charm.git
10 years agoAdaptive LB
Harshitha [Fri, 18 Nov 2011 18:01:28 +0000 (12:01 -0600)]
Adaptive LB

10 years agofix CMK_CPP_CHARM
Gengbin Zheng [Fri, 11 Nov 2011 19:21:44 +0000 (13:21 -0600)]
fix CMK_CPP_CHARM

10 years agodefine CMK_CPP_CHARM
Gengbin Zheng [Fri, 11 Nov 2011 19:15:49 +0000 (13:15 -0600)]
define CMK_CPP_CHARM

10 years agodefine CMK_CPP_CHARM
Gengbin Zheng [Fri, 11 Nov 2011 19:13:55 +0000 (13:13 -0600)]
define CMK_CPP_CHARM

10 years agodefine CMK_CPP_CHARM
Gengbin Zheng [Fri, 11 Nov 2011 19:09:46 +0000 (13:09 -0600)]
define CMK_CPP_CHARM

10 years agobuild state space search library
Gengbin Zheng [Fri, 11 Nov 2011 05:17:50 +0000 (23:17 -0600)]
build state space search library

10 years agomove state space search engine library to ck-libs
Gengbin Zheng [Fri, 11 Nov 2011 05:02:01 +0000 (23:02 -0600)]
move state space search engine library to ck-libs

10 years agoMesh Streamer: Deleting msg after reduction is still causing a segfault,
Lukasz Wesolowski [Thu, 10 Nov 2011 23:47:31 +0000 (17:47 -0600)]
Mesh Streamer: Deleting msg after reduction is still causing a segfault,
so I am commenting the line out for now. Also, since caching the location
information leads to a slowdown on BG/P, I am commenting out the define
which makes it used by default.

10 years agoMerge branch 'charm' of charmgit:charm into charm
Gengbin Zheng [Thu, 10 Nov 2011 20:48:58 +0000 (12:48 -0800)]
Merge branch 'charm' of charmgit:charm into charm

10 years agoincrease malloc test to 8MB to overcome the case when it uses large pages.
Gengbin Zheng [Thu, 10 Nov 2011 20:48:25 +0000 (12:48 -0800)]
increase malloc test to 8MB to overcome the case when it uses large pages.

10 years agoMerge branch 'charm' of charmgit:charm into charm
Gengbin Zheng [Thu, 10 Nov 2011 19:56:13 +0000 (14:56 -0500)]
Merge branch 'charm' of charmgit:charm into charm

10 years agosupport for intel compiler on cray
Gengbin Zheng [Thu, 10 Nov 2011 19:55:54 +0000 (14:55 -0500)]
support for intel compiler on cray

10 years agominor change to the stealing scheme
Gengbin Zheng [Thu, 10 Nov 2011 18:31:11 +0000 (10:31 -0800)]
minor change to the stealing scheme

10 years agoMerge branch 'charm' of charmgit:charm into charm
Gengbin Zheng [Thu, 10 Nov 2011 17:15:21 +0000 (09:15 -0800)]
Merge branch 'charm' of charmgit:charm into charm

10 years agosupport for intel compilers
Gengbin Zheng [Thu, 10 Nov 2011 17:13:16 +0000 (09:13 -0800)]
support for intel compilers

10 years agoremove print information
Yanhua Sun [Thu, 10 Nov 2011 04:58:18 +0000 (20:58 -0800)]
remove print information

10 years agoMerge branch 'charm' of charmgit:charm into charm
Yanhua Sun [Thu, 10 Nov 2011 04:12:22 +0000 (20:12 -0800)]
Merge branch 'charm' of charmgit:charm into charm

10 years agoadded statistic information about mempool
Yanhua Sun [Thu, 10 Nov 2011 04:12:15 +0000 (20:12 -0800)]
added statistic information about mempool

10 years agochange the default split size to 8K
Akhil Langer [Thu, 10 Nov 2011 03:08:04 +0000 (03:08 +0000)]
change the default split size to 8K

10 years agoRemove junk file added
Phil Miller [Thu, 10 Nov 2011 02:57:51 +0000 (20:57 -0600)]
Remove junk file added

10 years agochange the default split size to 8K
Akhil Langer [Thu, 10 Nov 2011 02:23:14 +0000 (02:23 +0000)]
change the default split size to 8K

10 years agoMeshStreamer: bug fix in final reduction.
Lukasz Wesolowski [Wed, 9 Nov 2011 23:34:26 +0000 (17:34 -0600)]
MeshStreamer: bug fix in final reduction.

10 years agofixed the registration
Yanhua Sun [Wed, 9 Nov 2011 20:07:31 +0000 (12:07 -0800)]
fixed the registration

10 years agofixed 100M smp ppn5 by fragmentation
Yanhua Sun [Wed, 9 Nov 2011 06:52:01 +0000 (01:52 -0500)]
fixed 100M smp ppn5 by fragmentation

10 years agofragmentation for large msg
Yanhua Sun [Wed, 9 Nov 2011 04:11:40 +0000 (20:11 -0800)]
fragmentation for large msg

10 years agoafter steal, if it fails, also linear search all memblocks to free some un-used ones...
Gengbin Zheng [Sun, 6 Nov 2011 20:53:05 +0000 (12:53 -0800)]
after steal, if it fails, also linear search all memblocks to free some un-used ones, and deregister the memory.

10 years agosome minor code cleanup and comments
Gengbin Zheng [Sat, 5 Nov 2011 05:00:47 +0000 (22:00 -0700)]
some minor code cleanup and comments

10 years agowhen steal mempool succeed, should just return
Gengbin Zheng [Sat, 5 Nov 2011 03:57:08 +0000 (20:57 -0700)]
when steal mempool succeed, should just return

10 years agoMeshStreamer: added hashing of coordinate information and refactored the code
Lukasz Wesolowski [Sat, 5 Nov 2011 02:45:21 +0000 (21:45 -0500)]
MeshStreamer: added hashing of coordinate information and refactored the code
to use coordinate objects which encapsulate row, column, plane, msgType

10 years agodoes not have to clear the buffer to 0.
Gengbin Zheng [Sat, 5 Nov 2011 02:10:15 +0000 (19:10 -0700)]
does not have to clear the buffer to 0.

10 years agoMerge branch 'charm' of charmgit:charm into charm
Gengbin Zheng [Sat, 5 Nov 2011 02:05:55 +0000 (19:05 -0700)]
Merge branch 'charm' of charmgit:charm into charm

10 years agoadd a mempool stealing scheme. When running out of registered memory, for SMP version...
Gengbin Zheng [Sat, 5 Nov 2011 02:04:36 +0000 (19:04 -0700)]
add a mempool stealing scheme. When running out of registered memory, for SMP version, try to steal free memblocks from other rank on the same node.

10 years agoinitialize cmiArgDebugFlag to 0
Gengbin Zheng [Sat, 5 Nov 2011 01:59:28 +0000 (18:59 -0700)]
initialize cmiArgDebugFlag to 0

10 years agockmulticast: more docs for sendToSection
Ramprasad Venkataraman [Fri, 4 Nov 2011 22:26:54 +0000 (17:26 -0500)]
ckmulticast: more docs for sendToSection

10 years agockmcast - avoid message copying when only 1 fragment; patch provided by Gengbin
Akhil Langer [Fri, 4 Nov 2011 20:06:55 +0000 (15:06 -0500)]
ckmcast - avoid message copying when only 1 fragment; patch provided by Gengbin

10 years agoSMP uses context thread
Gengbin Zheng [Fri, 4 Nov 2011 15:33:44 +0000 (08:33 -0700)]
SMP uses context thread

10 years agofix a size bug in LrtsAlloc (should minus mempool header),
Gengbin Zheng [Fri, 4 Nov 2011 09:04:17 +0000 (02:04 -0700)]
fix a size bug in LrtsAlloc (should minus mempool header),
also fix a bug in expand mempool with size greater than default. The size should consider mempool_block size.

10 years agockmulticast: Fix recent checkin with default args gone awry
Ramprasad Venkataraman [Fri, 4 Nov 2011 01:54:54 +0000 (20:54 -0500)]
ckmulticast: Fix recent checkin with default args gone awry

10 years agoMerge branch 'charm' of charmgit:charm into charm
Akhil Langer [Fri, 4 Nov 2011 01:06:55 +0000 (20:06 -0500)]
Merge branch 'charm' of charmgit:charm into charm

10 years agoAlways do Pipeling in CkMulticast && make SPLIT_SIZE and SPLIT_THRESHOLD as run time...
Akhil Langer [Fri, 4 Nov 2011 01:06:42 +0000 (20:06 -0500)]
Always do Pipeling in CkMulticast && make SPLIT_SIZE and SPLIT_THRESHOLD as run time variables rather than compile time constants

10 years agoMeshStreamer: Added the option to let users specify when each group member
Lukasz Wesolowski [Thu, 3 Nov 2011 23:04:14 +0000 (18:04 -0500)]
MeshStreamer: Added the option to let users specify when each group member
is done inserting, instead of using quiescence. When using periodic flushing,
added an optimization to prevent unnecessary flushing if recent progress
has been made.

10 years agoMeshStreamer: reintroduced a periodic flush function to help with
Lukasz Wesolowski [Thu, 3 Nov 2011 05:02:06 +0000 (00:02 -0500)]
MeshStreamer: reintroduced a periodic flush function to help with
the use case of applications which generate additional messages
in response to streamed messages.

10 years agoMerge branch 'charm' of charmgit:charm into charm
Gengbin Zheng [Thu, 3 Nov 2011 06:04:53 +0000 (01:04 -0500)]
Merge branch 'charm' of charmgit:charm into charm

10 years agotest gfortran lib in another possible location.
Gengbin Zheng [Thu, 3 Nov 2011 06:04:03 +0000 (01:04 -0500)]
test gfortran lib in another possible location.

10 years agoMerge branch 'charm' of charmgit:charm into charm
Yanhua Sun [Wed, 2 Nov 2011 20:28:53 +0000 (13:28 -0700)]
Merge branch 'charm' of charmgit:charm into charm

10 years agoadded error information for memory registration
Yanhua Sun [Wed, 2 Nov 2011 20:28:35 +0000 (13:28 -0700)]
added error information for memory registration

10 years agoMerge branch 'charm' of charmgit:charm into charm
Gengbin Zheng [Wed, 2 Nov 2011 19:22:51 +0000 (14:22 -0500)]
Merge branch 'charm' of charmgit:charm into charm

10 years agomake receiveCombinedMessage a virtual function.
Gengbin Zheng [Wed, 2 Nov 2011 18:55:14 +0000 (13:55 -0500)]
make receiveCombinedMessage a virtual function.

10 years agoremoving stupid binary that I pushed
Harshitha Menon [Wed, 2 Nov 2011 18:49:28 +0000 (13:49 -0500)]
removing stupid binary that I pushed

10 years agoExample usage of cksequence
Harshitha Menon [Wed, 2 Nov 2011 18:47:08 +0000 (13:47 -0500)]
Example usage of cksequence

10 years agoa flag to tell if yield or not
Gengbin Zheng [Wed, 2 Nov 2011 00:10:16 +0000 (19:10 -0500)]
a flag to tell if yield or not

10 years agosignificantly simply interface
Gengbin Zheng [Tue, 1 Nov 2011 21:23:29 +0000 (16:23 -0500)]
significantly simply interface
The user only needs to provide a function call process(data), library takes care of all streaming stuff.

10 years agoMeshStreamer: function rename to improve readability and avoid confusion
Lukasz Wesolowski [Tue, 1 Nov 2011 20:18:02 +0000 (15:18 -0500)]
MeshStreamer: function rename to improve readability and avoid confusion

10 years agoMeshStreamer: Modified the "start" function name to the more descriptive "callWhenFin...
Lukasz Wesolowski [Tue, 1 Nov 2011 20:15:44 +0000 (15:15 -0500)]
MeshStreamer: Modified the "start" function name to the more descriptive "callWhenFinished"

10 years agoMeshStreamer: commented out prototype for a function which is no longer being used.
Lukasz Wesolowski [Tue, 1 Nov 2011 19:25:26 +0000 (14:25 -0500)]
MeshStreamer: commented out prototype for a function which is no longer being used.

10 years agorestrict the hack for pxshm only for cray xe and smp.
Gengbin Zheng [Tue, 1 Nov 2011 16:43:19 +0000 (11:43 -0500)]
restrict the hack for pxshm only for cray xe and smp.

10 years agoMerge branch 'charm' of charmgit:charm into charm
Gengbin Zheng [Tue, 1 Nov 2011 14:37:26 +0000 (09:37 -0500)]
Merge branch 'charm' of charmgit:charm into charm

10 years agofix for VC++
Gengbin Zheng [Tue, 1 Nov 2011 14:32:51 +0000 (09:32 -0500)]
fix for VC++

10 years agoguard ft code by ft macro
Gengbin Zheng [Tue, 1 Nov 2011 14:30:30 +0000 (09:30 -0500)]
guard ft code by ft macro

10 years agoadd a start(callback) function to omit the need for a second QD from user code to...
Gengbin Zheng [Tue, 1 Nov 2011 04:12:51 +0000 (23:12 -0500)]
add a start(callback) function to omit the need for a second QD from user code to flush the buffered message

10 years agosimilar fix for cray-xe mallinfo
Gengbin Zheng [Mon, 31 Oct 2011 17:20:53 +0000 (12:20 -0500)]
similar fix for cray-xe mallinfo

10 years agominor change about a print
Gengbin Zheng [Mon, 31 Oct 2011 17:19:18 +0000 (12:19 -0500)]
minor change about a print

10 years agoSDAG: Move CCounter's pointers stars into its type string, to match changes for templ...
Phil Miller [Mon, 31 Oct 2011 13:49:18 +0000 (08:49 -0500)]
SDAG: Move CCounter's pointers stars into its type string, to match changes for template support

Now tested against the full set of Charm++ examples and tests, rather than the
couple I happened to look at.

10 years agoSDAG: simplify entry method matching logic
Phil Miller [Mon, 31 Oct 2011 07:44:47 +0000 (02:44 -0500)]
SDAG: simplify entry method matching logic

10 years agoSDAG: Consolidate handling of parameter types to improve template support
Phil Miller [Mon, 31 Oct 2011 07:43:56 +0000 (02:43 -0500)]
SDAG: Consolidate handling of parameter types to improve template support

10 years agoxi: rearrange inclusion a bit
Phil Miller [Mon, 31 Oct 2011 07:32:17 +0000 (02:32 -0500)]
xi: rearrange inclusion a bit

10 years agodon't print pid for mpi version. still print for net, which make it easy to pick...
Gengbin Zheng [Mon, 31 Oct 2011 00:50:01 +0000 (19:50 -0500)]
don't print pid for mpi version. still print for net, which make it easy to pick up a process to kill

10 years agoMerge branch 'charm' of charmgit:charm into charm
Gengbin Zheng [Mon, 31 Oct 2011 00:23:31 +0000 (19:23 -0500)]
Merge branch 'charm' of charmgit:charm into charm

10 years agoimplement CkDieNow function for net version too
Gengbin Zheng [Mon, 31 Oct 2011 00:23:07 +0000 (19:23 -0500)]
implement CkDieNow function for net version too

10 years agoclear
Xiang Ni [Mon, 31 Oct 2011 00:22:42 +0000 (19:22 -0500)]
clear

10 years agomissing initialization of total_pes when +wp is not present
Gengbin Zheng [Sun, 30 Oct 2011 22:32:04 +0000 (17:32 -0500)]
missing initialization of total_pes when +wp is not present

10 years agofix a bug in generating bufferMssage call for a when trigger function with no argument.
Gengbin Zheng [Sun, 30 Oct 2011 15:25:15 +0000 (10:25 -0500)]
fix a bug in generating bufferMssage call for a when trigger function with no argument.

10 years agoFT: print out which spare processor is used to replace the failed PE.
Gengbin Zheng [Sun, 30 Oct 2011 02:24:53 +0000 (21:24 -0500)]
FT: print out which spare processor is used to replace the failed PE.

10 years agoavoid TimerInit global synchronization under bgp for FT
Gengbin Zheng [Sun, 30 Oct 2011 01:33:13 +0000 (01:33 +0000)]
avoid TimerInit global synchronization under bgp for FT

10 years agoclean up
Gengbin Zheng [Sun, 30 Oct 2011 01:28:50 +0000 (01:28 +0000)]
clean up

10 years agoupdated
Gengbin Zheng [Sat, 29 Oct 2011 23:50:37 +0000 (18:50 -0500)]
updated

10 years agodon't test -lmpich for bgp builds
Gengbin Zheng [Sat, 29 Oct 2011 23:48:41 +0000 (23:48 +0000)]
don't test -lmpich for bgp builds

10 years agominor cleanup
Gengbin Zheng [Sat, 29 Oct 2011 22:23:01 +0000 (17:23 -0500)]
minor cleanup

10 years agoa minor optimization to find reverse buddy
Gengbin Zheng [Sat, 29 Oct 2011 22:16:00 +0000 (17:16 -0500)]
a minor optimization to find reverse buddy

10 years agoguard new function with mem checkpoint macro
Gengbin Zheng [Sat, 29 Oct 2011 22:11:05 +0000 (17:11 -0500)]
guard new function with mem checkpoint macro

10 years agoFurther reducing the size for memoryAccess to make sure that it runs on every machine
Nikhil Jain [Sat, 29 Oct 2011 16:41:17 +0000 (11:41 -0500)]
Further reducing the size for memoryAccess to make sure that it runs on every machine

10 years agoMerge branch 'charm' of charmgit:charm into charm
Harshitha Menon [Sat, 29 Oct 2011 07:05:03 +0000 (02:05 -0500)]
Merge branch 'charm' of charmgit:charm into charm

10 years agocksequence implementation with memory compaction
Harshitha Menon [Sat, 29 Oct 2011 06:52:19 +0000 (01:52 -0500)]
cksequence implementation with memory compaction

10 years agofix a bug when multiple crashes. The spare processors need to be updated with the...
Gengbin Zheng [Sat, 29 Oct 2011 06:27:11 +0000 (01:27 -0500)]
fix a bug when multiple crashes. The spare processors need to be updated with the petorank array.
added a function CkHasCheckpoints(0 which returns true when there is already checkpoints.

10 years agofix issue at end of program, where MPI_Barrier needs to be called on all processors...
Gengbin Zheng [Sat, 29 Oct 2011 03:04:20 +0000 (22:04 -0500)]
fix issue at end of program, where MPI_Barrier needs to be called on all processors including spare and crashed processors and MPI_Finalize is called to properly exit the MPI program.

10 years agoMeshStreamer: Tried to add const wherever possible and inserted some
Lukasz Wesolowski [Fri, 28 Oct 2011 22:33:20 +0000 (17:33 -0500)]
MeshStreamer: Tried to add const wherever possible and inserted some
missing parameter names in function prototypes.

10 years agoMerge branch 'charm' of charmgit:charm into charm
Gengbin Zheng [Fri, 28 Oct 2011 22:05:28 +0000 (17:05 -0500)]
Merge branch 'charm' of charmgit:charm into charm

10 years agodemonstrate inmem checkpoint restart on MPI layer:
Gengbin Zheng [Fri, 28 Oct 2011 22:04:08 +0000 (17:04 -0500)]
demonstrate inmem checkpoint restart on MPI layer:
1. CkDieNow()  to fake die a processor
2. +wp <numpe> to specify the number of charm processors,  +p fires with spared processors
3. a spared processor will be activated to replace the crashed one.

10 years agoMeshStreamer: Removed capacity from MeshStreamerMessage to decrease
Lukasz Wesolowski [Fri, 28 Oct 2011 21:56:48 +0000 (16:56 -0500)]
MeshStreamer: Removed capacity from MeshStreamerMessage to decrease
message size. Capacity can be obtained directly from the library.
Moved assert which checks that no messages remain after flush.
The check now occurs after all three flush phases are finished.

10 years agofix a bug in ckmulticast pipelining
Akhil Langer [Fri, 28 Oct 2011 19:09:50 +0000 (14:09 -0500)]
fix a bug in ckmulticast pipelining

10 years agoMerge branch 'charm' of charmgit:charm into charm
Akhil Langer [Fri, 28 Oct 2011 18:17:04 +0000 (13:17 -0500)]
Merge branch 'charm' of charmgit:charm into charm

10 years agockmulticast.C - SPLIT_NUM is redundant, set the FRAGMENT_SIZE for pipelining to be...
Akhil Langer [Fri, 28 Oct 2011 18:16:50 +0000 (13:16 -0500)]
ckmulticast.C - SPLIT_NUM is redundant, set the FRAGMENT_SIZE for pipelining to be 32k and increase the MAX_FRAGS to be 100 to allow large messages

10 years agoMerge branch 'charm' of charmgit:charm into charm
Gengbin Zheng [Fri, 28 Oct 2011 03:54:43 +0000 (20:54 -0700)]
Merge branch 'charm' of charmgit:charm into charm

10 years agotweak pxshm to work for SMP build
Gengbin Zheng [Fri, 28 Oct 2011 03:54:18 +0000 (20:54 -0700)]
tweak pxshm to work for SMP build

10 years agoMempool: Declare variables before other statements, to appease MSVC
Phil Miller [Fri, 28 Oct 2011 01:55:36 +0000 (20:55 -0500)]
Mempool: Declare variables before other statements, to appease MSVC

10 years agoreducing the test count
Nikhil Jain [Thu, 27 Oct 2011 17:15:29 +0000 (12:15 -0500)]
reducing the test count

10 years agoMerge branch 'charm' of charmgit:charm into charm
Yanhua Sun [Thu, 27 Oct 2011 06:23:47 +0000 (23:23 -0700)]
Merge branch 'charm' of charmgit:charm into charm

10 years agomake mempool thread-safe in smp; fix smp performance problem in gemini
Yanhua Sun [Thu, 27 Oct 2011 06:23:27 +0000 (23:23 -0700)]
make mempool thread-safe in smp; fix smp performance problem in gemini

10 years agoMerge branch 'charm' of charmgit:charm into charm
Gengbin Zheng [Thu, 27 Oct 2011 05:03:00 +0000 (00:03 -0500)]
Merge branch 'charm' of charmgit:charm into charm

10 years agofix CmiGetRestartPhase definition
Gengbin Zheng [Thu, 27 Oct 2011 05:02:44 +0000 (00:02 -0500)]
fix CmiGetRestartPhase definition