Project

General

Profile

Activity

From 08/25/2016 to 09/23/2016

09/24/2016

11:05 PM Revision 7b0a2de5 (charm): Documentation: clean up tuple reduction code example
Change-Id: I6450abf2400b1cfe37cf80077d3b02233e91f617 Sam White

09/23/2016

04:43 PM Revision b23d51b2 (charm): AMPI: separate AMPI error checking from Charm++ error checking
Change-Id: I59a37302ad1ade6abf47a35391f08c52c8c7a8b7 Sam White
04:40 PM Revision 0bcdc417 (charm): AMPI #1181: fix outcount handling in MPI_{Test,Wait}some routines
Change-Id: I78ac650f2c1b90082c34a65c68684e4be445c66f Nikhil Jain
04:21 PM Revision d8338da2 (charm): AMPI: update status with comm upon non-blocking send completion
Change-Id: I18c1a79f68a993b21ed3adcb40cd491b8e8fe29f Sam White
02:22 PM Bug #1197: Charmrun fails to connect to nodes on Taub in netlrts build.
Eric Bohm wrote:
> Is this still a problem?
If charm is built with gcc 4.9.2, doing the following fixes the issue...
Kavitha Chandrasekar
11:07 AM Feature #69: Get large messages sent via the ZC API to be received directly using a GET from the ...
Nitin, Vipul: Your recent work on RDMA support probably makes this and #68 easy to implement as well. Phil Miller
11:02 AM Feature #67 (Merged): Reduce the impact of large messages sent to unlocated or migrated chare arr...
This was merged with the 64-bit ID work. Phil Miller
08:26 AM Bug #1221 (Merged): Win build failures
Since merging 64bit IDs the *-win autobuild tests are failing with a n assertion from ckobjid.h:... Sam White
08:16 AM Feature #1213 (Merged): Use Charm++ built-in reducers for more predefined types/ops in AMPI
Sam White
08:16 AM Documentation #1140 (Merged): Document tuple/stats reducers
Sam White
08:16 AM Feature #557 (Merged): Built-in reducer types for logical and bitvec and/or/xor on bools and ints
Sam White
12:08 AM Revision 1a722a04 (charm): AMPI: only abort from ampiErrhandler if error checking is enabled
Change-Id: Ie0cde0df5a7606ec417269c2e6adcb9b5a8e7013 Sam White
12:07 AM Revision 746e8d10 (charm): Cleanup #1151: eliminate ROMIO on AMPI compiler warnings
Change-Id: I742220debbe64ff6ff655a409b7183fe0f5e19bd Sam White
10:58 PM Revision 0d953aa3 (charm): AMPI #1213: use built-in Charm++ reducers when possible for MPI_C_BOOL,
MPI_LOGICAL, MPI_BYTE, and MPI_(U)INT{8,16,32,64}_T datatypes.
Change-Id: If8e671343884e71916383f6ef39943dfb7e9b3eb
Sam White

09/22/2016

06:01 PM Revision 4159d512 (charm): Feature #557: expand the set of logical and bitvec reducers.
Add {logical,bitvec}_{and,or,xor}_{bool,int} reducer types, update the
documentation, and update existing uses of the...
Sam White
05:34 PM Bug #1215 (Merged): 64bit ID failures
Phil Miller
05:34 PM Feature #175 (Merged): objid_t: location cache should be keyed by objid_t
Phil Miller
05:34 PM Feature #174 (Merged): objid_t: Distributed lookup protocol
Phil Miller
05:34 PM Feature #170 (Merged): objid_t: Basic definitions and use in envelope
Phil Miller
05:32 PM Revision c3e6a809 (charm): Documentation: add tuple reducer info/example
Change-Id: I6e55d81c038cbfc07b8d24ad410f30b379dbeb24 Sam White
05:17 PM Revision 22f8019c (charm): New lockless queue implementation and integration,
Adds fallback for compilers that don't implement atomics.
Enabled for net and gni. Fixed CMIQueue / PCQueue operation...
Justin Miron
04:17 PM Bug #1220: AMPI: Support tlsglobals with dynamically linked objects
Notes from a conversation with a friend who works heavily on TLS implementation issues:

> Me
> So, I'm going to ...
Phil Miller
04:11 PM Bug #1220 (Merged): AMPI: Support tlsglobals with dynamically linked objects
Currently, the TLS-based privatization of global variables won't work in a program or library compiled as a dynamic o... Phil Miller
01:50 PM Documentation #1219 (Merged): Update SDAG forall documentation
SDAG's forall statement is barely explained in the manual. It should be clear that forall(..) { serial{ ... } } does ... Sam White
12:47 PM Feature #1218 (Merged): Move production features out of "experimental" part of Charm++ Manual
CkLoop, MPI interop, and Partitioning should all be moved out of the Experimental version to appropriate sections of ... Sam White

09/21/2016

02:07 PM Bug #1216 (Merged): Abort on message sends to out-of-bounds chare array elements
If a message is sent to a chare array element that is out-of-bounds for a statically sized chare array, we should abo... Sam White
02:02 PM Bug #1197: Charmrun fails to connect to nodes on Taub in netlrts build.
Is this still a problem? Eric Bohm
01:59 PM Bug #1200 (Merged): Openatom mixes CkMulticast managers and uses different instances to contribut...
Eric Bohm
12:17 PM Bug #1200 (In Progress): Openatom mixes CkMulticast managers and uses different instances to cont...
Eric Bohm
01:45 PM Bug #53: SDAG case constructs and Bigsim traces
Its been a year since the last update. If this is important someone should start work. If it isn't, the priority sh... Eric Bohm
01:42 PM Bug #829: CkLoop projections tracing doesn't reflect the origin/traceback of work
Has a consistent approach been implemented regarding projections tracing of our node level parallel schemes? Eric Bohm
01:40 PM Bug #828: Chare array construction semantics differ wrt readonly array proxy on PE 0
Is it worth taking another look at this after the recent infrastructure changes? Eric Bohm
01:37 PM Bug #756: CUDA build does not correctly find cuda location
is this still a problem? Eric Bohm
01:36 PM Bug #815: Makefile for hybrid API is not using the system OPTS
Is this still pertinent? Eric Bohm
01:36 PM Bug #684: Isomalloc PUP has invalid read memory errors
Probably. Need to re-investigate. Phil Miller
01:31 PM Bug #684: Isomalloc PUP has invalid read memory errors
Old bug, still open? Eric Bohm
01:35 PM Bug #807 (Rejected): redo performance comparison
not worth pursuing extensive analysis on this platform as pami efforts are now oriented to the newer versions of pami... Eric Bohm
01:33 PM Bug #707 (Closed): Net charmrun should round-robin simultaneous connections
scalable start is the default now Eric Bohm
01:32 PM Projections Bug #699: message sends from periodic calls attributed to previous entry
Any update on this? Eric Bohm
01:30 PM Projections Bug #414: Idle/Overhead Display Disagreement
Progress? Eric Bohm
01:26 PM Projections Bug #415: Performance
Break this up into subtasks with this as a parent task so that achievable and measurable units of progress can be mad... Eric Bohm
11:54 AM Revision 7e532fa3 (charm): Outputting features from MetaLB for classifying load balancing strategy
Kavitha Chandrasekar
10:49 AM Bug #1215 (Implemented): 64bit ID failures
Here's a fix for all the above issues, except for I don't know what is leading to the last runtime failure on net-win... Sam White
10:38 AM Bug #1215 (In Progress): 64bit ID failures
Sam White
08:44 AM Bug #1215: 64bit ID failures
Actually some of those build failures are occurring on uth-linux-x86_64, netlrts-linux-x86_64-syncft, and mpi-linux-x... Sam White
08:35 AM Bug #1215 (Merged): 64bit ID failures
Compiler errors building charm with the now merged 64bit ID on cygwin/win:... Sam White
10:48 AM Revision 164b6c53 (charm): Bug #1215: fix build failures from 64bit ID changes in cklocation
Change-Id: Id6d5f88cb760dd8cb6db7b5dc5a3a3401f8b5cf9 Sam White

09/20/2016

06:10 PM Documentation #1140 (In Progress): Document tuple/stats reducers
Sam White
05:13 PM Bug #1214: AMPI_Just_migrated callbacks break using tlsglobals/isomalloc
This is tricky, because we need to create an instance of TCharmAPIRoutine from the user's thread only after their abo... Sam White
11:36 PM Bug #1214: AMPI_Just_migrated callbacks break using tlsglobals/isomalloc
https://charm.cs.illinois.edu/gerrit/#/c/1865/ Sam White
11:33 PM Bug #1214 (New): AMPI_Just_migrated callbacks break using tlsglobals/isomalloc
AMPI doesn't properly take care of turning on/off isomalloc and tlsglobals before calling the user's justMigrated fun... Sam White
02:01 PM Revision 71a0f896 (charm): Replace CkArrayIndex with 64 bit IDs for internal location management.
Changing the CkArrayIndices to 64 bit IDs will make the message
envelopes smaller, as well as giving a single univers...
Phil Miller
02:01 PM Revision 847e1ce1 (charm): Remove CkLocRec hierarchy and use a map for non-local elements.
CkLocRec is now just a single class representing local records only.
Remote elements are found using a std::map from ...
Ramprasad Venkataraman
02:01 PM Revision bc10f561 (charm): Remove spring cleaning and the associated CkLocRec subclasses.
This also takes location lookup out of the CkLocRec hierarchy and
replaces it with a std::map mapping array elements ...
Ramprasad Venkataraman
02:01 PM Revision 57add175 (charm): Cleanup the relation between CkLocMgr, CkArray, and CkLocRec
This makes a small change to the populateInitial API by virtue of
merging the CkArrMgr class into its only subclass C...
Ramprasad Venkataraman
02:01 PM Revision 699d281a (charm): Cleanup #165: Annihilate CkMigratableList and temporarily typedef to s...
Much of our pre-stable-stl code grew its own mutant data structures
which could do all weird and fantastic things. Ck...
Ramprasad Venkataraman
12:13 PM Revision d6740f34 (charm): dynamic node table initial changes and work on stencil3ds elasticity
Justin Miron

09/19/2016

02:55 PM Bug #1188 (Implemented): AMPI Waitsome and Testsome bugs
https://charm.cs.illinois.edu/gerrit/#/c/1862/ Sam White
01:03 PM Revision 0ba474ad (charm): fix for multiple shrink expand requests
Justin Miron
11:13 PM Revision d802abdf (charm): changes for clang with c++ 14 on the mac
Jonathan Lifflander

09/16/2016

04:59 PM Bug #965: ampicc -swapglobals is broken for ld v2.24+
Using swapglobals with gcc-4.7.1 (on Cab at LLNL) results in aborts at runtime with the message:... Sam White
09:25 AM Bug #1097 (In Progress): Attributes are not duplicated in MPI_Comm_dup
This has come back to bite us in ROMIO. ROMIO duplicates a communicator and expects the attributes of the old one to ... Sam White
08:17 AM Cleanup #1151 (Implemented): Eliminate AMPI ROMIO compiler warnings
https://charm.cs.illinois.edu/gerrit/#/c/1853/ Sam White

09/15/2016

06:19 PM Bug #1207 (Merged): DDT predefined types are broken by missing C_BOOL type table entry
Sam White
11:34 AM Bug #1206: Building on OpenMPI fails due to mistaken need for C++ bindings library.
My above hypothesized fix is here:
https://charm.cs.illinois.edu/gerrit/1852
I have not even tested that this wor...
Phil Miller
11:19 AM Bug #1206: Building on OpenMPI fails due to mistaken need for C++ bindings library.
There's a simple solution to all this. We change our configure script to use @test_cc@ instead of @test_cxx@, since w... Phil Miller
11:16 AM Bug #1206: Building on OpenMPI fails due to mistaken need for C++ bindings library.
It looks like you can pass the '--disable-mpi-cxx' option to OpenMPI's configure script to avoid that if you are buil... Sam White
09:15 AM Bug #1206: Building on OpenMPI fails due to mistaken need for C++ bindings library.
I encountered the same issue in another project as well. It seems that OpenMPI's "mpi.h" pulls in the C++ MPI binding... Erik Schnetter
10:31 AM Feature #1213: Use Charm++ built-in reducers for more predefined types/ops in AMPI
https://charm.cs.illinois.edu/gerrit/#/c/1839/ Sam White
10:30 AM Feature #1213 (Merged): Use Charm++ built-in reducers for more predefined types/ops in AMPI
Quicksilver uses MPI_INT64_T and MPI_UINT64_T in reductions. Currently we use AMPI's custom reduction, which requires... Sam White
12:04 AM Feature #1114 (Merged): Build and link ROMIO on AMPI by default
Sam White
12:03 AM Bug #1198 (Merged): AMPI's mpi.h does not include ROMIO's mpio.h
Sam White

09/14/2016

03:58 PM Bug #1207 (Implemented): DDT predefined types are broken by missing C_BOOL type table entry
https://charm.cs.illinois.edu/gerrit/#/c/1850/ Sam White
01:55 PM Bug #1207 (In Progress): DDT predefined types are broken by missing C_BOOL type table entry
Sam White
01:46 PM Bug #1207 (Merged): DDT predefined types are broken by missing C_BOOL type table entry
When I added MPI_C_BOOL to AMPI and CkDDT_C_BOOL to DDT, I forgot to add the definition of that type to the type tabl... Sam White
03:56 PM Revision 5bf8ddf6 (charm): Bug #1207: update DDT's type table with C/C++ bool datatype
Change-Id: Iad1c81939ac599c63fbc3f293a9ebd28b6b6c223 Sam White
02:57 PM Revision 2ee691f5 (charm): Fix leaked reduction message in projections tracing
Change-Id: I7d9b8392ca75d7c4cf50024142aeb5b19ccbe117 Sam White
02:39 PM Bug #1206: Building on OpenMPI fails due to mistaken need for C++ bindings library.
Note: There are also references to external symbols ("U"), as in
libck.a:mpi-interoperate.o: U __...
Erik Schnetter
02:37 PM Bug #1206: Building on OpenMPI fails due to mistaken need for C++ bindings library.
I see symbols like these:
$ nm -A *.a | grep MPI | grep __ | grep Init
libck.a:mpi-interoperate.o: 00000000000009...
Erik Schnetter
02:11 PM Bug #1206: Building on OpenMPI fails due to mistaken need for C++ bindings library.
We do not use the MPI C++ bindings proper, we only use MPI's C language bindings. We obviously need a C++ compiler to... Sam White
02:07 PM Bug #1206: Building on OpenMPI fails due to mistaken need for C++ bindings library.
The C++ bindings of MPI are deprecated/removed, so if they are in Charm++ they need to go.
See http://blogs.cisco.co...
Jim Phillips
11:09 AM Bug #1206: Building on OpenMPI fails due to mistaken need for C++ bindings library.
These commands indeed don't mention the "mpi_cxx" library. Here is the pkgconfig file:
$ cat ../lib/pkgconfig/ompi...
Erik Schnetter
09:38 AM Bug #1206: Building on OpenMPI fails due to mistaken need for C++ bindings library.
Erik, what do "mpicc --showme:compile" and "mpicc --showme:link" return on your system?
It seems strange that Charm+...
Jim Phillips
11:31 PM Bug #1206 (Implemented): Building on OpenMPI fails due to mistaken need for C++ bindings library.
Will post for review and merge shortly. Thank you for the contribution! Phil Miller
11:57 AM Revision f3937a55 (charm): Cleanup: use nop reductions for NULL contributions
Change-Id: I7d9030422fd7eb54f93b72582d4a6c5e48aef3ec Sam White
11:55 AM Revision 76d6c43e (charm): AMPI: improve error handling by aborting with more information
Change-Id: I26f166fa93288a5da5b832cc0745971a3e668de9 Sam White
11:37 AM Bug #872: smp standalone inconsistency
The hwloc changes being worked on at Charmworks will presumably address these and other issues (if I'm not mistaken). Juan Galvez
10:54 AM Bug #872: smp standalone inconsistency
Ping. Jim Phillips
01:01 AM Feature #1088 (In Progress): Trace MPI_ functions in AMPI
Sam White
12:01 AM Feature #967: AMPI support for MPI-3's MPI_Comm_create_group
Here's the paper that originally proposed this feature for MPI-3: http://www.mcs.anl.gov/papers/P1914.pdf Sam White
11:53 PM Bug #1191: AMPI is missing MPI_Type_*_{attr,keyval} routines
No-op implementations of these routines: https://charm.cs.illinois.edu/gerrit/#/c/1840/ Sam White
11:53 PM Bug #1191 (Implemented): AMPI is missing MPI_Type_*_{attr,keyval} routines
Sam White

09/13/2016

05:48 PM Bug #1206 (Merged): Building on OpenMPI fails due to mistaken need for C++ bindings library.
I tried building Charm++ with MPI support, using OpenMPI. This did not work, as the Charm++ configure stage could not... Erik Schnetter
05:48 PM Feature #1095 (In Progress): Improve AMPI error handling
Made AMPI abort on more types of errors and to be more informative when aborting: https://charm.cs.illinois.edu/gerri... Sam White
02:32 PM Feature #557 (Implemented): Built-in reducer types for logical and bitvec and/or/xor on bools and...
https://charm.cs.illinois.edu/gerrit/#/c/1841/ Sam White

09/12/2016

04:47 PM Cleanup #1034: Constant comparison warnings from xlC compiler
It would have been an SMP build. Jim Phillips
03:55 PM Revision 7a5869c7 (charm): Apply Tim's patch, change ifdef to back to if
Change-Id: Ice010d597681814caa31577f4d6769ddc210c16c Redmine Admin
02:49 PM Bug #1205 (New): AMPI's -tlsglobals option is only supported by GCC
When compiling AMPI with clang and passing the -tlsglobals flag, charmc attempts to pass the -mno-tls-direct-seg-refs... Tarun Prabhu
08:51 PM Bug #1017 (Merged): charmxi: using reserved keywords fails silently
Change merged into charm by Sam. Nitin Bhat

09/11/2016

08:54 PM Bug #961: AMPI support for cancelling recv requests/statuses
Deassign and mark Unscheduled. Phil Miller
08:53 PM Bug #961 (New): AMPI support for cancelling recv requests/statuses
Fine to mark it low priority, and not planed for an upcoming release. Re-opening, because the issue still exists, and... Phil Miller

09/09/2016

01:03 PM Bug #1188: AMPI Waitsome and Testsome bugs
Any update on this? Sam White

09/08/2016

02:56 PM Bug #1190 (Merged): AMPI fails to rename C++ main with signature 'int main()'
Sam White
02:55 PM Cleanup #1203 (Merged): AMPI forces builds to be serial for ROMIO
AMPI's Makefile include ".NOTPARALLEL:" in order to force parallel builds to serialize. It does this because the curr... Sam White
02:53 PM Revision ee84e087 (charm): AMPI #1114: build and link ROMIO by default on AMPI
Change-Id: If3af056677f704212a69759d5564f9c0f0f635f8 Sam White
02:49 PM Revision 8df8f153 (charm): AMPI cleanup: silence unused variable compiler warnings
Change-Id: I962065d9b7c6b8ea2be1821f65eeb130bab697e6 Sam White
02:43 PM Revision c0f0f9c8 (charm): nodeGPU: Pre-allocate additional pinned memory
Change-Id: Iad1085df230e0723ca532bf025ed3de21249aac9 Redmine Admin
02:43 PM Revision 9dcb825a (charm): Single instance of GPU Manager, shared across all PEs within a node
Change-Id: I15d0808a4d6f38fbacfe30cb0642dc6d56cc944c Harshit Dokania
02:43 PM Revision e7dedfaa (charm): exitHybridAPI(): delete extra CmiDestroyLock().
Thomas Quinn
02:43 PM Revision 1989ddf3 (charm): nodeGPU: refactor recursive function and extend lock
Redmine Admin
02:43 PM Revision e0df78e1 (charm): Single instance of GPU Manager, shared across all PEs within a node
Change-Id: I15d0808a4d6f38fbacfe30cb0642dc6d56cc944c Harshit Dokania
11:37 AM Revision 0589a1cf (charm): Allow Central and Distributed LBs to be called seqeuntially
Change-Id: If58ffbff73dbf90cafe590b8338b898a81ee2d6d Kavitha Chandrasekar
10:31 AM Cleanup #1151: Eliminate AMPI ROMIO compiler warnings
With ICC-16.0.3 on Cab @ LLNL I get a bunch of warnings like this too:... Sam White
08:06 AM Bug #1202 (New): Memory leaks in converse's cldb
Running valgrind on netlrts-linux-x86_64 examples reveals two memory leaks in initialization routines. One is in src/... Sam White
07:55 AM Feature #1095: Improve AMPI error handling
Added some rudimentary support for this here: https://charm.cs.illinois.edu/gerrit/#/c/1828/ Sam White
11:38 PM Revision a9573324 (charm): AMPI: fix ROMIO build on Clang
Change-Id: If7e40808371e802fb1c4938670f545dc20007f64 Sam White
09:03 PM Feature #1114 (Implemented): Build and link ROMIO on AMPI by default
Sam White

09/07/2016

05:13 PM Feature #947: AMPI support for MPI-2's MPI_Type_* routines
These 2 would be good to support:
MPI_Type_dup
MPI_Type_get_true_extent
And after you've added MPI_Type_get_tr...
Sam White
04:58 PM Feature #1114 (In Progress): Build and link ROMIO on AMPI by default
ROMIO is now built by default when building AMPI. You can specify '--without-romio' to not build it. It is always lin... Sam White
03:51 PM Revision ad2b9ea3 (charm): AMPI #1198: include ROMIO's mpio headers in AMPI's mpi headers
Change-Id: I9249eb6b0a5c274e692ee6104f1e42a25e433aaf Sam White
03:46 PM Bug #1198 (Implemented): AMPI's mpi.h does not include ROMIO's mpio.h
https://charm.cs.illinois.edu/gerrit/#/c/1832/ Sam White
02:34 PM Bug #1198 (In Progress): AMPI's mpi.h does not include ROMIO's mpio.h
Two proposed solutions:
A) configure ROMIO (to generate its headers) before building AMPI, then make ROMIO after A...
Sam White
03:28 PM Revision 6e41e5e4 (charm): AMPI #1190: rename C++ main routines without args correctly
Change-Id: I350173ec85a15fd5f004647b13bf0f6b8df2aaf5 Sam White
03:21 PM Feature #574 (Merged): Windws builds: don't pollute /usr/bin
The win32 build targets are now all gone, so that part is moot.
The change linked above plus https://charm.cs.illi...
Phil Miller
03:18 PM Bug #1190 (Implemented): AMPI fails to rename C++ main with signature 'int main()'
https://charm.cs.illinois.edu/gerrit/#/c/1831/ Sam White
02:18 PM Bug #1190: AMPI fails to rename C++ main with signature 'int main()'
Let's talk about this directly - I can probably figure out a way around it. Phil Miller
03:11 PM Bug #228 (Rejected): net-win32 configuration runs in a semi-smp mode
@net-win32@ got deleted some months back. This is now moot. Phil Miller
02:50 PM Bug #1201 (Rejected): SMP builds segfault on NULL lock in tests/charm++/chkpt
http://ppl-jenkins.cs.illinois.edu:8080/job/Nightly-Build/940/label=precise,platform=netlrts-linux-x86_64-smp-clang/c... Phil Miller
02:15 PM Bug #1200: Openatom mixes CkMulticast managers and uses different instances to contribute and bro...
This needs way more explanation of why the described behavior does/will illustrate a bug in Charm++. Phil Miller
09:36 PM Bug #1200 (Merged): Openatom mixes CkMulticast managers and uses different instances to contribut...
The working of CkMulticast Managers is such that it doesn't get exposed now, but it does get affected in the 64-bit b... Vipul Harsh
02:12 PM Bug #923 (Merged): Compilation failing in net-cygwin
This was the _ftime/ftime issue. Fixed with a patch contributed by Ed Karrells, from Sanjay's CS598 course. Phil Miller
02:10 PM Feature #1175: Don't require autoconf to be installed on user systems for hwloc build
We can revisit this later, then. The other code changes shouldn't interact with it. Phil Miller
02:09 PM Feature #1177: Support variable numbers of processes per host, each with identical thread counts
The expectation is that we'd potentially move to a mode of usage where we always ask the system process launcher for ... Phil Miller
02:05 PM Feature #1177: Support variable numbers of processes per host, each with identical thread counts
This is useful, but the Cray XE/XK case isn't relevant as we use aprun on that platform. Jim Phillips
12:22 PM Revision c3cfad09 (charm): cygwin: replace calls to _ftime() with ftime()
Reported fix by Ed Karrels
Change-Id: Iee6a1321de50d2f168c613c6352c880dec847a84
Phil Miller
11:43 AM Revision 42f74e6d (charm): AMPI: make MPI_C_BOOL definition portable to MSVC
Change-Id: I1edaf810d466e956fa82589b27f5cbc18c5c3fdc Sam White

09/05/2016

06:18 PM Revision 78bf698b (charm): Post working, polishing needed
Change-Id: Ia8b6f684ea75ea698a4e1eeedd40afdecb724a40 Vipul Harsh

09/04/2016

01:49 PM Revision 94c5265d (charm): made changes in xi to generate a seperate unmarshalling function for rdma
Change-Id: I7821099c336a1bd40c71e8c35aa75f4f46d3ccaa Vipul Harsh

09/02/2016

04:47 PM Revision 8301f5fa (charm): AMPI: abort on errors by default
Change-Id: I3a05b5f1babec31f9ad067b417a46a9262151d1a Sam White
04:04 PM Revision 46cf40c0 (charm): AMPI: add MPI_OFFSET datatype definition for ROMIO
Change-Id: Ib169397b6b630fa417c0f3eac673649b122cc4a8 Sam White
04:04 PM Revision 2a7fcb65 (charm): AMPI: add MPI_C_BOOL type to DDT and AMPI
Change-Id: Icc15b2db1a3aeb5494074a1c1bcc6a0f20ac1388 Sam White
04:03 PM Revision e7a07a78 (charm): AMPI: only check arguments for errors on ranks they are needed
ROMIO was hanging in AMPI_Gatherv because the non-root ranks passed NULL
for the recv buffer, displs, and counts.
Ch...
Sam White
04:01 PM Revision 8b63b791 (charm): AMPI: remove definition of MPI_OFFSET_KIND until we support MPI_Offset
Change-Id: I7c02217dd8e391faeab229ee711064f86be7f7b6 Sam White
03:56 PM Bug #1199 (Merged): Fix hash function "hash" in "ckarrayindex.h" to include the first element in ...
This was suggested by Alexander Frolov. For 1d arrays, this makes for a continuous sequencing. Needs performance test... Vipul Harsh
03:23 PM Documentation #1136 (Merged): Document ability of [sync] entry methods to return any type that su...
https://charm.cs.illinois.edu/gerrit/#/c/1822/ Sam White
12:14 PM Bug #1198 (Merged): AMPI's mpi.h does not include ROMIO's mpio.h
AMPI's mpi.h and mpif.h should include ROMIO's mpio.h and mpiof.h, respectively. Otherwise applications cannot use ty... Sam White
12:08 PM Revision 57f414ef (charm): Bug #1136: Document ability of [sync] entry methods to return any type...
Change-Id: Id5ced35bb0f81bd03227f9d1f549ce282f900367 Vipul Harsh
11:42 AM Bug #1197: Charmrun fails to connect to nodes on Taub in netlrts build.
Those are just memory leaks. They should be cleaned up, but they are not a cause of launch failures. Phil Miller
11:32 AM Bug #1197: Charmrun fails to connect to nodes on Taub in netlrts build.
Nevermind, I was getting an allocation on Golub before, and it works there. On Taub, ++mpiexec does not work with gcc... Sam White
11:19 AM Bug #1197: Charmrun fails to connect to nodes on Taub in netlrts build.
netlrts-linux-x86_64 runs on Golub using ++mpiexec with the gcc/4.9.2 and mvapich2/1.6-gcc modules loaded. It also ru... Sam White
10:02 AM Bug #1197: Charmrun fails to connect to nodes on Taub in netlrts build.
Please try ++mpiexec Bilge Acun
09:52 AM Bug #1197: Charmrun fails to connect to nodes on Taub in netlrts build.
I've found this issue when I run a simple exercise in CS 598. It doesn't crash but just be terminated with the messag... Seonmyeong Bak
09:44 AM Bug #1197: Charmrun fails to connect to nodes on Taub in netlrts build.
Please try running a standalone version of some small program, and see if that crashes. If it doesn't crash, run it u... Phil Miller
09:32 AM Bug #1197 (Rejected): Charmrun fails to connect to nodes on Taub in netlrts build.
The netlrts version of Charm++ is built and executed well on taub but when it is built with gcc/4.9 (or later), it do... Seonmyeong Bak
09:15 PM Feature #1175: Don't require autoconf to be installed on user systems for hwloc build
it seems some systems requires autoconf. It seems that there is some issue with hwloc's default configure script that... Gengbin Zheng

09/01/2016

07:29 PM Revision 06293a33 (charm): charmxi #1017 - Added error messages when keywords are used for identi...
Change-Id: Ic9d1b7997312817d33d0a608f2d5bd73ad325dfe Nitin Bhat

08/31/2016

01:24 PM Documentation #1196 (Merged): Document and exemplify chare class inheritance
There's nothing in the manual or examples about chare classes having parent class relationships from which the child ... Phil Miller
11:33 PM Bug #1194 (Merged): AMPI Gather and its variants don't support non-contiguous contributions
Sam White
11:25 PM Bug #1104: AMPI instances may change if migrated while suspended
Fixes a couple newly created instances of this bug here: https://charm.cs.illinois.edu/gerrit/#/c/1812/
Still need...
Sam White
10:19 PM Revision 59e3a224 (charm): AMPI #1194: support non-contiguous send types in Gathers
Change-Id: Ie62ccf4fde39c73beedb8c1eb7100ec3ff8397f4 Sam White
10:18 PM Revision ce10c145 (charm): AMPI: update ampi instance ptr after blocking in wait routines
Change-Id: Id982e4ad2453d655e7cded1eb4d1c11fed74bf61 Sam White
10:10 PM Revision d349229b (charm): AMPI cleanup: deduplicate posting of nonblocking requests
Change-Id: Id8a0d06d73380698eb123403711b32444d865843 Sam White

08/30/2016

12:16 PM Revision 68436beb (charm): AMPI cleanup: eliminate variable length array usage
Change-Id: Ib4cd20dcf64392272f4eaa8db73ba74be7d05ec3 Sam White
11:13 AM Feature #387: Pack variable-envelope structures to reduce overhead bytes
We should revisit the size and layout of the envelope with this kind of change in mind. I'd suggest using the @pahole... Phil Miller
10:43 AM Feature #387: Pack variable-envelope structures to reduce overhead bytes
Should we revisit compiler support for this now that 64bit ID is just about ready to be merged? Sam White
10:41 AM Feature #557: Built-in reducer types for logical and bitvec and/or/xor on bools and ints
Supporting more datatypes for logical reduction operations in Charm will benefit AMPI too. Sam White
10:36 AM Bug #1169 (Merged): Command line argument +balancer for multiple balancers does not work
Sam White
10:33 PM Feature #1116 (In Progress): AMPI broadcasts should be [nokeep]
Sam White
10:32 PM Bug #1193 (Merged): AMPI assumes all user-defined Ops are non-commutative
Sam White
10:32 PM Bug #1185 (Merged): AMPI support for non-commutative reduction operations
Sam White
09:01 PM Revision 9f4a65d3 (charm): AMPI #1193: do not assume all user-defined Ops are non-commutative
Change MPI_Op from a function pointer to an int, used as an index into
a table of Ops. The table also keeps track of ...
Sam White
08:59 PM Revision 01139d76 (charm): AMPI #1185: support non-commutative MPI_Op's, assume user-defined op's...
Change-Id: Ib276359971cc362d5ae2c5fa45d8cde85c388897 Sam White
08:44 PM Bug #1186 (Merged): AMPI non-blocking routines don't always create actual requests
Sam White

08/29/2016

06:25 PM Revision ba0b7ce9 (charm): AMPI #1186: create actual requests for all non-blocking communication ...
Instead of setting the value of request to MPI_REQUEST_NULL.
Change-Id: I16136b9fced610a38d67008acd0eaf363ac78e6d
Sam White
05:00 PM Bug #1194 (Implemented): AMPI Gather and its variants don't support non-contiguous contributions
https://charm.cs.illinois.edu/gerrit/#/c/1820/ Sam White
04:55 PM Bug #1194 (Merged): AMPI Gather and its variants don't support non-contiguous contributions
AMPI_(I)(all)Gather(v) should all support contributions of non-contiguous datatypes, but currently the first 'typeSiz... Sam White
04:06 PM Revision 57326ffe (charm): Fixed the following issues:
- Receiver not posting for 4GB buffer on psm2 provider
- Packetization bug (send sender's nodeNo instead of receiver's)
Nitin Kundapur Bhat
11:35 AM Revision c4a9a274 (charm): Bug #1169: Allow duplicate arguments like +balancer, reject other dupl...
Change-Id: Ibe9f71f745b94251e718816de79873d4b371e277 Kavitha Chandrasekar
11:34 AM Bug #1186: AMPI non-blocking routines don't always create actual requests
Another instance of this bug besides in AMPI_Isend was in all collective routines we were setting the request to MPI_... Sam White
11:00 AM Bug #1161 (Closed): charm-6.7.1.tar.gz untars to charm
Marking 'closed' so that it doesn't show up under 'open' issues. Sam White
07:48 AM Bug #1189: AMPI resumeOnRecv should be a property of the thread, not each comm
Currently resumeOnRecv is inside the ampi class. Two proposals:
1. Move it inside ampiParent since the thread that i...
Sam White
07:40 AM Bug #1193 (Implemented): AMPI assumes all user-defined Ops are non-commutative
NOTE: this breaks ABI compatibility with older AMPIs, because this changes MPI_Op to an int. If you get a bunch of un... Sam White
10:08 PM Bug #1193 (Merged): AMPI assumes all user-defined Ops are non-commutative
The patch to support non-commutative reduction operations also made the assumption that all user-defined Ops are non-... Sam White
03:29 AM Revision 9b5b632d (charm): Rdma support for pamilrts machine layer
Change-Id: I825bbf7d1ddc9f5460f8ae3ebaa66d5b4fbde78a Nitin Bhat
03:22 AM Revision 991aeca9 (charm): Rdma API at the Lrts layer
Change-Id: I36a90b305eea0450f1620702636ee938c3afde96 Nitin Bhat
03:02 AM Bug #1017: charmxi: using reserved keywords fails silently
Gerrit link: https://charm.cs.illinois.edu/gerrit/#/c/1816/ Nitin Bhat
03:01 AM Bug #1017 (Implemented): charmxi: using reserved keywords fails silently
I wasn't able to entirely reproduce this bug. For the code in the bug description, I get a slightly informative comma... Nitin Bhat
09:53 PM Revision 633d3519 (charm): Rdma support at the generic layer
Changes include:
Charmxi changes for supporting rdma parameter
Charm and converse support for rdma parameters
Added a...
Nitin Bhat
09:07 PM Revision a138a102 (charm): Rdma support at the generic layer
Changes include:
Charmxi changes for supporting rdma parameter
Charm and converse support for rdma parameters
Added a...
Nitin Bhat
07:05 PM Bug #1038 (Merged): Add support for srun to charmrun script for gni layer
Should be all good now Michael Robson

08/28/2016

05:02 PM Bug #1045: Array pointers in messages can be overwritten by generated code in the message class c...
More details on the optimization here:
https://gcc.gnu.org/gcc-6/porting_to.html
Phil Miller
04:16 PM Bug #1045: Array pointers in messages can be overwritten by generated code in the message class c...
This crashes in megatest... Phil Miller
03:59 PM Bug #1045: Array pointers in messages can be overwritten by generated code in the message class c...
I need to investigate more, but I think this has come to the fore. Phil Miller
02:39 PM Bug #1185 (Implemented): AMPI support for non-commutative reduction operations
Support non-commutative reduction op's and assume all user-defined op's are non-commutative: https://charm.cs.illinoi... Sam White
10:01 AM Bug #1185: AMPI support for non-commutative reduction operations
For now we will assume all user-defined MPI_Ops are non-commutative, and we'll implement them like a gather where at ... Sam White
09:57 AM Bug #1185 (In Progress): AMPI support for non-commutative reduction operations
Sam White
09:36 AM Cleanup #1192: AMPI is missing const qualifiers on buffer arguments to communication routines
AMPI currently doesn't use const on sendbuf arguments so that we can handle MPI_BOTTOM and MPI_IN_PLACE in the easies... Sam White
11:19 PM Revision 6100bf4a (charm): AMPI: set freed MPI_Info objects to MPI_INFO_NULL
Change-Id: I46371d1a4c2a0ca01e204a10775e4599dcff999a Sam White
11:15 PM Feature #954 (Implemented): Update AMPI's version of MPICH test suite
Updated to mpich-3.2 and added no-op implementations of AMPI's unimplemented features: https://charm.cs.illinois.edu/... Sam White

08/27/2016

12:09 PM Cleanup #1192 (Merged): AMPI is missing const qualifiers on buffer arguments to communication rou...
All Gather, Scatter, Alltoall, Reduce, and Scan variants, as well as MPI_Sendrecv and possibly others, are missing co... Sam White
10:38 AM Bug #1191: AMPI is missing MPI_Type_*_{attr,keyval} routines
Missing routines:... Sam White
10:33 AM Bug #1191 (Merged): AMPI is missing MPI_Type_*_{attr,keyval} routines
Came up when trying to build MPICH-3.2's test suite on AMPI. We already support these on Comm's and Win's. Sam White
10:22 AM Feature #947: AMPI support for MPI-2's MPI_Type_* routines
I didn't have MPI_Type_dup on this list earlier, which was added in MPI-2.1. We'll need to eventually implement that ... Sam White
09:57 AM Feature #954 (In Progress): Update AMPI's version of MPICH test suite
To build mpich3.2's mpi test suite on AMPI:... Sam White

08/26/2016

05:22 PM Bug #1190 (Merged): AMPI fails to rename C++ main with signature 'int main()'
The problem is main routines without (int argc, char **argv) arguments. This commonly appears in build system tests f... Sam White
05:12 PM Bug #1186 (Implemented): AMPI non-blocking routines don't always create actual requests
Create actual request structs: https://charm.cs.illinois.edu/gerrit/#/c/1811/
Related cleanup of request insertion...
Sam White
03:03 PM Bug #1186 (In Progress): AMPI non-blocking routines don't always create actual requests
Sam White
04:48 PM Revision 7cc8eeb1 (charm): Rdma support at the generic layer
Changes include:
Charmxi changes for supporting rdma parameter
Charm and converse support for rdma parameters
Added a...
Nitin Bhat
03:04 PM Cleanup #1103 (Implemented): Refactor AMPI virtual topologies
Sam White
02:45 PM Bug #1188 (In Progress): AMPI Waitsome and Testsome bugs
Phil Miller
02:11 PM Bug #1188 (Merged): AMPI Waitsome and Testsome bugs
Nikhil found errors in AMPI's MPI_Waitsome and MPI_Testsome implementations related to the outcount parameter and han... Sam White
02:14 PM Bug #1189 (Merged): AMPI resumeOnRecv should be a property of the thread, not each comm
Two intertwined issues:
1. resumeOnRecv and resumeOnColl belong to the ampi class, of which there can be multiple in...
Sam White
12:41 PM Revision eaf65781 (charm): compiler error fixes
Change-Id: I64bed4b952ef7a4af2584dcc44ec12a2891e49d4 Justin Miron
12:32 PM Revision 72b4eaf4 (charm): compiler error fixes
Change-Id: I1463422a9a071fb90a17be9f62f19225559cfc14 Justin Miron
12:08 PM Revision f3caaabd (charm): gni rdma code without callbacks
Change-Id: I6ad0afac08ae3fc03ee4771ae96710dd245ae6a2 Justin Miron
11:14 AM Revision a931a82d (charm): AMPI cleanup: add comment for implicit function declaration
Comment will show up in the compiler warning's line
Change-Id: I25aa3cc27e41da5d9a03f82353459df3ab23d43e
Sam White
10:05 PM Revision 77f8a38e (charm): Rdma support at the generic layer
Changes include:
Charmxi changes for supporting rdma parameter
Charm and converse support for rdma parameters
Added a...
Nitin Bhat

08/25/2016

05:42 PM Revision add73521 (charm): Merge commit 'refs/changes/41/1341/5' of http://charm.cs.illinois.edu/...
Justin Miron
02:14 PM Feature #1187 (Implemented): Automatic delegation of section work to CkMulticastMgr
Vipul Harsh
02:14 PM Feature #1187 (Merged): Automatic delegation of section work to CkMulticastMgr
Vipul Harsh
01:52 PM Bug #1186 (Merged): AMPI non-blocking routines don't always create actual requests
Nikhil reported issues with Yorick on AMPI because AMPI always sets AMPI_Isend requests to MPI_REQUEST_NULL rather th... Sam White
12:40 PM Bug #1185 (Merged): AMPI support for non-commutative reduction operations
AMPI currently aborts if a user tries to create an Op that is non-commutative, and it doesn't keep track of the commu... Sam White
08:42 AM Bug #1038: Add support for srun to charmrun script for gni layer
Issue is fixed for gni-crayxc but gni-crayxc-smp has a parsing issue in ./instead_test.sh:... Sam White
 

Also available in: Atom