jjenista [Tue, 17 Mar 2009 02:27:22 +0000 (02:27 +0000)]
Build aliased parameter models--still a partial implementation of the new model, but stable.
jjenista [Mon, 16 Mar 2009 23:26:56 +0000 (23:26 +0000)]
partway to new parameter model
jjenista [Mon, 16 Mar 2009 18:13:49 +0000 (18:13 +0000)]
Stable capture while moving towards typed heap regions, type and field on edges, and more precise parameter models
adash [Mon, 16 Mar 2009 10:07:11 +0000 (10:07 +0000)]
add makefile
adash [Mon, 16 Mar 2009 08:36:35 +0000 (08:36 +0000)]
most recent versions
bdemsky [Sat, 14 Mar 2009 17:26:13 +0000 (17:26 +0000)]
hack to avoid reprefetching the same things
bdemsky [Sat, 14 Mar 2009 17:19:56 +0000 (17:19 +0000)]
change
bdemsky [Sat, 14 Mar 2009 01:42:16 +0000 (01:42 +0000)]
align everything to 8 byte boundaries
bdemsky [Sat, 14 Mar 2009 01:38:31 +0000 (01:38 +0000)]
align to 8 byte boundaries
jzhou [Sat, 14 Mar 2009 00:39:27 +0000 (00:39 +0000)]
add -abcclose into buildscript
jjenista [Fri, 13 Mar 2009 22:49:19 +0000 (22:49 +0000)]
Be more precise about enqueing dependent method contexts, and use topological sorted order of the call graph to give method contexts priority for analysis, targetting leaf methods first.
adash [Fri, 13 Mar 2009 22:27:07 +0000 (22:27 +0000)]
bug fix for getting the correct offset when prefetching
bdemsky [Fri, 13 Mar 2009 03:50:41 +0000 (03:50 +0000)]
batch communications for prefetches
bdemsky [Fri, 13 Mar 2009 03:47:48 +0000 (03:47 +0000)]
takes parameter that specifies how many items to prefetch
bdemsky [Fri, 13 Mar 2009 03:47:18 +0000 (03:47 +0000)]
new 3d version
jjenista [Fri, 13 Mar 2009 00:45:28 +0000 (00:45 +0000)]
Bug fixes:
1. rewriteCallerReachability was not rewriting parameter "star" tokens.
2. globalSweep fixes:
a. initialize alpha prime for all nodes to the empty reachability set, nothing else
b. when propagating contraint for bold_B, and encountering a new edge, changed initial
result from just the intersection value to union of new edge's old beta and intersection
c. when using bold_B to reduce alpha states, now checking all edges for the state before
marking token for removal, instead of removing if we find one negative match
jjenista [Fri, 13 Mar 2009 00:31:42 +0000 (00:31 +0000)]
little benchmark cleanup changes
jzhou [Thu, 12 Mar 2009 22:59:28 +0000 (22:59 +0000)]
add FastCheck version classes in ClassLibrary
bdemsky [Thu, 12 Mar 2009 21:28:33 +0000 (21:28 +0000)]
check in working version
bdemsky [Thu, 12 Mar 2009 21:27:45 +0000 (21:27 +0000)]
changes
jjenista [Thu, 12 Mar 2009 21:16:47 +0000 (21:16 +0000)]
Rewrote global sweep to match paper, keeping ugly debug functionality for further testing, will remove later
jjenista [Thu, 12 Mar 2009 21:14:31 +0000 (21:14 +0000)]
improve report and debug for analysis interface
jjenista [Thu, 12 Mar 2009 21:13:04 +0000 (21:13 +0000)]
allows application of change set to reachability set toggle whether to keep the source states or not
jjenista [Thu, 12 Mar 2009 21:12:10 +0000 (21:12 +0000)]
another test
adash [Thu, 12 Mar 2009 20:06:40 +0000 (20:06 +0000)]
Brian's changes to manual prefetching for array and chase
adash [Thu, 12 Mar 2009 20:04:21 +0000 (20:04 +0000)]
bug fix and make tabbing
bdemsky [Thu, 12 Mar 2009 19:42:55 +0000 (19:42 +0000)]
change size
adash [Thu, 12 Mar 2009 18:42:47 +0000 (18:42 +0000)]
machine ip addr change and small changes
adash [Thu, 12 Mar 2009 18:41:27 +0000 (18:41 +0000)]
machine ip change
adash [Thu, 12 Mar 2009 18:40:02 +0000 (18:40 +0000)]
new manual prefetching scheme
adash [Thu, 12 Mar 2009 17:46:51 +0000 (17:46 +0000)]
add latest machine ips and modify the script for these machines
adash [Thu, 12 Mar 2009 17:45:11 +0000 (17:45 +0000)]
Manual prefetch versions of Array and Chase
microbenchmarks
bdemsky [Thu, 12 Mar 2009 06:29:38 +0000 (06:29 +0000)]
less hacky fix
bdemsky [Thu, 12 Mar 2009 06:02:13 +0000 (06:02 +0000)]
fix gcc bug through evil hack....
adash [Thu, 12 Mar 2009 01:54:11 +0000 (01:54 +0000)]
added fprintf options to output to trans stats to a file
adash [Thu, 12 Mar 2009 01:36:32 +0000 (01:36 +0000)]
bug fixes for LookUpServerThread.java
compile error fixes for MatrixMultiplyN.java
small changes to script
changes to JGFMolDynBenchSizeC.java for new parameter for Moldyn
bdemsky [Wed, 11 Mar 2009 06:22:56 +0000 (06:22 +0000)]
change to use less memory
bdemsky [Wed, 11 Mar 2009 06:22:41 +0000 (06:22 +0000)]
change
bdemsky [Wed, 11 Mar 2009 05:42:22 +0000 (05:42 +0000)]
no point in prefetching local objects
bdemsky [Wed, 11 Mar 2009 05:21:45 +0000 (05:21 +0000)]
prefetch less in advance
bdemsky [Wed, 11 Mar 2009 05:16:15 +0000 (05:16 +0000)]
go for smaller batches
bdemsky [Wed, 11 Mar 2009 00:40:33 +0000 (00:40 +0000)]
change
bdemsky [Tue, 10 Mar 2009 20:16:58 +0000 (20:16 +0000)]
changes to pipeline prefetches for fft2d
bdemsky [Tue, 10 Mar 2009 20:05:49 +0000 (20:05 +0000)]
prfetch change
bdemsky [Tue, 10 Mar 2009 20:03:26 +0000 (20:03 +0000)]
change
bdemsky [Tue, 10 Mar 2009 19:58:56 +0000 (19:58 +0000)]
next obvious improvement
bdemsky [Tue, 10 Mar 2009 19:57:45 +0000 (19:57 +0000)]
make sure not to prefetch stuff we already have...
bdemsky [Tue, 10 Mar 2009 19:48:56 +0000 (19:48 +0000)]
changes to prefetch...hope it makes a difference...
jjenista [Tue, 10 Mar 2009 19:39:56 +0000 (19:39 +0000)]
changed analysis public interface to report a set of common heap regions rather than a boolean for alias indication. Changed multicore code builder to check for non-empty set rather than boolean value
jjenista [Tue, 10 Mar 2009 19:37:32 +0000 (19:37 +0000)]
Crack at tuning the punt number between keeping problem tractable and not introducing multi-tokens too soon
adash [Tue, 10 Mar 2009 19:36:29 +0000 (19:36 +0000)]
Manual prefetch version for 2DFFT
bdemsky [Tue, 10 Mar 2009 19:29:19 +0000 (19:29 +0000)]
check in Java code
adash [Tue, 10 Mar 2009 17:52:05 +0000 (17:52 +0000)]
added to
script to generate the stats on remote machine and print to /tmp/client_stats.txt
still has some problems but can generate necessary info for now
adash [Tue, 10 Mar 2009 08:48:10 +0000 (08:48 +0000)]
new benchmarks in benchmark arguments file
adash [Tue, 10 Mar 2009 08:46:09 +0000 (08:46 +0000)]
make output files more legible in script
adash [Tue, 10 Mar 2009 08:04:58 +0000 (08:04 +0000)]
Manual prefetch version of Moldyn
bdemsky [Tue, 10 Mar 2009 07:41:40 +0000 (07:41 +0000)]
bug fix
bdemsky [Tue, 10 Mar 2009 07:30:58 +0000 (07:30 +0000)]
some changes...
adash [Tue, 10 Mar 2009 07:10:34 +0000 (07:10 +0000)]
fix operator precedence + options for makefile
adash [Tue, 10 Mar 2009 01:13:03 +0000 (01:13 +0000)]
bug fix for lookupservice java version( now works fine for 8 threads)
change makefiles
get rid of print statements
jjenista [Mon, 9 Mar 2009 23:46:24 +0000 (23:46 +0000)]
Bug fix, field matching for fields that model array element references weren't matching properly before
jjenista [Mon, 9 Mar 2009 23:37:41 +0000 (23:37 +0000)]
Bug fix, was not growing the mapping of descriptors->method contexts. Bug fix, handling array element references as quasi-fields properly.
bdemsky [Mon, 9 Mar 2009 23:29:02 +0000 (23:29 +0000)]
fix bug
bdemsky [Mon, 9 Mar 2009 22:08:24 +0000 (22:08 +0000)]
bug fix for alokika
adash [Mon, 9 Mar 2009 20:30:56 +0000 (20:30 +0000)]
Manual prefetcing for 2DConv
jjenista [Mon, 9 Mar 2009 18:10:52 +0000 (18:10 +0000)]
Pass around pointer to D2 as an argument to everything, rather than stash the pointer in almost every class in the benchmark. Also explicitly rewrote non-void methods with no return statement to have them
bdemsky [Mon, 9 Mar 2009 18:06:15 +0000 (18:06 +0000)]
bug fix
jjenista [Mon, 9 Mar 2009 18:03:29 +0000 (18:03 +0000)]
A small example with allocated regions propagating through various method contexts and properly mapping into the top-level caller. Does not expose bug that appears in direct-to benchmark
jjenista [Mon, 9 Mar 2009 18:01:58 +0000 (18:01 +0000)]
Bug fix in applying change sets, have to use the source reachability state in every rule and not just the first
jjenista [Mon, 9 Mar 2009 18:00:59 +0000 (18:00 +0000)]
changed toString so method context's produce unique names and therefore write unique DOT graphs during output after analysis
adash [Mon, 9 Mar 2009 17:53:19 +0000 (17:53 +0000)]
script to run manual prefetches
bm.txt file has the arguments for each benchmark
adash [Mon, 9 Mar 2009 17:52:10 +0000 (17:52 +0000)]
temporary fix for createNewSocket error...add config files to create a link
to the dstm.conf. e.g. when running 1 thread use dstm1.conf, when running 2 threads
use dstm2.conf etc...
adash [Mon, 9 Mar 2009 15:09:00 +0000 (15:09 +0000)]
new script files and some bug fixes...LookUpservice java still not working well
adash [Mon, 9 Mar 2009 03:09:00 +0000 (03:09 +0000)]
Script to run java + jvm versions of benchmarks
adash [Mon, 9 Mar 2009 01:36:41 +0000 (01:36 +0000)]
change script files
adash [Mon, 9 Mar 2009 00:02:14 +0000 (00:02 +0000)]
Add manual prefetch files
adash [Sun, 8 Mar 2009 02:06:09 +0000 (02:06 +0000)]
bug fix
adash [Sat, 7 Mar 2009 20:08:42 +0000 (20:08 +0000)]
small changes and add comments
adash [Sat, 7 Mar 2009 03:49:52 +0000 (03:49 +0000)]
Working and stable version of Java
bdemsky [Sat, 7 Mar 2009 01:30:00 +0000 (01:30 +0000)]
anges
bdemsky [Sat, 7 Mar 2009 01:29:49 +0000 (01:29 +0000)]
guard stuff
bdemsky [Sat, 7 Mar 2009 01:26:45 +0000 (01:26 +0000)]
guard header files
adash [Sat, 7 Mar 2009 01:16:42 +0000 (01:16 +0000)]
changes and some bug fixes
jjenista [Sat, 7 Mar 2009 00:47:20 +0000 (00:47 +0000)]
changes
jjenista [Sat, 7 Mar 2009 00:45:35 +0000 (00:45 +0000)]
changes
jjenista [Fri, 6 Mar 2009 18:41:05 +0000 (18:41 +0000)]
Support cast nodes
adash [Fri, 6 Mar 2009 05:10:53 +0000 (05:10 +0000)]
java version
jjenista [Fri, 6 Mar 2009 00:38:16 +0000 (00:38 +0000)]
Fix bug where element access fields have wrong type, and printing of array types
jjenista [Thu, 5 Mar 2009 23:30:24 +0000 (23:30 +0000)]
fiddling with debug util
jjenista [Thu, 5 Mar 2009 21:15:46 +0000 (21:15 +0000)]
Alter token propagation from store so that computed change sets are applied to alpha prime only after change sets have propagated everywhere they can
jjenista [Thu, 5 Mar 2009 18:24:29 +0000 (18:24 +0000)]
fix bug in alias reporting for Java programs
jjenista [Wed, 4 Mar 2009 20:55:14 +0000 (20:55 +0000)]
accidentally checked in makefile line with one option pasted in between another option and its argument
jjenista [Wed, 4 Mar 2009 20:54:29 +0000 (20:54 +0000)]
Topological sort actually helps some benchmarks quite a bit
jjenista [Wed, 4 Mar 2009 18:21:26 +0000 (18:21 +0000)]
during token propagation for store, change to, don't just add, newly computed reachability states
jjenista [Wed, 4 Mar 2009 18:15:36 +0000 (18:15 +0000)]
Add a remove() operation
jjenista [Wed, 4 Mar 2009 18:14:37 +0000 (18:14 +0000)]
topological sort has little impact in overall analysis time; time spent sorting actually adds a few fractions of a second; removing the sort
jjenista [Wed, 4 Mar 2009 00:50:30 +0000 (00:50 +0000)]
change global sweep to shrink reachability sets rather than trash them
jjenista [Wed, 4 Mar 2009 00:03:19 +0000 (00:03 +0000)]
Topologically sort callee leaves to front of analysis work set
jjenista [Wed, 4 Mar 2009 00:02:51 +0000 (00:02 +0000)]
Return empty sets instead of null
adash [Tue, 3 Mar 2009 23:37:46 +0000 (23:37 +0000)]
bug fixes to prevent starvation of a lumberjack