charm.git
12 years agoThe new version of comlib! This version passed "make test" in charm/tests on order...
Isaac Dooley [Thu, 28 May 2009 20:15:15 +0000 (20:15 +0000)]
The new version of comlib! This version passed "make test" in charm/tests on order.cs.uiuc.edu.

12 years agoAdding a command line option for specifying additional commands that get passed into...
Isaac Dooley [Thu, 28 May 2009 19:41:31 +0000 (19:41 +0000)]
Adding a command line option for specifying additional commands that get passed into gdb when running with ++debug.

12 years agoImplementing a two-phase critical path detection scheme. All the critical path detect...
Isaac Dooley [Thu, 28 May 2009 19:12:36 +0000 (19:12 +0000)]
Implementing a two-phase critical path detection scheme. All the critical path detection code is disabled by a preprocessor macro by default. The new scheme is supported in SDAG and Charisma now.

12 years agoRefactoring control point and critical path detection code.
Isaac Dooley [Thu, 28 May 2009 18:19:34 +0000 (18:19 +0000)]
Refactoring control point and critical path detection code.

12 years agoUpdated some of the binary operations to correct errors.
David Kunzman [Thu, 28 May 2009 17:09:29 +0000 (17:09 +0000)]
Updated some of the binary operations to correct errors.

12 years agoFixed some bugs with the AltiVec version of the code.
David Kunzman [Wed, 27 May 2009 21:17:54 +0000 (21:17 +0000)]
Fixed some bugs with the AltiVec version of the code.

12 years agoFixed not operations.
David Kunzman [Wed, 27 May 2009 20:08:50 +0000 (20:08 +0000)]
Fixed not operations.

12 years agorevert Phil's previous check in since it crashes on solaris and windows.
Gengbin Zheng [Wed, 27 May 2009 19:44:57 +0000 (19:44 +0000)]
revert Phil's previous check in since it crashes on solaris and windows.

12 years agoupdated TODO
Aaron Becker [Tue, 26 May 2009 20:37:10 +0000 (20:37 +0000)]
updated TODO

12 years agoadded some more test cases
Aaron Becker [Tue, 26 May 2009 20:37:07 +0000 (20:37 +0000)]
added some more test cases

12 years agofixed arg name in main, simplified option handling
Aaron Becker [Tue, 26 May 2009 20:37:02 +0000 (20:37 +0000)]
fixed arg name in main, simplified option handling

12 years agotweaked template test, added test for use of reserved words
Aaron Becker [Tue, 26 May 2009 20:36:59 +0000 (20:36 +0000)]
tweaked template test, added test for use of reserved words

12 years agoimproved vcs ignoring
Aaron Becker [Tue, 26 May 2009 20:36:53 +0000 (20:36 +0000)]
improved vcs ignoring

12 years agofixed up some very long lines
Aaron Becker [Tue, 26 May 2009 20:36:51 +0000 (20:36 +0000)]
fixed up some very long lines

12 years agonew tests, todo list
Aaron Becker [Tue, 26 May 2009 20:36:47 +0000 (20:36 +0000)]
new tests, todo list

12 years agosimple makefile for unit tests
Aaron Becker [Tue, 26 May 2009 20:36:42 +0000 (20:36 +0000)]
simple makefile for unit tests

12 years agoadded .gitignore
Aaron Becker [Tue, 26 May 2009 20:36:39 +0000 (20:36 +0000)]
added .gitignore

12 years agoadded unit test directory
Aaron Becker [Tue, 26 May 2009 20:36:35 +0000 (20:36 +0000)]
added unit test directory

12 years agoallow net version to work with ccs on PE0
Gengbin Zheng [Tue, 26 May 2009 16:50:18 +0000 (16:50 +0000)]
allow net version to work with ccs on PE0

12 years agoset NODE_0_IS_CONVHOST to allow ccs working through PE0
Gengbin Zheng [Tue, 26 May 2009 16:46:31 +0000 (16:46 +0000)]
set NODE_0_IS_CONVHOST to allow ccs working through PE0

12 years agochange back so the BgStartLogByName when creating a dummy event does not automaticall...
Gengbin Zheng [Tue, 26 May 2009 16:27:25 +0000 (16:27 +0000)]
change back so the BgStartLogByName when creating a dummy event does not automatically assign objID. It is done selectively in trace-bluegene now

12 years agomark creating a sdag event as splitting an existing event
Gengbin Zheng [Tue, 26 May 2009 16:03:28 +0000 (16:03 +0000)]
mark creating a sdag event as splitting an existing event

12 years agoadded a parameter in creating a dummy event, which indicate if the new event is just...
Gengbin Zheng [Tue, 26 May 2009 16:02:38 +0000 (16:02 +0000)]
added a parameter in creating a dummy event, which indicate if the new event is just a split of an existing event.

12 years agoQuiet a bunch more warnings
Phil Miller [Mon, 25 May 2009 18:51:08 +0000 (18:51 +0000)]
Quiet a bunch more warnings

12 years agoremoved comments
Pritish Jetley [Mon, 25 May 2009 17:19:49 +0000 (17:19 +0000)]
removed comments

12 years agoRemove unused variables found by -Wall
Phil Miller [Mon, 25 May 2009 16:51:29 +0000 (16:51 +0000)]
Remove unused variables found by -Wall

They were even declared 'register', despite the non-criticality of
this code for performance.

12 years agoCorrect format string issues identified by -Wall
Phil Miller [Mon, 25 May 2009 16:41:59 +0000 (16:41 +0000)]
Correct format string issues identified by -Wall

12 years agolog
Pritish Jetley [Sun, 24 May 2009 01:26:33 +0000 (01:26 +0000)]
log

12 years agoFix variable names in header to match source, calls
Phil Miller [Fri, 22 May 2009 23:44:42 +0000 (23:44 +0000)]
Fix variable names in header to match source, calls

12 years agoa new command line option "++no-xrdb" that tells charmrun not to find xrdb. This...
Gengbin Zheng [Fri, 22 May 2009 03:37:59 +0000 (03:37 +0000)]
a new command line option "++no-xrdb" that tells charmrun not to find xrdb. This is to make ++debug works on turing compute nodes.

12 years agoassign objID from parent log
Gengbin Zheng [Thu, 21 May 2009 20:37:05 +0000 (20:37 +0000)]
assign objID from parent log

12 years agochanges to propagate objID among SDAG depedence events.
Gengbin Zheng [Thu, 21 May 2009 20:36:08 +0000 (20:36 +0000)]
changes to propagate objID among SDAG depedence events.

12 years agoUpdated malloc calls to use Charm++ aligned malloc instead of malloc_aligned.
David Kunzman [Thu, 21 May 2009 20:31:25 +0000 (20:31 +0000)]
Updated malloc calls to use Charm++ aligned malloc instead of malloc_aligned.

12 years agomake "group" in BgMsgEntry public again
Gengbin Zheng [Thu, 21 May 2009 13:49:59 +0000 (13:49 +0000)]
make "group" in BgMsgEntry public again

12 years agoMaking sum detail tracing only get intialized once.
Isaac Dooley [Thu, 21 May 2009 01:32:31 +0000 (01:32 +0000)]
Making sum detail tracing only get intialized once.

12 years agoget rid of compiler warnings for compiling SDAG code for bigsim.
Gengbin Zheng [Wed, 20 May 2009 20:58:21 +0000 (20:58 +0000)]
get rid of compiler warnings for compiling SDAG code for bigsim.

12 years agoCrayNeighborTable is no longer used
Abhinav Bhatele [Wed, 20 May 2009 17:59:49 +0000 (17:59 +0000)]
CrayNeighborTable is no longer used

12 years agosync of charm and standalone versions
Abhinav Bhatele [Wed, 20 May 2009 17:52:29 +0000 (17:52 +0000)]
sync of charm and standalone versions

12 years agonew Charm++ port for SiCortex and other mips64 machines
Abhinav Bhatele [Tue, 19 May 2009 19:44:08 +0000 (19:44 +0000)]
new Charm++ port for SiCortex and other mips64 machines

12 years agominor changes
Pritish Jetley [Tue, 19 May 2009 19:35:19 +0000 (19:35 +0000)]
minor changes

12 years agoFixing bugs in the Sum Detail CCS streaming.
Isaac Dooley [Mon, 18 May 2009 22:32:21 +0000 (22:32 +0000)]
Fixing bugs in the Sum Detail CCS streaming.

12 years agoproj
Pritish Jetley [Mon, 18 May 2009 17:41:02 +0000 (17:41 +0000)]
proj

12 years agomodified makefile so that CHARM_DIR can be changed
Pritish Jetley [Mon, 18 May 2009 17:39:50 +0000 (17:39 +0000)]
modified makefile so that CHARM_DIR can be changed

12 years agoadded code for load balancing.
Pritish Jetley [Mon, 18 May 2009 17:30:27 +0000 (17:30 +0000)]
added code for load balancing.

12 years agoupdated vmi path
Gengbin Zheng [Mon, 18 May 2009 17:05:31 +0000 (17:05 +0000)]
updated vmi path

12 years ago*** empty log message ***
Abhinav Bhatele [Fri, 15 May 2009 22:41:46 +0000 (22:41 +0000)]
*** empty log message ***

12 years agoRemoving bzero call which broke autobuild. Bzero is not portable.
Isaac Dooley [Fri, 15 May 2009 18:35:29 +0000 (18:35 +0000)]
Removing bzero call which broke autobuild. Bzero is not portable.

12 years agoNow the summary detail gathering mechanism supports multiple types for the fields...
Isaac Dooley [Fri, 15 May 2009 17:32:37 +0000 (17:32 +0000)]
Now the summary detail gathering mechanism supports multiple types for the fields (to adjust level compression).

12 years ago*** empty log message ***
Abhinav Bhatele [Fri, 15 May 2009 04:41:13 +0000 (04:41 +0000)]
*** empty log message ***

12 years agoAdd DEBUG_POSE_INVOKE as a compile time ifdef which will output verbose
Eric Bohm [Fri, 15 May 2009 00:55:00 +0000 (00:55 +0000)]
Add DEBUG_POSE_INVOKE as a compile time ifdef which will output verbose
parameter information for each POSE invoke.  Useful for figuring out
when your invoke results in an mismatch between method message and object.

12 years agobug fixes
Abhinav Bhatele [Fri, 15 May 2009 00:02:41 +0000 (00:02 +0000)]
bug fixes

12 years agoImproving the CCS sum detail streaming reporting tool.
Isaac Dooley [Thu, 14 May 2009 20:45:38 +0000 (20:45 +0000)]
Improving the CCS sum detail streaming reporting tool.

12 years agoimplemented local barriers to get around bigsim problem
Abhinav Bhatele [Thu, 14 May 2009 05:45:31 +0000 (05:45 +0000)]
implemented local barriers to get around bigsim problem

12 years agoproper BgPrintfs
Abhinav Bhatele [Thu, 14 May 2009 02:51:54 +0000 (02:51 +0000)]
proper BgPrintfs

12 years agothis wouldn't compile with gnu compilers on XT4
Abhinav Bhatele [Wed, 13 May 2009 21:07:07 +0000 (21:07 +0000)]
this wouldn't compile with gnu compilers on XT4

12 years agothis was needed for AMPI to compile correctly by using Charm's mpi.h
Abhinav Bhatele [Wed, 13 May 2009 21:05:59 +0000 (21:05 +0000)]
this was needed for AMPI to compile correctly by using Charm's mpi.h

12 years agohide cuda as build option
Gengbin Zheng [Tue, 12 May 2009 18:58:04 +0000 (18:58 +0000)]
hide cuda as build option

12 years agowrong comment style
Filippo Gioachin [Tue, 12 May 2009 16:38:17 +0000 (16:38 +0000)]
wrong comment style

12 years agoSplit nInts of CkArrayIndex into two shorts: the first one still named nInts, the...
Filippo Gioachin [Tue, 12 May 2009 04:15:45 +0000 (04:15 +0000)]
Split nInts of CkArrayIndex into two shorts: the first one still named nInts, the second one containing the dimensionality of the index. Valid only for 1-6D array indices.
Deleted some duplicated code (idx2str)

12 years agofixed problem with parallel make
Filippo Gioachin [Tue, 12 May 2009 04:00:12 +0000 (04:00 +0000)]
fixed problem with parallel make

12 years agoSwitching reduction of sum detail data to floating point utilization values and integ...
Isaac Dooley [Tue, 12 May 2009 03:11:54 +0000 (03:11 +0000)]
Switching reduction of sum detail data to floating point utilization values and integer sizes/EPs. A single char isn't big enough for these reductions.

12 years agoFixed define for VMI layers
Filippo Gioachin [Mon, 11 May 2009 23:40:58 +0000 (23:40 +0000)]
Fixed define for VMI layers

12 years agofind aprun first, using mpirun does not work on kraken
Gengbin Zheng [Mon, 11 May 2009 20:01:32 +0000 (20:01 +0000)]
find aprun first, using mpirun does not work on kraken

12 years agoNew functionality in CCS streaming of sum detail data for projections. The reduction...
Isaac Dooley [Mon, 11 May 2009 19:10:23 +0000 (19:10 +0000)]
New functionality in CCS streaming of sum detail data for projections. The reduction combiner needs to be fixed, and then the sum detail information will be correct.

12 years agoAdding CCS handlers that export trace information to projections as the program runs...
Isaac Dooley [Sun, 10 May 2009 23:50:01 +0000 (23:50 +0000)]
Adding CCS handlers that export trace information to projections as the program runs. Currently simple summary utilization over time is exported, and the beginnings of some handlers that export sum-detail per EP utilization.

12 years agoadded a comment about seqno in BgWriteTimelines
Gengbin Zheng [Sat, 9 May 2009 06:50:35 +0000 (06:50 +0000)]
added a comment about seqno in BgWriteTimelines

12 years agominor changes in API writing bgtrace summary file, and fixed couple memory leaks
Gengbin Zheng [Sat, 9 May 2009 06:44:23 +0000 (06:44 +0000)]
minor changes in API writing bgtrace summary file, and fixed couple memory leaks

12 years agobug fix for net layer
Filippo Gioachin [Sat, 9 May 2009 03:51:09 +0000 (03:51 +0000)]
bug fix for net layer

12 years agoAdded an implementation for all cores only having inter-node communication (#msgs...
Chao Mei [Fri, 8 May 2009 23:04:49 +0000 (23:04 +0000)]
Added an implementation for all cores only having inter-node communication (#msgs is roughly O(p^2))

12 years agoimplemented forwarding of output streams for all machine layers that do not reimpleme...
Filippo Gioachin [Fri, 8 May 2009 21:47:11 +0000 (21:47 +0000)]
implemented forwarding of output streams for all machine layers that do not reimplement CMK_PRINTF (includes MPI). Given this, moved some common code from charmrun.c to ccs-server.c

12 years agofixed bug on CkConditional using erroneous destructor
Filippo Gioachin [Fri, 8 May 2009 21:44:48 +0000 (21:44 +0000)]
fixed bug on CkConditional using erroneous destructor

12 years agofixed bug on CkConditional using erroneous destructor
Filippo Gioachin [Fri, 8 May 2009 21:38:26 +0000 (21:38 +0000)]
fixed bug on CkConditional using erroneous destructor

12 years agocode now relies on QD to start tree building. it also allocates cells and leaves
Pritish Jetley [Thu, 7 May 2009 23:57:58 +0000 (23:57 +0000)]
code now relies on QD to start tree building. it also allocates cells and leaves
from contiguous blocks within treepieces, instead of calling new each time makecell/leaf
is invoked.

bugs still remain. look up TMS logs.

12 years agomake sure xlc generate 32 bit code by using -q32
Gengbin Zheng [Thu, 7 May 2009 20:29:55 +0000 (20:29 +0000)]
make sure xlc generate 32 bit code by using -q32

12 years agomodifications.
Pritish Jetley [Thu, 7 May 2009 06:56:39 +0000 (06:56 +0000)]
modifications.

Code mostly revised.

* Have to insert QD code and callback to doBuildTree().
* Also, pendingChildren isn't correctly set. Currently, it is incremented in
  sendParticlesToChildren(), which can be called multiple times for a single
  child (first time when created, and next time when it is sent a message).
* Otherwise fine: if node is too fat and don't have parent, buffer messages
  till haveParent is set from recvRootFromParent. then, because wantToSplit,
  myRoot is created, all self and buffered particles are processed and sent
  to children. if node is too fat and have parent, we directly do the latter
  steps.

12 years agocode to continue simulation for multiple iterations. treepieces are reused
Pritish Jetley [Thu, 7 May 2009 03:18:55 +0000 (03:18 +0000)]
code to continue simulation for multiple iterations. treepieces are reused
and must be given roots by their parents. treepieces can only split their
roots when they have received their parents' roots. the splitting and treebuilding
must be delayed until they do.

12 years agobgprintf added
Abhinav Bhatele [Wed, 6 May 2009 20:48:17 +0000 (20:48 +0000)]
bgprintf added

12 years agoMoved freezing condition for CharmDebug into the charmdebug specific CpdPython group
Filippo Gioachin [Wed, 6 May 2009 16:11:48 +0000 (16:11 +0000)]
Moved freezing condition for CharmDebug into the charmdebug specific CpdPython group

12 years agorenamed cuda iterators
Aaron Becker [Wed, 6 May 2009 00:26:58 +0000 (00:26 +0000)]
renamed cuda iterators

12 years agochanged the bigsim mapping of processors to coordinates from ZYX to XYZ
Abhinav Bhatele [Tue, 5 May 2009 23:24:21 +0000 (23:24 +0000)]
changed the bigsim mapping of processors to coordinates from ZYX to XYZ
this now matches BigNetSim's mapping

12 years agoAdded node cyclic mapping for elements to test how well comm thread behaves when...
Chao Mei [Tue, 5 May 2009 22:40:57 +0000 (22:40 +0000)]
Added node cyclic mapping for elements to test how well comm thread behaves when inter-node comm happens more frequently.

12 years agoadded module dep file
Aaron Becker [Tue, 5 May 2009 22:24:37 +0000 (22:24 +0000)]
added module dep file

12 years agoAdded ParFUM iterators layer
Aaron Becker [Tue, 5 May 2009 22:20:45 +0000 (22:20 +0000)]
Added ParFUM iterators layer

12 years agodefault optimization level changed
Aaron Becker [Tue, 5 May 2009 19:07:02 +0000 (19:07 +0000)]
default optimization level changed

12 years agoadded success assertions
Aaron Becker [Tue, 5 May 2009 19:06:34 +0000 (19:06 +0000)]
added success assertions

12 years agoMinor changes to makefile.
Isaac Dooley [Tue, 5 May 2009 18:53:44 +0000 (18:53 +0000)]
Minor changes to makefile.

12 years agocompilation error fixed
Abhinav Bhatele [Tue, 5 May 2009 17:45:41 +0000 (17:45 +0000)]
compilation error fixed

12 years agowhen reduction client is CkReductionClientBundle, pup storedCallback is not enough...
Gengbin Zheng [Tue, 5 May 2009 07:27:16 +0000 (07:27 +0000)]
when reduction client is CkReductionClientBundle, pup storedCallback is not enough, because the c function param point to CkReductionClientBundle, which needs to be separately pup'ed.

12 years agofixed a typo
Gengbin Zheng [Tue, 5 May 2009 07:10:07 +0000 (07:10 +0000)]
fixed a typo

12 years agodisable scalable-start's fork setting when restart a crashed node.
Gengbin Zheng [Tue, 5 May 2009 07:03:50 +0000 (07:03 +0000)]
disable scalable-start's fork setting when restart a crashed node.

12 years agofixed USE_REDUCTION macro redefinition when using FT
Gengbin Zheng [Tue, 5 May 2009 06:56:41 +0000 (06:56 +0000)]
fixed USE_REDUCTION macro redefinition when using FT

12 years ago*** empty log message ***
Abhinav Bhatele [Tue, 5 May 2009 06:17:55 +0000 (06:17 +0000)]
*** empty log message ***

12 years agothe unscalable way of collecting topology will not be used on BG and XT machines
Abhinav Bhatele [Tue, 5 May 2009 02:47:32 +0000 (02:47 +0000)]
the unscalable way of collecting topology will not be used on BG and XT machines
anymore.
I also changed nodenum to nodeIDs

12 years agomoved new function BgSplitEntry out of bigsim API lib because otherwise it can not...
Gengbin Zheng [Sat, 2 May 2009 13:18:55 +0000 (13:18 +0000)]
moved new function BgSplitEntry out of bigsim API lib because otherwise it can not compile a standalone bigsim program due to calls to timers

12 years agoa new way to split broadcast, instead of a chain, all broadcast delivery has a backwa...
Gengbin Zheng [Sat, 2 May 2009 06:03:34 +0000 (06:03 +0000)]
a new way to split broadcast, instead of a chain, all broadcast delivery has a backward dep to the head event

12 years agoa new split log bigsim API calls that takes an array of bglogs for backward deps
Gengbin Zheng [Sat, 2 May 2009 06:02:34 +0000 (06:02 +0000)]
a new split log bigsim API calls that takes an array of bglogs for backward deps

12 years agomade changes to child treepiece spawn procedure. Only pieces that are actually needed
Pritish Jetley [Sat, 2 May 2009 00:18:31 +0000 (00:18 +0000)]
made changes to child treepiece spawn procedure. Only pieces that are actually needed
are ever created.

12 years agoInitial check-ins for multicore-linux32
Chao Mei [Fri, 1 May 2009 06:10:59 +0000 (06:10 +0000)]
Initial check-ins for multicore-linux32

12 years agoAll phases work for single iteration. Remove inserted treepieces and free
Pritish Jetley [Fri, 1 May 2009 00:09:31 +0000 (00:09 +0000)]
All phases work for single iteration. Remove inserted treepieces and free
allocated memory in ParticleChunk::cleanup, etc.

12 years agoprint teh current configuration target
Gengbin Zheng [Thu, 30 Apr 2009 20:35:39 +0000 (20:35 +0000)]
print teh current configuration target