| Wed, 03 Jun 2015 15:47:54 +0200 |
Gilles Duboscq |
Introduce Compiler interface in jvmci. Use it from jvmci.hotspot.CompilationTask
|
| Fri, 29 May 2015 19:11:39 -0700 |
Christian Wimmer |
Make BytecodeParser a top-level class to avoid excessive indentation of two class nesting levels
|
| Fri, 29 May 2015 17:01:31 -0700 |
Christian Wimmer |
Allow multiple ParameterPlugin and InlineInvokePlugin in graph builder plugins; cleanup InlineInvokePlugin and implementations, including in Truffle
|
| Fri, 29 May 2015 13:19:05 -0700 |
Christian Wimmer |
Cleanup and generalize graph builder plugins
|
| Thu, 28 May 2015 17:00:59 -0700 |
Christian Wimmer |
Do not rely on single/double slot information from the Kind of the value in a local variable or expression stack entry. With word type rewriting during parsing, long values can be single slot values when they are rewritten from Object values.
|
| Fri, 29 May 2015 00:19:05 +0200 |
Doug Simon |
Merge.
|
| Wed, 27 May 2015 13:32:18 +0200 |
Doug Simon |
Merge.
|
| Wed, 27 May 2015 00:36:16 +0200 |
Doug Simon |
renamed com.oracle.graal.api[meta|code] modules to com.oracle.jvmci.[meta|code] (JBS:GRAAL-53)
|
| Tue, 26 May 2015 23:21:15 +0200 |
Doug Simon |
renamed com.oracle.graal.[debug|options|hotspotvmconfig]* modules to com.oracle.jvmci.[debug|options|hotspotvmconfig]* modules (JBS:GRAAL-53)
|
| Mon, 25 May 2015 23:30:34 +0200 |
Doug Simon |
moved GraalInternalError to com.oracle.jvmci.common and renamed it to JVMCIError (JBS:GRAAL-53)
|
| Thu, 28 May 2015 21:25:21 +0200 |
Thomas Wuerthinger |
Apply type check plugin only when parsing intrinsics.
|
| Tue, 26 May 2015 21:22:00 +0200 |
Thomas Wuerthinger |
Create different nodes in the graph builder when instanceof and checkcast have a single type profile.
|
| Tue, 26 May 2015 20:21:07 +0200 |
Thomas Wuerthinger |
Disable inlining during parsing.
|
| Tue, 26 May 2015 18:57:51 +0200 |
Thomas Wuerthinger |
Enable inlining during parsing by default.
|
| Sun, 24 May 2015 00:21:20 +0200 |
Doug Simon |
compute compiled bytecodes using method inlining recording instead of a DebugMetric
|
| Thu, 14 May 2015 16:14:17 -0700 |
Christian Wimmer |
Always do liveness analysis, but allow non-clearing of local variables; centralize omitAssertion code in bytecode parser
|
| Wed, 13 May 2015 13:11:17 +0200 |
Doug Simon |
cleaned up and improved documentation for IntrinsicScope
|
| Wed, 13 May 2015 10:54:14 +0200 |
Doug Simon |
rename GraphBuilderPhase.getFrameState to getFrameStateBuilder
|
| Wed, 13 May 2015 10:31:19 +0200 |
Doug Simon |
rename HIRFrameStateBuilder to FrameStateBuilder
|
| Tue, 12 May 2015 23:26:16 +0200 |
Doug Simon |
consolidated AbstractBytecodeParser into BytecodeParser
|
| Tue, 12 May 2015 16:05:32 +0200 |
Doug Simon |
don't create infopoint nodes within intrinsics
|
| Tue, 12 May 2015 15:57:40 +0200 |
Doug Simon |
removed unnecessary Intrinsic interface
|
| Tue, 12 May 2015 13:56:11 +0200 |
Doug Simon |
consolidated ReplacementContext and IntrinsicContext
|
| Mon, 11 May 2015 17:56:16 -0700 |
Tom Rodriguez |
Merge
|
| Mon, 11 May 2015 10:29:42 -0700 |
Tom Rodriguez |
Improve accuracy of info point states
|
| Mon, 11 May 2015 17:12:15 +0200 |
Doug Simon |
consolidate frame state creation and processing for intrinsics into graph parsing, removing need for CollapseFrameForSingleSideEffectPhase
|
| Wed, 22 Apr 2015 16:01:59 +0200 |
Doug Simon |
changed GraphBuilderContext.createStateAfter() to .setStateAfter(StateSplit)
|
| Sat, 09 May 2015 12:35:54 +0200 |
Doug Simon |
ensure RegisterFinalizerNode is created when Object::<init> is the compilation root
|
| Tue, 21 Apr 2015 16:18:23 -0700 |
Christian Wimmer |
Move PEGraphDecoder from truffle to graal project; factor out SimplifyingGraphDecoder that performs just canonicalization but not method inlining; fix bugs
|
| Tue, 21 Apr 2015 01:27:08 +0200 |
Thomas Wuerthinger |
Modify semantics of StressInvokeWithExceptionNode.
|
| Tue, 21 Apr 2015 00:47:48 +0200 |
Thomas Wuerthinger |
Add StressInvokeWithExceptionNode flag.
|
| Fri, 17 Apr 2015 18:24:59 +0200 |
Doug Simon |
loop safepoints are removed during graph parsing of intrinsics and snippets
|
| Fri, 10 Apr 2015 13:21:33 +0200 |
Doug Simon |
made ReplacementContext and IntrinsicContext top level classes
|
| Thu, 09 Apr 2015 10:26:28 +0200 |
Doug Simon |
InvocationPlugins can only be applied for direct invokes
|
| Thu, 09 Apr 2015 09:52:33 +0200 |
Doug Simon |
remove GraphBuilderContext.getSnippetReflection()
|
| Tue, 07 Apr 2015 17:32:26 +0200 |
Doug Simon |
added MethodSubstitutionPlugin as (eventual) replacement for @MethodSubstitution mechanism
|
| Mon, 06 Apr 2015 13:14:01 +0200 |
Doug Simon |
changed InvocationPlugin.execute to be a default method instead of a static method
|
| Wed, 08 Apr 2015 22:38:40 -0700 |
Christian Wimmer |
New partial evaluator that works on encoded graphs (instead of on bytecodes)
|
| Sat, 04 Apr 2015 21:43:21 +0200 |
Doug Simon |
removed some redundant GraphBuilderContext methods or gave them default implementations
|
| Sat, 04 Apr 2015 19:52:26 +0200 |
Doug Simon |
removed GraphBuilderContext.getRootMethod()
|
| Sat, 04 Apr 2015 19:47:24 +0200 |
Doug Simon |
removed commented out code
|
| Fri, 03 Apr 2015 17:49:43 +0200 |
Stefan Anzinger |
Ignore ResolvedJavaMethod.canBeInlined when parsing Snippets or Replacements
|
| Fri, 27 Mar 2015 13:29:08 +0100 |
Doug Simon |
converted all @MacroSubstitution uses to InvocationPlugins
|
| Thu, 26 Mar 2015 17:06:08 +0100 |
Doug Simon |
removed uses of StringBuffer, Hashtable and Stack
|
| Tue, 24 Mar 2015 11:56:10 +0100 |
Doug Simon |
check whether a StateSplit node being added via GraphBuilderContext has a null stateAfter before setting one
|
| Mon, 23 Mar 2015 20:57:21 +0100 |
Doug Simon |
added helper methods to GraphBuilderContext for common operation combinations
|
| Sun, 22 Mar 2015 17:04:03 +0100 |
Doug Simon |
made partial intrinsification work (i.e., intrinsics that call the original method to handle slow/uncommon paths)
|
| Fri, 20 Mar 2015 16:05:24 +0100 |
Doug Simon |
removed Receiver.isNullConstant() since InvocationPlugins will never be applied if the receiver is the null constant
|
| Thu, 19 Mar 2015 22:31:42 +0100 |
Doug Simon |
added null checking for the receiver when inlining or applying an InvocationPlugin during graph parsing
|
| Thu, 19 Mar 2015 12:47:06 +0100 |
Doug Simon |
Merge.
|
| Thu, 19 Mar 2015 12:46:06 +0100 |
Doug Simon |
added support for checking nodes added to the graph by an Invocation plugin and used this to check that only legal constants are added under ImmutableCode
|
| Wed, 18 Mar 2015 11:46:06 +0100 |
Doug Simon |
factored out assertion checks around application of an InvocationPlugin and added a check that all StateSplit nodes added by a plugin have a non-null frame state
|
| Wed, 18 Mar 2015 23:26:23 +0100 |
Thomas Wuerthinger |
Fix an issue when the result of canonicalization is appended in the graph builder. Introduce GraphBuilderPhase#recursiveAppend.
|
| Tue, 17 Mar 2015 12:13:24 +0100 |
Doug Simon |
Merge.
|
| Sun, 15 Mar 2015 11:48:14 +0100 |
Doug Simon |
refactored graph builder plugins to be top level interfaces
|
| Sun, 15 Mar 2015 11:36:54 +0100 |
Doug Simon |
refactored GraphBuilderConfiguration (and its component classes) into a separate project
|
| Sat, 14 Mar 2015 22:45:52 +0100 |
Thomas Wuerthinger |
Remove dead code in GraphBuilderPhase.
|
| Sat, 14 Mar 2015 19:15:51 +0100 |
Thomas Wuerthinger |
Assign the state after when appending the node and not automatically in the main graph builder loop.
|
| Sat, 14 Mar 2015 17:59:33 +0100 |
Thomas Wuerthinger |
Small simplification in GraphBuilderPhase.
|
| Sat, 14 Mar 2015 02:04:40 +0100 |
Thomas Wuerthinger |
Merge.
|
