Project

General

Profile

Activity

From 06/26/2016 to 07/25/2016

07/22/2016

08:33 AM Bug #1149 (Merged): Cray CC builds are broken
Build failure due to warning while testing C++11 options has been an issue for quite some time, especially since it f... Sam White

07/21/2016

03:06 PM Bug #1148 (Merged): Define 'thisIndex' for Groups
Chare arrays all have a member variable 'thisIndex', while Groups and NodeGroups have a semantically similar ID calle... Sam White
01:48 PM Bug #1146 (Merged): TCharm calls CthAwaken twice on the same thread after migration
Sam White
10:51 AM Bug #1145: PathHistory breaks chkpt test on multicore-linux64
Looking at autobuild, several targets show up good, and I explicitly checked that they run that test to completion. I... Phil Miller
10:41 AM Bug #1145: PathHistory breaks chkpt test on multicore-linux64
It happens in a netlrts-linux-x86_64-syncft build I just tried... Sam White
10:33 AM Bug #1145: PathHistory breaks chkpt test on multicore-linux64
So, this crashes the same test on @multicore-darwin-x86_64@ too. What I'm confused about is why it doesn't crash the ... Phil Miller
09:18 AM Bug #1145: PathHistory breaks chkpt test on multicore-linux64
Strange that this only fails for multicore-linux64 Phil Miller
08:31 AM Bug #1145: PathHistory breaks chkpt test on multicore-linux64
This is from the recent critical path header changes for the PICS merge: https://charm.cs.illinois.edu/gerrit/#/c/893/ Sam White
10:47 AM Bug #1147: pathHistory.C fails to build on Windows
Rough guess at a fix here: https://charm.cs.illinois.edu/gerrit/1328
Phil Miller
08:28 AM Bug #1147 (Merged): pathHistory.C fails to build on Windows
The critical path header changes for PICS merge has resulted in multicore/net/mpi-win builds failing: http://charm.cs... Sam White
09:09 AM Bug #1143: AMPI's Hindexed datatype is broken
Datatype related tests in tests/ampi/mpich-test/pt2pt that currently fail:
hindexed
hvectest2
typeub
typeub3
d...
Sam White
11:23 PM Bug #1143 (In Progress): AMPI's Hindexed datatype is broken
Indexed datatype fix: https://charm.cs.illinois.edu/gerrit/#/c/1324/ Sam White

07/20/2016

06:47 PM Bug #1084: Eliminate extra copies in AMPI reduce/gather(v) receive paths
This patch was held up on this issue: #1146 Sam White
06:40 PM Bug #1146 (Implemented): TCharm calls CthAwaken twice on the same thread after migration
https://charm.cs.illinois.edu/gerrit/#/c/1323/ Sam White
06:20 PM Bug #1146: TCharm calls CthAwaken twice on the same thread after migration
The issue is only with TCHARM_Migrate, not TCHARM_Async_Migrate. Sam White
06:01 PM Bug #1146 (Merged): TCharm calls CthAwaken twice on the same thread after migration
After migration, TCharm::ckJustMigrated and TCharm::ResumeFromSync both call CthAwaken on the user-level thread. The ... Sam White
08:29 AM Bug #1145 (Merged): PathHistory breaks chkpt test on multicore-linux64
Multicore-linux64 autobuild failure:... Sam White

07/19/2016

05:57 PM Charm-NG Feature #1144: Batched message delivery to objects for better cache behavior
I realize that it may be possible to 'trick' Xiang's out-of-core scheduling code to do this, by having each object sp... Phil Miller
05:54 PM Charm-NG Feature #1144 (New): Batched message delivery to objects for better cache behavior
Transcript of a slack conversation indicating the multiple applications with a potential need for better cache utiliz... Phil Miller
02:48 PM Bug #1143 (Merged): AMPI's Hindexed datatype is broken
Running the hindexed test in charm/tests/ampi/mpich-test/pt2pt/ results in a segfault.
I am suspicious of the way ...
Sam White
10:25 AM Bug #1142 (Merged): AMPI's main() broken for C main's not matching its signature
Sam White
10:25 AM Bug #1141 (Merged): MPI_Datatype_iscontig undefined in AMPI ROMIO
Sam White

07/18/2016

04:04 PM Bug #1142 (Implemented): AMPI's main() broken for C main's not matching its signature
https://charm.cs.illinois.edu/gerrit/#/c/1316/ Sam White
02:43 PM Bug #1142 (Merged): AMPI's main() broken for C main's not matching its signature
AMPI redefines the user's main function, assuming that the user's main function is "int main(int argc, char *argv);"
...
Sam White
04:04 PM Bug #1141 (Implemented): MPI_Datatype_iscontig undefined in AMPI ROMIO
https://charm.cs.illinois.edu/gerrit/#/c/1317/ Sam White
01:55 PM Bug #1141 (Merged): MPI_Datatype_iscontig undefined in AMPI ROMIO
This call should be changed to AMPI_Type_is_contiguous Sam White
02:55 PM Feature #1133: PMPI_ interface for AMPI
Full support for the PMPI_ interface in AMPI would require some invasive changes to AMPI:
1. Define each MPI_ routin...
Sam White
09:25 PM Documentation #1140 (Merged): Document tuple/stats reducers
Ch. 4 of the Charm++ manual is out of date, missing info on tuple and stats reducers, as well as built-in support for... Sam White

07/15/2016

02:28 PM Feature #1114 (New): Build and link ROMIO on AMPI by default
Sam White

07/14/2016

02:31 PM Bug #1050 (In Progress): AMPI collectives support for MPI_IN_PLACE
MPI_IN_PLACE for everything but MPI_Alltoall added here. Note that we just need a different algorithm for large messa... Sam White
01:54 PM Feature #1111 (Implemented): Avoid sender-side copy in AMPI for large contiguous messages
Initial implementation here: https://charm.cs.illinois.edu/gerrit/#/c/1312/
The biggest issue here is migration: w...
Sam White
12:12 PM Feature #1060: AMPI compliance with MPI-3.1 standard
MPI_Get_accumulate, MPI_Fetch_and_op, and MPI_Compare_and_swap added here: https://charm.cs.illinois.edu/gerrit/#/c/1... Sam White
12:11 PM Bug #1118 (Merged): syncft: sporadic ampi/jacobi3d crash due to mis-matched reduction callback
Sam White

07/13/2016

01:29 PM Bug #1118 (Implemented): syncft: sporadic ampi/jacobi3d crash due to mis-matched reduction callback
https://charm.cs.illinois.edu/gerrit/#/c/1309/ Sam White
11:56 AM Bug #1118 (In Progress): syncft: sporadic ampi/jacobi3d crash due to mis-matched reduction callback
Sam White
11:56 AM Bug #1118: syncft: sporadic ampi/jacobi3d crash due to mis-matched reduction callback
This is the code for the contribution to the reduction that starts the checkpoint and sometimes results in "mis-match... Sam White

07/12/2016

04:29 PM Bug #1138 (Merged): Buffer overflow in AMPI_Reduce_scatter_block
https://charm.cs.illinois.edu/gerrit/#/c/1305/ Sam White
03:29 PM Bug #1138 (Merged): Buffer overflow in AMPI_Reduce_scatter_block
Reduce_scatter_block has an error in it currently, discovered when debugging MPI_Dist_graph_create. The problem is th... Sam White
09:10 PM Bug #1137 (Merged): Templated derived chares throw compilation error for ckCreateEmptyArrayAsyc m...
Phil Miller
09:07 PM Bug #1137 (Implemented): Templated derived chares throw compilation error for ckCreateEmptyArrayA...
Fix: https://charm.cs.illinois.edu/gerrit/#/c/1304/
The fix included modification to charmxi where the definition ...
Nitin Bhat
07:43 PM Bug #1137 (Merged): Templated derived chares throw compilation error for ckCreateEmptyArrayAsyc m...
The bug is because it is required in C++ to specify the object/class in order to perform a name lookup for function c... Nitin Bhat

07/11/2016

04:49 PM Documentation #1136: Document ability of [sync] entry methods to return any type that supports PUP
The manual still says that @[sync]@ entry methods can only return messages Phil Miller
04:48 PM Documentation #1136 (Merged): Document ability of [sync] entry methods to return any type that su...
Phil Miller
04:46 PM Feature #1135 (New): Pass a future into or return one from [sync] entry methods, so that the call...
Right now, the generated code in the CProxy object for a @[sync]@ entry method dispatches the call to the recipient, ... Phil Miller
03:14 PM Bug #1050 (Implemented): AMPI collectives support for MPI_IN_PLACE
Sam White

07/10/2016

09:49 AM Bug #1118: syncft: sporadic ampi/jacobi3d crash due to mis-matched reduction callback
Another occurrence: https://charm.cs.illinois.edu/autobuild/old.2016_07_10__01_01/netlrts-linux-x86_64-syncft.txt Phil Miller

07/07/2016

03:30 PM Feature #331 (Merged): Return any type that has a PUP routine from [sync] entry methods
Vipul Harsh
03:30 PM Feature #336 (Merged): Add "return value marshalling" to entry methods
Vipul Harsh
10:22 AM Projections Feature #1057 (In Progress): Just user event tracing
Ronak Buch

07/06/2016

02:39 PM Support #1079 (In Progress): Remove deprecated machine layers and retired machines from Autobuild
1- darwin 32 builds are removed.
2- net-linux-x86_64-icc on Finessse is removed, uth-linux-x86_64 is migrated to Bea...
Bilge Acun
10:33 AM Feature #1105 (Implemented): AMPI support for large counts
Sam White

07/01/2016

03:31 PM Feature #1133 (Merged): PMPI_ interface for AMPI
See section 14.2.6 of the MPI-3.1 standard here: https://www.mpi-forum.org/docs/mpi-3.1/mpi31-report.pdf Sam White
02:15 PM Feature #870: SDAG methods marked as [sync] should only return when run to completion
Reassigning to Vipul since he has more experience in this area than Seon, and Robert Pavel from LANL emailed me with ... Sam White
12:14 PM Bug #1118 (Merged): syncft: sporadic ampi/jacobi3d crash due to mis-matched reduction callback
https://charm.cs.illinois.edu/autobuild/old.2016_06_29__01_02/netlrts-linux-x86_64-syncft.txt... Phil Miller
07:02 PM Bug #1063: Error in buddy assignment code for checkpointing
There is performance difference between the two modes depending on checkpoint size and applications. The original mod... Xiang Ni

06/30/2016

06:05 PM Bug #1063: Error in buddy assignment code for checkpointing
Eric and I found this issue when we were reading through the code for other purposes. Is there a reason this version ... Phil Miller
05:55 PM Bug #1063 (In Progress): Error in buddy assignment code for checkpointing
I think this part of code is not used by default (when CMK_CHKP_ALL is 1).
The CMK_CHKP_ALL mode is designed to av...
Xiang Ni

06/29/2016

05:58 PM Feature #331: Return any type that has a PUP routine from [sync] entry methods
https://charm.cs.illinois.edu/gerrit/1294
Linking implementation.
Phil Miller
05:12 PM Feature #331 (Implemented): Return any type that has a PUP routine from [sync] entry methods
Vipul Harsh
02:24 PM Bug #1115: verbs slower than mpi on Omni-Path
Recommendation from Intel is that OFI should beat Verbs on Omni-Path. Jim Phillips
09:02 AM Feature #1114 (In Progress): Build and link ROMIO on AMPI by default
Note: parallel make with ROMIO tries to run ROMIO's configure in parallel, resulting in failure. Sam White
08:28 AM Feature #1117 (New): Shrink-expand malleable jobs should be able to change node set
As originally implemented, shrink-expand works with a fixed list of nodes specified at startup. In many realistic env... Phil Miller
03:04 AM Feature #1088: Trace MPI_ functions in AMPI
Discussing this with Ronak, realized we should use system events rather than user events like Charm does for entry me... Sam White
03:01 AM Bug #668 (In Progress): ampi/megampi test fails with randomized queues
Sam White
02:58 AM Bug #1112 (Merged): AMPI thread migration fails on Edison gni-crayxc
Sam White

06/28/2016

07:34 AM Feature #1116 (Merged): AMPI broadcasts should be [nokeep]
AMPI_Bcast currently calls ampi::generic with an AmpiMsg. For broadcasts, we should invoke a version of ampi::generic... Sam White

06/27/2016

04:11 PM Bug #668: ampi/megampi test fails with randomized queues
I built as above and ran megampi for 1000 iterations 10 times (>1 hour), and got no failures. None from mpich-tests/c... Sam White
01:53 PM Bug #668: ampi/megampi test fails with randomized queues
Per the parent task, ... Phil Miller
03:51 PM Bug #1048: Verbs on Bridges at PSC crashes or hangs
Jim Phillips wrote:
> Adding --with-qlogic to the Charm++ build line appears to fix the issue for NAMD.
>
> There...
Jim Phillips
01:40 PM Feature #236: Charm++ Shrink-Expand (malleable jobs, dynamically varying node count)
Make the subject more explicit, so that readers who don't know the lingo can understand this at a glance. Phil Miller
09:41 PM Bug #1115: verbs slower than mpi on Omni-Path
Note that building Charm++ "--with-qlogic" is necessary for verbs or net-ibverbs to run. Jim Phillips
09:39 PM Bug #1115 (New): verbs slower than mpi on Omni-Path
On PSC Bridges, NAMD running apoa1 on 8 nodes on mpi (OpenMPI) is 20% faster than verbs or net-ibverbs for Charm++ 6.... Jim Phillips

06/26/2016

10:00 AM Bug #1038 (In Progress): Add support for srun to charmrun script for gni layer
Getting there, but still not quite:... Phil Miller
 

Also available in: Atom