charm.git
15 years agoAdded a new gcc4 configuration for the XT3 machine option only.
Chee Wai Lee [Tue, 1 Aug 2006 20:07:43 +0000 (20:07 +0000)]
Added a new gcc4 configuration for the XT3 machine option only.

The XT3 pcqueue hack is apparently no longer required. calloc fails to do
the right job on the XT3 and is currently changed to malloc/memset on the
platform. All other platforms will continue to use calloc.

A decision to use malloc/memset is prefered over the original hack because
it is unclear how the original even helped in the first place, raising
concerns that some bug remains unexercised on the system. malloc/memset
on the other hand behaves semantically the same as calloc though performance
is expected to suffer due to the result of making 2 system calls as opposed
to 1.

15 years agocomments
Gengbin Zheng [Tue, 1 Aug 2006 18:43:43 +0000 (18:43 +0000)]
comments

15 years agoadded file for globally renumbering nodes
Sayantan Chakravorty [Wed, 26 Jul 2006 21:19:53 +0000 (21:19 +0000)]
added file for globally renumbering nodes

15 years agosmarter constructor
Eric Bohm [Wed, 26 Jul 2006 15:11:42 +0000 (15:11 +0000)]
smarter constructor

15 years agoadded function to destroy all allocated data
Filippo Gioachin [Fri, 21 Jul 2006 23:06:55 +0000 (23:06 +0000)]
added function to destroy all allocated data

15 years agoThese modifications allow for compatibility with the VMI 2.2 patch for
Greg Koenig [Mon, 17 Jul 2006 21:02:17 +0000 (21:02 +0000)]
These modifications allow for compatibility with the VMI 2.2 patch for
VMI_RDMA_Publish_Buffer_With_Callback() which allows better overlapping of
computation with communication, especially for high-latency environments
like Grids.

15 years agoAdded alignment attribute to the dmalist stack variable.
David Kunzman [Mon, 17 Jul 2006 19:41:36 +0000 (19:41 +0000)]
Added alignment attribute to the dmalist stack variable.

15 years ago*** empty log message ***
Chao Huang [Fri, 14 Jul 2006 17:11:29 +0000 (17:11 +0000)]
*** empty log message ***

15 years ago*** empty log message ***
Chao Huang [Fri, 14 Jul 2006 06:37:37 +0000 (06:37 +0000)]
*** empty log message ***

15 years ago*** empty log message ***
Chao Huang [Fri, 14 Jul 2006 06:37:37 +0000 (06:37 +0000)]
*** empty log message ***

15 years agoremoved two unnecesary includes to get it compiled with gcc 4
Gengbin Zheng [Thu, 13 Jul 2006 20:24:44 +0000 (20:24 +0000)]
removed two unnecesary includes to get it compiled with gcc 4

15 years agoChanged some characters in the comments to get rid of a "multi-line comment" warning.
David Kunzman [Thu, 13 Jul 2006 15:56:20 +0000 (15:56 +0000)]
Changed some characters in the comments to get rid of a "multi-line comment" warning.

15 years ago*** empty log message ***
Sameer Kumar [Wed, 12 Jul 2006 23:07:48 +0000 (23:07 +0000)]
*** empty log message ***

15 years ago*** empty log message ***
Sameer Kumar [Wed, 12 Jul 2006 23:07:48 +0000 (23:07 +0000)]
*** empty log message ***

15 years agofixed lib name
Gengbin Zheng [Tue, 11 Jul 2006 19:00:45 +0000 (19:00 +0000)]
fixed lib name

15 years agofixed file name.
Gengbin Zheng [Tue, 11 Jul 2006 18:58:50 +0000 (18:58 +0000)]
fixed file name.

15 years agolanguage bluegene => bigsim
Gengbin Zheng [Tue, 11 Jul 2006 16:04:29 +0000 (16:04 +0000)]
language bluegene => bigsim

15 years agoblue gene => bigsim
Gengbin Zheng [Tue, 11 Jul 2006 06:29:54 +0000 (06:29 +0000)]
blue gene => bigsim

15 years agoBlue Gene => bigsim
Gengbin Zheng [Tue, 11 Jul 2006 05:35:41 +0000 (05:35 +0000)]
Blue Gene => bigsim

15 years agobluegene => bigsim
Gengbin Zheng [Tue, 11 Jul 2006 04:47:34 +0000 (04:47 +0000)]
bluegene => bigsim

15 years agomore bluegene => bigsim
Gengbin Zheng [Tue, 11 Jul 2006 04:36:24 +0000 (04:36 +0000)]
more bluegene => bigsim

15 years agofile name changes from bluegene to bigsim
Gengbin Zheng [Tue, 11 Jul 2006 04:26:04 +0000 (04:26 +0000)]
file name changes from bluegene to bigsim

15 years agobluegene => bigsim
Gengbin Zheng [Tue, 11 Jul 2006 04:22:12 +0000 (04:22 +0000)]
bluegene => bigsim

15 years agochanged blue_logs.h into bigsim_logs.h
Sayantan Chakravorty [Mon, 10 Jul 2006 22:58:22 +0000 (22:58 +0000)]
changed blue_logs.h into bigsim_logs.h

15 years agoMoved special.sh again (so charmc can execute within it).
David Kunzman [Mon, 10 Jul 2006 22:45:51 +0000 (22:45 +0000)]
Moved special.sh again (so charmc can execute within it).

15 years agonow includes bigsim_logs.h instead of blue_logs.h
Sayantan Chakravorty [Mon, 10 Jul 2006 22:31:38 +0000 (22:31 +0000)]
now includes bigsim_logs.h instead of blue_logs.h

15 years agoMoved the execution of special.sh to a later time in the build process. The net...
David Kunzman [Mon, 10 Jul 2006 22:08:00 +0000 (22:08 +0000)]
Moved the execution of special.sh to a later time in the build process.  The net-linux-cell
special.sh file relies on "[charmDir]/[buildDir]/include" existing so special.sh is now
being executed after it is created.

15 years agoModified define for reserved stack size (increased for NAMD). Added defines for...
David Kunzman [Sat, 8 Jul 2006 05:38:52 +0000 (05:38 +0000)]
Modified define for reserved stack size (increased for NAMD).  Added defines for user defined
INIT and CLOSE functions (called through funcLookup).  Added defines for DMA tag values usable
by user code.

15 years agoAdded branch hints. Added calls to user defined INIT and CLOSE functions through...
David Kunzman [Sat, 8 Jul 2006 05:35:41 +0000 (05:35 +0000)]
Added branch hints.  Added calls to user defined INIT and CLOSE functions through funcLookup().

15 years agoModified ROUNDUP_16 macro.
David Kunzman [Sat, 8 Jul 2006 05:22:54 +0000 (05:22 +0000)]
Modified ROUNDUP_16 macro.

15 years agochanges to smp to improve performance of smp.
Sayantan Chakravorty [Fri, 7 Jul 2006 23:17:24 +0000 (23:17 +0000)]
changes to smp to improve performance of smp.
computation threads do not sleep anymore by default.
this is fine if the communication thread has a processor to itself.
If it doesnt, computation threads can be made to sleep by turning on a
 flag called +CmiNoProcForComThread

15 years agofixed a bug that could cause nan in total walltime
Gengbin Zheng [Fri, 7 Jul 2006 20:24:07 +0000 (20:24 +0000)]
fixed a bug that could cause nan in total walltime

15 years agoadded more debugging
Gengbin Zheng [Fri, 7 Jul 2006 16:27:58 +0000 (16:27 +0000)]
added more debugging

15 years agoAdded Tecplot output for tet meshes.
Terry L. Wilmarth [Fri, 7 Jul 2006 06:03:32 +0000 (06:03 +0000)]
Added Tecplot output for tet meshes.

15 years agoadd canonical blockmap with procnum example and proper idx.data()
Eric Bohm [Mon, 3 Jul 2006 23:57:47 +0000 (23:57 +0000)]
add canonical blockmap with procnum example and proper idx.data()
to manual

15 years agoModified charmgrid.py to set VMI_MMAP_MAX=0 when launching a job. This
Greg Koenig [Mon, 26 Jun 2006 18:36:55 +0000 (18:36 +0000)]
Modified charmgrid.py to set VMI_MMAP_MAX=0 when launching a job.  This
activates a patch I made to VMI 2.2 that disables using mmap() for large
memory allocations (sbrk is used instead).  Calling mmap() means that the
only available option to the runtime system is to register/deregister each
memory allocation.  This causes large buffer sends to be 10-20x slower on
vmi-linux than on net-linux.  Fixing this causes our performance to be on
the same scale.

15 years agouse GreedyLB instead of MetisLB
Gengbin Zheng [Fri, 23 Jun 2006 18:53:36 +0000 (18:53 +0000)]
use GreedyLB instead of MetisLB

15 years agoshrink data starting at 8K
Gengbin Zheng [Fri, 23 Jun 2006 17:37:13 +0000 (17:37 +0000)]
shrink data starting at 8K

15 years agochanged // to /* */
Gengbin Zheng [Wed, 21 Jun 2006 07:06:55 +0000 (07:06 +0000)]
changed // to /* */

15 years agofixed compiler error on Cray Xt3
Gengbin Zheng [Wed, 21 Jun 2006 07:06:24 +0000 (07:06 +0000)]
fixed compiler error on Cray Xt3

15 years agofixed some minor compiler warnings.
Gengbin Zheng [Wed, 21 Jun 2006 06:45:59 +0000 (06:45 +0000)]
fixed some minor compiler warnings.

15 years agofixed a weird bug on cray XT3 of the copy function of CkArrayIndex to CkArrayIndexMax.
Gengbin Zheng [Wed, 21 Jun 2006 06:43:15 +0000 (06:43 +0000)]
fixed a weird bug on cray XT3 of the copy function of CkArrayIndex to CkArrayIndexMax.

15 years agoRDMA operations with callback
Nilesh Choudhury [Fri, 16 Jun 2006 15:51:08 +0000 (15:51 +0000)]
RDMA operations with callback

15 years agoifixed for win32 build.
Gengbin Zheng [Thu, 15 Jun 2006 03:48:25 +0000 (03:48 +0000)]
ifixed for win32 build.

15 years agowas using malloc & cmialloc interchangably which was bad for some architectures,
Nilesh Choudhury [Wed, 14 Jun 2006 20:03:30 +0000 (20:03 +0000)]
was using malloc & cmialloc interchangably which was bad for some architectures,
this was also the cause of the cmifrees causing double frees on turing.

15 years agohad to remove a couple of CmiFree in put_callback and get_callback because
Nilesh Choudhury [Tue, 13 Jun 2006 18:02:17 +0000 (18:02 +0000)]
had to remove a couple of CmiFree in put_callback and get_callback because
on turing this was becoming a double free... don't understand where the other
free is??

15 years agoadded test code for the one sided comm operations with callback
Nilesh Choudhury [Tue, 13 Jun 2006 17:21:58 +0000 (17:21 +0000)]
added test code for the one sided comm operations with callback

15 years agoadded two new functions CmiPutCb and CmiGetCb which are the versions
Nilesh Choudhury [Tue, 13 Jun 2006 17:21:18 +0000 (17:21 +0000)]
added two new functions CmiPutCb and CmiGetCb which are the versions
of the one sided RDMA operations with callbacks and do not return a
handle, so the user is notified of completion not by polling but
by the callback.

15 years ago*** empty log message ***
Nilesh Choudhury [Sun, 11 Jun 2006 22:42:57 +0000 (22:42 +0000)]
*** empty log message ***

15 years ago*** empty log message ***
Nilesh Choudhury [Sun, 11 Jun 2006 22:42:16 +0000 (22:42 +0000)]
*** empty log message ***

15 years ago*** empty log message ***
Nilesh Choudhury [Sun, 11 Jun 2006 04:09:44 +0000 (04:09 +0000)]
*** empty log message ***

15 years agoFixed VN mode bug. The zsize is doubled by VN mode, but only nzsize is
Eric Bohm [Sat, 10 Jun 2006 02:38:10 +0000 (02:38 +0000)]
Fixed VN mode bug.  The zsize is doubled by VN mode, but only nzsize is
checked in absz().  So now we halve the input if VN, which gives us a
more realistic hopcount in VN mode.  It also doesn't abort in corner cases.
Which is why I found the bug in the first place.

15 years ago*** empty log message ***
Nilesh Choudhury [Fri, 9 Jun 2006 15:37:44 +0000 (15:37 +0000)]
*** empty log message ***

15 years agoadded a new section on converse one sided communication interface.
Nilesh Choudhury [Fri, 9 Jun 2006 15:28:11 +0000 (15:28 +0000)]
added a new section on converse one sided communication interface.

15 years agolink -lz only when linking projections
Gengbin Zheng [Fri, 9 Jun 2006 04:32:12 +0000 (04:32 +0000)]
link -lz only when linking projections

15 years agouse assembly version CMK_QT="i386-gcc"
Gengbin Zheng [Fri, 9 Jun 2006 04:31:25 +0000 (04:31 +0000)]
use assembly version CMK_QT="i386-gcc"

15 years ago*** empty log message ***
Nilesh Choudhury [Thu, 8 Jun 2006 18:45:37 +0000 (18:45 +0000)]
*** empty log message ***

15 years agochange order of variable declaration to match constructor initialization
Eric Bohm [Thu, 8 Jun 2006 01:03:42 +0000 (01:03 +0000)]
change order of variable declaration to match constructor initialization

15 years agoExample program to test the new shadow array class and the new idxl interface
Nilesh Choudhury [Thu, 8 Jun 2006 01:01:14 +0000 (01:01 +0000)]
Example program to test the new shadow array class and the new idxl interface
functions, could be extended to test the adaptivity functions we write later

15 years agothere are two important changes, one is the addition of the new shadow array
Nilesh Choudhury [Thu, 8 Jun 2006 00:59:14 +0000 (00:59 +0000)]
there are two important changes, one is the addition of the new shadow array
and the other is that we have now made a parfum module and the other modules
Parfum_adapt and Parfum_SA are included in this, earlier we had a mess, fem, parfum...
People should do a make clean; make

15 years agoThis introduces a parfum shadow array and the idea is to make all future
Nilesh Choudhury [Thu, 8 Jun 2006 00:57:15 +0000 (00:57 +0000)]
This introduces a parfum shadow array and the idea is to make all future
modifications for communication using this shadow array. This shodow array
currently provides the interface to idxl functions that we discussed over email
recently

15 years agoadded a lock to idxl_list and a bunch of interface functions to
Nilesh Choudhury [Thu, 8 Jun 2006 00:54:45 +0000 (00:54 +0000)]
added a lock to idxl_list and a bunch of interface functions to
support locking and unlocking idxlLists and also adding new entries
and removing them and looking them up in various ways.

15 years agofixed a return value in operator= for net-sol-cc
Gengbin Zheng [Wed, 7 Jun 2006 14:23:50 +0000 (14:23 +0000)]
fixed a return value in operator= for net-sol-cc

15 years agoassignment operator should return value
Sayantan Chakravorty [Tue, 6 Jun 2006 00:40:27 +0000 (00:40 +0000)]
assignment operator should return value

15 years agoFixed an initialization bug in netfem that seems to show up only in
Sayantan Chakravorty [Tue, 6 Jun 2006 00:03:51 +0000 (00:03 +0000)]
Fixed an initialization bug in netfem that seems to show up only in
the netfem mode. netfem_state was CpvInitialized and the code
expected it to be set to NULL. However it did not work that way.
I had to add a CpvAccess(netfem_state) to set it to NULL perforce.

15 years agoconfiguration files to switch on gm one sided communication
Nilesh Choudhury [Sat, 3 Jun 2006 18:35:36 +0000 (18:35 +0000)]
configuration files to switch on gm one sided communication

15 years agoexample code for one sided communication
Nilesh Choudhury [Fri, 2 Jun 2006 22:13:44 +0000 (22:13 +0000)]
example code for one sided communication

15 years agoone sided communication for GM added (where we use gm_directed_send as put and the get
Nilesh Choudhury [Fri, 2 Jun 2006 22:02:45 +0000 (22:02 +0000)]
one sided communication for GM added (where we use gm_directed_send as put and the get
is emulated by using put from the destination machine).

15 years agoadded a lot of documentation
Nilesh Choudhury [Thu, 1 Jun 2006 02:19:41 +0000 (02:19 +0000)]
added a lot of documentation

15 years agoExplicit cast to enable comipiling with xlc.
Sameer Kumar [Thu, 1 Jun 2006 00:48:46 +0000 (00:48 +0000)]
Explicit cast to enable comipiling with xlc.

15 years ago*** empty log message ***
Sameer Kumar [Thu, 1 Jun 2006 00:47:23 +0000 (00:47 +0000)]
*** empty log message ***

15 years agoModified so it can be compiled as C instead of C++ (other minor changes).
David Kunzman [Wed, 31 May 2006 13:45:02 +0000 (13:45 +0000)]
Modified so it can be compiled as C instead of C++ (other minor changes).

15 years agoUpdated funcLookup() function prototype.
David Kunzman [Wed, 31 May 2006 13:43:24 +0000 (13:43 +0000)]
Updated funcLookup() function prototype.

15 years agoAdded code to display memory needed by each work request.
David Kunzman [Wed, 31 May 2006 13:42:05 +0000 (13:42 +0000)]
Added code to display memory needed by each work request.

15 years agoRemoved jacobi_shared.cpp from being compiled into PPE code (because of sim_printf...
David Kunzman [Wed, 31 May 2006 13:41:21 +0000 (13:41 +0000)]
Removed jacobi_shared.cpp from being compiled into PPE code (because of sim_printf).  Modified SPE compile rule so SPE code is compiled as C instead of C++ (left old commands in, for now, as reference).

15 years agoAdded some comments.
David Kunzman [Wed, 31 May 2006 13:36:53 +0000 (13:36 +0000)]
Added some comments.

15 years agoAdded "extern C" for includes. Changed printf's to sim_printf's. Modified code...
David Kunzman [Wed, 31 May 2006 13:36:19 +0000 (13:36 +0000)]
Added "extern C" for includes.  Changed printf's to sim_printf's.  Modified code so it could be compiled as C instead of C++.

15 years agoUpdated funcLookup() function prototype.
David Kunzman [Wed, 31 May 2006 13:34:53 +0000 (13:34 +0000)]
Updated funcLookup() function prototype.

15 years agoModified SPE compile rule so the SPE code is compiled as C instead of C++.
David Kunzman [Wed, 31 May 2006 13:34:12 +0000 (13:34 +0000)]
Modified SPE compile rule so the SPE code is compiled as C instead of C++.

15 years agoAdded code to get error code from completed work request. Started adding code for...
David Kunzman [Wed, 31 May 2006 13:31:04 +0000 (13:31 +0000)]
Added code to get error code from completed work request.  Started adding code for calculating the overall amount of memory needed by a work request before the work request is passed to the SPE Runtime.

15 years agoAdded several defines to control and get feedback from SPE Runtime. Also, added...
David Kunzman [Wed, 31 May 2006 13:29:48 +0000 (13:29 +0000)]
Added several defines to control and get feedback from SPE Runtime.  Also, added error state and error codes for work requests to be returned to the PPE.

15 years agoMany changes. Diff the file versions to see.
David Kunzman [Wed, 31 May 2006 13:28:23 +0000 (13:28 +0000)]
Many changes.  Diff the file versions to see.

15 years agoRemoved alloca_aligned() because the SPE Runtime is being linked against a different...
David Kunzman [Wed, 31 May 2006 13:26:44 +0000 (13:26 +0000)]
Removed alloca_aligned() because the SPE Runtime is being linked against a different libc.

15 years agoAdded "extern C". Commented out a function (because the SPE Runtime is now being...
David Kunzman [Wed, 31 May 2006 13:25:07 +0000 (13:25 +0000)]
Added "extern C".  Commented out a function (because the SPE Runtime is now being linked against a different libc).

15 years agoModified the include directory path. Made a change to force the SPE code to be compi...
David Kunzman [Wed, 31 May 2006 13:23:34 +0000 (13:23 +0000)]
Modified the include directory path.  Made a change to force the SPE code to be compiled as C instead of C++ (at least for now).

15 years agoadded a copy constructor to tupledata with const tupledata as
Sayantan Chakravorty [Tue, 30 May 2006 19:20:45 +0000 (19:20 +0000)]
added a copy constructor to tupledata with const tupledata as
an argument

15 years agoChanged stuff so that parallel partitoner can now be used
Sayantan Chakravorty [Tue, 30 May 2006 07:06:42 +0000 (07:06 +0000)]
Changed stuff so that parallel partitoner can now be used
without recompiling stuff.
The parallel partitioner can be invoked by using the
+Parfum_parallel_partition flag

15 years agoone sided communication initialization
Nilesh Choudhury [Mon, 29 May 2006 20:47:46 +0000 (20:47 +0000)]
one sided communication initialization

15 years agoadded defines to switch on one sided communication
Nilesh Choudhury [Mon, 29 May 2006 20:45:54 +0000 (20:45 +0000)]
added defines to switch on one sided communication

15 years agosupport for one-sided communication in converse
Nilesh Choudhury [Mon, 29 May 2006 20:44:41 +0000 (20:44 +0000)]
support for one-sided communication in converse

15 years agoone sided communication API
Nilesh Choudhury [Mon, 29 May 2006 20:43:07 +0000 (20:43 +0000)]
one sided communication API

15 years agofixed a bug in isomallocfree which involves freeing a remote slot. no converse handle...
Gengbin Zheng [Sun, 28 May 2006 16:50:26 +0000 (16:50 +0000)]
fixed a bug in isomallocfree which involves freeing a remote slot. no converse handler was set.

15 years agoadded threaded array
Gengbin Zheng [Fri, 26 May 2006 17:00:40 +0000 (17:00 +0000)]
added threaded array

15 years agochare reuse msg also has payload
Gengbin Zheng [Wed, 24 May 2006 19:59:12 +0000 (19:59 +0000)]
chare reuse msg also has payload

15 years agomake payload a parameter, group message now has payload.
Gengbin Zheng [Wed, 24 May 2006 19:41:55 +0000 (19:41 +0000)]
make payload a parameter, group message now has payload.

15 years agoplace holders
Gengbin Zheng [Tue, 23 May 2006 03:57:38 +0000 (03:57 +0000)]
place holders

15 years agoBackward compatability typedef ckcomplex complex.
Eric Bohm [Mon, 22 May 2006 22:17:41 +0000 (22:17 +0000)]
Backward compatability typedef ckcomplex complex.
Allows old codes which used ckcomplex (like leanCP) to just work.
Define CKCOMPLEX_ISNOT_COMPLEX to turn off the typedef and use
something else as the definition of complex.

Otherwise compilation for leanCP (and presumably anything else which
relied on ckcomplex) will fail horribly.

15 years agochange timer to mpi timer
Gengbin Zheng [Sat, 20 May 2006 02:56:17 +0000 (02:56 +0000)]
change timer to mpi timer

15 years agominor changes
Gengbin Zheng [Fri, 19 May 2006 19:19:20 +0000 (19:19 +0000)]
minor changes

15 years agofixed some common compiler warnings reported by bgl/xlc
Gengbin Zheng [Fri, 19 May 2006 19:18:17 +0000 (19:18 +0000)]
fixed some common compiler warnings reported by bgl/xlc