bdemsky [Sat, 3 May 2008 22:05:57 +0000 (22:05 +0000)]
Changes:
1) nojava option to the buildscript to built the executable in tmpbuilddir
2) fixed prefetch analysis
bdemsky [Sat, 3 May 2008 20:07:04 +0000 (20:07 +0000)]
various changes...
bug fix for transactions with a large number of objects
bdemsky [Sat, 3 May 2008 09:14:50 +0000 (09:14 +0000)]
hacks to speed up prefetching...doesn't really help though..
adash [Fri, 2 May 2008 22:51:57 +0000 (22:51 +0000)]
bug fix
adash [Fri, 2 May 2008 21:46:47 +0000 (21:46 +0000)]
check LOCK versions
bdemsky [Fri, 2 May 2008 20:30:16 +0000 (20:30 +0000)]
bug fix
bdemsky [Fri, 2 May 2008 20:18:29 +0000 (20:18 +0000)]
check in code
bdemsky [Fri, 2 May 2008 20:16:51 +0000 (20:16 +0000)]
Add multiple thread version
adash [Fri, 2 May 2008 19:51:45 +0000 (19:51 +0000)]
Em3d versions for 1,2, 4 threads
bdemsky [Fri, 2 May 2008 19:45:05 +0000 (19:45 +0000)]
new makefile and rewritten benchmark
adash [Fri, 2 May 2008 19:11:15 +0000 (19:11 +0000)]
new makefile for running 1, 2 and 4 threads
adash [Fri, 2 May 2008 19:10:08 +0000 (19:10 +0000)]
Add script
adash [Fri, 2 May 2008 01:46:02 +0000 (01:46 +0000)]
bug fixes and some more changes
adash [Wed, 30 Apr 2008 09:15:56 +0000 (09:15 +0000)]
barrier bug fix
adash [Wed, 30 Apr 2008 09:14:02 +0000 (09:14 +0000)]
barrier bug -> new hack
adash [Wed, 30 Apr 2008 09:11:43 +0000 (09:11 +0000)]
barrier bug fix -> new hack and align prints
adash [Wed, 30 Apr 2008 01:48:58 +0000 (01:48 +0000)]
SOR benchmark
adash [Mon, 28 Apr 2008 22:52:12 +0000 (22:52 +0000)]
bug fixes and changes to the matrixmultiply benchmark
adash [Sun, 27 Apr 2008 18:34:47 +0000 (18:34 +0000)]
add java option to increase memory available to the JVM
bdemsky [Sun, 27 Apr 2008 07:12:31 +0000 (07:12 +0000)]
Add correct atomic blocks to benchmark
adash [Sat, 26 Apr 2008 20:08:44 +0000 (20:08 +0000)]
New benchmark
bdemsky [Fri, 25 Apr 2008 22:23:51 +0000 (22:23 +0000)]
fix compiler bug
bdemsky [Fri, 25 Apr 2008 21:51:35 +0000 (21:51 +0000)]
Debug statements
bdemsky [Fri, 25 Apr 2008 21:49:37 +0000 (21:49 +0000)]
add valueof method for booleans
bdemsky [Fri, 25 Apr 2008 21:49:17 +0000 (21:49 +0000)]
Debug Barrier
jzhou [Fri, 25 Apr 2008 16:59:55 +0000 (16:59 +0000)]
add multi-thread simulator for multi-core version codes. Also add two new files in Runtime directory to hold multi-core related codes.
bdemsky [Fri, 25 Apr 2008 08:58:35 +0000 (08:58 +0000)]
changes to the barrier and rearrange atomic statements in the run method
bdemsky [Fri, 25 Apr 2008 08:39:32 +0000 (08:39 +0000)]
add our barrier
bdemsky [Fri, 25 Apr 2008 08:04:30 +0000 (08:04 +0000)]
Add yield
adash [Fri, 25 Apr 2008 07:52:10 +0000 (07:52 +0000)]
bug fixes
bdemsky [Thu, 24 Apr 2008 22:11:33 +0000 (22:11 +0000)]
bugs
adash [Thu, 24 Apr 2008 21:19:30 +0000 (21:19 +0000)]
Add new benchmark...still have compile errors
jjenista [Wed, 23 Apr 2008 18:14:17 +0000 (18:14 +0000)]
Test construction of disjoint graphs and let parameters have references
to a node in them. These should not result in an alias.
adash [Fri, 18 Apr 2008 00:16:43 +0000 (00:16 +0000)]
Add java version of Crypt benchmark
adash [Fri, 18 Apr 2008 00:14:53 +0000 (00:14 +0000)]
fully compiled version of Crypt benchmark
bdemsky [Thu, 17 Apr 2008 00:29:31 +0000 (00:29 +0000)]
make clearprefetch cache commands into barriers for prefetches
bdemsky [Thu, 17 Apr 2008 00:27:59 +0000 (00:27 +0000)]
changes and bug fixes
bdemsky [Wed, 16 Apr 2008 23:26:53 +0000 (23:26 +0000)]
more options
bdemsky [Wed, 16 Apr 2008 23:22:24 +0000 (23:22 +0000)]
things to do
bdemsky [Wed, 16 Apr 2008 22:41:29 +0000 (22:41 +0000)]
more changes
bdemsky [Wed, 16 Apr 2008 22:28:10 +0000 (22:28 +0000)]
code changes
jjenista [Wed, 16 Apr 2008 17:54:30 +0000 (17:54 +0000)]
Properly use transitive closure of allocation sites when resolving
a FlatCall node, and related extensions to the node classes.
Needs some cleanup though!
adash [Wed, 16 Apr 2008 01:26:40 +0000 (01:26 +0000)]
Clean up code
jzhou [Tue, 15 Apr 2008 21:23:49 +0000 (21:23 +0000)]
add new option "-robustroot" to enable setting up ROBUSTROOT. Defaulty it is set as ~/research/Robust/src.
bdemsky [Tue, 15 Apr 2008 20:50:26 +0000 (20:50 +0000)]
Comment out code that is breaking compile....please check in the code this needs to compile.
jjenista [Tue, 15 Apr 2008 20:50:25 +0000 (20:50 +0000)]
Whenever running ownership analysis, spit out a file with
all detected aliases between parameters and/or allocation
sites.
jjenista [Tue, 15 Apr 2008 20:48:57 +0000 (20:48 +0000)]
Before writing a dot file, remove all symbols that confuse
the dot syntax from the graph's name.
bdemsky [Tue, 15 Apr 2008 20:44:46 +0000 (20:44 +0000)]
bug fixes to prefetch printing code
jzhou [Tue, 15 Apr 2008 17:48:01 +0000 (17:48 +0000)]
add codes for generating multi-core version binary. Also add options -multicore & -raw into the buildscript
bdemsky [Tue, 15 Apr 2008 16:59:47 +0000 (16:59 +0000)]
changes to bigraph....Alokika's are included in this commit
bdemsky [Tue, 15 Apr 2008 05:03:56 +0000 (05:03 +0000)]
patches to:
1) fix locality analysis bug
2) remove work around from em3d
3) rewrite em3d parsing
4) generate syntactically good code for native method calls inside atomic regions...no guarantees about whether the code is safe...
bdemsky [Tue, 15 Apr 2008 00:15:15 +0000 (00:15 +0000)]
hack to work around analysis bug
bdemsky [Mon, 14 Apr 2008 22:49:45 +0000 (22:49 +0000)]
Add join stuff
bdemsky [Mon, 14 Apr 2008 22:21:59 +0000 (22:21 +0000)]
bug fixes
adash [Mon, 14 Apr 2008 21:46:24 +0000 (21:46 +0000)]
java version for MatrixMultiply
bdemsky [Mon, 14 Apr 2008 21:41:21 +0000 (21:41 +0000)]
try to add join support
bdemsky [Mon, 14 Apr 2008 21:41:07 +0000 (21:41 +0000)]
various changes
adash [Mon, 14 Apr 2008 21:38:58 +0000 (21:38 +0000)]
add native method for clearing prefetch cache to System.java
adash [Mon, 14 Apr 2008 21:27:28 +0000 (21:27 +0000)]
changed fillTable() to static method
bdemsky [Mon, 14 Apr 2008 20:07:55 +0000 (20:07 +0000)]
Print out warning message instead of throwing exception if someone calls static method on an object.
adash [Mon, 14 Apr 2008 17:57:14 +0000 (17:57 +0000)]
compile error fixes
bdemsky [Mon, 14 Apr 2008 07:16:27 +0000 (07:16 +0000)]
major changes to prefetching code
adash [Sun, 13 Apr 2008 04:40:23 +0000 (04:40 +0000)]
java version of Em3d benchmark
adash [Sun, 13 Apr 2008 04:24:02 +0000 (04:24 +0000)]
New benchmark
adash [Sun, 13 Apr 2008 04:11:39 +0000 (04:11 +0000)]
New benchmark
bdemsky [Fri, 11 Apr 2008 08:54:06 +0000 (08:54 +0000)]
fix bugs in sockpool...test and set has to be atomic
bdemsky [Fri, 11 Apr 2008 07:53:08 +0000 (07:53 +0000)]
changes
jjenista [Tue, 8 Apr 2008 21:03:36 +0000 (21:03 +0000)]
A little code cleanup and support for different options when writing
ownership graphs to dot files.
jjenista [Tue, 8 Apr 2008 20:50:50 +0000 (20:50 +0000)]
Fixed a bug in assigning temp to new allocation. Clear other references
from the destination temp before assigning a reference to the newest heap
region for the allocation site.
adash [Mon, 7 Apr 2008 21:26:26 +0000 (21:26 +0000)]
complete hashtable implementation with separate socket pools for read and
prefetch.
commented out array implementation of socket pools
adash [Fri, 4 Apr 2008 19:49:45 +0000 (19:49 +0000)]
changes
adash [Fri, 4 Apr 2008 18:20:49 +0000 (18:20 +0000)]
single socketpool version with spin locks
jjenista [Mon, 31 Mar 2008 20:47:54 +0000 (20:47 +0000)]
Altered ownership graph to dot file by removing label nodes, marking allocation
site nodes with the object type, numbering heap regions allocated for a parameter
by the parameter index and tested that initial parameter relfexive edges get
clobbered by code that introduces a self-reference.
adash [Mon, 31 Mar 2008 01:12:35 +0000 (01:12 +0000)]
bug fix for Atomic5.java testcase
jjenista [Fri, 28 Mar 2008 17:57:21 +0000 (17:57 +0000)]
Retain properties of edges that are mapped from callee to a caller graph.
jjenista [Fri, 28 Mar 2008 17:50:13 +0000 (17:50 +0000)]
When performing the operations:
assignTempToTemp
assignTempToField
assignFieldToTemp
the newly created reference edge was copying the properties of the
edges from labels (or other heap regions) that led to the src, dst
pair. It creates bugs like when creating a reference from one
parameter's heap region to another, it would copy the special
"initial parameter reflexive edge" property to what is otherwise a
normal edge. All new edges using these three basic assignment
operations are now created with no special properties. In other
words, the context of the new edge alone should determine its
properties.
jjenista [Thu, 27 Mar 2008 21:31:29 +0000 (21:31 +0000)]
Stable capture.
Method call resolution algorithm is drafted, but clearly there are bugs
because a simple program with an alias-creating method doesn't reflect
the correct heap structure in the caller graph.
jzhou [Thu, 27 Mar 2008 19:43:22 +0000 (19:43 +0000)]
small change to make the naming of output dot files generated by analysis stage and simulation stage consistent
jzhou [Thu, 27 Mar 2008 19:34:54 +0000 (19:34 +0000)]
fix bug in ScheduleAnalysis.splitSNode()
bdemsky [Thu, 27 Mar 2008 08:00:58 +0000 (08:00 +0000)]
code change
bdemsky [Thu, 27 Mar 2008 07:51:54 +0000 (07:51 +0000)]
small changes
jjenista [Wed, 26 Mar 2008 21:55:55 +0000 (21:55 +0000)]
Stable capture of work on method call resolution, specifically the guts
of it in OwnershipGraph.java. Added a new attribute to the
ReferenceEdgeProperties class indicating an initial reflexive param edge.
adash [Wed, 26 Mar 2008 18:18:02 +0000 (18:18 +0000)]
completed and tested socket reuse code for trans read and trans prefetch() messages
bdemsky [Wed, 26 Mar 2008 02:51:21 +0000 (02:51 +0000)]
remove debug code
bdemsky [Wed, 26 Mar 2008 02:50:58 +0000 (02:50 +0000)]
fixed other small issues
bdemsky [Wed, 26 Mar 2008 02:36:36 +0000 (02:36 +0000)]
bugs
bdemsky [Wed, 26 Mar 2008 02:22:17 +0000 (02:22 +0000)]
loop exit detection
adash [Wed, 26 Mar 2008 00:29:22 +0000 (00:29 +0000)]
rewrite arrays for efficiency
bdemsky [Tue, 25 Mar 2008 22:50:08 +0000 (22:50 +0000)]
code change for bug Jim posted about
jjenista [Tue, 25 Mar 2008 20:36:06 +0000 (20:36 +0000)]
Got clarification on how to resolve method calls, introduced concept of
a set of possible callees on virtual dispatch, have to take the merge of
those ownership graphs and allocation sites to be correct. Stable, but
only a draft of the OwnershipAnalysis portion. OwnershipGraph mechanics
to determine mapping of references in callee merge-o-tron to caller graph
not added yet.
bdemsky [Tue, 25 Mar 2008 06:46:53 +0000 (06:46 +0000)]
revert this change
bdemsky [Tue, 25 Mar 2008 06:44:04 +0000 (06:44 +0000)]
changes
bdemsky [Tue, 25 Mar 2008 06:28:39 +0000 (06:28 +0000)]
mistakes in changes
bdemsky [Tue, 25 Mar 2008 05:43:51 +0000 (05:43 +0000)]
more changes
bdemsky [Tue, 25 Mar 2008 05:43:10 +0000 (05:43 +0000)]
changes to prefetch analysis to simplify code
bdemsky [Tue, 25 Mar 2008 03:10:04 +0000 (03:10 +0000)]
clean up some equal methods
bdemsky [Tue, 25 Mar 2008 02:48:08 +0000 (02:48 +0000)]
fix some bugs
bdemsky [Tue, 25 Mar 2008 02:13:14 +0000 (02:13 +0000)]
more changes
bdemsky [Tue, 25 Mar 2008 01:48:08 +0000 (01:48 +0000)]
correct bug in new code
bdemsky [Tue, 25 Mar 2008 01:08:41 +0000 (01:08 +0000)]
*** empty log message ***