Project

General

Profile

Activity

From 10/31/2016 to 11/29/2016

11/29/2016

05:10 PM Bug #1299 (Merged): SDAG_RTS is missing object attribution in Projections traces
Sam White
03:44 PM Documentation #1256: Document '-tracemode perfReport' in the manual
Assigning to Ronak because tracing Sam White
03:37 PM Bug #1277 (Merged): Multiple shrink expand requests lead to segmentation fault
Sam White
03:36 PM Feature #1295 (Merged): Reduce overheads of message creation in the OpenMP integration by adjusti...
Sam White
01:35 PM Feature #1295 (Implemented): Reduce overheads of message creation in the OpenMP integration by ad...
https://charm.cs.illinois.edu/gerrit/#/c/1982/ Sam White
03:28 PM Feature #1309: Use CkMulticast for collectives on AMPI subcommunicators
A couple issues we'd like to see resolved in CkMulticast first:
1. Distributed creation (reducing the cost of recrea...
Sam White
02:50 PM Feature #1184: SMP-safe CmiReference and CmiFree
Reassigning to Vipul since this is mostly targeted at broadcasts.
See the note in conv-config.sh about CMK_BROADCA...
Sam White
09:26 AM Feature #1098: Better CMake support
Other calls for CMake support (based on GMail searching my mail archive):
https://lists.cs.illinois.edu/lists/arc/ch...
Phil Miller

11/28/2016

05:55 PM Bug #1316 (Merged): AMPI Isomalloc/Mempool abort with PUP sizing mismatch
When built without production:... Sam White
05:08 PM Cleanup #1012: Replace deprecated auto_ptr with unique_ptr in generated SDAG code
The fix linked above depends on libstdc++ corresponding to GCC 4.4+, but that should be fine now. We've merged some o... Phil Miller
04:56 PM Cleanup #1012 (Implemented): Replace deprecated auto_ptr with unique_ptr in generated SDAG code
https://charm.cs.illinois.edu/gerrit/#/c/2002/ Sam White
04:45 PM Bug #1308 (Implemented): AMPI message sequencing queues use O(n^2) memory
Sam White
04:05 PM Bug #1307: AMPI_Comm_free should free the ampi instance
Here's the failure from running tests/ampi/megampi on the above patch:... Sam White
12:36 PM Bug #1307 (In Progress): AMPI_Comm_free should free the ampi instance
AMPI-level implementation here, but needs underlying fix in Charm++'s location manager to support bound array deletio... Sam White
03:46 PM Cleanup #1315 (New): examples/charm++/jacobi*d are non-exemplary HPC code, using nested arrays
... Phil Miller
03:38 PM Cleanup #1314 (In Progress): Replace widespread dynamic allocated arrays with std::vector
There are loads of places in the runtime system and our examples where the code manually allocates an array member us... Phil Miller
12:49 PM Bug #1313 (New): Support MPI_Info's on a per-{comm,win} basis
Implement MPI_Comm_{set,get,dup_with}_info and MPI_Win_{set,get}_info routines. These are currently no-ops.
Suppor...
Sam White

11/27/2016

02:09 PM Bug #1312: Deleting an array disables reclamation for all arrays bound to that location manager
Interesting that we independently stumbled on this potential problem from opposite directions almost simultaneously. Phil Miller
02:08 PM Bug #1312: Deleting an array disables reclamation for all arrays bound to that location manager
I recently opened bug #1307 for AMPI_Comm_free (currently a no-op). I made a quick effort at implementing it, just ca... Sam White
01:53 PM Bug #1312: Deleting an array disables reclamation for all arrays bound to that location manager
Actually, it looks like element deletion may destroy its corresponding CkLocRec and trigger this path regardless of w... Phil Miller
01:47 PM Bug #1312 (Merged): Deleting an array disables reclamation for all arrays bound to that location ...
https://charm.cs.illinois.edu/gerrit/1114
The above change disabled sending @reclaimRemote()@ messages to location m...
Phil Miller

11/26/2016

11:36 AM Bug #1257: AMPI missing support for PSCW synchronization of RMA
MPI_Win_post, MPI_Win_start, MPI_Win_wait, and MPI_Win_complete are also no-op implementations current in AMPI, along... Sam White
10:36 PM Bug #1002: Changa's Final CkWaitQD() hangs after AtSync deletion-counting changes
The original (somewhat incorrect) changa code now runs without hanging with the latest charm branch. However, there ... Thomas Quinn

11/25/2016

03:32 PM Bug #1002: Changa's Final CkWaitQD() hangs after AtSync deletion-counting changes
Pinged change 1115 to ChaNGa, hopefully Tom can clear this up for us. Phil Miller

11/23/2016

03:20 PM Bug #1310: Isomalloc hangs in CpvInitialize during startup for multicore/smp builds
Updated 12-20-16: the builds affected by this issue are:... Sam White
01:49 PM Bug #1310: Isomalloc hangs in CpvInitialize during startup for multicore/smp builds
netlrts-linux-x86_64-smp is broken in the same way. Sam White
12:32 PM Bug #1310 (Merged): Isomalloc hangs in CpvInitialize during startup for multicore/smp builds
On a multicore-linux64 build, an AMPI program hangs in Charm's startup when using isomalloc. This is reproducible on ... Sam White
01:23 PM Cleanup #1311 (Closed): Align XL-specific conditional compilation TRAM to relevant versions
We need testing of the XL compiler on Crest to see whether it references standard libraries that contain @std::unorde... Phil Miller
01:16 PM Feature #185 (Rejected): provide backup implementations of std::unordered containers
We don't care about supporting platforms that are so far behind C++11 any more. The only one of concern was XL (possi... Phil Miller
01:15 PM Cleanup #1064 (Merged): Replace std::maps with a better implementation
https://charm.cs.illinois.edu/gerrit/1978
I'll open a follow-up bug about resolving which XLC versions/platforms we ...
Phil Miller
12:59 PM Bug #1299 (Implemented): SDAG_RTS is missing object attribution in Projections traces
https://charm.cs.illinois.edu/gerrit/#/c/1995/ Sam White
12:59 PM Bug #1294: Darwin SMP failure in exit
This is failing every night in autobuild. Sam White
10:37 AM Revision fe4b831d (charm): Bug #1239: Cleanup reduction uses in ck-perf
Change-Id: I430e679505a9e0adfd794972cb922926bb3cc375 Karthik Senthil

11/22/2016

03:24 PM Revision 1d9930a8 (charm): Loop-level parallelism section in the manual is modified to add OpenMP
integration
Change-Id: I174f532032c16dffc3f1ccc757eb5b24ea42b2a4
Seonmyeong Bak

11/21/2016

04:26 PM Bug #1308: AMPI message sequencing queues use O(n^2) memory
Use a std::{map,unordered_map} instead of dense array: https://charm.cs.illinois.edu/gerrit/#/c/1994/ Sam White
12:30 AM Bug #1308 (In Progress): AMPI message sequencing queues use O(n^2) memory
The first minor part of this, to allocate the queue with length proportional to the size of the actual communicator i... Sam White
12:28 AM Bug #1308 (Merged): AMPI message sequencing queues use O(n^2) memory
AMPI defines a message sequencing queue per VP per comm. Each queue has 'n' entries, where 'n' is the size of MPI_COM... Sam White
01:36 PM Feature #1309 (New): Use CkMulticast for collectives on AMPI subcommunicators
Currently, AMPI implements all communicators as their own separate chare array instances, and all collectives on thos... Sam White
01:30 PM Feature #1292 (In Progress): Alternatives to linked-list message queues in AMPI
Sam White
01:30 PM Feature #1292: Alternatives to linked-list message queues in AMPI
Pulled Cmm into AMPI as 'Amm' in order to support wildcard values that are not -1 (MPI_ANY_TAG is now MPI_TAG_UB_VALU... Sam White
09:59 PM Revision fce7c243 (charm): Remove all remaining RTH code
Change-Id: I10f72e02ef09b043af5621a01f7b501a3060c8b1 Eric Mikida
06:59 PM Bug #1307: AMPI_Comm_free should free the ampi instance
Also, this is important for AMPI users of MPI_Comm_split_type, because those communicators can be invalidated by migr... Sam White
06:38 PM Revision ac84e2be (charm): Change std::maps to std::unordered_maps in CkLocMgr when possible
Change-Id: I247671fd24799cb25a9533a64a21b543685d5cd5 Eric Mikida
06:38 PM Revision 974a6d34 (charm): Disable compilation/execution of tests & examples that won't compile w...
Change-Id: I5ebeb92f6b571f801eb37dab5533ece526de60c2 Phil Miller
06:37 PM Revision 6b4a9e9c (charm): Flag builds using BG XLC
It doesn't support std::unordered_map and fails to compile TRAM.
We want to disable those features and/or tests there...
Phil Miller

11/20/2016

04:46 PM Bug #1307 (Merged): AMPI_Comm_free should free the ampi instance
Right now, AMPI_Comm_free is a no-op. This could be bad for memory usage of applications that create and destroy comm... Sam White
04:40 PM Bug #961 (In Progress): AMPI support for cancelling recv requests/statuses
Sam White
04:40 PM Feature #967 (Implemented): AMPI support for MPI-3's MPI_Comm_create_group
Sam White
04:39 PM Bug #1160 (Merged): AMPI intercommunicator creation is broken
Sam White
04:35 PM Bug #1191 (Merged): AMPI is missing MPI_Type_*_{attr,keyval} routines
Sam White
04:26 PM Cleanup #1012: Replace deprecated auto_ptr with unique_ptr in generated SDAG code
After 6.8.0 we are going to require full C++11 support from compilers, so this can be changed to unique_ptr after the... Sam White
04:24 PM Bug #1299: SDAG_RTS is missing object attribution in Projections traces
Related: https://charm.cs.illinois.edu/redmine/issues/805 Sam White
01:35 PM Revision ebe2ba02 (charm): net(lrts) and verbs: fix annoying valgrind uninitialized value error
Change-Id: Iecc0ccfb4b58bd1f868aa36d0d4244a367c0bb0f Phil Miller
01:21 PM Revision 7c9b0c92 (charm): Projections tracing: Record array element index from SDAG statements w...
Change-Id: Ia56ef2131eb173f654943456bd05af7f8cc200de Phil Miller (Charmworks)
01:21 PM Revision 338ecb27 (charm): Projections tracing: Remove unused argument to getProjectionID
Change-Id: I6f5f040b301b1918cee628f66518fd1f6d1c9c3c Ronak Buch
01:21 PM Revision 97f76bba (charm): charmxi: set entry method pointer on SDAG constructs recursively
Change-Id: I3aa53fd26a3f0b1f181f7686bdea5988a72bfd66 Phil Miller
01:18 PM Revision 612140f1 (charm): MetaLB: record strategy cost regardless of debug setting
Change-Id: Ibea6e92b7212784ee5c497ba2cd308bd90135085 Phil Miller (Charmworks)

11/19/2016

11:13 PM Bug #1226 (Merged): Support namespaces in generated code from TRAM [aggregate] entry method attri...
Phil Miller

11/16/2016

02:39 PM Bug #1306 (New): PythonCCS-client needs to be compiled without CONVERSE
libs/ck-libs/pythonCCS/PythonCCS-client.C is meant to be used by non-charm programs, but commit a316ff3 made a change... Thomas Quinn

11/14/2016

04:47 PM Feature #1305 (New): Implement MPI-R debugging hooks in RTS startup
Phil Miller
04:46 PM Feature #1304 (New): Charmrun: implement MPI-R debugging hooks
Phil Miller
04:46 PM Feature #1303 (New): Implement MPI-R debugging hooks to support Allinea DDT and Rogue Wave Totalview
http://mpi-forum.org/docs/mpir-specification-10-11-2010.pdf
If we implement the minimal elements of the MPI-R spec...
Phil Miller
11:23 AM Documentation #1302 (Merged): pass arbitrary arguments to mpiexec via charmrun
It would potentially be useful to be able to pass arguments to mpiexec via the charmrun command line when necessary o... Jim Phillips

11/11/2016

09:19 AM Bug #1301 (Merged): AMPI_Issend can complete before message is matched
ampi::generic() currently sends an acknowledgement regardless of whether the message was matched or not.
The standar...
Sam White
06:58 PM Feature #1300: Integrated OpenMP should co-exist with MPI interop
OK. I'll test it and think of how to make the integration compatible with this environment. Seonmyeong Bak

11/10/2016

05:40 PM Feature #1300: Integrated OpenMP should co-exist with MPI interop
It should be straightforward to modify the current MPI interop example we already have in the repo, to add OpenMP pra... Phil Miller
03:43 PM Feature #1300: Integrated OpenMP should co-exist with MPI interop
Can I get any example to test my patch for this issue? Seonmyeong Bak
03:07 PM Feature #1300 (In Progress): Integrated OpenMP should co-exist with MPI interop
I expect that if we were to compile, link, and run a mixed Charm++/MPI application with OpenMP pragmas in both parts,... Phil Miller
12:01 PM Revision a28868e2 (charm): AMPI: match messages w/o comm since it is already done per comm
Change-Id: I267bf9171c9da8383b09583004fbe2e6bfaf683f Sam White
11:34 AM Bug #1299 (Merged): SDAG_RTS is missing object attribution in Projections traces
We recently merged a change to attribute user-defined blocks of work in SDAG to the objects running them. However, th... Phil Miller
11:12 PM Revision 8c4dd7bf (charm): Add documentation for the new TRAM interface
Change-Id: I641a379f96b492e4aaaef223ed1707651dff72bd Eric Mikida
10:42 PM Bug #1097 (New): Attributes are not duplicated in MPI_Comm_dup
Sam White
06:41 PM Revision 0d7cb2bb (charm): Feature #1060: Implementation of MPI_Comm_dup_with_info
Change-Id: I235fe476099618d2d5e988a058a7b232bcaa1fdf Karthik Senthil

11/09/2016

01:53 PM Bug #1267: Integrate METIS graph partitioning library patched for global variable privatization
When you add METIS into the makefiles, please include it as a dependency of ParFUM. Phil Miller
01:38 PM Bug #1267: Integrate METIS graph partitioning library patched for global variable privatization
Please go ahead with the re-integration there, with suitable Makefile changes. I'll add a couple follow-on tasks. Phil Miller
01:27 PM Bug #1267: Integrate METIS graph partitioning library patched for global variable privatization
h3. Is it okay to add the library to src/libs/ck-libs with appropriate changes in the Makefile? Kavitha Chandrasekar
01:40 PM Feature #1298 (Merged): Build MetisLB by default
Since we can now ship METIS in-tree, we can build MetisLB unconditionally. Phil Miller
01:39 PM Feature #1297 (Merged): Patch METIS to privatize global variables
Phil Miller
12:37 PM Feature #1295 (Merged): Reduce overheads of message creation in the OpenMP integration by adjusti...
OpenMP integration into Charm++ split each Chare's work into fine-grained OpenMP tasks.
Because most Charm++ applic...
Seonmyeong Bak
12:20 PM Bug #1045 (Merged): Array pointers in messages can be overwritten by generated code in the messag...
It'll be nice to see this get through autobuild, but the fix is in. Phil Miller
12:11 PM Bug #1272 (Implemented): Node GPU Manager hangs on exit
Michael Robson
09:07 AM Bug #1294: Darwin SMP failure in exit
The failure is consistently reproducible Sam White
09:03 AM Bug #1294 (Merged): Darwin SMP failure in exit
Autobuild shows a failure for netlrts-darwin-x86_64-smp in tests/charm++/sdag/case/, seemingly during exit:... Sam White
08:26 AM Feature #1292: Alternatives to linked-list message queues in AMPI
Minor enhancement to our current CmmGet: there is no need to match on comm, since we already do our matching on a per... Sam White

11/08/2016

05:51 PM Bug #1240: MPI_COMM_SELF handling issues in AMPI
The MPI_Comm_split failures have been observed in XPACC. Sam White
05:04 PM Revision 7b45f719 (charm): Minor fixes
Change-Id: Ia1c69e075a17095ba450d0f0fe472a7b24c52f57 Seonmyeong Bak
04:38 PM Bug #1189: AMPI resumeOnRecv should be a property of the thread, not each comm
Nikhil's fix for this issue is here: https://charm.cs.illinois.edu/gerrit/#/c/1867/ Sam White
03:33 PM Support #442 (Closed): Performance regressions in Episimdemics from 6.5-6.6
There's now a much more intensive effort within the development team to optimize fine-grain application performance. ... Phil Miller
03:14 PM Bug #901 (Merged): Threads awoken by CthAwaken don't let Projections trace back to the event that...
Phil Miller
03:07 PM Projections Bug #1293 (Closed): Projections summary shows 0s for EP Exec Time for a single core large ChaNGa ...
EPExecTime has a large number of zeros.
To reproduce this issue:
1) Compile charm with enable-tracing.
./buil...
Nitin Bhat
02:23 PM Revision 68990667 (charm): Cleanup: update Converse msgmgr to use modern C function prototypes
Change-Id: Ibaa6a7351b667dcd5fb942c74724f026d0a96140 Sam White
02:23 PM Revision c057d2fd (charm): Cleanup: make timing of task stealing conditional on tracing
Change-Id: I6329d74b3e4d006248e8f62cbe220794933a2ac4 Sam White
12:48 PM Feature #1292 (Closed): Alternatives to linked-list message queues in AMPI
Trade-offs in MPI message queue designs have been the subject of many papers.
Currently, AMPI relies on Converse's C...
Sam White
12:27 PM Bug #1045 (Implemented): Array pointers in messages can be overwritten by generated code in the m...
Subsequent fixes pushed, awaiting review. Phil Miller
06:27 PM Bug #1045: Array pointers in messages can be overwritten by generated code in the message class c...
On examination, there is no case where the flags we're detecting in configure should actually be passed to the native... Phil Miller
12:10 PM Revision 5d10d7ed (charm): Bug #1045 Follow-up: ignore configure-detected compiler options when t...
Change-Id: Ia05319060372364a51b29edcdc1456c445eead84 Phil Miller
12:10 PM Revision aaffa833 (charm): megatest/Make.depends: Don't regenerate automatically
Macs with older toolchain versions (gcc not aliasing to hacked-up clang)
don't support the C++11 flag to their ancien...
Phil Miller
12:10 PM Revision d177c530 (charm): Bug #1045 follow-up: recognize IBM XL compilers's idiosyncratic warnin...
This prevents the configure script from incorrectly passing the
GCC-specific flag -fno-lifetime-dse to a compiler tha...
Phil Miller
11:40 AM Bug #1272: Node GPU Manager hangs on exit
ping Thomas Quinn
10:32 AM Revision 783127f0 (charm): AMPI #947: no-op implementation of MPI_Type_***_attr and MPI_Type_***_...
Change-Id: Ia8504e149d43178201ec43bfe9d0bb81c6cbf994 Edward Hutter
09:20 PM Revision a678a96f (charm): Remove unnecessary codes
Change-Id: I604094bbb1d5496e3ad2b7cfd9c1dd8c35aa0b59 Seonmyeong Bak
09:16 PM Revision 49305263 (charm): Remove unnecessary codes
Change-Id: I604094bbb1d5496e3ad2b7cfd9c1dd8c35aa0b59 Seonmyeong Bak
08:14 PM Bug #1276: net-cygwin: multiple testing failures
The list of substasks reporting individual failures using @--with-production --enable-error-checking@ should be exhau... Phil Miller
07:20 PM Bug #1276: net-cygwin: multiple testing failures
In a non-production build, all seems to run successfully. That contradicts autobuild, but I haven't manually tested @... Phil Miller
08:13 PM Bug #1291 (Rejected): net-cygwin crashes in examples/bigsim/sdag
Phil Miller
08:12 PM Bug #1290 (Rejected): net-cygwin crashes in examples/bigsim/emulator
Phil Miller
08:01 PM Bug #1289 (Rejected): net-cygwin crashes in megampi
Phil Miller
07:41 PM Bug #1287: net-cygwin crashes in megatest/{queens,tempotest,synctest} under --with-production
Changing the thread package to qt or back to uJcontext doesn't seem to help these. There's something else wrong, perh... Phil Miller
06:32 PM Bug #1287 (Rejected): net-cygwin crashes in megatest/{queens,tempotest,synctest} under --with-pro...
Phil Miller
07:17 PM Bug #1288 (Rejected): net-cygwin: megacon crashes when blkinhand and posixth are enabled
I cut the enabled tests down to just blkinhand and posixth in megacon.c, and got the following failure:... Phil Miller
06:31 PM Bug #1270 (Merged): net-cygwin crashes in threaded entry methods
Phil Miller

11/07/2016

04:52 PM Projections Bug #1283 (Closed): Usage Profile is blank and shows incorrect selection range for application ru...
This happened because there was a .projrc file in Nitin's logs that had cached an erroneous time. This was fixed in ... Ronak Buch
12:43 PM Revision 3e1aa789 (charm): Bug #1226: Fix [aggregate] entry method pup generation in namespaces
Change-Id: I0fab73fc5c235487778c82e3b252e1f844f3e40a Eric Mikida
12:43 PM Bug #1226 (Implemented): Support namespaces in generated code from TRAM [aggregate] entry method ...
https://charm.cs.illinois.edu/gerrit/#/c/1959/ Phil Miller
12:11 PM Bug #1269 (New): multicore-win64, netlrts-win-x86_64: megampi fails when +vp < +p
Still failing on multicore-win64 in autobuild
https://charm.cs.illinois.edu/autobuild/old.2016_11_07__01_01/multicor...
Phil Miller
07:29 PM Revision c33ed0e6 (charm): Bug #1160: Fix intercommunicator creation and pt2pt messaging in AMPI
Change-Id: Iab2af46d0be439caea387c58fbe4d82b31a2818b Edward Hutter

11/06/2016

04:00 PM Bug #1282: win64 builds with tracing enabled are broken
Not release critical - we're treating Windows as a deployment platform for NAMD, not a development platform. In that ... Phil Miller
11:10 AM Bug #1282 (Merged): win64 builds with tracing enabled are broken
The autobuild tests are run with --with-production so tracing is not built.
PICS doesn't compile on MSVC. There seem...
Sam White
01:41 PM Revision 2a90145b (charm): pls
Change-Id: I1449c39c69cf8307e8425f6153ed0e8c712b0890 Justin Miron
12:47 PM Bug #1269 (Merged): multicore-win64, netlrts-win-x86_64: megampi fails when +vp < +p
Fixing a race condition in trying to create array elements for not-yet-constructed managers seems to have solved this... Sam White
11:27 AM Bug #1269: multicore-win64, netlrts-win-x86_64: megampi fails when +vp < +p
It seems to work for me with current master on dexterity, using './build AMPI multicore-win64 -j8 --with-production -... Sam White
11:19 AM Projections Bug #1283: Usage Profile is blank and shows incorrect selection range for application run with +s...
The leanmd.0.sum file seems to have been generated correctly. It is almost the same as the leanmd.0.sum generated wit... Nitin Bhat
11:16 AM Projections Bug #1283 (Closed): Usage Profile is blank and shows incorrect selection range for application ru...
I compiled leanmd with +tracemode summary and ran it with +sumDetail. On opening the leanmd.sum.sts file with project... Nitin Bhat
08:32 AM Projections Bug #1230: Excessive memory utilization when ingesting files into Timeline
These logs which take really long to load and on loading become unresponsive are from a single core ChaNGa run (SMP m... Nitin Bhat
08:06 PM Bug #1045: Array pointers in messages can be overwritten by generated code in the message class c...
Phil Miller wrote:
> Ugh, the configure test is supposed to catch that. I'll fix.
Config does catch it, but on BG...
Nikhil Jain
07:42 PM Revision 6fd10623 (charm): AMPI: do not rebuild ROMIO if it was already built
Change-Id: I647dcd79449e11597cb2f30ff83214c20b0dcb80 Sam White

11/05/2016

03:13 PM Bug #1278: ReductionStarting overhead in SMP mode
For note, the code on that branch at least mildly conflicts with the more recent streaming reduction changes, and I d... Phil Miller
02:36 PM Bug #1278: ReductionStarting overhead in SMP mode
Harshitha seems to have started on code to fix this in branch @harshitha/reductionstartingnode@ but not finished it. Phil Miller
02:14 PM Bug #1278: ReductionStarting overhead in SMP mode
In projections traces, I noticed that each contribution was only triggering sends of ReductionStarting messages to ot... Phil Miller
01:39 PM Bug #1278: ReductionStarting overhead in SMP mode
The fix in #29 should have covered this Phil Miller
02:25 PM Feature #1114: Build and link ROMIO on AMPI by default
Patch to not rebuild AMPI-ROMIO when it has already been built: https://charm.cs.illinois.edu/gerrit/#/c/1971/ Sam White
01:16 PM Revision ab467cc0 (charm): Update emails in LICENSE from @cs.uiuc.edu to @illinois.edu
Change-Id: Iee7cb653cafff0911e5a8aba4e2b07d94c283389 Ronak Buch
01:16 PM Revision 7cd9d2b5 (charm): Fix misspelling in LICENSE
Change-Id: If2f180bf0d4a473132855618bea7842ed37e90b0 Ronak Buch
01:16 PM Revision 19c86b8f (charm): Change LBPeriod in every PE in stencil3d example
Since this program uses AtSync to initiate LB, and the time between AtSync's
can be much lower than the default LBPer...
Juan Galvez
10:11 AM Bug #1280 (Closed): CkLoop gives bad results on > 64 pes per process
Nevermind. This turned out to be bad CkLoop code in NAMD protected by a "CmiAssert(first==last)" that was preprocesse... Jim Phillips

11/04/2016

05:19 PM Bug #1281: Bigsim crash in SDAG code in examples/charm++/load_balancing/stencil3d
Possibly related to #82 Phil Miller
05:18 PM Bug #1281 (New): Bigsim crash in SDAG code in examples/charm++/load_balancing/stencil3d
... Phil Miller
05:11 PM Revision 3fc7d668 (charm): Bug #901: Added tracing for thread context switch events
- Add a thread listener for CthAwaken() calls
- Attach listeners for scheduler threads (original and stand-in)
- Reco...
Steve Hoelle
04:06 PM Revision 2ef9a34a (charm): Applying dynamic number of messages for OpenMP
Change-Id: Ie88e600fc9929c9267538aef2c2fb40cbf3ff08d Seonmyeong Bak
03:29 PM Bug #961: AMPI support for cancelling recv requests/statuses
Note that recent moves in the MPI forum suggest deprecating the ability to cancel send requests, since the semantics ... Phil Miller
11:58 AM Revision 42ee026a (charm): GTopoCommLB can now run multiple instances in parallel
Change-Id: If10ddbda5f729b37745cc68a5d25ee4b17447844 Juan Galvez
10:57 AM Bug #1280: CkLoop gives bad results on > 64 pes per process
Assigning to Seonmyeong based on module ownership Sam White
10:22 AM Bug #1280 (Rejected): CkLoop gives bad results on > 64 pes per process
If you run NAMD on apoa1 with --pmeProcessors 20 --useCkLoop 3 on multicore +p65 you get bad force errors. +p64 work... Jim Phillips
10:36 AM Revision 3fc1ebbe (charm): Mark a method 'const' to broaden when it can be called
Change-Id: Iafc35334a7d5429b049e74c4ee4d25112fe4a366 Phil Miller
10:26 AM Bug #1045 (In Progress): Array pointers in messages can be overwritten by generated code in the m...
Ugh, the configure test is supposed to catch that. I'll fix. Phil Miller
10:23 AM Bug #1045: Array pointers in messages can be overwritten by generated code in the message class c...
XLC actually gives an error for this, so the build aborts:... Sam White
10:49 PM Revision 308a2fb9 (charm): Rdma support for gni machine layer
Change-Id: I7414b07c7a86d0a109f4c8c6a07c02a5bc7e571b Justin Miron
09:15 PM Revision 9d2300e5 (charm): Add initial changes for streaming projections
Change-Id: I550dc1044c9dc86bd8c42b29da3993c792160987 Ronak Buch
09:09 PM Revision 1f456208 (charm): Update emails in LICENSE from @cs.uiuc.edu to @illinois.edu
Change-Id: Iee7cb653cafff0911e5a8aba4e2b07d94c283389 Ronak Buch
09:08 PM Revision c4f15810 (charm): Fix misspelling in LICENSE
Change-Id: If2f180bf0d4a473132855618bea7842ed37e90b0 Ronak Buch

11/03/2016

06:38 PM Revision 4cec42a7 (charm): fix msg size
Change-Id: I91f76cc6421cde29101ba4edb050f5260bdba2d7 Justin Miron
06:12 PM Revision 6981598b (charm): Rdma support for gni machine layer
Change-Id: I7414b07c7a86d0a109f4c8c6a07c02a5bc7e571b Justin Miron
04:30 PM Revision b8ad9cc1 (charm): Applying dynamic number of messages for OpenMP
Change-Id: Ie88e600fc9929c9267538aef2c2fb40cbf3ff08d Seonmyeong Bak
12:31 PM Bug #1279: Proactive fault tolerance fails due to sending message to dead node.
Thanks, that helps a lot.
Reinserting the getNextPE code works when finding the next PE off of the evacuated dest...
Justin Miron
11:25 AM Bug #1279: Proactive fault tolerance fails due to sending message to dead node.
This was changed in the 64bit ID changes. Look at line 2635 here:
https://charm.cs.illinois.edu/gerrit/#/c/1217/5/sr...
Sam White
09:52 AM Bug #1279: Proactive fault tolerance fails due to sending message to dead node.
This check may not be neccessary. If the PE was previously on a node that is now dead, then it should call DeliverUnk... Justin Miron
09:49 AM Bug #1279 (New): Proactive fault tolerance fails due to sending message to dead node.
CkLocMgr::deliverMsg attempts to send messages to evacuated chares.
Fails this check:
if((!CmiNodeAlive(destPE) &...
Justin Miron
12:02 PM Cleanup #1249 (Merged): Deprecate setReductionClient and CkSetReductionClient
Phil Miller
12:01 PM Bug #1255 (In Progress): Race condition in demand creation causing collidethread example failures
https://charm.cs.illinois.edu/gerrit/#/c/1964
There's a race condition in demand creation that it can induce the loc...
Phil Miller
11:46 AM Revision 4d493fbb (charm): Fix more undefined behavior of writing to object fields before the obj...
Change-Id: Id07180a2dbd0e4d42715cdff6255a2b7cac67ef7 Phil Miller
11:46 AM Revision 0d78c282 (charm): Parenthesize arguments in macro definitions to protect expression from...
Change-Id: I4bd4b2790467b24c9d4076d9b00da8e22339b824 Phil Miller
11:46 AM Revision d1dc7474 (charm): tests & examples Makefiles: fix dependency generation for C++11 support
Change-Id: Idc8ffed0cf0d175ce6face271b26f67b787af360 Phil Miller
11:46 AM Revision a3fbda14 (charm): megatest: Fix and regenerate Make dependencies
Change-Id: I473a3c852622adfe8fb59c816f3e81978e016609 Phil Miller
11:46 AM Revision 2e5e07c2 (charm): Avoid undefined call 'memcpy(dest, NULL, 0);'
Change-Id: Ib76451b2fc4b67b6c4a3516a80f978341cf10c30 Phil Miller
11:46 AM Revision c22034b2 (charm): Remove archaic assignment that was taken care of with Cpv currentChareIdx
Change-Id: Icd61214b9c2d4891955d4a1d93dce4fbc7bcf221 Phil Miller
11:46 AM Revision 115c9fe0 (charm): Delete archaic (commented) assignments to chareIdx field of unconstruc...
Change-Id: Ifca983573ecd1fa66ea9a41924ce516bd2b8fd2b Phil Miller
11:46 AM Revision 441220e9 (charm): Bug #1255: Avoid race condition in trying to create array elements for...
Change-Id: I716d0537fa067c26e34e5069d8201ef24996dafc Phil Miller
11:46 AM Revision 363834ac (charm): examples/charm++/*/Makefile: Add missing $(OPTS) to compiler and linke...
Change-Id: I0ea1bbacd5b9a7551622e3f51fa4af621775d196 Phil Miller
11:37 PM Feature #1238: Onesided sender side implementation for Verbs layer
https://charm.cs.illinois.edu/gerrit/#/c/1960/ Sam White
11:34 PM Bug #1277: Multiple shrink expand requests lead to segmentation fault
https://charm.cs.illinois.edu/gerrit/#/c/1961/ Sam White
07:17 PM Revision 0bd333d7 (charm): Improve grammar and language for sections documentation
Change-Id: Ie4c49270ef74b3c7a064c5591e55c761236ab021 Ronak Buch

11/02/2016

06:49 PM Feature #858: improve efficiency of exclusive entry methods
Moving this to Nitin since he has more experience with charmxi. I think it should be easy to do what Jim has pointed ... Sam White
06:14 PM Bug #1278 (Merged): ReductionStarting overhead in SMP mode
On a benchmark where every PE has array elements, I'm still seeing a lot of @CkReductionMgr::ReductionStarting@ messa... Phil Miller
05:52 PM Bug #1277 (Implemented): Multiple shrink expand requests lead to segmentation fault
The shrink expand restarts the application using a new command line string. This new string would copy the old comman... Justin Miron
05:49 PM Bug #1277 (Merged): Multiple shrink expand requests lead to segmentation fault
If a CCS client issues multiple shrink expand requests, the application segmentation faults.
The issue is with com...
Justin Miron
05:52 PM Cleanup #1249 (Implemented): Deprecate setReductionClient and CkSetReductionClient
Implemented in https://charm.cs.illinois.edu/gerrit/#/c/1962/. Ronak Buch
02:51 PM Cleanup #1249 (In Progress): Deprecate setReductionClient and CkSetReductionClient
Ronak Buch
05:50 PM Revision 953363e0 (charm): Cleanup #1249: Remove deprecated documentation on setReductionClient
Change-Id: Id9d96598652c86f62642dc4d36f42589c70ad28a Ronak Buch
05:39 PM Cleanup #1064: Replace std::maps with a better implementation
I just missed it being assigned to me. Working on it now. The only "complexity" is that without 'using' we will can't... Eric Mikida
05:34 PM Bug #1275: DistributedLB: Objects not migrating after strategy runs
The potential to hang on 2 PEs seems like it should also be fixed.
I was running on 40 PEs when I observed the iss...
Phil Miller
05:24 PM Bug #1275: DistributedLB: Objects not migrating after strategy runs
The migration phase after strategy phase seems to work okay for me. In the Makefile, the test uses 2 PEs. I believe w... Kavitha Chandrasekar
02:00 PM Bug #1275: DistributedLB: Objects not migrating after strategy runs
Neither of the pending patches in Gerrit resolves this issue.
Demonstration code attached.
Phil Miller
01:46 PM Bug #1275: DistributedLB: Objects not migrating after strategy runs
Eric M just submitted a patch for an issue with DistributedLB: https://charm.cs.illinois.edu/gerrit/#/c/1951/ Sam White
01:21 PM Bug #1275 (Merged): DistributedLB: Objects not migrating after strategy runs
Will attach a demonstration case shortly Phil Miller
03:34 PM Feature #1238 (Implemented): Onesided sender side implementation for Verbs layer
Jaemin Choi
02:47 PM Revision fb42cebb (charm): Test pingpongrdma, clean up code, fix memory leak
Change-Id: I6fb6d8713a530308ed0a383e3395231ded650cb6 Jaemin Choi
02:12 PM Bug #1276 (Rejected): net-cygwin: multiple testing failures
Please open sub-issues for the currently observed individual failures Phil Miller
02:06 PM Revision a0ebf4a5 (charm): net-cygwin: switch thread default from uJcontext to context to fix cra...
Change-Id: I091207e43167faf17aca8b975d8ba0e18b2c1e68 Phil Miller
01:22 PM Revision 1fc2482e (charm): Add weak symbol pragmas
Change-Id: Ib33cc04819921a14f5e0b7df712f20f5c659ee13 Sam White
12:26 PM Revision 8dc374c6 (charm): Rdma support for gni machine layer
Change-Id: I7414b07c7a86d0a109f4c8c6a07c02a5bc7e571b Justin Miron
11:56 AM Revision 1deb9f64 (charm): Rdma support for gni machine layer
Change-Id: I7414b07c7a86d0a109f4c8c6a07c02a5bc7e571b Justin Miron
11:48 AM Bug #1274 (Merged): Tracemode utilization produces incorrect results for EP utilization metric
Even if a Charm++ program is successfully connected to Projections via the Streaming CCS Tool,
two graphs showing th...
Jaemin Choi
11:18 AM Revision 9566df69 (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
11:17 AM Bug #1273 (Closed): Tracemode utilization crashes in production build of Charm++
Charm++ programs compiled in production mode with tracemode utilization yield in a segmentation fault. The error occu... Karthik Senthil
10:30 AM Revision 997a8409 (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
10:30 AM Bug #854: RRMap broken for >1D chare arrays
Abandoned patch for this: https://charm.cs.illinois.edu/gerrit/#/c/1053/
BlockLB similarly does not take into acco...
Sam White
10:29 AM Revision d72068ba (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
10:21 AM Revision 7f138332 (charm): Fix sizing checks to not go beyond 2^32 on 32-bit systems
Change-Id: I25253ac98c5d207927554f750cfb5ea7155f9cc1 Phil Miller
09:02 PM Revision 001eb877 (charm): Rename AMPI_* routines to MPI_*
Change-Id: I711ff64ab0f271349d5a3d84456ea0e3fb6092cb Sam White

11/01/2016

06:21 PM Revision f37d10e6 (charm): Fix bugs and successfully run simple rdma test program
Change-Id: Iadaa87281ea964256f2d795bf511d14605217086 Jaemin Choi
03:52 PM Revision b98c9161 (charm): convcore: Update to modern C function prototypes from K&R style
Change-Id: I01ca53e6293911bbe09a8e16576c510985e4c45a Phil Miller

10/31/2016

12:28 PM Revision 8f64c83d (charm): Fix constants to be well-defined on LLP64 (Windows) systems
Change-Id: I8685ddf05abf01be2298a672d86474c61b22864e Phil Miller
 

Also available in: Atom