adash [Wed, 31 Dec 2008 20:18:08 +0000 (20:18 +0000)]
new files for new keyword "getoffset"
adash [Wed, 31 Dec 2008 20:16:53 +0000 (20:16 +0000)]
modified compiler files for adding new keyword "getoffset" and adding
hack for manual prefetches
check in some files that needed tabbing
runtime still not working but compiling works!
bdemsky [Sun, 21 Dec 2008 05:37:08 +0000 (05:37 +0000)]
inline the gc check
bdemsky [Sun, 21 Dec 2008 05:31:04 +0000 (05:31 +0000)]
better error message
adash [Mon, 15 Dec 2008 18:20:37 +0000 (18:20 +0000)]
scripts for automatically running javasingle versions of benchmarks
bdemsky [Thu, 11 Dec 2008 19:44:30 +0000 (19:44 +0000)]
change
bdemsky [Tue, 9 Dec 2008 00:07:25 +0000 (00:07 +0000)]
change
bdemsky [Mon, 8 Dec 2008 23:45:11 +0000 (23:45 +0000)]
more changes to allow an override
bdemsky [Mon, 8 Dec 2008 23:42:04 +0000 (23:42 +0000)]
make afid work easily
navid [Wed, 26 Nov 2008 19:42:56 +0000 (19:42 +0000)]
*** empty log message ***
adash [Tue, 25 Nov 2008 21:19:07 +0000 (21:19 +0000)]
more changes to support range prefetching :
TODO
1. optimizing boundary condition for arrays
jzhou [Mon, 24 Nov 2008 22:30:36 +0000 (22:30 +0000)]
bug fix in threadsimulator
navid [Mon, 24 Nov 2008 22:25:12 +0000 (22:25 +0000)]
*** empty log message ***
adash [Mon, 24 Nov 2008 21:13:57 +0000 (21:13 +0000)]
more changes to range prefetches
navid [Mon, 24 Nov 2008 20:46:53 +0000 (20:46 +0000)]
*** empty log message ***
navid [Mon, 24 Nov 2008 19:02:15 +0000 (19:02 +0000)]
*** empty log message ***
adash [Mon, 24 Nov 2008 18:06:52 +0000 (18:06 +0000)]
new files for range prefetching
More work needs to be done to test it
navid [Sat, 22 Nov 2008 02:33:13 +0000 (02:33 +0000)]
*** empty log message ***
navid [Sat, 22 Nov 2008 00:42:38 +0000 (00:42 +0000)]
*** empty log message ***
navid [Fri, 21 Nov 2008 19:07:26 +0000 (19:07 +0000)]
*** empty log message ***
navid [Fri, 21 Nov 2008 09:52:41 +0000 (09:52 +0000)]
*** empty log message ***
navid [Fri, 21 Nov 2008 03:10:31 +0000 (03:10 +0000)]
*** empty log message ***
navid [Fri, 21 Nov 2008 03:01:40 +0000 (03:01 +0000)]
*** empty log message ***
adash [Thu, 20 Nov 2008 00:37:01 +0000 (00:37 +0000)]
changes for the new build
bdemsky [Wed, 19 Nov 2008 23:19:43 +0000 (23:19 +0000)]
changes to build SOR java version
bdemsky [Wed, 19 Nov 2008 23:02:32 +0000 (23:02 +0000)]
changes
bdemsky [Wed, 19 Nov 2008 22:58:57 +0000 (22:58 +0000)]
more hacks
bdemsky [Wed, 19 Nov 2008 09:24:20 +0000 (09:24 +0000)]
changes
bdemsky [Wed, 19 Nov 2008 09:11:14 +0000 (09:11 +0000)]
another build error
bdemsky [Wed, 19 Nov 2008 09:09:14 +0000 (09:09 +0000)]
more changes
bdemsky [Wed, 19 Nov 2008 08:56:21 +0000 (08:56 +0000)]
build script for everything
bdemsky [Wed, 19 Nov 2008 08:53:53 +0000 (08:53 +0000)]
standardize the benchmarks so we script them easily
bdemsky [Wed, 19 Nov 2008 08:38:24 +0000 (08:38 +0000)]
change mid arrays
bdemsky [Wed, 19 Nov 2008 07:09:41 +0000 (07:09 +0000)]
new option for alex
bdemsky [Wed, 19 Nov 2008 06:07:52 +0000 (06:07 +0000)]
print a better error message
jzhou [Tue, 18 Nov 2008 23:20:44 +0000 (23:20 +0000)]
Fix bugs in scheduling simulator. Also add -robustroot setting in Tests/dotest script
adash [Tue, 18 Nov 2008 23:08:09 +0000 (23:08 +0000)]
check in all changes to the benchmarks for the pldi2009 paper
adash [Fri, 14 Nov 2008 08:23:43 +0000 (08:23 +0000)]
Add new results for bandwidth
adash [Thu, 13 Nov 2008 00:23:24 +0000 (00:23 +0000)]
bug fix in 2DConv and other small changes to makefile
jjenista [Wed, 12 Nov 2008 06:12:29 +0000 (06:12 +0000)]
trim outputs for benchmarking
adash [Wed, 12 Nov 2008 04:09:34 +0000 (04:09 +0000)]
latest changes to 2DFFT and LUFact benchmark
adash [Wed, 12 Nov 2008 03:59:45 +0000 (03:59 +0000)]
Changes to increase garbage collector heap
Runtime changes for a single threaded transaction commit
adash [Wed, 12 Nov 2008 03:54:33 +0000 (03:54 +0000)]
checking file for matrix size 8192 X 8192 that causes a segfault for java single version
jzhou [Wed, 12 Nov 2008 03:34:14 +0000 (03:34 +0000)]
changes
jzhou [Tue, 11 Nov 2008 23:02:32 +0000 (23:02 +0000)]
add the JGF MolDyn benchmark
jjenista [Tue, 11 Nov 2008 05:38:34 +0000 (05:38 +0000)]
add new class file
jjenista [Tue, 11 Nov 2008 05:33:11 +0000 (05:33 +0000)]
ran all benchmarks with new method context version, alias reports are the same
jjenista [Tue, 11 Nov 2008 04:44:34 +0000 (04:44 +0000)]
added concept of method context
jjenista [Mon, 10 Nov 2008 23:55:42 +0000 (23:55 +0000)]
eliminate two erroneous statements in callee->caller node creation
jjenista [Mon, 10 Nov 2008 22:52:40 +0000 (22:52 +0000)]
eliminate alias
jjenista [Mon, 10 Nov 2008 20:23:26 +0000 (20:23 +0000)]
added new benchmark result to summary
bdemsky [Mon, 10 Nov 2008 04:56:59 +0000 (04:56 +0000)]
small changes...
adash [Mon, 10 Nov 2008 04:02:24 +0000 (04:02 +0000)]
new 2D convolution benchmark
jzhou [Mon, 10 Nov 2008 01:18:11 +0000 (01:18 +0000)]
use SLGCC
jzhou [Sun, 9 Nov 2008 22:33:19 +0000 (22:33 +0000)]
change to randomly generate inputs
jjenista [Sun, 9 Nov 2008 21:50:27 +0000 (21:50 +0000)]
make sure strong updates do not affect benchmark results, currently
jzhou [Sun, 9 Nov 2008 03:30:00 +0000 (03:30 +0000)]
*** empty log message ***
jzhou [Sun, 9 Nov 2008 03:28:01 +0000 (03:28 +0000)]
*** empty log message ***
jzhou [Sun, 9 Nov 2008 03:20:46 +0000 (03:20 +0000)]
use SLGCC instead of GCC and use -o2 for raw version
jzhou [Sun, 9 Nov 2008 02:50:15 +0000 (02:50 +0000)]
remove redundant files
jzhou [Sun, 9 Nov 2008 02:44:19 +0000 (02:44 +0000)]
Using -o2 for raw version binary causes problems. Change it to -o0.
jzhou [Wed, 5 Nov 2008 23:00:01 +0000 (23:00 +0000)]
bug fixing
jzhou [Wed, 5 Nov 2008 22:25:11 +0000 (22:25 +0000)]
bug fixing
adash [Wed, 5 Nov 2008 02:21:20 +0000 (02:21 +0000)]
improved version for better scaling results
jzhou [Wed, 5 Nov 2008 01:08:08 +0000 (01:08 +0000)]
new benchmark -- fluidanimate for Scheduling
adash [Tue, 4 Nov 2008 20:24:56 +0000 (20:24 +0000)]
make 1D fft objects local and clean up code
jzhou [Tue, 4 Nov 2008 18:44:41 +0000 (18:44 +0000)]
add precise profiling data for multicore version profiling
adash [Mon, 3 Nov 2008 03:36:47 +0000 (03:36 +0000)]
add java files
adash [Mon, 3 Nov 2008 03:34:50 +0000 (03:34 +0000)]
working version ....
global array allocation in fft needs to address scaling problems
adash [Fri, 31 Oct 2008 23:29:10 +0000 (23:29 +0000)]
Added few more Math functionality for 2DFFT benchmark
adash [Fri, 31 Oct 2008 23:28:23 +0000 (23:28 +0000)]
New benchmark for 2D FFT ... still has compile problems with global data
jjenista [Mon, 27 Oct 2008 23:16:57 +0000 (23:16 +0000)]
realized strong updates has incorrect conditions, made comment to fix, will do this later
adash [Mon, 27 Oct 2008 19:55:34 +0000 (19:55 +0000)]
Java non transactional version for Em3d
jzhou [Wed, 22 Oct 2008 22:35:30 +0000 (22:35 +0000)]
Integrate Jim's ownership ananlysis and fix some bugs in scheduling simulator
bdemsky [Wed, 22 Oct 2008 00:53:50 +0000 (00:53 +0000)]
extra hash table implementation...might be worth trying
adash [Tue, 21 Oct 2008 00:31:41 +0000 (00:31 +0000)]
Scripts and Benchmarks changed to function with 8 machines
jjenista [Fri, 17 Oct 2008 19:57:44 +0000 (19:57 +0000)]
test doesn't work\!
jjenista [Fri, 17 Oct 2008 17:45:13 +0000 (17:45 +0000)]
new test for paper example
jzhou [Fri, 10 Oct 2008 19:43:59 +0000 (19:43 +0000)]
let scheduling analysis be able to use profile data
jjenista [Thu, 9 Oct 2008 21:58:36 +0000 (21:58 +0000)]
Added specific method call improvement test case for global sweep
jjenista [Wed, 8 Oct 2008 23:03:38 +0000 (23:03 +0000)]
Global sweep added
jjenista [Wed, 8 Oct 2008 18:30:48 +0000 (18:30 +0000)]
Committing a stable version of global sweep that works for simple strong updates, does not work correctly after some method calls
jjenista [Mon, 6 Oct 2008 23:54:38 +0000 (23:54 +0000)]
Added hooks for global sweep and made sure that strong update logic without effecting the actual strong update does not affect the final result. Next step is to do the strong update edge removal and implement global sweep after strong updates and all method calls.
jjenista [Fri, 3 Oct 2008 21:47:05 +0000 (21:47 +0000)]
Fixed a hack that was bugging me with a proper solution.
jzhou [Fri, 3 Oct 2008 19:16:15 +0000 (19:16 +0000)]
add profile codes for multicore version
jjenista [Thu, 2 Oct 2008 21:43:55 +0000 (21:43 +0000)]
Verified accuracy of results by manual inspection.
jjenista [Wed, 1 Oct 2008 22:25:57 +0000 (22:25 +0000)]
Fixed up another canonical misuse, changed wording of alias reporting to see the phrase 'no aliases' first, reran all benchmarks
jjenista [Wed, 1 Oct 2008 21:51:48 +0000 (21:51 +0000)]
Tighten up use of canonical objects and halt system if a canonical object's hashcode changes.
navid [Tue, 30 Sep 2008 21:06:59 +0000 (21:06 +0000)]
*** empty log message ***
navid [Mon, 29 Sep 2008 22:05:04 +0000 (22:05 +0000)]
*** empty log message ***
navid [Mon, 29 Sep 2008 21:35:02 +0000 (21:35 +0000)]
*** empty log message ***
navid [Mon, 29 Sep 2008 19:14:24 +0000 (19:14 +0000)]
*** empty log message ***
jjenista [Mon, 29 Sep 2008 18:43:31 +0000 (18:43 +0000)]
Introduced ZEROORMORE arity, something appears to be broken so another change is coming
jjenista [Thu, 25 Sep 2008 22:10:15 +0000 (22:10 +0000)]
Fix implementation to use little d when possible, and big D only if there's no other choice
jjenista [Wed, 24 Sep 2008 22:49:38 +0000 (22:49 +0000)]
Bug fix where two isolated heap regions with empty reachability sets have
a reference edge created between them. Actually, the reachability sets
arent' empty, they contain the empty token tuple set, but the edge that
gets created has the empty reachability set. Changed the field set method
to detect this and fill it in. This is not a good fix, find the reason
why it happens in the first place.
navid [Tue, 23 Sep 2008 20:13:56 +0000 (20:13 +0000)]
*** empty log message ***
navid [Tue, 23 Sep 2008 19:03:38 +0000 (19:03 +0000)]
*** empty log message ***
jjenista [Tue, 23 Sep 2008 00:42:02 +0000 (00:42 +0000)]
Reran analysis on benchmarks with newest updates
jjenista [Tue, 23 Sep 2008 00:06:13 +0000 (00:06 +0000)]
First bug fix is that the "unshadow" token conversion mistakenly transforms the oldest shadow token back into a shadow token.
Second bug fix is that callee nodes from an allocation site don't use the parameter rewrite rules (H) and similarly edges in the callee that are added to the caller don't use the J or K rewrite rules either. Those reachability sets should use their callee alpha() or beta() as the input "rules" to the rewrite method.
adash [Mon, 22 Sep 2008 22:10:02 +0000 (22:10 +0000)]
get rid of unnecessary recv_data , SUCCESSFUL/UNSUCESSFUL ctrl msg
add new debug info