| Mon, 01 Oct 2012 15:53:03 +0200 |
Gilles Duboscq |
Fix problem in MemoryPhi creation during FloatingRead, add an assert to detect this kind of problems
|
| Sun, 30 Sep 2012 21:44:03 +0200 |
Doug Simon |
reverted LoadHubNode to be a fixed node
|
| Sun, 30 Sep 2012 13:21:20 +0200 |
Doug Simon |
removed XIR
|
| Sun, 30 Sep 2012 13:19:38 +0200 |
Doug Simon |
removed support for extended bytecodes
|
| Fri, 28 Sep 2012 15:33:29 +0200 |
Doug Simon |
renamed ReadHubNode to LoadHubNode to be imply higher level operation (c.f. UnsafeLoadNode vs ReadNode)
|
| Fri, 28 Sep 2012 15:27:42 +0200 |
Doug Simon |
made ReadHubNode floating
|
| Fri, 28 Sep 2012 11:50:37 +0200 |
Doug Simon |
made all @NodeIntrinsic annotated methods be native (this is a requirement now enforced during snippet intrinsification)
|
| Thu, 27 Sep 2012 21:37:46 +0200 |
Doug Simon |
removed IsTypeNode and replaced its usage with an ObjectEqualsNode
|
| Thu, 27 Sep 2012 20:29:17 +0200 |
Doug Simon |
removed need for genTypeBranch() in XIR by replacing an IsTypeNode with an ObjectEqualsNode during lowering
|
| Thu, 27 Sep 2012 16:58:01 +0200 |
Doug Simon |
made position calculation for VMErrorNode more robust when debug info is not available
|
| Thu, 27 Sep 2012 16:56:50 +0200 |
Doug Simon |
added missing stub registration for VMErrorNode
|
| Thu, 27 Sep 2012 14:23:04 +0200 |
Doug Simon |
added VMErrorNode intrinsic to support handling fatal errors in snippets
|
| Wed, 26 Sep 2012 21:48:07 +0200 |
Doug Simon |
fixed wrong path in README_GRAAL.txt
|
| Wed, 26 Sep 2012 21:47:42 +0200 |
Doug Simon |
fixed bug in monitor snippets; they are now used for general lowering of monitor operations
|
| Wed, 26 Sep 2012 21:45:20 +0200 |
Doug Simon |
replaced magic constant with configuration value
|
| Wed, 26 Sep 2012 21:44:04 +0200 |
Doug Simon |
added sanity assertion to snippet instantiation
|
| Wed, 26 Sep 2012 21:43:31 +0200 |
Doug Simon |
added test for deopt while holding lock(s)
|
| Wed, 26 Sep 2012 21:42:11 +0200 |
Doug Simon |
minor restructuring of GraalCompilerTest
|
| Wed, 26 Sep 2012 18:15:11 +0200 |
Doug Simon |
added Log.printf() for more flexibility in debugging snippets
|
| Wed, 26 Sep 2012 13:56:27 +0200 |
Doug Simon |
added 'name' field to RuntimeCall so that RuntimeCall objects still format nicely in traces and logs as they did when RuntimeCall was an enum
|
| Wed, 26 Sep 2012 12:15:52 +0200 |
Lukas Stadler |
fix bailouts due to insufficient balanced monitor checks
|
| Tue, 25 Sep 2012 09:28:03 -0700 |
Christian Wimmer |
Make RuntimeCall more extensible by changing it from an enum to a class
|
| Tue, 25 Sep 2012 09:23:45 -0700 |
Christian Wimmer |
Add MetaAccessProvider.getResolvedJavaField to convert from reflection field to metadata field
|
| Tue, 25 Sep 2012 09:22:22 -0700 |
Christian Wimmer |
Add a NodeIntrinsic factory method
|
| Tue, 25 Sep 2012 09:21:52 -0700 |
Christian Wimmer |
MembarNode does not need state and has no side effects, so it needs not be a StateSplit
|
| Tue, 25 Sep 2012 09:21:16 -0700 |
Christian Wimmer |
Invoke constructor required too specific call target class
|
| Tue, 25 Sep 2012 09:20:18 -0700 |
Christian Wimmer |
Allow calls without a FrameState in the backend
|
| Tue, 25 Sep 2012 09:19:46 -0700 |
Christian Wimmer |
Add conversion from java.lang.Class to Kind
|
| Tue, 25 Sep 2012 17:50:01 +0200 |
Lukas Stadler |
move monitors into FrameState (fixes subtle issues for tail duplication and other optimizations)
|
| Tue, 25 Sep 2012 16:35:27 +0200 |
Lukas Stadler |
fix for InliningUtil (dropped outer frame states while inlining)
|
