bdemsky [Sat, 16 Aug 2008 22:58:34 +0000 (22:58 +0000)]
change
bdemsky [Sat, 16 Aug 2008 19:39:57 +0000 (19:39 +0000)]
try to lower lock overhead
bdemsky [Sat, 16 Aug 2008 03:30:52 +0000 (03:30 +0000)]
bug fix
bdemsky [Sat, 16 Aug 2008 03:30:18 +0000 (03:30 +0000)]
bug in the Vector class
jjenista [Fri, 15 Aug 2008 22:34:17 +0000 (22:34 +0000)]
strong updates, everything ready to do method calls
jjenista [Fri, 15 Aug 2008 18:47:49 +0000 (18:47 +0000)]
HeapRegionNode and ReferenceEdge both have had the equals() and hashCode()
methods updated to exclude transient attributes like reachability information.
If these objects are put into work sets and those attributes are changed they
have become different objects, according to the hash set. So equals() was changed
and other methods are used for determining equality in other situations, like
asking whether heap region nodes in different graphs are equivalent.
Also, a few other minor changes.
jjenista [Fri, 15 Aug 2008 00:26:02 +0000 (00:26 +0000)]
Retooled edges and basic stuff is working again (finally) but reachability needs to be reintegrated
jjenista [Wed, 13 Aug 2008 23:59:03 +0000 (23:59 +0000)]
Node and edge equality and hash updates
jjenista [Wed, 13 Aug 2008 23:05:51 +0000 (23:05 +0000)]
token classes combed over and tested thoroughly
adash [Wed, 13 Aug 2008 22:45:45 +0000 (22:45 +0000)]
latest files and new changes to the script for collected non-cache and
non-prefetch data
adash [Wed, 13 Aug 2008 01:31:32 +0000 (01:31 +0000)]
changes to enable no caching and no prefetching by adding DCACHE flag to
buildscript. Tested on Em3d
jjenista [Wed, 13 Aug 2008 01:05:39 +0000 (01:05 +0000)]
updated TokenTuple and TokenTupleSet, tested
jjenista [Tue, 12 Aug 2008 23:12:32 +0000 (23:12 +0000)]
Bug fix and makefile update
jjenista [Tue, 12 Aug 2008 23:09:48 +0000 (23:09 +0000)]
Made a big change to reference edges, touched a lot of code.
This commit is stable, but reachability needs to be added back
in, a little bit at a time with testing to make sure everything
gets put back together.
adash [Tue, 12 Aug 2008 21:46:44 +0000 (21:46 +0000)]
commit latest files
adash [Tue, 12 Aug 2008 21:38:33 +0000 (21:38 +0000)]
latest files
adash [Tue, 12 Aug 2008 21:35:52 +0000 (21:35 +0000)]
Add new file for increasing datasize to 4000 X 4000
adash [Tue, 12 Aug 2008 00:17:42 +0000 (00:17 +0000)]
change benchmark
jjenista [Mon, 11 Aug 2008 23:17:47 +0000 (23:17 +0000)]
start of new file
jjenista [Mon, 11 Aug 2008 23:10:45 +0000 (23:10 +0000)]
equals() and hashCode() methods are bunk
jzhou [Mon, 11 Aug 2008 20:16:49 +0000 (20:16 +0000)]
bug fixing in Scheduling analysis(combination stage)
bdemsky [Mon, 11 Aug 2008 20:13:42 +0000 (20:13 +0000)]
rewritten SOR benchmark
adash [Mon, 11 Aug 2008 19:23:06 +0000 (19:23 +0000)]
bug fix for invalidation (also allow invalidation of objects when these objects are
commited inside handlelocalRequest)
add few more debug messages for runtime debugging
adash [Sat, 9 Aug 2008 22:48:47 +0000 (22:48 +0000)]
bug fixes
changes to trans.c to get object type without doing a prefetch but a transRead instead
bdemsky [Sat, 9 Aug 2008 08:07:15 +0000 (08:07 +0000)]
do sane array allocation now
adash [Fri, 8 Aug 2008 00:31:05 +0000 (00:31 +0000)]
Latest version of Em3d
adash [Fri, 8 Aug 2008 00:22:58 +0000 (00:22 +0000)]
Sor files with out new barrier implementation
jjenista [Thu, 7 Aug 2008 22:49:53 +0000 (22:49 +0000)]
I erroneously added this sanity check and wondered why I see the output without crashing the analysis. Actually, when an edge in the callee HRN1->HRN2 is being mapped into the caller, if HRN1 is a callee parameter object then it's ID will definitely not be in the caller already. And the method getHRNSetThatPossiblyMapToCalleeHRN() recognizes that and does the right thing already, so the sanity check was reporting a non-existent problem.
bdemsky [Thu, 7 Aug 2008 21:57:26 +0000 (21:57 +0000)]
changes to support gc with dstm
jjenista [Thu, 7 Aug 2008 18:37:09 +0000 (18:37 +0000)]
Bug fix that param2id tables need to initialized for every method before starting analysis
jjenista [Wed, 6 Aug 2008 19:33:14 +0000 (19:33 +0000)]
Added fields to ReferenceEdgeProperties and combed over all classes that need proper equals and hashcode methods
jjenista [Wed, 6 Aug 2008 19:32:15 +0000 (19:32 +0000)]
Added a toStringBrief()
bdemsky [Wed, 6 Aug 2008 19:13:06 +0000 (19:13 +0000)]
change
bdemsky [Wed, 6 Aug 2008 19:08:15 +0000 (19:08 +0000)]
bug fix
adash [Wed, 6 Aug 2008 18:47:39 +0000 (18:47 +0000)]
fully compiled with new Barrier
adash [Wed, 6 Aug 2008 08:49:21 +0000 (08:49 +0000)]
still have some compile errors . Use Em3d with new Barrier class
bdemsky [Wed, 6 Aug 2008 07:39:07 +0000 (07:39 +0000)]
correction
adash [Wed, 6 Aug 2008 07:34:11 +0000 (07:34 +0000)]
compile error fixes for new Barrier.java
adash [Wed, 6 Aug 2008 07:31:31 +0000 (07:31 +0000)]
add new Barrier Class
bdemsky [Wed, 6 Aug 2008 06:01:36 +0000 (06:01 +0000)]
reduce the number of spins...functions that call lock are taking a long time
bdemsky [Wed, 6 Aug 2008 05:42:03 +0000 (05:42 +0000)]
make sure the other thread doesn't just fall asleep
bdemsky [Wed, 6 Aug 2008 05:33:35 +0000 (05:33 +0000)]
New strategy... If the prefetching thread gets behind us, we don't issue a prefetch.
bdemsky [Wed, 6 Aug 2008 04:12:00 +0000 (04:12 +0000)]
new Barrier class
bdemsky [Tue, 5 Aug 2008 22:03:21 +0000 (22:03 +0000)]
changes
jjenista [Tue, 5 Aug 2008 18:55:55 +0000 (18:55 +0000)]
Break aging into smaller procedures and improve graphing for method call work.
adash [Tue, 5 Aug 2008 08:32:53 +0000 (08:32 +0000)]
minor changes for update cache call(where is it called from)
bdemsky [Tue, 5 Aug 2008 04:21:10 +0000 (04:21 +0000)]
bug fix
jjenista [Tue, 5 Aug 2008 01:28:15 +0000 (01:28 +0000)]
Prune change sets during token prop by beta info only, not whether the rule was applied
jzhou [Tue, 5 Aug 2008 01:03:17 +0000 (01:03 +0000)]
fix a bug in combination stage of scheduling analysis
jjenista [Mon, 4 Aug 2008 23:23:33 +0000 (23:23 +0000)]
x.f = y now prunes new edge's beta by alpha at x
jjenista [Mon, 4 Aug 2008 23:09:09 +0000 (23:09 +0000)]
Fixed x.f = y operation
bdemsky [Mon, 4 Aug 2008 23:07:39 +0000 (23:07 +0000)]
bug fix
bdemsky [Mon, 4 Aug 2008 08:03:48 +0000 (08:03 +0000)]
remove people from email list...they're probably tired of the spam by now
bdemsky [Mon, 4 Aug 2008 08:00:54 +0000 (08:00 +0000)]
Bug correction
bdemsky [Mon, 4 Aug 2008 07:59:19 +0000 (07:59 +0000)]
Parallelized Version...Haven't tested
adash [Sun, 3 Aug 2008 19:47:33 +0000 (19:47 +0000)]
makefile with transstats option enabled for finding number of remote reads
adash [Sun, 3 Aug 2008 19:45:17 +0000 (19:45 +0000)]
Added new header file for debug only
adash [Sun, 3 Aug 2008 19:43:58 +0000 (19:43 +0000)]
added dsmdebug file for debugging using macros and setting flags in buildscript
added code to verify number of reads in transaction cache, main object store etc
compile error fix for clookup.h(static makes chashSearch local to file)
jzhou [Sun, 3 Aug 2008 16:43:39 +0000 (16:43 +0000)]
some benchmarks for scheduling
bdemsky [Sun, 3 Aug 2008 06:23:35 +0000 (06:23 +0000)]
unused variable
bdemsky [Sun, 3 Aug 2008 06:13:49 +0000 (06:13 +0000)]
code changes...
jzhou [Sun, 3 Aug 2008 00:55:15 +0000 (00:55 +0000)]
small changes
bdemsky [Fri, 1 Aug 2008 19:11:21 +0000 (19:11 +0000)]
Fix sockpool. The current version would trash the linked list if getSock2WithLock ever opened a socket.
adash [Fri, 1 Aug 2008 17:48:33 +0000 (17:48 +0000)]
remove the system.clearPrefetchCache call
bug fix for openning large number of sockets (use socket pool and recycle sockets)
bug fix for updating prefetch cache using version increment for modified objects
fix compile errors by moving #include for plookup.h
jzhou [Fri, 1 Aug 2008 07:15:27 +0000 (07:15 +0000)]
fix some bugs in multi-core version codes generation and runtime
jzhou [Fri, 1 Aug 2008 07:12:04 +0000 (07:12 +0000)]
fix compile problem
adash [Thu, 31 Jul 2008 07:33:12 +0000 (07:33 +0000)]
small changes for improving prefetch
adash [Thu, 31 Jul 2008 02:27:50 +0000 (02:27 +0000)]
new files for dynamic prefetch enabling
and signal file for getting new stats
adash [Thu, 31 Jul 2008 02:26:59 +0000 (02:26 +0000)]
Changes for dynamic prefetch enabling
adash [Tue, 29 Jul 2008 01:49:11 +0000 (01:49 +0000)]
bug fixes in gCollect.c and gCollect.h
changes to trans.c that aids in prefetch cache allocation
adash [Mon, 28 Jul 2008 19:50:37 +0000 (19:50 +0000)]
cache flushing for preventing memory leakage in SOR
jjenista [Sun, 27 Jul 2008 23:46:39 +0000 (23:46 +0000)]
Reachability bug fixes. Tokens propagate and age in allocation sites,
and the example programs produce results that are almost correct. One
thing to note is that token propagation through many allocation sites
in a loop doesn't work properly, where the same field-by-field construction of
a single complex object does propagate the tokens.
jjenista [Sun, 27 Jul 2008 22:17:25 +0000 (22:17 +0000)]
Token propagation for normal assignments fixed. Tokens do not "age" with
edges properly during assignment to new allocation yet.
adash [Thu, 24 Jul 2008 21:22:20 +0000 (21:22 +0000)]
Segfault fix in prefetch queue + additional macros for debugging
jzhou [Tue, 22 Jul 2008 19:10:04 +0000 (19:10 +0000)]
fix up a bug in multicore RAW interruption version codes
jjenista [Fri, 18 Jul 2008 19:29:25 +0000 (19:29 +0000)]
Working on allocating with reachability, mostly the "aging" of tokens.
Implementation is stable but it is not working correctly.
jjenista [Fri, 18 Jul 2008 00:34:31 +0000 (00:34 +0000)]
Added some functionality to reachability classes that is apparently
not very helpful. They've been tested though, so might as well check
it in, but the methods that are not of obvious use are committed but
commented out.
jjenista [Thu, 17 Jul 2008 18:32:15 +0000 (18:32 +0000)]
Fixed propagation, different change sets for the x and y branches before a new edge is added
jzhou [Thu, 17 Jul 2008 01:56:10 +0000 (01:56 +0000)]
fix up interruption related bug in RAW version. Use our own interruption setup codes instead of using libints which is out of date.
jjenista [Wed, 16 Jul 2008 19:45:08 +0000 (19:45 +0000)]
Fixed a few minor bugs in token propagation, and major bug that every
reachability set should have the empty token tuple set as a member.
Also, propagation does not go back up edges yet.
jjenista [Tue, 15 Jul 2008 23:24:06 +0000 (23:24 +0000)]
Token propagation implemented, stable but incorrect. This is just a capture.
jzhou [Tue, 15 Jul 2008 22:35:25 +0000 (22:35 +0000)]
fix a bug in thread simulator
jzhou [Tue, 15 Jul 2008 21:55:03 +0000 (21:55 +0000)]
add makefiles for building RAW version binary
jjenista [Tue, 15 Jul 2008 21:42:41 +0000 (21:42 +0000)]
More reachability implementation, no token propagation yet.
jzhou [Tue, 15 Jul 2008 21:40:15 +0000 (21:40 +0000)]
make multicore version PERT benchmark work on RAW(without interruption)
jzhou [Tue, 15 Jul 2008 21:39:19 +0000 (21:39 +0000)]
add multi-core version for PERT benchmark
jjenista [Tue, 15 Jul 2008 19:22:32 +0000 (19:22 +0000)]
Added reachability to simple edge cases.
adash [Thu, 10 Jul 2008 01:03:02 +0000 (01:03 +0000)]
changes not to break the compiler
adash [Thu, 10 Jul 2008 00:52:54 +0000 (00:52 +0000)]
bug fixes for udp broadcast
flags and code added to collect transaction commit+abort statistics
jjenista [Wed, 9 Jul 2008 21:07:04 +0000 (21:07 +0000)]
Added reachability classes, all of which are extensions of Canonical such that
after makeCanonical() two equivalent reachability classes are actually the same
class. This was the bug that kept the orginial analysis tests from terminating
and now they run fine.
jjenista [Wed, 2 Jul 2008 18:37:05 +0000 (18:37 +0000)]
Starting to integrate reachability classes into ownership.
jjenista [Tue, 1 Jul 2008 21:49:10 +0000 (21:49 +0000)]
special union of reachability sets works correctly now
jjenista [Tue, 1 Jul 2008 18:48:04 +0000 (18:48 +0000)]
More reachability set functionality, but not all there yet
jjenista [Fri, 27 Jun 2008 20:24:54 +0000 (20:24 +0000)]
Broadening implementation for reachability in ownership analysis.
jjenista [Thu, 26 Jun 2008 20:56:11 +0000 (20:56 +0000)]
Straightened out TokenTuple equality and hash codes so token sets
work properly.
jjenista [Tue, 24 Jun 2008 18:51:27 +0000 (18:51 +0000)]
Some initial implementation of reachability for ownership analysis.
jjenista [Tue, 24 Jun 2008 18:50:01 +0000 (18:50 +0000)]
some slight testing changes
adash [Tue, 10 Jun 2008 17:18:36 +0000 (17:18 +0000)]
changes to buildscript
adash [Tue, 10 Jun 2008 17:17:53 +0000 (17:17 +0000)]
Udp invalidation of objects
adash [Sat, 31 May 2008 01:28:18 +0000 (01:28 +0000)]
changes for prefetch objects on a transaction abort and updating the prefetch cache
when transactions commit