:00 +0100 |
Doug Simon |
added ResolvedJavaType.isClass(Class c) to replace some uses of ResolvedJavaType.toJava()
|
Fri, 23 Nov 2012 09:02:12 +0100 |
Doug Simon |
removed AddressMap utility
|
Fri, 23 Nov 2012 14:01:02 +0100 |
Lukas Stadler |
Merge
|
Fri, 23 Nov 2012 13:44:03 +0100 |
Lukas Stadler |
add test for escape analysis of InstanceOf and CheckCast
|
Fri, 23 Nov 2012 13:38:53 +0100 |
Lukas Stadler |
make InstanceOfNode Virtualizable
|
Tue, 20 Nov 2012 16:38:09 +0100 |
Lukas Stadler |
more small fixes to PEA
|
Tue, 20 Nov 2012 15:57:48 +0100 |
Lukas Stadler |
small fix to multi-allocation PEA
|
Tue, 20 Nov 2012 15:57:32 +0100 |
Lukas Stadler |
remove MethodCallTargetNodes while inlining
|
Tue, 20 Nov 2012 13:36:22 +0100 |
Lukas Stadler |
provide MetaAccessProvider to EscapeAnalyzable
|
Tue, 20 Nov 2012 13:25:28 +0100 |
Lukas Stadler |
refactor EscapeAnalyzeable (remove EscapeOp)
|
Thu, 22 Nov 2012 23:36:04 +0100 |
Doug Simon |
fixed bug in the resetting of static call stubs that caused infinite, safepoint free loops in non-product builds of the VM
|