Project

General

Profile

Activity

From 04/24/2016 to 05/23/2016

05/24/2016

08:00 PM Revision 302f9a97 (charm): Cleanup: Change atomic -> serial, Cmi* -> Ck* in examples and tests
Change-Id: Icdb19308b405dc74199247674daf032a141853ef Sam White

05/23/2016

01:25 PM Feature #982: Gather reduction operation using Tuple/Set reducers
This could be implemented using the new Tuple reducer type with two CkReduction::set reducers, one containing the arr... Sam White
11:50 AM Bug #1076: failure to exit after long run on PSC Bridges
According to Rick Costa of PSC User Services
"The same hanging issue has surfaced for other packages."
There is a...
Jim Phillips
07:36 PM Bug #1080 (In Progress): multicore projections tracing runs hang at startup on 129 pes
I'm not sure how the status and target version were changed in my earlier update, I'm setting them back here. Ronak Buch

05/20/2016

06:15 PM Bug #1080 (New): multicore projections tracing runs hang at startup on 129 pes
I was able to replicate this on triton, but I was also able (somewhat oddly) to successfully run a 129 PE job with Pr... Ronak Buch
11:52 AM Revision ad829962 (charm): AMPI cleanup: make member functions const
Change-Id: Icdabd5ca350d8776f6588f8de2ec499bb43d27f0 Sam White
09:10 AM Bug #1081 (Merged): Converse command line arguments produce false warnings
Converse command line arguments are not properly registered during initialization, so using them generates a warning ... Sam White
09:13 PM Bug #1078 (Merged): AMPIF doesn't handle Fortran strings correctly
Sam White
07:14 PM Revision 912878e3 (charm): SDAG changes in xi-Entry.C
Vipul Harsh

05/19/2016

06:49 PM Revision c2bab017 (charm): fixed SDAG closure for Rdma to do the same thing as for arrays
Vipul Harsh
05:14 PM Revision 6778513e (charm): freeing msg using CkFreeMsg
Vipul Harsh
04:41 PM Revision 244befe1 (charm): AMPI: set default names for predefined communicators
Change-Id: I7ce0f63547d64e38e93a39586caefc135b90d551 Sam White
03:55 PM Revision 4de0a720 (charm): Fix makefiles and move user_call out of the API
Reuben Tan
02:59 PM Revision dd4e4269 (charm): Merge
Reuben Tan
02:37 PM Bug #937: [local] entry methods don't set tracing event dependencies
Necessary change is in xi-Entry.C in Entry::genArrayDefs and analogs for chare/group/nodegroup entry methods.
At ...
Phil Miller
02:20 PM Bug #503 (New): Replace use of obsolete memalign function with posix_memalign
This is clearly not 'High' priority, since it hasn't been worked on in over a year and no one has cared.
Also, no ...
Phil Miller
01:38 PM Bug #1080 (In Progress): multicore projections tracing runs hang at startup on 129 pes
Ronak Buch
11:00 AM Bug #1080 (Merged): multicore projections tracing runs hang at startup on 129 pes
... Jim Phillips
11:59 AM Revision b179c31f (charm): AMPIF #1078: Terminate Fortran strings correctly.
Strings passed from Fortran into AMPI need to be NULL terminated.
We now define 'MPI_' routines that take string argu...
Sam White
10:15 PM Feature #969: AMPI support for collectives on inter-communicators
AMPI communicators, including intercommunicators, are implemented as chare arrays, so an intercomm has 3 proxies: loc... Sam White

05/18/2016

06:44 PM Revision fe9a38ab (charm): fixing rdma example
Vipul Harsh
06:25 PM Revision 0ff41407 (charm): fixing pingpong rdma, don't free msg in callback
Vipul Harsh
06:09 PM Revision a977bbb5 (charm): Shrink-expand (malleable jobs): Fix build issues with earlier merge
Change-Id: Icd337a18a79e4989287f3b4866bfcb86cd7ebcb4 Phil Miller
03:48 PM Bug #1078 (Implemented): AMPIF doesn't handle Fortran strings correctly
Sam White
03:40 PM Revision 84ae73f0 (charm): Cleanup: Remove stray print statement from Shrink-Expand #236
Change-Id: I3a867bef3a21e640564551937b4691ae70208af9 Sam White
09:28 AM Bug #1062: userDrivenMode undefined in converse/lrts exit
Note that the commit above does not fix the issue for non-LRTS layer builds (net- and pami-). Sam White
09:21 AM Support #1079 (Closed): Remove deprecated machine layers and retired machines from Autobuild
Autobuild has the following tests that should be removed/migrated:
1. 32-bit Darwin builds
2. Builds on Finesse (...
Sam White
06:33 AM Bug #1076: failure to exit after long run on PSC Bridges
Stampede MPI version exits normally. Jim Phillips
09:43 PM Revision a7a26a7f (charm): AMPI cleanup: make inclusion of ckliststring.h in ampiimpl.h conditional
on AMPIMSGLOG
Change-Id: I603b219107268826a0495c3b37cdcab3efc0ead6
Sam White

05/17/2016

06:01 PM Revision 2e6b3017 (charm): done
Reuben Tan
03:55 PM Bug #1078 (In Progress): AMPIF doesn't handle Fortran strings correctly
Sam White
09:19 PM Bug #1078: AMPIF doesn't handle Fortran strings correctly
We were also not correctly blanking out all characters after the end of strings before handing them back to Fortran u... Sam White
12:54 PM Revision 6b3548f0 (charm): DDT: make copy/assignment ctors private for all datatypes
This also fixes failures on Windows stemming from commit #1144.
Change-Id: I89eed2a72c37c9b9d9f38045aa9b272e48baf820
Sam White
12:43 PM Bug #788 (Merged): Out of tree build broken for CUDA
Phil Miller
12:33 PM Bug #1061 (Merged): Generated unmarshalling code for [reductiontarget] inappropriately offsets in...
Phil Miller
12:32 PM Feature #236 (Merged): Charm++ Shrink-Expand (malleable jobs, dynamically varying node count)
Phil Miller
11:01 PM Feature #236: Charm++ Shrink-Expand (malleable jobs, dynamically varying node count)
Set target version, since the implementation is finally ready to merge thanks to Ralf and Bilge tirelessly addressing... Phil Miller
12:32 PM Bug #1062 (Merged): userDrivenMode undefined in converse/lrts exit
Phil Miller
09:35 AM Revision b2061c5e (charm): AMPI_Scatter(v) use bcast instead of pt2pt messages
Change-Id: I1cbe425bb30a4f739453530d4a3786d95d765242 Sam White
08:07 AM Revision a9e07d44 (charm): AMPI collectives: copy buffer directly rather than sending messages to...
Change-Id: Ic7c16bcf2d942ead0b20e8674385e6f36573f529 Sam White
11:46 PM Bug #1076: failure to exit after long run on PSC Bridges
I get the same symptom even for NAMD 2.10 with Charm++
@v6.6.1-rc1-1-gba7c3c3-namd-charm-6.6.1-build-2014-Dec-08-289...
Jim Phillips
11:07 PM Revision a112b91f (charm): Bug #1062: Fix linking errors by moving definition of userDrivenMode t...
Change-Id: I4e9e985d40b2b903b9b0bab408a253d4a13411d2 Eric Mikida
11:05 PM Revision abf3ac80 (charm): Shrink-Expand #236: Dynamic malleable job resizing at runtime
Shrink-expand allows users to dynamically and transparently change the
number of PEs during runtime at load balancing...
Abhishek Gupta
11:04 PM Revision 5c00bf0f (charm): Bug #1061: Correct code generation for [reductiontarget] methods takin...
Change-Id: Ie12acab18b9161c0deacc73d9a8c0b241badc1e7 Phil Miller
10:57 PM Bug #1026 (Rejected): Running ChaNGa on latest build of charm gives a crash
Phil Miller
10:32 PM Revision f07b5764 (charm): Update smart-build.pl and install.tex for machine layer removals
Change-Id: I0eef11dbc83c21f96a99ba0b8953e6aecde71550 Sam White
10:07 PM Revision 448be1df (charm): AMPI cleanup: simplify mpi_comm_worlds class
Change-Id: I3131a6832deb8b95589c1d904df840a64359579e Sam White
10:04 PM Bug #1077 (Merged): AMPI 'names' are not stripped of trailing whitespace
Juan Galvez
10:04 PM Bug #1047 (Merged): AMPI_Migrate doesn't PUP info key/val and name strings correctly
Juan Galvez
10:02 PM Feature #1067 (Merged): Make all AMPI entry methods [expedited]
Juan Galvez
09:29 PM Revision eaf9f53d (charm): AMPI #1077: Strip trailing whitespace from comm/win/type names
Change-Id: I8318c43688f184c820ccaeacd106ec5a23b97046 Sam White
09:29 PM Revision 895bd079 (charm): AMPI #1047: Initialize and PUP strings with explicitly added NULL
terminators correctly.
Change-Id: I050b412f21f0b69e787aa8123a23c8c3488988c8
Sam White
09:27 PM Revision 7daaf9c1 (charm): AMPI cleanup: use MPI_Comm instead of int for function parameters
Change-Id: I301b0c14acdf6121e3c87c6194af73bd2c7248e8 Sam White

05/16/2016

03:20 PM Bug #1077 (Implemented): AMPI 'names' are not stripped of trailing whitespace
https://charm.cs.illinois.edu/gerrit/#/c/1207/ Sam White
12:27 PM Bug #1077 (Merged): AMPI 'names' are not stripped of trailing whitespace
The MPI standard dictates that names on communicators, types, and windows should be stripped of trailing whitespace (... Sam White
03:12 PM Bug #1047: AMPI_Migrate doesn't PUP info key/val and name strings correctly
The above commit resolves the issue for C AMPI programs, but there are deeper issues for Fortran AMPI programs: https... Sam White
03:11 PM Bug #1078: AMPIF doesn't handle Fortran strings correctly
Abandoned commit #1160 is a hacky workaround for this issue in AMPI_Migrate for Fortran programs: https://charm.cs.i... Sam White
03:09 PM Bug #1078 (Merged): AMPIF doesn't handle Fortran strings correctly
AMPIF currently has no way of correctly identifying the length of a Fortran string that is passed into it, since Fort... Sam White
06:20 AM Bug #1048: Verbs on Bridges at PSC crashes or hangs
Ping. Jim Phillips
10:16 PM Bug #1076: failure to exit after long run on PSC Bridges
This seems reproducible. 1M-step jobs exit fine, 10M-step jobs hang as above. Jim Phillips

05/15/2016

09:45 AM Bug #1076 (Rejected): failure to exit after long run on PSC Bridges
Long MPI run on 9 nodes of PSC Bridges failed to exit, hung until job was killed. Note the lack of "[Partition 0][No... Jim Phillips

05/14/2016

04:48 PM Feature #1060 (In Progress): AMPI compliance with MPI-3.1 standard
Sam White
03:48 PM Bug #1047 (Implemented): AMPI_Migrate doesn't PUP info key/val and name strings correctly
The problem was simple, we were not correctly pupping the NULL terminator that we explicitly add to MPI_Info key/valu... Sam White
03:13 PM Feature #967: AMPI support for MPI-3's MPI_Comm_create_group
It looks like in the current method of group creation for MPI_Comm_create() all members of the input communicator con... Sam White
09:42 AM Feature #1067 (Implemented): Make all AMPI entry methods [expedited]
https://charm.cs.illinois.edu/gerrit/#/c/1202/ Sam White
09:40 AM Revision 03a22afe (charm): AMPI #1067: Tag all AMPI entry methods with 'expedited' attribute
Change-Id: If68ac336f1d5fc14857ad16a7b558ebd01c8d30c Sam White

05/13/2016

11:33 AM Bug #1035: Idle PEs compete with comm thread for node queue lock
I tried the above variant, as well as adding a lock to restrict node queue access to a single consumer at a time. Nei... Jim Phillips

05/12/2016

05:58 PM Revision 8a32386f (charm): Remove deprecated network layers from the doxygen configuration file
Change-Id: I435d4f8c8b435dab99cbc20d6a1e57ab530c6ce7 Ronak Buch
02:56 PM Feature #1075 (Rejected): Migrate solaris builds from net to netlrts
See parent issue
net-sol or net-sol-x86_64, which ever you have access to
Phil Miller
02:55 PM Feature #1074 (Merged): Migrate net-linux-ppc to netlrts
See parent Phil Miller
02:55 PM Feature #1072 (Merged): Migrate net-linux-arm7 to netlrts
See parent Phil Miller
02:55 PM Feature #1071 (Merged): Migrate windows builds from net to netlrts
see parent Phil Miller
02:54 PM Feature #1070 (Closed): Migrate lagging 'net' builds to 'netlrts'
Based on net-linux-x86_64 and net-darwin-x86_64 vs their netlrts equivalents, the conv-mach files can be copied strai... Phil Miller
02:28 PM Feature #906: support ++mpiexec with poe
https://charm.cs.illinois.edu/gerrit/#/c/942/ Sam White
02:18 PM Feature #1067 (Merged): Make all AMPI entry methods [expedited]
AMPI does not use message priorities, so we can tag all of its entry methods as expedited to avoid the RTS checking t... Sam White
02:17 PM Bug #1035: Idle PEs compete with comm thread for node queue lock
Jim,
Bilge mentioned that you had implemented this and sent a patch to Antti-Pekka but it did not show benefit. Can...
Harshitha Menon
02:03 PM Projections Feature #515 (Implemented): Add new image output formats
Ronak Buch
01:50 PM Cleanup #1064: Replace std::maps with a better implementation
It looks like unordered_map is the clear winner in just about every case. Sam suggested using a typedef to use unorde... Eric Mikida
08:44 PM Cleanup #1064: Replace std::maps with a better implementation

Performance comparisons of map, unordered_map and ckhashtable: https://charm.cs.illinois.edu/private/tms/listlog.ph...
Vipul Harsh
08:33 PM Revision fe974b8f (charm): Delete archaic support/references to GCC3 and KAI compilers
Change-Id: I586a399239d5bd3f99715f3cba1d7f76d5147336 Phil Miller
08:30 PM Revision 4a1956a6 (charm): Delete machine layers for obsolete IA-64 processors
Change-Id: I14c0860a49faf0d5be4f80743f8c840e39b43a9d Phil Miller
08:30 PM Revision 8d97a886 (charm): Delete machine layers for 32-bit x86 processors on non-Linux OS's
Change-Id: I34652fff736eff446a2617fb89259d50b0e0173d Phil Miller
08:30 PM Revision 39643aa4 (charm): Delete machine layers for obsolete AIX/PPC systems
Change-Id: Iad5b12eac170d4844dfb84221795267c22e263f2 Phil Miller
08:30 PM Revision aa535b68 (charm): build: clear out references to many deleted machine layers
Change-Id: I5eee6741d2abd7c7adf475fdc7ec3018a4b3b87c Phil Miller
08:20 PM Cleanup #898 (Implemented): Drop Blue Gene P code
Doing this now. Phil Miller
08:19 PM Revision cd07e07b (charm): Delete obsolete support for Cell processors
Change-Id: I5acc6d978b30375aeb78ebb52e03f5c98f3ae99e Phil Miller
08:19 PM Revision f585db75 (charm): Delete obsolete Cray XT machine layers
Change-Id: I2b64a768669a20c7c8c63fd4a685a1e45b303308 Phil Miller
08:19 PM Revision be3db1d3 (charm): Cleanup #898: Delete obsolete Blue Gene/P machine layers
Change-Id: Iacda9ac40147c8a90584147bc899e7702e782b44 Phil Miller

05/11/2016

04:19 PM Bug #1062 (Implemented): userDrivenMode undefined in converse/lrts exit
https://charm.cs.illinois.edu/gerrit/#/c/1189/ Sam White
08:37 AM Bug #1062 (Merged): userDrivenMode undefined in converse/lrts exit
Several builds (net, netlrts, mpi, multicore}-{linux, darwin}-{smp} fail autobuild because the first tests fail to bu... Sam White
04:14 PM Cleanup #1064: Replace std::maps with a better implementation
Related: https://charm.cs.illinois.edu/redmine/issues/185 Sam White
04:08 PM Cleanup #1064 (Merged): Replace std::maps with a better implementation
During the 64bit merge, some one-off data structures (such as CkMigratableList) were deleted and replaced by std::map... Eric Mikida
04:12 PM Cleanup #1065 (In Progress): Create a more efficient caching structure for location lookup
In the 64 bit merge, spring cleaning was eliminated in order to clean up the code and remove the large CkLocRec class... Eric Mikida
02:29 PM Bug #1063 (New): Error in buddy assignment code for checkpointing
In void ArrayElement::init_checkpt() defined in src/ck-core/ckmemcheckpoint.C, array elements are assigned buddy PEs.... Eric Mikida

05/10/2016

04:17 PM Bug #1061: Generated unmarshalling code for [reductiontarget] inappropriately offsets into messag...
https://charm.cs.illinois.edu/gerrit/1188 Phil Miller
04:12 PM Bug #1061 (Merged): Generated unmarshalling code for [reductiontarget] inappropriately offsets in...
Originally reported to me by Robert Pavel via email.
In the case where the entry method argument was a single arra...
Phil Miller
04:13 PM Revision fc22294a (charm): AMPI cleanup: remove AMPI_COUNTER support
Change-Id: Ibd578e8052371eb4ba132a5c74328b15486b86d7 Sam White
12:50 PM Revision 9c9dac03 (charm): AMPI cleanup: remove unused bcast counter
Change-Id: I4b9b8ab4dbf241592dc181c10f23bd75faf6fa47 Sam White
12:48 PM Revision f265f364 (charm): AMPI: fix AMPI_Ibcast to be non-blocking
Change-Id: I6d8eba73814ca08a4eb7f4fd62e73fdefae64e6a Sam White
12:36 PM Revision 66d9c0d9 (charm): AMPI: Cleanup whitespaces and newlines
Change-Id: Ief60d8054480045b0f753ca8c24010af5ed012ca Sam White
10:32 AM Revision be8c1131 (charm): Extend interop code to support user driven Charm++ execution.
Change-Id: I610130518347fd4d239272039fe0c273dbae7d06 Nikhil Jain
10:28 AM Revision ea0feb41 (charm): AMPIF: Fix Fortran interfaces to some MPI_ functions
Change-Id: I433525e5e5beebbc6105421ba53f28f64df139a4 Sam White
10:28 AM Revision 7a21850a (charm): AMPI cleanup: use applyOp where applicable
Change-Id: I8a6bc8a04c06074e681e9e0ed6cb2f8fc94a724b Sam White
10:28 AM Revision 7f37b414 (charm): AMPIF: MPI_IN_PLACE support for reduce routines
Change-Id: I9aa92d855822ca0246c28064e1568ebcb7b809fe Sam White

05/09/2016

06:41 PM Revision a66aee1e (charm): AMPI: remove commented out code
Change-Id: Ieb4a0d261b4ed878eae58594a720e7ddcb7cd356 Sam White
11:45 PM Revision 0e2d6852 (charm): AMPI: remove ampirun.in script
Change-Id: Iae7fa3155163fccba09ebb93e20d6a85dd5512c8 Sam White
10:56 PM Bug #1050: AMPI collectives support for MPI_IN_PLACE
ampif.C needs to test if buffers are NULL and set them to MPI_IN_PLACE if so. Currently, this is only done for MPI_R... Sam White

05/08/2016

12:22 PM Feature #1060: AMPI compliance with MPI-3.1 standard
Running list of unimplemented routines:
MPI_Win_allocate_shared
MPI_Win_shared_query
MPI_Win_attach
MPI_Win_det...
Sam White
12:01 PM Feature #1060 (In Progress): AMPI compliance with MPI-3.1 standard
This is a top level task to track AMPI's progress toward complying with MPI-3.1's standard. Sam White
10:26 AM Feature #947: AMPI support for MPI-2's MPI_Type_* routines
Subarray (initial commit missing Fortran support and will break ROMIO unless we change ROMIO's configure): https://ch... Sam White

05/06/2016

11:36 AM Revision 4d115311 (charm): AMPI #947: Add support for MPI types indexed_block and hindexed_block
Change-Id: Ice33ce916398fc6fe8e24d1c01d304a2f0be2b43 Edward Hutter
10:28 AM Cleanup #1059 (New): Unify Data Collection in Charm++
Currently, things like load balancing, PICS, BigSim, and tracing (for Projections) all collect data in their own uniq... Ronak Buch
08:31 AM Cleanup #1058 (Merged): Make tracing warning about flushing logs more informative
Currently, Projections warns users at the end of a run in which it had to flush its logs, but it's not clear how to s... Sam White

05/05/2016

06:30 PM Projections Feature #1057 (In Progress): Just user event tracing
It would be useful to be able to use projections just for user event tracing. For very finegrained applications like ... Eric Mikida
03:57 PM Revision dcee5b1e (charm): added correct initializations of smp queue
Change-Id: Id75ec59e7e6b15d7c266838c04265ad834055ca8 Justin Miron
03:51 PM Revision 6e337c18 (charm): replace machine-common-core recieving queue with lockless mpmc queue
Change-Id: I6465e71a1a5f6b9fc5b9f682d1925db16f7523e8 Justin Miron
02:41 PM Revision 6ac0e598 (charm): one more call change to mpsc
Change-Id: Icc521f3417ed68ca24a0feda9af8c82dd8a0b636 Justin Miron
02:27 PM Revision 0839331f (charm): replaced one send queues with mpsc lockless queues
Change-Id: I482ce9104599af7cd1dfae9c1bafedebf697b922 Justin Miron
02:21 PM Bug #1038: Add support for srun to charmrun script for gni layer
From Core: Attack problem by just failing over to srun if aprun and mpirun both fail to run Michael Robson
02:08 PM Revision 2930c94f (charm): mpsc and mpmc lockless queues added to pcqueue.h
Change-Id: I6808e1b19a989d16fe25b164f45fcb79fa2feff8 Justin Miron
01:54 PM Feature #1056: make trivial SDAG nodegroup entries reentrant
I would lean more toward this sort of thing being an explicit developer-directed choice. Harshitha is adding @task@ d... Phil Miller
01:08 PM Feature #1056 (New): make trivial SDAG nodegroup entries reentrant
Writing the following in the .ci file results in an unnecessarily and surprisingly exclusive entry method:
nodegr...
Jim Phillips
11:49 AM Revision fd096388 (charm): LBDatabase: make ResumeClients an entry method
Shrink-expand needs to create a checkpoint callback to ResumeFromSync so
that restarted processes can properly restar...
Ralf Gunter
11:45 AM Bug #948 (In Progress): CkExit() crashes called from main chare constructor with `-tracemode proj...
Posted patch needs substantial further work. Phil Miller

05/04/2016

03:10 PM Bug #1054 (Closed): Darwin builds fail on Wit
Sam White
03:09 PM Bug #1054 (Resolved): Darwin builds fail on Wit
The problem seems to be that autobuild doesn't clean up after itself on Wit. Workaround is to kill its leftover proce... Sam White
09:45 AM Cleanup #1052 (Merged): AMPI error handling is inconsistent, broken in places
Sam White
09:45 AM Cleanup #1053 (Merged): Remove copyglobals-specific code in AMPI_Alltoallv and AMPI_Ialltoallv
Sam White
09:41 AM Bug #1051 (Merged): Some AMPI collectives on MPI_COMM_SELF are broken
Sam White
09:25 AM Revision 0b51bccf (charm): AMPI: cleanup CDECL function declarations
Change-Id: I95c66d5698540d7d5a5d32ee8e2aea38c71f415b Sam White
09:25 AM Revision c38d73a8 (charm): AMPI: cleanup AMPIAPI calls
Change-Id: I42e6f67015eb7535cc03cc51daaee1a2ed29e3ff Sam White
09:25 AM Revision 202de7e0 (charm): AMPI cleanup #1053: remove copyglobals-specific code in Alltoallv
Change-Id: Ibde12d24f2e374c947f05c478cf672597002c503 Sam White
09:21 AM Revision ca014618 (charm): AMPI #1051: Consistent handling of collectives to MPI_COMM_SELF
Non-blocking collectives need to set the request ptr, and some
collectives were not copying the send buffer to the re...
Sam White

05/03/2016

02:36 PM Revision 8a4696de (charm): empty queue check
Change-Id: I712dd8340581c2111ad6778b7dccd66aa880e01d Justin Miron
02:21 PM Revision 51bde7ce (charm): fixed bad header included
Change-Id: I4045aa31f501bec41953022c29b0a5d50a7e0b94 Justin Miron
02:02 PM Revision ef085482 (charm): mpmc pcqueue for machine-common-core
Change-Id: I3ac91c0cdf2a0d42e04bf110f7e275bb41130651 Justin Miron
09:06 AM Revision f8ff1af8 (charm): AMPI cleanup: AMPI_Finalized has wrong AMPIAPI string name
Change-Id: Ic049d8267340387fd554796847cb4c2db8839827 Sam White
09:05 AM Revision f19c6ed4 (charm): AMPI cleanup #1052: Consistent error handling for send/receive buffers,
MPI_IN_PLACE, and Inter-communicators
Change-Id: Ibd59a33d849f71a082094a903de97560c62616d7
Sam White
08:29 AM Bug #1054 (Closed): Darwin builds fail on Wit
The Darwin tests in autobuild are failing to build on Wit.... Sam White

05/02/2016

02:38 PM Documentation #884 (Merged): document physical node API
Sam White
02:25 PM Support #989 (Resolved): Tutorial codes are not accessible
It looks like this has been fixed Sam White
02:23 PM Bug #1049 (Merged): Autobuild tests with +p>5 hang on Darwin
Sam White
12:18 PM Bug #374: Use new topology ordering function on BGQ
https://charm.cs.illinois.edu/gerrit/#/c/422/ Sam White
10:17 AM Revision 070b3adb (charm): Cleanup: remove deprecated +nodesize option in kNeighbor pxshm tests
Change-Id: Iee8d076d7a3085856a5be52f549cc745dc126e8d Sam White

05/01/2016

10:17 AM Bug #1048: Verbs on Bridges at PSC crashes or hangs
Sam White wrote:
> ChaNGa is seeing crashes/hangs for some runs on verbs, which are fixed with this commit: https://...
Jim Phillips

04/30/2016

09:37 PM Cleanup #1053 (Implemented): Remove copyglobals-specific code in AMPI_Alltoallv and AMPI_Ialltoallv
https://charm.cs.illinois.edu/gerrit/#/c/1170/ Sam White
09:33 PM Cleanup #1053 (Merged): Remove copyglobals-specific code in AMPI_Alltoallv and AMPI_Ialltoallv
I don't see any reason why AMPI_Alltoallv and AMPI_Ialltoallv should have some special support for copyglobals, when ... Sam White
09:23 PM Bug #1051 (Implemented): Some AMPI collectives on MPI_COMM_SELF are broken
https://charm.cs.illinois.edu/gerrit/#/c/1169/ Sam White

04/29/2016

06:21 PM Revision 7d62ba6c (charm): Bug #1049: Decrease +p params in tests to avoid autobuild hangs on
Darwin because of ssh connection limits
Change-Id: I1f6d820879a162f7cbd3ea525a708a38a4befdb9
Sam White
03:14 PM Cleanup #1052 (Implemented): AMPI error handling is inconsistent, broken in places
https://charm.cs.illinois.edu/gerrit/#/c/1165/ Sam White
03:06 PM Cleanup #1052 (Merged): AMPI error handling is inconsistent, broken in places
AMPI's error handling, particularly for collectives, is inconsistent and broken in places.
For instance, MPI_Scatter...
Sam White
03:01 PM Bug #1051 (Merged): Some AMPI collectives on MPI_COMM_SELF are broken
All non-blocking collectives need to set the request to MPI_REQUEST_NULL before copying the buffer and returning MPI_... Sam White
02:55 PM Bug #1050 (Merged): AMPI collectives support for MPI_IN_PLACE
AMPI currently implements MPI_IN_PLACE only for MPI_Reduce, MPI_Ireduce, MPI_Allreduce, MPI_Iallreduce, MPI_Reduce_sc... Sam White
12:37 PM Revision dc80f309 (charm): one more lock added
Change-Id: I6889a2751b657e80b25a15fafec97ffda2bbb7fe Justin Miron
12:29 PM Revision 9193afe0 (charm): add one lock to machine core
Change-Id: I25760a0be48e455c602c0a7b0c6519d476745683 Justin Miron
12:20 PM Revision 483933cc (charm): remove locks from machine-common-core.c
Change-Id: I9c6fc6a74340bf7c697335fe0b6dbb3508fa6ce5 Justin Miron
09:47 AM Bug #1043 (Rejected): ChaNGa crashes on verbs for dwf.6144 on 128 nodes
Same issue as 1042. Sam White
09:47 AM Bug #1049 (Implemented): Autobuild tests with +p>5 hang on Darwin
The problem is that on Wit the maximum number of ssh connections that can be established at once is 5.
kNeighbor a...
Sam White
08:46 AM Bug #1049 (Merged): Autobuild tests with +p>5 hang on Darwin
kNeighbor and possibly other tests use +p >= 6, which hangs on Wit. Sam White
08:45 AM Bug #1027 (Merged): Chkpt tests hang on Darwin in autobuild
Sam White
08:30 AM Bug #1048: Verbs on Bridges at PSC crashes or hangs
ChaNGa is seeing crashes/hangs for some runs on verbs, which are fixed with this commit: https://charm.cs.illinois.ed... Sam White
08:28 AM Bug #1046 (Merged): AMPI's Info keys/values should be stripped of leading/trailing whitespace
Sam White

04/28/2016

06:05 PM Revision a8dc734c (charm): AMPI 1046: Strip leading and trailing whitespaces from MPI_Info keys
and values to conform to MPI standard
Change-Id: I35a229e51262774ca112505c1462847f1e7dd932
Sam White
05:17 PM Revision ab6ce273 (charm): AMPIF cleanup: Internal calls should use AMPI instead of MPI calls
Change-Id: I18fcaad4acdf1a4c8ce0b59a71b21160c2e44090 Sam White
02:24 PM Bug #1048 (Rejected): Verbs on Bridges at PSC crashes or hangs
Megatest sometimes hangs with:... Bilge Acun
12:30 PM Bug #1047 (In Progress): AMPI_Migrate doesn't PUP info key/val and name strings correctly
Sam White
12:28 PM Bug #1047: AMPI_Migrate doesn't PUP info key/val and name strings correctly
Building AMPI with Address Sanitizer gives the following report with a warning about possible false positives:
<pr...
Sam White
11:05 AM Bug #1047: AMPI_Migrate doesn't PUP info key/val and name strings correctly
A simple test with a for-loop containing a call to AMPI_Migrate followed by a barrier shows the MPI_Info value that w... Sam White
08:44 AM Revision 0efd82bd (charm): new machine.c changes for lockless queue
Change-Id: Ie3339cef745488ed11f5fa2cbef451c2a8e0d23c Justin Miron

04/27/2016

06:01 PM Cleanup #165: object location services: separate location caching from msg delivery and buffering
Reassign and make this non-private, since there's nothing to be scooped here. Phil Miller
02:29 PM Bug #1047 (Implemented): AMPI_Migrate doesn't PUP info key/val and name strings correctly
The issue seems to be some sort of memory corruption, but it does not reproduce under valgrind. Comparing only the pa... Sam White
02:23 PM Bug #1047 (Merged): AMPI_Migrate doesn't PUP info key/val and name strings correctly
AMPI_Migrate sometimes fails to compare strings correctly. This didn't show up in autobuild or Jenkins since it seems... Sam White
02:12 PM Bug #1046 (Implemented): AMPI's Info keys/values should be stripped of leading/trailing whitespace
Implemented fix here: https://charm.cs.illinois.edu/gerrit/#/c/1158/ Sam White
02:11 PM Bug #1046 (Merged): AMPI's Info keys/values should be stripped of leading/trailing whitespace
According to the standard, MPI_Info_{set, get, delete, get_valuelen} should strip the leading and trailing whitespace... Sam White
01:29 PM Revision 672d1b28 (charm): revert machine.c, set ONE_SEND_QUEUE 1
Change-Id: Ie9c3d74bd0aba95bfdefa6b5f341886fd3514a5e Justin Miron
11:50 AM Bug #1045 (Merged): Array pointers in messages can be overwritten by generated code in the messag...
https://charm.cs.illinois.edu/private/tms/listlog.php?param=1111
Upcoming changes in GCC 6 optimizer capabilities ...
Phil Miller
08:32 PM Revision bed0fe43 (charm): fix on valid_q push
Change-Id: I3b159bb0a19dd2d37e54a5eb121b78534aabae91 Justin Miron
08:23 PM Revision 2365c4bc (charm): fix
Change-Id: I60af712c41be35ed14398ded353ffbca843790f6 Justin Miron
08:12 PM Revision 08360cea (charm): extra c
Change-Id: Ib8c50aa93e06ed714e5bc0173999e506a2037773 Justin Miron
08:04 PM Revision 4c0a35cd (charm): lockless queue changes
Change-Id: If2148119984a4032b8c923640179a8e8eae5fa8b Justin Miron
07:14 PM Bug #1041 (In Progress): AMPI's struct datatype is broken
The issue is in CkDDT_Struct::serialize not handling different sizes/counts of types in structs properly. Sam White

04/26/2016

03:23 PM Revision c3e3e404 (charm): Adds the output of data by default for PICS
Reuben Tan

04/25/2016

03:35 PM Revision 6276060b (charm): Bug 1027: Decrease +p param in chkpt tests to avoid autobuild hangs on...
Change-Id: Iac8f21db4dd6a4ae20362b8cd21a35c2723577ca Sam White
01:24 PM Bug #1007: charmxi: redundant semicolons lead to erroneous error messages
I ran into a similar issue when compiling amr with latest branch of charm
to reproduce
git clone charmgit:users...
Harshitha Menon
11:36 AM Revision f3dba012 (charm): Add hwloc-1.11.2 to src/hwloc
Project: https://www.open-mpi.org/projects/hwloc/
Archive: https://www.open-mpi.org/software/hwloc/v1.11/downloads/hw...
Steve Hoelle
11:36 AM Revision 8d83e114 (charm): Add hwloc to charm build
Change-Id: I95db11047dbb4819829d4c312ef5f351a8e6b6d3 Steve Hoelle
11:36 AM Revision d65ab62b (charm): Add Readme
Change-Id: I0dc44f8fca660b5dbc1a9cfc60fae00588f1da80 Steve Hoelle
11:23 PM Bug #1027 (Implemented): Chkpt tests hang on Darwin in autobuild
Decrease the +p parameter in the Makefile so that we don't hit the ssh connection limit on Wit but we still test chkp... Sam White
 

Also available in: Atom