| Mon, 28 Nov 2011 18:20:46 -0800 |
Christian Wimmer |
Add newlines at the end of files, otherwise my version of gcc complains
|
| Fri, 25 Nov 2011 23:09:07 +0100 |
Thomas Wuerthinger |
Return a RiResolvedField array when looking up the fields of a class.
|
| Thu, 24 Nov 2011 17:45:38 +0100 |
Thomas Wuerthinger |
Merge.
|
| Thu, 24 Nov 2011 17:45:19 +0100 |
Thomas Wuerthinger |
Set catch all to true if the caught class is java.lang.Throwable.
|
| Thu, 24 Nov 2011 17:25:39 +0100 |
Gilles Duboscq |
fix componentType for the primitive array case
|
| Tue, 22 Nov 2011 10:41:05 +0100 |
Lukas Stadler |
Merge
|
| Fri, 18 Nov 2011 17:20:38 +0100 |
Lukas Stadler |
Merge
|
| Fri, 18 Nov 2011 13:34:10 +0100 |
Lukas Stadler |
Merge
|
| Fri, 18 Nov 2011 11:54:07 +0100 |
Lukas Stadler |
store RiCompiledMethod in nmethod and:
|
| Mon, 21 Nov 2011 14:31:28 +0100 |
Thomas Wuerthinger |
Removed debug output.
|
| Mon, 21 Nov 2011 13:56:38 +0100 |
Thomas Wuerthinger |
Fixed a regression.
|
| Fri, 18 Nov 2011 16:23:26 +0100 |
Thomas Wuerthinger |
put back in thread transitions.
|
| Fri, 18 Nov 2011 14:14:37 +0100 |
Thomas Wuerthinger |
No longer use shared ciFactory. Remove make_global usage in code installer.
|
| Fri, 18 Nov 2011 13:07:25 +0100 |
Thomas Wuerthinger |
Removed a safepoint possibility in the code installer when creating CiVirtualObject objects.
|
| Thu, 17 Nov 2011 18:33:33 +0100 |
Thomas Wuerthinger |
Fixed another possible safepoint in the code installer.
|
| Thu, 17 Nov 2011 18:02:12 +0100 |
Thomas Wuerthinger |
Fixed wrong type transition.
|
| Thu, 17 Nov 2011 16:08:14 +0100 |
Gilles Duboscq |
Fix typeProfile : collect all receivers and counts before making java calls
|
| Thu, 17 Nov 2011 00:52:17 +0100 |
Thomas Wuerthinger |
Do not use result when resolve_field throws an exception.
|
| Thu, 17 Nov 2011 00:01:56 +0100 |
Thomas Wuerthinger |
Handlize to fix GC issue.
|
| Wed, 16 Nov 2011 21:27:28 +0100 |
Thomas Wuerthinger |
Pass-by compilation broker.
|
| Wed, 16 Nov 2011 16:46:32 +0100 |
Thomas Wuerthinger |
Remove ci usage on field lookup.
|
| Wed, 16 Nov 2011 16:02:04 +0100 |
Thomas Wuerthinger |
Remove ci usage on method lookup.
|
| Wed, 16 Nov 2011 16:00:11 +0100 |
Thomas Wuerthinger |
Remove ci usage from switchProbability.
|
| Wed, 16 Nov 2011 15:58:18 +0100 |
Thomas Wuerthinger |
Remove ci usage for accessing branch probability.
|
| Wed, 16 Nov 2011 13:56:32 +0100 |
Thomas Wuerthinger |
Fixed a problem when looking up class constants.
|
| Wed, 16 Nov 2011 01:52:31 +0100 |
Thomas Wuerthinger |
Fixed an issue with the runtime queried escape analysis field array.
|
| Wed, 16 Nov 2011 01:48:25 +0100 |
Thomas Wuerthinger |
Use GraalEnv to look up types.
|
| Tue, 15 Nov 2011 22:06:02 +0100 |
Thomas Wuerthinger |
Use GraalEnv for installing code.
|
| Mon, 14 Nov 2011 22:33:15 +0100 |
Thomas Wuerthinger |
Remove the need for VmId (last usage was the constant pool).
|
| Wed, 09 Nov 2011 15:49:51 +0100 |
Lukas Stadler |
fix build warnings
|
| Wed, 09 Nov 2011 11:36:20 +0100 |
Lukas Stadler |
Merge
|
| Wed, 09 Nov 2011 11:27:38 +0100 |
Lukas Stadler |
support for new RiRuntime features: add code without making it the default for the method, executing Java tasks on the compile thread, communicate nmethod reference to Java code as HotSpotCompiledMethod
|
| Sat, 05 Nov 2011 16:14:56 +0100 |
Thomas Wuerthinger |
Fix crash in branch prediction if there's no current environment.
|
| Fri, 14 Oct 2011 15:39:40 +0200 |
Thomas Wuerthinger |
Only return declared fields on fields() call.
|
| Fri, 07 Oct 2011 16:49:07 +0200 |
Gilles Duboscq |
rename com.oracle.max.graal.runtime to com.oracle.max.graal.hotspot
|
| Thu, 06 Oct 2011 17:11:04 +0200 |
Gilles Duboscq |
Remove GRAAL env var requirement; Add ResourceMarks in code installer to avoid leaks
|
| Tue, 13 Sep 2011 19:59:10 +0200 |
Thomas Wuerthinger |
Added new method that loads and initializes a class that is referenced by a cp entry.
|
| Tue, 06 Sep 2011 11:18:53 +0200 |
Peter Hofer |
Factor out runtime call target address determination to avoid redundancy between CodeInstaller::site_Call() and VMEntries.getMaxCallTargetOffset()
|
| Sat, 03 Sep 2011 22:08:00 +0200 |
Thomas Wuerthinger |
Add two VM configs for intrinsifications.
|
| Fri, 02 Sep 2011 18:29:16 +0200 |
Thomas Wuerthinger |
Merge.
|
| Fri, 02 Sep 2011 18:28:45 +0200 |
Thomas Wuerthinger |
New runtime parameter: runtimeCallStackSize (size on the stack that must be left empty in case of runtime calls)
|
| Thu, 01 Sep 2011 09:52:58 +0200 |
Lukas Stadler |
Merge
|
| Thu, 01 Sep 2011 09:43:10 +0200 |
Lukas Stadler |
more work on ci removal
|
| Wed, 31 Aug 2011 09:58:35 +0200 |
Lukas Stadler |
remove some ci-dependencies
|
| Wed, 31 Aug 2011 14:55:40 +0200 |
Peter Hofer |
Support runtime calls to targets that don't fit in a 32-bit immediate: allow to query the maximum offset of a CiRuntimeCall target to determine the required immediate size and patch call sites with a mov/call instruction pair for indirect calls
|
| Wed, 24 Aug 2011 17:46:51 +0200 |
Lukas Stadler |
remove some dependencies on ci interface
|
| Wed, 24 Aug 2011 01:05:02 +0200 |
Thomas Wuerthinger |
Various hacks to be able to install machine code from a Java thread.
|
| Mon, 22 Aug 2011 14:40:08 +0200 |
Thomas Wuerthinger |
better error message of field not found
|
| Fri, 19 Aug 2011 17:28:37 +0200 |
Gilles Duboscq |
Add support for polling mark in code installer
|
| Thu, 04 Aug 2011 18:52:50 +0200 |
Lukas Stadler |
special handling for VirtualObject(Field) in IdentifyBlocksPhase (don't schedule out of loops), better toString for HotSpotMethod/Field
|
| Mon, 01 Aug 2011 13:45:25 +0200 |
Peter Hofer |
When a method that has been compiled before is considered for inlining, use its compiled (native) code size for size checks. This avoids inlining compiled methods that already contain many other inlined methods, reducing generated code size and usually execution time as well. Also, add GraalMetrics bookkeeping for inlining.
|
| Tue, 12 Jul 2011 17:00:25 +0200 |
Lukas Stadler |
global absolute probability analysis:
|
| Wed, 29 Jun 2011 19:52:51 +0200 |
Lukas Stadler |
fix for unresolved exception bug, exactType and declaredType logic for inlining, more on escape analysis
|
| Tue, 28 Jun 2011 19:14:28 +0200 |
Lukas Stadler |
more escape analysis work: debug info
|
| Tue, 21 Jun 2011 11:13:52 +0200 |
Lukas Stadler |
runtime profiles exceptions at the call site
|
| Sat, 18 Jun 2011 21:23:40 +0200 |
Thomas Wuerthinger |
Changed from method/reflect object back to methodOop, because using method/reflection can create deadlock with the user application (test with -Xcomp). Disabled method caching by default and created a flag -G:StoreResultGraph (otherwise eclipse will fail because of being out of memory).
|
| Fri, 17 Jun 2011 17:30:35 +0200 |
Lukas Stadler |
better inlining logic, DCE removes If with constant comparison
|
| Thu, 16 Jun 2011 20:36:17 +0200 |
Lukas Stadler |
inlining and runtime interface related changes:
|
| Wed, 15 Jun 2011 16:49:46 +0200 |
Lukas Stadler |
inlining of monomorphic profiled callsites with quick type checks
|
| Thu, 09 Jun 2011 17:33:08 +0200 |
Lukas Stadler |
RiMethod provides branch prediction information
|
