Project

General

Profile

Activity

From 07/26/2016 to 08/24/2016

08/25/2016

07:26 PM Bug #1157 (Merged): DDT support for MPI_BOTTOM with derived datatypes
Sam White

08/24/2016

06:39 PM Bug #1157 (Implemented): DDT support for MPI_BOTTOM with derived datatypes
Sam White
03:44 PM Bug #1183: megatest and megacon should work for large node counts
There are some extremely slow tests even for 64 nodes with +ppn 60:... Jim Phillips
02:24 PM Bug #1183 (New): megatest and megacon should work for large node counts
Some test runtimes appear to scale as O(P) or worse, which makes tests useless for testing large machines. Test comp... Jim Phillips
03:08 PM Feature #1173: Automatic process launching, thread spawning, and hardware binding
Two missing capabilities:
1) reserve one (more?) core or PU for the OS - equivalent to -r 1 on Cray's aprun
2) ...
Jim Phillips
02:52 PM Feature #1173: Automatic process launching, thread spawning, and hardware binding

*CAVEATS AND LIMITATIONS*
Before getting into how it works, there are a few caveats of the current implementation ...
Phil Miller
10:48 AM Feature #1173: Automatic process launching, thread spawning, and hardware binding
For new readers: the implementation on that branch currently works for runs spawned with charmrun for netlrts and ver... Phil Miller
10:46 AM Feature #1173: Automatic process launching, thread spawning, and hardware binding
Existing implementation is on branch hwloc: https://charm.cs.illinois.edu/gerrit/gitweb?p=charm.git;a=shortlog;h=refs... Phil Miller
02:50 PM Feature #1184 (Implemented): SMP-safe CmiReference and CmiFree
Two related issues that could be broken out into separate issues once this work is begun:
1. Currently CmiReferenc...
Sam White

08/23/2016

06:46 PM Projections Bug #1170 (Merged): Avoid loading log files beyond what is specified in .sts file
Ronak Buch
06:13 PM Feature #1182: Support automated thread spawn/bind for MPI SMP builds
Just realized, this may need a subtask of detecting hazardous non-uniformity as well. Phil Miller
06:03 PM Feature #1182 (New): Support automated thread spawn/bind for MPI SMP builds
Phil Miller
06:11 PM Bug #1168 (Merged): Test that number of PEs running actually matches expectations (often an issue...
Phil Miller
06:01 PM Feature #1181 (Merged): Support automated process launch on a single host for standalone runs
If the user is running on one host but wants to be lazy about using charmrun, it should still be possible for them to... Phil Miller
06:00 PM Feature #1180 (Merged): Support automated thread spawn/bind for standalone runs
When running on a single host without using charmrun (in a build of netlrts/verbs/multicore), it should still be poss... Phil Miller
05:54 PM Feature #1179 (New): Support automated launch/spawn/bind on gni-crayxc/gni-crayxe systems
I guess they'll need to use PMI-level data and communication at startup to stand in for what charmrun provides netlrt... Phil Miller
05:53 PM Feature #1178 (Merged): Support automated launch/spawn/bind when using charmrun ++mpiexec
This may already even work. I haven't tested it yet. Phil Miller
05:52 PM Feature #1177 (New): Support variable numbers of processes per host, each with identical thread c...
This would allow us to support systems like the XK/XE mix on Blue Waters, without disrupting the current assumption t... Phil Miller
05:50 PM Feature #1176 (Merged): Detect unsupported non-uniformity of processes/threads in charmrun, and e...
Phil Miller
05:43 PM Feature #1175 (Merged): Don't require autoconf to be installed on user systems for hwloc build
Phil Miller
05:42 PM Bug #1174 (Merged): Use hwloc data from compute host, rather than assuming they're identical to t...
Phil Miller
05:41 PM Feature #1173 (In Progress): Automatic process launching, thread spawning, and hardware binding
Top-level tracking issue for work based on integrating hwloc with the RTS and using it to make life easier for users,... Phil Miller
03:06 PM Bug #1150 (Merged): Specialize pup for std::vector to vector<bool>
Phil Miller
10:08 AM Feature #1158 (Implemented): AMPI scatter(v) performance is poor
The following will work once we have [nokeep] broadcasts working in AMPI: https://charm.cs.illinois.edu/gerrit/#/c/1807/ Sam White
08:07 AM Feature #1158: AMPI scatter(v) performance is poor
This is not so nice for Scatterv because the displs argument is only relevant at the root (users can pass NULL from t... Sam White
11:42 PM Feature #1158 (In Progress): AMPI scatter(v) performance is poor
Sam White
10:06 AM Feature #1116 (Implemented): AMPI broadcasts should be [nokeep]
https://charm.cs.illinois.edu/gerrit/#/c/1806/ Sam White
09:31 AM Feature #1171 (Closed): Add option to make all TCharm entry methods [expedited]
Upon closer inspection, there are no entry methods in TCharm that are called from AMPI not during init/exit, so there... Sam White
09:20 AM Feature #1171 (Closed): Add option to make all TCharm entry methods [expedited]
AMPI entry methods are by default all [expedited] because we don't use message priorities at all in AMPI.
AMPI build...
Sam White

08/22/2016

06:02 PM Feature #1116 (In Progress): AMPI broadcasts should be [nokeep]
Sam White
04:20 PM Bug #1169: Command line argument +balancer for multiple balancers does not work
The TCharm/AMPI option '+vp' is also supposed to be accepted multiple times (in the case of running multiple modules). Sam White
03:50 PM Projections Bug #1170 (Merged): Avoid loading log files beyond what is specified in .sts file
Currently, when logs numbered beyond the number specified in the sts file exist in a folder, Projections attempts to ... Ronak Buch
10:01 AM Feature #921 (In Progress): Entry tag [inline] is unable to optimize away most of the overhead
Phil Miller
09:57 AM Bug #1168 (Implemented): Test that number of PEs running actually matches expectations (often an ...
http://charm.cs.illinois.edu/gerrit/1804 Phil Miller
09:47 AM Bug #1145 (Merged): PathHistory breaks chkpt test on multicore-linux64
Phil Miller
07:37 PM Feature #1107 (Merged): Use built-in reducer types for basic types/operations in AMPI
Sam White

08/21/2016

03:47 PM Bug #1084 (Merged): Eliminate extra copies in AMPI reduce/gather(v) receive paths
Sam White

08/19/2016

12:00 PM Bug #1084: Eliminate extra copies in AMPI reduce/gather(v) receive paths
I think the issue here might actually a bug in AMPI_Wait(all) and BigSim that manifests as a failure in the reduction... Sam White
10:49 AM Bug #1169 (Merged): Command line argument +balancer for multiple balancers does not work
Specifying the command line argument +balancer multiple times, to allow different balancers to be called in subsequen... Kavitha Chandrasekar

08/18/2016

05:42 PM Bug #1149: Cray CC builds are broken
I'm also getting this trying to build with Cray compilers:... Jim Phillips
05:37 PM Bug #1149: Cray CC builds are broken
Talk to Phil if you need specifics now but there are a couple patches related to this in gerrit review (authored by h... Sam White
05:30 PM Bug #1149: Cray CC builds are broken
Sitting with someone from Cray now. Is there a reason this isn't committed yet? Jim Phillips
02:33 PM Bug #1168 (Merged): Test that number of PEs running actually matches expectations (often an issue...
On mpi-win64 from autobuild:... Phil Miller
02:15 PM Support #1167: Scripted tests for all mini-apps
http://charmplusplus.org/miniApps/ Phil Miller
02:12 PM Support #1167 (New): Scripted tests for all mini-apps
Enable automatic testing of the various mini-appls that don't live as examples/tests in side the main Charm++ repository Phil Miller

08/17/2016

04:56 PM Bug #1161 (Resolved): charm-6.7.1.tar.gz untars to charm
Sam White
04:56 PM Bug #1161 (Merged): charm-6.7.1.tar.gz untars to charm
I changed it to untar to a directory named 'charm-6.7.1' Sam White
04:46 PM Bug #1166 (Rejected): pathHistoryManager doesn't PUP
Sam White
02:00 PM Bug #1166: pathHistoryManager doesn't PUP
Things work with the latest version. Sorry for not checking this first. Thomas Quinn
01:33 PM Bug #1166: pathHistoryManager doesn't PUP
Do you still see problems on current master, since the following commit was merged? https://charm.cs.illinois.edu/ger... Sam White
07:14 PM Bug #1166 (Rejected): pathHistoryManager doesn't PUP
Commit b240e238385ba2d0aa08748880b3b34ce59d03a2
Author: Yanhua Sun <sun51@illinois.edu>
Date: Mon Nov 2 11:21:03 ...
Thomas Quinn
04:37 PM Bug #1084: Eliminate extra copies in AMPI reduce/gather(v) receive paths
I can reproduce it on 1 PE with the examples/ampi/creduce program, which is just one MPI_Reduce followed by one MPI_I... Sam White
01:45 PM Bug #1084: Eliminate extra copies in AMPI reduce/gather(v) receive paths
Can you reduce that test case further? Fewer PEs? One PE? Fewer VPs? Phil Miller
01:19 PM Bug #1084: Eliminate extra copies in AMPI reduce/gather(v) receive paths
This patch is now held up because it breaks BigSim, with aborts for recv'ing unexpected reductions in ampi::rednResul... Sam White
02:10 PM Feature #882 (Merged): Change type to indicate non-participation in reductions
Juan Galvez

08/16/2016

04:06 PM Feature #1164 (Implemented): AMPI support for MPI-3's request-based RMA routines
Initial support, without creating actual requests and without using true RDMA, is here: https://charm.cs.illinois.edu... Sam White
11:40 AM Feature #1164 (Merged): AMPI support for MPI-3's request-based RMA routines
MPI-3 added request-based RMA routines Rput, Rget, Raccumulate, and Rget_accumulate. They are analogous to the two-si... Sam White
01:13 PM Bug #1165 (Resolved): avoid -lm with Intel compiler
It is advised to not explicitly link to -lm with the Intel compiler as this may override the faster Intel math librar... Jim Phillips
10:55 AM Feature #882 (Implemented): Change type to indicate non-participation in reductions
Patch here: https://charm.cs.illinois.edu/gerrit/#/c/1800/
Juan Galvez

08/15/2016

02:53 PM Bug #1163 (Merged): AMPI_Put should use targdisp from creation
MPI_Put takes a targdisp parameter, but at the remote window the displacement used should be the one that was specifi... Sam White
02:51 PM Bug #1110 (Merged): Cray builds break if explicitly passed compilers
Seonmyeong Bak
02:50 PM Bug #1096 (Rejected): PGI CC builds fail to find C++11 flags
This issue doesn't happen on BlueWaters with PGI CC 16.3.0. Seonmyeong Bak

08/14/2016

05:35 PM Bug #1162: tracing runs segfault while writing logs
I'm consistently seeing segfaults or, most informatively, free(): invalid pointer, at the end of smp tracing runs on ... Jim Phillips
04:59 PM Bug #1162 (Closed): tracing runs segfault while writing logs
Jim Phillips
02:13 PM Bug #1161 (Closed): charm-6.7.1.tar.gz untars to charm
The file distrib/charm-6.7.1.tar.gz file untars to the directory charm, which is not what the user would expect. Jim Phillips

08/11/2016

05:34 PM Bug #1159 (Implemented): AMPI should check for truncated messages
For now, just abort on recv'ing a message larger than the matching recvbuf from ampi::processMessage. the mpich-test/... Sam White
02:15 PM Bug #1160 (Merged): AMPI intercommunicator creation is broken
All of the tests in tests/ampi/mpich-test/context/ for intercommunicator creation and duplication are currently broke... Sam White

08/10/2016

08:41 AM Bug #1038: Add support for srun to charmrun script for gni layer
Now the error is this:... Sam White

08/09/2016

10:55 AM Bug #1159 (In Progress): AMPI should check for truncated messages
tests/ampi/mpich-test/pt2pt/trunc fails because AMPI doesn't check for truncation on recv at all right now.
Add a ch...
Sam White
10:00 AM Feature #1105: AMPI support for large counts
Support for large counts in DDT and all MPI-3 large count APIs except get_true_extent_x: https://charm.cs.illinois.ed... Sam White
09:30 AM Bug #1149: Cray CC builds are broken
Improve configure test for 128-bit integers so that it fails correctly for CCE: https://charm.cs.illinois.edu/gerrit/... Sam White
09:26 AM Bug #1038: Add support for srun to charmrun script for gni layer
https://charm.cs.illinois.edu/gerrit/#/c/1343/ Sam White
09:21 AM Bug #1143 (Merged): AMPI's Hindexed datatype is broken
Sam White
09:19 AM Bug #1110: Cray builds break if explicitly passed compilers
Follow up patches:
https://charm.cs.illinois.edu/gerrit/#/c/1340/
https://charm.cs.illinois.edu/gerrit/#/c/1345/
Sam White
11:02 PM Bug #1153 (Merged): MPI_BOTTOM support in AMPI communication routines
Sam White
11:01 PM Bug #1157 (In Progress): DDT support for MPI_BOTTOM with derived datatypes
https://charm.cs.illinois.edu/gerrit/#/c/1342/ Sam White

08/05/2016

09:31 AM Feature #982 (Merged): Gather reduction operation using Tuple/Set reducers
Closing this now since there is an open issue for the documentation, #1140. Sam White

08/04/2016

04:41 PM Feature #896: AMPI support for compiler-automated static variable privatization
We should get into contact with Intel's compiler people about this Sam White
04:24 PM Bug #1110 (In Progress): Cray builds break if explicitly passed compilers
Sam White
02:00 PM Bug #1110 (Merged): Cray builds break if explicitly passed compilers
Sam White
03:12 PM Feature #1158 (New): AMPI scatter(v) performance is poor
This is a performance bug: we currently do the naive method of sending 'p' individual messages from the root to the '... Sam White
02:55 PM Bug #1157 (Merged): DDT support for MPI_BOTTOM with derived datatypes
This change adds support for using MPI_BOTTOM in AMPI communication routines: https://charm.cs.illinois.edu/gerrit/#/... Sam White
01:57 PM Bug #833: mpi smp build is locked to one core per node by default
I sort of agreed with Eric to wait until hwloc changes are finalized, since this is going to bring changes to cputopo... Juan Galvez
01:39 PM Bug #833: mpi smp build is locked to one core per node by default
Bump: this has been stuck in gerrit since March Sam White
01:54 PM Bug #1152 (Merged): examples/charm++/namespace/Makefile: missing dependencies
Sam White
01:35 PM Bug #1042 (Merged): verbs layer runs out of memory regions and segfaults
Sam White
01:34 PM Cleanup #197 (In Progress): Replace use of signal() with sigaction()
This has been sitting in review since December, please address the comments and get it merged: https://charm.cs.illin... Sam White
01:11 PM Documentation #1156 (In Progress): Add example showing refnum's on callbacks
Sam White
10:35 PM Documentation #1156 (Merged): Add example showing refnum's on callbacks
Jonathan Lifflander emailed the [ppl] mailing list with a suggestion to add this example, as well as the code itself.... Sam White
09:36 PM Cleanup #798: Add tests and documentation for sparse-array bulk construction
For ease of cross-reference: https://charm.cs.illinois.edu/gerrit/879 Phil Miller
08:05 PM Bug #1155 (New): AMPI's non-blocking collectives are not sequenced
AMPI's non-blocking collectives should be sequenced/ordered just like non-blocking pt2pt messages are. Sam White

08/03/2016

03:48 PM Bug #1153 (Implemented): MPI_BOTTOM support in AMPI communication routines
https://charm.cs.illinois.edu/gerrit/#/c/1336/ Sam White
02:58 PM Bug #1153 (Merged): MPI_BOTTOM support in AMPI communication routines
AMPI defines MPI_BOTTOM but does nothing to support its use in communication operations.
tests/ampi/mpich-test/pt2...
Sam White
03:07 PM Bug #1147 (Merged): pathHistory.C fails to build on Windows
Sam White
02:59 PM Bug #1150 (Implemented): Specialize pup for std::vector to vector<bool>
https://charm.cs.illinois.edu/gerrit/#/c/1332/ Sam White
02:59 PM Bug #1152 (Implemented): examples/charm++/namespace/Makefile: missing dependencies
Sam White
11:40 AM Feature #984 (Implemented): AMPI support for missing MPI-3 non-blocking collectives
Sam White
11:37 AM Feature #1018 (Implemented): AMPI support for MPI-3's distributed graph interface
Sam White

08/02/2016

06:50 PM Bug #1152 (Merged): examples/charm++/namespace/Makefile: missing dependencies
... Phil Miller
12:28 PM Cleanup #1151 (Merged): Eliminate AMPI ROMIO compiler warnings
The ROMIO on AMPI build is full of these warnings:... Sam White
09:46 PM Bug #1110 (Feedback): Cray builds break if explicitly passed compilers
Seonmyeong Bak

08/01/2016

06:21 PM Bug #1145 (Implemented): PathHistory breaks chkpt test on multicore-linux64
http://charm.cs.illinois.edu/gerrit/1331
There will still be an incompatibility between checkpoint/restart and cri...
Phil Miller
04:02 PM Bug #1147 (In Progress): pathHistory.C fails to build on Windows
Linked patch merged. We'll see if more are needed. Phil Miller
02:21 PM Bug #1150 (Merged): Specialize pup for std::vector to vector<bool>
> I get a compiler warning "taking address of temporary" when I pup a std::vector<bool>. There is no warning when I ... Phil Miller

07/31/2016

12:50 AM Cleanup #1034: Constant comparison warnings from xlC compiler
Can you write down some details of the environment where you found this bug? I tried to build charm++ on ANL Mira(BG/... Seonmyeong Bak

07/29/2016

04:43 PM Bug #1110 (Upstream): Cray builds break if explicitly passed compilers
Seonmyeong Bak

07/28/2016

01:51 PM Bug #1149: Cray CC builds are broken
Even with that patch to configure, I still get consistent but inscrutable build failures in early object files, while... Phil Miller
01:49 PM Bug #1149: Cray CC builds are broken
For the @int128_t@ crap, this is easier and cleaner:... Phil Miller

07/27/2016

09:45 AM Bug #1143: AMPI's Hindexed datatype is broken
Patch set #6 fixes most of these tests, the exceptions are listed below: http://charm.cs.illinois.edu/gerrit/#/c/1321... Sam White
09:40 AM Feature #947: AMPI support for MPI-2's MPI_Type_* routines
MPI_Type_create_resized: http://charm.cs.illinois.edu/gerrit/#/c/1321/ Sam White
09:29 AM Bug #1149: Cray CC builds are broken
Cray CC 8.5 doesn't build even with the warnings turned off. Nikhil found a hack to make it work:
Apparently, tell...
Sam White
 

Also available in: Atom