Soc2011 Log

Commit Date  
[3d066e] by Dave Grove

RVM-817 WIP: Change various LICENSE files from CPL to EPL.

2009-04-24 14:02:49 Tree
[143f95] by Dave Grove

fix bug in recomputeNormalOut.

If a branch instruction or fall through has been introduced whose target is an ExceptionHandlerBasicBlock,then we must set the bit on the target exception handler basic block that indicates it has a normal in edge. Failure to do this results in improper computation of the normal out edges of the source block.

2009-04-22 20:47:39 Tree
[0f4f1a] by Dave Grove

a couple LogicallyUninterruptible ==> Unpreemptible changes that were missed when the annotations on the overriding methods was changed.

2009-04-16 13:22:25 Tree
[097e55] by Steve Blackburn

Rationalize triggering mechanisms to allow sane triggering of defrag and emergency GCs for Immix collectors.

2009-04-16 06:58:08 Tree
[9688e2] by Steve Blackburn

Clean up of Immix's marking mechanism. Now use a rolling mark state for line marks. Avoids previous bugs and complexity without requiring additional meta-data. Superceeds work in r15651.

2009-04-16 06:55:00 Tree
[ef3e27] by Steve Blackburn

Switch to FastAdaptive for now since FullAdaptive can't build due to boot image bloat (see RVM-809)

2009-04-16 06:43:56 Tree
[74da52] by Steve Blackburn

Numerous fixes to defrag. Cleaned up code. Added an option to explicitly duplicate mark bytes so that marking and availability are distinct, with state copied from mark bytes to avail bytes at sweep time, at the cost of 1/128 overhead, but with a big advantiage in clarity (turned on by default). Removed silly heuristic which resulted in every full heap GC being a defrag GC for GenImmix.

2009-04-08 06:01:59 Tree
[50f405] by Steve Blackburn

Some extra debugging info for immix, plus some trivial code clean-ups

2009-04-07 11:32:14 Tree
[05e885] by Steve Blackburn

Clean-up and fix to address RVM-811

2009-04-07 03:57:44 Tree
[7bd00c] by Filip Pizlo

renamed dangerously to NoHandshake, and nicely to WithHandshake

2009-04-06 19:33:03 Tree
[e19e14] by Steve Blackburn

Clean-ups related to defragmentation in immix

2009-04-05 04:56:25 Tree
[d8db04] by Steve Blackburn

Another shot at turning GenImmix on. The bugs I'm aware of have been fixed. Still sorting out a performance issue with defrag, but want to see whether we're in shape for GenImmix being produciton in the next release. Again, feel free to back out if you think it appropriate....

2009-04-05 02:43:49 Tree
[ca0013] by Filip Pizlo

fixed RVMThread.join and

2009-04-04 05:26:48 Tree
[f52913] by Filip Pizlo

added more documentation for acctLock

2009-04-04 03:05:28 Tree
[a1f339] by Filip Pizlo

added docs to BlockAdapter

2009-04-04 00:35:34 Tree
[14802d] by Filip Pizlo

cleaned up createVM; fix for RVM-752

2009-04-04 00:00:32 Tree
[ae3fde] by Filip Pizlo

final touch-ups for RVM-763

2009-04-03 23:33:30 Tree
[de4a36] by Steve Blackburn

Minor fix. Move this stats counter to the right place.

2009-03-31 06:22:08 Tree
[b9d028] by Steve Blackburn

Improve handling of LOS thresholds. Addresses part of RVM-808. Differentiate the various reasons for allocating or copying to LOS on a per plan basis.

2009-03-29 10:01:00 Tree
[ba376e] by Daniel Frampton

Fix for osx/ia32 that folds in the hacky fix for aix. This also normalizes how the harmony defines are passed around somewhat.

2009-03-25 05:52:46 Tree
[b3726a] by Steve Blackburn

Add some debug verbosity to Immix

2009-03-19 10:50:20 Tree
[38f788] by Steve Blackburn

Another fix for correct defrag in GenImmix

2009-03-19 10:30:33 Tree
[2661ea] by Steve Blackburn

A variety of fixes to get GenImmix working properly. Mostly related to defrag, which apparently was never tested in the generational setting, where it rarely arises

2009-03-18 05:11:58 Tree
[a48795] by Steve Blackburn

Avoid dynamic check in hot path unless assertions are on

2009-03-18 05:00:12 Tree
[668358] by Dave Grove

RVM-800. allow baseline compiler to resolve the array ref for newarrary using similar logic as in opt compiler. There are slight differences due to the way in which we instantiate java.lang.Object, but otherwise I believe the code is now doing the same thing in both compilers.

2009-03-12 04:34:06 Tree
Older >