charm.git
16 years agoadded back missing NO_PUP for isomalloc version.
Gengbin Zheng [Tue, 20 Sep 2005 18:33:20 +0000 (18:33 +0000)]
added back missing NO_PUP for isomalloc version.
fixed Makefile for compilation.

16 years agoAdded test for iso and use RotateLB
Chao Huang [Tue, 20 Sep 2005 18:28:59 +0000 (18:28 +0000)]
Added test for iso and use RotateLB

16 years agoRoll back to 12/04 to correct my accidental checkin of test code.
Chao Huang [Tue, 20 Sep 2005 18:22:09 +0000 (18:22 +0000)]
Roll back to 12/04 to correct my accidental checkin of test code.

16 years agoFixed a bug. Worked correctly now for cases like: size=40, pencilsize=320 --> splitti...
Yan Shi [Tue, 20 Sep 2005 03:32:33 +0000 (03:32 +0000)]
Fixed a bug. Worked correctly now for cases like: size=40, pencilsize=320 --> splitting into
40*8, not 20*16

16 years agotorus manager initialized.
Sameer Kumar [Tue, 20 Sep 2005 02:42:22 +0000 (02:42 +0000)]
torus manager initialized.

16 years agoFixed pencil square blocking. Now pencil cluster size could be
Yan Shi [Tue, 20 Sep 2005 01:35:10 +0000 (01:35 +0000)]
Fixed pencil square blocking. Now pencil cluster size could be
as large as N*N --> only 1 pencil is created.

16 years agoAdded workaround for immediate msg for groups on smp:
Chao Huang [Mon, 19 Sep 2005 21:48:39 +0000 (21:48 +0000)]
Added workaround for immediate msg for groups on smp:
in that case, immedate msg should be sent out as normal msg.

16 years agofixed for uth version
Gengbin Zheng [Sun, 18 Sep 2005 16:09:46 +0000 (16:09 +0000)]
fixed for uth version

16 years agoAdded a couple of functions to CkMigratable for the purpose of tracing ckLocal functi...
Amit Sharma [Sat, 17 Sep 2005 20:28:25 +0000 (20:28 +0000)]
Added a couple of functions to CkMigratable for the purpose of tracing ckLocal function calls correctly in the LB and to do away with the inline message overhead in cosmology

16 years ago*** empty log message ***
Amit Sharma [Sat, 17 Sep 2005 20:27:10 +0000 (20:27 +0000)]
*** empty log message ***

16 years agoRefinement algorithm is working, sequentially as well as in parallel. Coarsen algori...
Terry L. Wilmarth [Fri, 16 Sep 2005 22:09:53 +0000 (22:09 +0000)]
Refinement algorithm is working, sequentially as well as in parallel.  Coarsen algorithm is implemented but not yet working.
Task: FEM

16 years agoLatest refine algortihm.
Terry L. Wilmarth [Thu, 15 Sep 2005 20:51:33 +0000 (20:51 +0000)]
Latest refine algortihm.

16 years agoI got careless and left out two #if statements that isolated the scenarios
Chee Wai Lee [Thu, 15 Sep 2005 18:38:21 +0000 (18:38 +0000)]
I got careless and left out two #if statements that isolated the scenarios
where an explicit buffer is not required (hence breaking all the net versions)

16 years agoAdded a tentative #define and machine flag for the XT3 in order to
Chee Wai Lee [Wed, 14 Sep 2005 23:04:41 +0000 (23:04 +0000)]
Added a tentative #define and machine flag for the XT3 in order to
implement a pcqueue workaround the prevents a strange memory corruption
error on the XT3. This workaround depends on the correctness of Q->len
in the PCQueue implementation which is not guaranteed on almost every other
machine, hence the need for a special #define until the code can be fully
debugged.

16 years agoadded locking code for idxl lists, and added some more tests for structure of the
Nilesh Choudhury [Wed, 14 Sep 2005 20:39:43 +0000 (20:39 +0000)]
added locking code for idxl lists, and added some more tests for structure of the
refined and coarsened mesh

16 years agopriority of 1 for output works better.
Eric Bohm [Wed, 14 Sep 2005 14:44:42 +0000 (14:44 +0000)]
priority of 1 for output works better.

16 years agoC++ style comments are often unacceptable to more pedantic compilers
Chee Wai Lee [Wed, 14 Sep 2005 10:39:45 +0000 (10:39 +0000)]
C++ style comments are often unacceptable to more pedantic compilers
(as in the case of the cross compiler on the xt3). Changed to use
old style c comments.

16 years ago * Added a mechanism to allow some control over machines with extremely
Chee Wai Lee [Wed, 14 Sep 2005 10:27:49 +0000 (10:27 +0000)]
 * Added a mechanism to allow some control over machines with extremely
 * inefficient terminal IO mechanisms. Case in point: the XT3 has a
 * 20ms flush overhead along with about 25MB/s bandwidth for IO. This,
 * coupled with a default setup using unbuffered stdout introduced
 * severe overheads (and hence limiting scaling) for applications like
 * NAMD.

Features:
conv-mach flags: CMI_IO_BUFFER_EXPLICIT - converse explicitly provides
    a io buffer to the kernel
                 CMI_IO_FLUSH_EXPLICIT - machine requires users to
    either manually flush stdout as required or entrust flushes to
    kernel management of the io buffer.
runtime flags: +io_buffer_size <size> - user tells converse to
    create an io buffer of size <size> (default 250k) if
    CMI_IO_BUFFER_EXPLICIT is set. Ignored otherwise.
               +io_flush_explicit - user tells converse not to flush
    after each CmiPrintf, giving the user full control over stdout
    flushing.

16 years agoSuperficial formatting changes.
Terry L. Wilmarth [Tue, 13 Sep 2005 21:40:18 +0000 (21:40 +0000)]
Superficial formatting changes.

16 years agofixed allocation problem in setup
Eric Bohm [Tue, 13 Sep 2005 15:55:43 +0000 (15:55 +0000)]
fixed allocation problem in setup

16 years agoFix initialization bug in setup. Object needs to be allocated for use
Eric Bohm [Tue, 13 Sep 2005 14:50:10 +0000 (14:50 +0000)]
Fix initialization bug in setup.  Object needs to be allocated for use
of its members to be sensible.

16 years agoModified code to take sequence numbers into account when ordering events.
David Kunzman [Mon, 12 Sep 2005 20:56:06 +0000 (20:56 +0000)]
Modified code to take sequence numbers into account when ordering events.

16 years agoModified code to take sequence numbers into account when ordering events.
David Kunzman [Mon, 12 Sep 2005 20:46:56 +0000 (20:46 +0000)]
Modified code to take sequence numbers into account when ordering events.

16 years agoBGL Torous utility file added to Makefile
Sameer Kumar [Mon, 12 Sep 2005 20:04:48 +0000 (20:04 +0000)]
BGL Torous utility file added to Makefile

16 years agoA topology utility class for BGL. Has to be in src/util so that the Makefile can...
Sameer Kumar [Mon, 12 Sep 2005 20:00:54 +0000 (20:00 +0000)]
A topology utility class for BGL. Has to be in src/util so that the Makefile can include it for
both bluegene and mpi-bluegene versions.

16 years agoremovenode must make the entry -1
Nilesh Choudhury [Mon, 12 Sep 2005 16:18:08 +0000 (16:18 +0000)]
removenode must make the entry -1

16 years agomodifications to support refinement coarsening for any type of partition.
Nilesh Choudhury [Mon, 12 Sep 2005 16:17:29 +0000 (16:17 +0000)]
modifications to support refinement coarsening for any type of partition.
map.C & parallel_part.C need to be copied from my directory. Those are not being checked in,
as they now generate a different set of ghost lists.

16 years ago*** empty log message ***
Chao Huang [Mon, 12 Sep 2005 00:58:52 +0000 (00:58 +0000)]
*** empty log message ***

16 years agoAdded a mesh sizing attribute for mesh adaptivity.
Terry L. Wilmarth [Fri, 9 Sep 2005 20:51:46 +0000 (20:51 +0000)]
Added a mesh sizing attribute for mesh adaptivity.
Task: FEM

16 years agodefault to having the secondphase load balancing off
Eric Bohm [Fri, 9 Sep 2005 03:49:24 +0000 (03:49 +0000)]
default to having the secondphase load balancing off
it sometimes produces invalid results

16 years agoenable gspacesum for the asymmetric pc
Eric Bohm [Thu, 8 Sep 2005 20:32:49 +0000 (20:32 +0000)]
enable gspacesum for the asymmetric pc

16 years agoAdded #define for CmiMemcpy. Currently just an alias to memcpy.
Eric Bohm [Wed, 7 Sep 2005 16:14:48 +0000 (16:14 +0000)]
Added #define for CmiMemcpy.  Currently just an alias to memcpy.
This will permit local memcpy's to be introduced at the machine layer,
such as BGLmemcpy, which should be faster than the default memcpy.

16 years agobetter handling of making nodes and cpus out of ++p cpus. Ctl-c get rid of the job.
Gengbin Zheng [Wed, 7 Sep 2005 02:56:53 +0000 (02:56 +0000)]
better handling of making nodes and cpus out of ++p cpus. Ctl-c get rid of the job.

16 years agouse macro CmiNodeAlive to replace direct access to _validProcessors, this fixed blueg...
Gengbin Zheng [Tue, 6 Sep 2005 03:50:21 +0000 (03:50 +0000)]
use macro CmiNodeAlive to replace direct access to _validProcessors, this fixed bluegene test

16 years agoreplace alloca by malloc which could overflow bigsim's small stack
Gengbin Zheng [Mon, 5 Sep 2005 19:45:00 +0000 (19:45 +0000)]
replace alloca by malloc which could overflow bigsim's small stack

16 years agouse macro CmiNodeAlive to replace direct access to _validProcessors, this fixed blueg...
Gengbin Zheng [Mon, 5 Sep 2005 19:38:39 +0000 (19:38 +0000)]
use macro CmiNodeAlive to replace direct access to _validProcessors, this fixed bluegene test

16 years agodefined macro CmiNodeAlive to replace direct access to_validProcessors.
Gengbin Zheng [Mon, 5 Sep 2005 19:29:51 +0000 (19:29 +0000)]
defined macro CmiNodeAlive to replace direct access to_validProcessors.

16 years agoadded pathscale in help page
Gengbin Zheng [Mon, 5 Sep 2005 04:19:37 +0000 (04:19 +0000)]
added pathscale in help page

16 years agoconfig for pathscale compiler suite
Gengbin Zheng [Mon, 5 Sep 2005 04:17:46 +0000 (04:17 +0000)]
config for pathscale compiler suite

16 years agoproperly keep expedited message when sending message unknown.
Gengbin Zheng [Sun, 4 Sep 2005 23:05:58 +0000 (23:05 +0000)]
properly keep expedited message when sending message unknown.

16 years agoCMK_NO_DL=true which fixed the rutime error on mac with mpich with malloc hook test...
Gengbin Zheng [Sun, 4 Sep 2005 06:59:05 +0000 (06:59 +0000)]
CMK_NO_DL=true which fixed the rutime error on mac with mpich with malloc hook test error.

16 years agoupdated.
Gengbin Zheng [Sun, 4 Sep 2005 06:58:06 +0000 (06:58 +0000)]
updated.

16 years agowhen defined CMK_NO_DL, skip -ldl
Gengbin Zheng [Sun, 4 Sep 2005 06:57:37 +0000 (06:57 +0000)]
when defined CMK_NO_DL, skip -ldl

16 years agomove -lm to the end of link command line (this fixed Mac link with mpich and malloc...
Gengbin Zheng [Sun, 4 Sep 2005 06:55:56 +0000 (06:55 +0000)]
move -lm to the end of link command line (this fixed Mac link with mpich and malloc hook check problem)

16 years agoremoved debugging statement
Gengbin Zheng [Sun, 4 Sep 2005 06:40:49 +0000 (06:40 +0000)]
removed debugging statement

16 years agoModified the regulaor expressions that detect function declarations. Changed the
David Kunzman [Sat, 3 Sep 2005 17:47:17 +0000 (17:47 +0000)]
Modified the regulaor expressions that detect function declarations.  Changed the
regular expressions so they no longer consider the "*" apart of a parameter's type if
there is no space between the parameter's type and the "*".

16 years agomore stricter structure tests and new coord and boundary print functions for debugging
Nilesh Choudhury [Sat, 3 Sep 2005 16:28:07 +0000 (16:28 +0000)]
more stricter structure tests and new coord and boundary print functions for debugging

16 years agosome bug fixes and test cases to check for sane connectivity and slivers
Nilesh Choudhury [Sat, 3 Sep 2005 06:15:44 +0000 (06:15 +0000)]
some bug fixes and test cases to check for sane connectivity and slivers

16 years agochanged mode of copying headers from ln to cp
Filippo Gioachin [Sat, 3 Sep 2005 00:37:05 +0000 (00:37 +0000)]
changed mode of copying headers from ln to cp

16 years agoadded a comment so that the compiler does not give the annoying warning about empty...
Gengbin Zheng [Fri, 2 Sep 2005 18:23:55 +0000 (18:23 +0000)]
added a comment so that the compiler does not give the annoying warning about empty file.

16 years agoturn off delegation when running on Bluegene/L
Eric Bohm [Fri, 2 Sep 2005 06:39:20 +0000 (06:39 +0000)]
turn off delegation when running on Bluegene/L
it does awful things to performance

16 years agoifdef protect the memalign call so it doesn't happen on BG/L
Eric Bohm [Fri, 2 Sep 2005 00:12:59 +0000 (00:12 +0000)]
ifdef protect the memalign call so it doesn't happen on BG/L
memalign isn't supported there and produces weird unaligned results

16 years agoXLC didnt like my malloc change. I gave some temporary redefs for mm_malloc. Hopefull...
Sameer Kumar [Thu, 1 Sep 2005 14:19:22 +0000 (14:19 +0000)]
XLC didnt like my malloc change. I gave some temporary redefs for mm_malloc. Hopefully it should be all good now.

16 years agoFixing autobuild crash. I was using metamalloc before it was declared.
Sameer Kumar [Thu, 1 Sep 2005 13:47:09 +0000 (13:47 +0000)]
Fixing autobuild crash. I was using metamalloc before it was declared.
Switching that to mm_malloc. Also I put a define for mm_malloc in memory OS version.

16 years agominor changes.
Gengbin Zheng [Thu, 1 Sep 2005 04:47:18 +0000 (04:47 +0000)]
minor changes.

16 years agoremoved semicolumn
Filippo Gioachin [Thu, 1 Sep 2005 04:42:57 +0000 (04:42 +0000)]
removed semicolumn

16 years agoguess where gm is installed.
Gengbin Zheng [Thu, 1 Sep 2005 04:13:21 +0000 (04:13 +0000)]
guess where gm is installed.

16 years agoMaking following changes for BGL machine layer.
Sameer Kumar [Wed, 31 Aug 2005 23:18:16 +0000 (23:18 +0000)]
Making following changes for BGL machine layer.

1) Getting rid of the preallocate hack error. It was getting quite annoying.
2) Changing call to meta_malloc in CmiOutOfMemoryInit, it was bug which was
breaking memory gnu. Basically when malloc failed, it called cmiabort.

3) Make memory gnu return 16 byte aligned pointers.

16 years agoMaking a network progress period of 16 the default.
Sameer Kumar [Wed, 31 Aug 2005 23:11:59 +0000 (23:11 +0000)]
Making a network progress period of 16 the default.

16 years agomodified the coarsening algorithm
Nilesh Choudhury [Wed, 31 Aug 2005 21:04:55 +0000 (21:04 +0000)]
modified the coarsening algorithm

16 years agoupdated.
Gengbin Zheng [Wed, 31 Aug 2005 17:27:29 +0000 (17:27 +0000)]
updated.

16 years agofixed area calculation routine for slivers and added updating shared node attributes...
Nilesh Choudhury [Wed, 31 Aug 2005 17:14:19 +0000 (17:14 +0000)]
fixed area calculation routine for slivers and added updating shared node attributes across chunks

16 years agofixed to display the exact linking options.
Gengbin Zheng [Wed, 31 Aug 2005 16:46:21 +0000 (16:46 +0000)]
fixed to display the exact linking options.

16 years agoudpated.
Gengbin Zheng [Wed, 31 Aug 2005 16:41:11 +0000 (16:41 +0000)]
udpated.

16 years agoMac specific conf file for xlc++.
Gengbin Zheng [Wed, 31 Aug 2005 16:40:39 +0000 (16:40 +0000)]
Mac specific conf file for xlc++.

16 years agoset OPTS_CC="$OPTS_CC $USER_OPTS_CC", etc so that user provided libdir and incdir...
Gengbin Zheng [Wed, 31 Aug 2005 16:39:42 +0000 (16:39 +0000)]
set OPTS_CC="$OPTS_CC $USER_OPTS_CC", etc so that user provided libdir and incdir get through.

16 years agodefined user provided libdir and incdir to USER_OPTS_LD, etc so that it gets through...
Gengbin Zheng [Wed, 31 Aug 2005 16:38:35 +0000 (16:38 +0000)]
defined user provided libdir and incdir to USER_OPTS_LD, etc so that it gets through properly.

16 years agoparallel ref & coarsening should work now in tandem. I have added sliver/flip tests
Nilesh Choudhury [Tue, 30 Aug 2005 22:54:57 +0000 (22:54 +0000)]
parallel ref & coarsening should work now in tandem. I have added sliver/flip tests
for coarsening. There are still some issues with adjacencies, which we will resolve soon.

16 years agoadded operator new for (int, void*)
Filippo Gioachin [Tue, 30 Aug 2005 19:07:10 +0000 (19:07 +0000)]
added operator new for (int, void*)

16 years agoAdded ckpool.h to include directory
Amit Sharma [Tue, 30 Aug 2005 19:02:51 +0000 (19:02 +0000)]
Added ckpool.h to include directory

16 years agoFixes.
Yan Shi [Tue, 30 Aug 2005 16:48:54 +0000 (16:48 +0000)]
Fixes.

16 years agosupport new decomposition with NEW_DECOMP ifdef.
Eric Bohm [Tue, 30 Aug 2005 15:01:12 +0000 (15:01 +0000)]
support new decomposition with NEW_DECOMP ifdef.
use priority flag on forward path

16 years agocorrection in syntax
Filippo Gioachin [Sun, 28 Aug 2005 22:51:37 +0000 (22:51 +0000)]
correction in syntax

16 years agoadded basic class CkOStream to unify CkOutStream and CkErrStream when redefining
Filippo Gioachin [Sun, 28 Aug 2005 21:27:08 +0000 (21:27 +0000)]
added basic class CkOStream to unify CkOutStream and CkErrStream when redefining
the operator<<

16 years agoBug fixes for Poll Mode.
Isaac Dooley [Sat, 27 Aug 2005 05:20:16 +0000 (05:20 +0000)]
Bug fixes for Poll Mode.

16 years agochanged order on python locking
Filippo Gioachin [Sat, 27 Aug 2005 01:11:08 +0000 (01:11 +0000)]
changed order on python locking

16 years agoFixed reference to vector2d.h. This may not be the most elegant solution
Chee Wai Lee [Thu, 25 Aug 2005 20:15:51 +0000 (20:15 +0000)]
Fixed reference to vector2d.h. This may not be the most elegant solution
as it depends on the way build places the code.

16 years agoFixed capitalization of NodeGroup
Isaac Dooley [Thu, 25 Aug 2005 19:52:16 +0000 (19:52 +0000)]
Fixed capitalization of NodeGroup

16 years agoupdated test program, uses FEM_Mesh_smooth
Dmitriy Ofman [Thu, 25 Aug 2005 18:37:24 +0000 (18:37 +0000)]
updated test program, uses FEM_Mesh_smooth

16 years agolooks like my changes didnt go through last time. Remivong void from the function...
Sameer Kumar [Thu, 25 Aug 2005 18:15:13 +0000 (18:15 +0000)]
looks like my changes didnt go through last time. Remivong void from the function calls.

16 years agoSwitch from alloca to new for array sections do avoid lemieux
Eric Bohm [Thu, 25 Aug 2005 17:36:35 +0000 (17:36 +0000)]
Switch from alloca to new for array sections do avoid lemieux
compiler hostility

16 years agoupdate version of CkPoolQueue templated only with the class type
Filippo Gioachin [Thu, 25 Aug 2005 01:01:11 +0000 (01:01 +0000)]
update version of CkPoolQueue templated only with the class type

16 years agoadded FEM_Mesh_smooth function
Dmitriy Ofman [Wed, 24 Aug 2005 21:15:17 +0000 (21:15 +0000)]
added FEM_Mesh_smooth function

16 years agovector2d - 2d vectors of doubles
Dmitriy Ofman [Wed, 24 Aug 2005 21:13:55 +0000 (21:13 +0000)]
vector2d - 2d vectors of doubles
: ----------------------------------------------------------------------

16 years agoadded CkPool implementation
Filippo Gioachin [Tue, 23 Aug 2005 22:48:42 +0000 (22:48 +0000)]
added CkPool implementation

16 years agoallow --basedir, etc to take multiple directories with space
Gengbin Zheng [Tue, 23 Aug 2005 15:56:07 +0000 (15:56 +0000)]
allow --basedir, etc to take multiple directories with space

16 years agoparallel coarsening works, but there seems to be some issues with the new nodes added...
Nilesh Choudhury [Tue, 23 Aug 2005 01:10:52 +0000 (01:10 +0000)]
parallel coarsening works, but there seems to be some issues with the new nodes added, the attributes are not correctly updated while refinement.

16 years agoAdding Dima's quality measure.
Terry L. Wilmarth [Mon, 22 Aug 2005 23:06:55 +0000 (23:06 +0000)]
Adding Dima's quality measure.

16 years agoFixed remaining coarsening bugs. Hopefully BC bugs are fixed too. Will test that...
Terry L. Wilmarth [Mon, 22 Aug 2005 20:54:28 +0000 (20:54 +0000)]
Fixed remaining coarsening bugs.  Hopefully BC bugs are fixed too. Will test that next.

16 years agoadded back missing output.
Gengbin Zheng [Mon, 22 Aug 2005 19:56:17 +0000 (19:56 +0000)]
added back missing output.

16 years agofixed for uth version.
Gengbin Zheng [Mon, 22 Aug 2005 19:03:12 +0000 (19:03 +0000)]
fixed for uth version.

16 years agoMore of the main Refine algorithm is in place.
Terry L. Wilmarth [Mon, 22 Aug 2005 16:41:31 +0000 (16:41 +0000)]
More of the main Refine algorithm is in place.

16 years agofixed some memory bugs
Nilesh Choudhury [Mon, 22 Aug 2005 09:37:32 +0000 (09:37 +0000)]
fixed some memory bugs

16 years agofixed more memory leaks
Nilesh Choudhury [Mon, 22 Aug 2005 07:08:52 +0000 (07:08 +0000)]
fixed more memory leaks
CV: ----------------------------------------------------------------------

16 years agofixed memory leaks
Nilesh Choudhury [Mon, 22 Aug 2005 06:18:49 +0000 (06:18 +0000)]
fixed memory leaks
CV: ----------------------------------------------------------------------

16 years agofixed some bugs related to parallel coarsening. Now, I can run coarsening till 10...
Nilesh Choudhury [Sun, 21 Aug 2005 22:57:31 +0000 (22:57 +0000)]
fixed some bugs related to parallel coarsening. Now, I can run coarsening till 10 steps easily on the bar example, without any error.

16 years agoUpdated liveViz Poll section to match the new Poll mode code
Isaac Dooley [Sat, 20 Aug 2005 19:32:57 +0000 (19:32 +0000)]
Updated liveViz Poll section to match the new Poll mode code

16 years agoFixed boundary condition bug, but there are other problems with this code. Checking...
Terry L. Wilmarth [Sat, 20 Aug 2005 00:18:22 +0000 (00:18 +0000)]
Fixed boundary condition bug, but there are other problems with this code.  Checking in a broken version!

16 years agoFixed Line FFT. Correct for both forward and backward direction.
Yan Shi [Fri, 19 Aug 2005 20:46:06 +0000 (20:46 +0000)]
Fixed Line FFT. Correct for both forward and backward direction.

16 years agomodification to "operator new" for varsize messages. now it reflects the manual
Filippo Gioachin [Fri, 19 Aug 2005 18:12:12 +0000 (18:12 +0000)]
modification to "operator new" for varsize messages. now it reflects the manual