Tue, 08 May 2012 15:52:01 +0200 |
Doug Simon |
pulled AbstractStateSplit into FixedWithNextNode and made it no longer implement StateSplit. StateSplit is now implemented only by nodes that really have side effects.
|
Mon, 07 May 2012 11:40:44 +0200 |
Doug Simon |
lowering checkcasts with Java snippets (incomplete)
|
Mon, 07 May 2012 11:34:16 +0200 |
Doug Simon |
moved creation of a RiCodeInfo to the runtime side of the CRI
|
Mon, 07 May 2012 11:08:58 +0200 |
Doug Simon |
added thread name to debug info message
|
Fri, 04 May 2012 16:54:27 +0200 |
Doug Simon |
canonicalize load from a compile-constant, immutable array
|
Fri, 04 May 2012 16:52:58 +0200 |
Doug Simon |
extended canonicalization so that a caller can specify a CiConstant value is an immutable object or array
|
Fri, 04 May 2012 16:38:38 +0200 |
Doug Simon |
removed product and debug configurations from hotspot CDT project config
|
Fri, 04 May 2012 12:08:05 +0200 |
Gilles Duboscq |
Add verification in NullCheckNode and AccessFieldNode
|
Fri, 04 May 2012 12:08:00 +0200 |
Gilles Duboscq |
Fix : in peeling, while merging early exits, transfer all anchored nodes from the exit to the merge
|
Thu, 03 May 2012 15:43:03 +0200 |
Gilles Duboscq |
Fix : There can be proxies in the virtual chain
|
Thu, 03 May 2012 15:19:16 +0200 |
Gilles Duboscq |
Merge
|
Thu, 03 May 2012 15:19:11 +0200 |
Gilles Duboscq |
Fix merge
|
Thu, 03 May 2012 11:05:46 +0200 |
Gilles Duboscq |
Merge
|
Fri, 27 Apr 2012 13:13:06 +0200 |
Gilles Duboscq |
rename CiKind.toUnboxedJavaClass to toBoxedJavaClass
|
Fri, 27 Apr 2012 12:14:47 +0200 |
Gilles Duboscq |
Merge
|
Fri, 27 Apr 2012 12:14:38 +0200 |
Gilles Duboscq |
Fix a peeling-escape analysis related problem that was causin LinearScan bailouts (merge virtual chains when merging ends of the peeled code)
|
Thu, 03 May 2012 13:39:45 +0200 |
Doug Simon |
added mechanism (enabled by -G:PICache and -G:PiFilter) for saving/loading method profiling info to/from disk
|
Wed, 02 May 2012 18:23:12 +0200 |
Doug Simon |
fixed bug in use of CheckcastMinHintHitProbability
|
Wed, 02 May 2012 18:04:36 +0200 |
Doug Simon |
moved RiResolvedMethod.dumpProfile() to CiUtil.profileAsString()
|
Wed, 02 May 2012 17:09:00 +0200 |
Doug Simon |
added GraalOptions.CheckcastMinHintHitProbability to better guide use of hints for checkcasts
|
Wed, 02 May 2012 16:32:00 +0200 |
Lukas Stadler |
test case for the InstanceOfNode canonicalizer fix
|
Wed, 02 May 2012 15:08:41 +0200 |
Lukas Stadler |
Merge
|
Wed, 02 May 2012 14:56:07 +0200 |
Lukas Stadler |
changes to canonicalizer log output
|
Wed, 02 May 2012 14:55:51 +0200 |
Lukas Stadler |
use PiNodes instead of CheckCastNodes to pin inlining receivers, remove emitCode flag
|
Wed, 02 May 2012 14:53:20 +0200 |
Lukas Stadler |
fix canonicalization of InstanceOfNode
|
Wed, 02 May 2012 14:47:04 +0200 |
Lukas Stadler |
change PiNode to anchor on any FixedNode, not only BeginNodes
|
Wed, 02 May 2012 14:45:56 +0200 |
Lukas Stadler |
re-enable graph caching and the priority compilation queue
|
Wed, 02 May 2012 14:39:45 +0200 |
Doug Simon |
types profiles are now sorted in descending order of each profiled type's probability
|
Wed, 02 May 2012 12:59:59 +0200 |
Lukas Stadler |
turn off priority compilation queue
|
Wed, 02 May 2012 11:31:53 +0200 |
Doug Simon |
expanded analysis of compiled checkcast code paths
|