Project

General

Profile

Activity

From 07/19/2018 to 08/17/2018

08/17/2018

05:29 PM Bug #1949: Ensure that 'End of Program' message is printed consistently for every charm program e...
Does this happen with and/or without Charmrun? Evan Ramos
05:24 PM Bug #1954 (Merged): Verbs Init code hangs when it tries connecting an inactive device
Evan Ramos
04:41 PM Cleanup #1967 (Implemented): Section creation API should take its arguments as std::vector's inst...
https://charm.cs.illinois.edu/gerrit/#/c/charm/+/4487/ Sam White
12:26 PM Cleanup #1967: Section creation API should take its arguments as std::vector's instead of arrays
Fixed memory leaks in example here: https://charm.cs.illinois.edu/gerrit/#/c/charm/+/4482/ Sam White
12:11 PM Cleanup #1967 (Merged): Section creation API should take its arguments as std::vector's instead o...
The API currently takes several C-style arrays, but could be cleaned up to take std::vector's instead (or as well, ca... Sam White

08/16/2018

01:29 PM Bug #1966 (New): ofi non smp fails when #PUs > #cores
Get this error with a 128 cpus per task, 1 task per node batch file AND with a 128 task per node, 1 cpus per task bat... Joseph Hutter

08/15/2018

03:04 PM Bug #1395 (Merged): AMPI_Get_elements returns incorrect results for struct type
Sam White

08/14/2018

02:54 PM Bug #1965 (Merged): LAMMPS on AMPI fails after migration due to AMPI not wrapping posix_memalign()
The application's data stuructures appear to be corrupted after migrating when using -memory isomalloc, though the is... Sam White
11:07 AM Feature #1446 (Merged): AMPI support for generalized requests
Sam White

08/13/2018

05:48 PM Bug #1954 (Implemented): Verbs Init code hangs when it tries connecting an inactive device
https://charm.cs.illinois.edu/gerrit/4474 Evan Ramos
02:44 PM Feature #1964 (New): Enable +isomalloc_sync by default for AMPI
If running an AMPI application, it's often necessary to pass the runtime option '+isomalloc_sync' to ensure correctne... Sam White
12:28 PM Feature #177 (Merged): objid_t: load balancing infrastructure should use objid_t
Sam White

08/12/2018

11:51 AM Feature #1953 (In Progress): cmake-based build system
Matthias Diener
11:37 AM Bug #1962 (Merged): AMPI aborts from objid.h when run with >= 8192 ranks in non-production builds
Sam White

08/11/2018

02:38 PM Bug #1811 (Merged): Cross-array section reduction and multicast with callback set at contribute time
Sam White
02:38 PM Bug #1950 (Merged): Isomalloc produces misleading memory usage numbers
Sam White

08/10/2018

06:05 PM Feature #934 (Implemented): Isomalloc: refactor to associate migratable allocations with entities...
https://charm.cs.illinois.edu/gerrit/4472
With my latest push implementing lazy allocation (which only takes place...
Evan Ramos
05:08 PM Feature #934: Isomalloc: refactor to associate migratable allocations with entities other than pe...
If I break on mempool_init in maxReduce, the crashing PE hits the breakpoint 16 times, and the other three PEs hit it... Evan Ramos
04:07 PM Feature #934: Isomalloc: refactor to associate migratable allocations with entities other than pe...
Patch revival: https://charm.cs.illinois.edu/gerrit/4472
The remaining issue from the initial autobuild failures, ...
Evan Ramos
02:05 PM Cleanup #1872 (Implemented): Move performance tests and benchmarks from "make test" to a new "mak...
https://charm.cs.illinois.edu/gerrit/4218 Evan Ramos
11:25 AM Bug #1961 (Merged): Unbounded memory usage in LBDB
Evan Ramos

08/09/2018

05:33 PM Bug #1950 (Implemented): Isomalloc produces misleading memory usage numbers
https://charm.cs.illinois.edu/gerrit/4470... Evan Ramos
04:13 PM Bug #1950 (In Progress): Isomalloc produces misleading memory usage numbers
What is happening is that @-memory isomalloc@ defines its own CmiMemoryUsage that reports a statistic that is interna... Evan Ramos
02:44 PM Feature #1963: AMPI implements subcommunicators in an unscalable fashion
An alternative would be to use sections to implement subcommunicators, with additional data structures to keep track ... Sam White
02:39 PM Feature #1963 (New): AMPI implements subcommunicators in an unscalable fashion
This is the real issue behind the following: https://charm.cs.illinois.edu/redmine/issues/1962
MPI_COMM_SELF is th...
Sam White

08/08/2018

04:47 PM Bug #1962 (Implemented): AMPI aborts from objid.h when run with >= 8192 ranks in non-production b...
The workaround should be sufficient for 6.9.0 but longer term I'll open new redmine issues for the remaining work on ... Sam White
04:46 PM Bug #1962: AMPI aborts from objid.h when run with >= 8192 ranks in non-production builds
Here's the workaround, in which I somewhat arbitrarily chose the new ratio of collection vs element bits so that we c... Sam White
03:59 PM Bug #1962: AMPI aborts from objid.h when run with >= 8192 ranks in non-production builds
Core discussion:
- MPI_COMM_SELF should not be implemented using single-element chare arrays. The alternative is to ...
Sam White
02:34 PM Bug #1962: AMPI aborts from objid.h when run with >= 8192 ranks in non-production builds
This issue should have been showing up since this change to implement MPI_COMM_SELF as chare arrays: https://charm.cs... Sam White
02:27 PM Bug #1962: AMPI aborts from objid.h when run with >= 8192 ranks in non-production builds
So we could change the implementation of MPI_COMM_SELF from a 1-element chare array to just a C++ object, though then... Sam White
02:20 PM Bug #1962: AMPI aborts from objid.h when run with >= 8192 ranks in non-production builds
I think I know the problem here. When you launch an AMPI program with `n` VPs, you get a TCharm chare array with `n` ... Sam White
01:35 PM Bug #1962 (Merged): AMPI aborts from objid.h when run with >= 8192 ranks in non-production builds
Even the simplest AMPI program can fail at startup with the following:... Sam White
02:58 PM Feature #693 (In Progress): add CcdCallBacksReset() to header file (or improve callback frequency)
Michael Robson
02:02 PM Bug #1960: tests/charm++/megatest hangs on netlrts-win-x86_64-smp build
I was unable to reproduce this with a non-production debug build (-g -O0) Nitin Bhat
10:20 AM Bug #1960 (Merged): tests/charm++/megatest hangs on netlrts-win-x86_64-smp build
... Nitin Bhat
01:44 PM Bug #1961 (Implemented): Unbounded memory usage in LBDB
https://charm.cs.illinois.edu/gerrit/#/c/charm/+/4460/ Ronak Buch
01:33 PM Bug #1961: Unbounded memory usage in LBDB
I think it would be good to have this fixed for 6.9.0 since it should be simple and non-API-breaking Sam White
01:33 PM Bug #1961 (Merged): Unbounded memory usage in LBDB
Ronak mentioned finding a vector in LBDB that stores location records, which we simply do 'push_back()' on always, ev... Sam White
12:51 PM Bug #1956: tests/charm++/sdag/migration and tests/charm++/sdag/anytimeMigration fail on mpi-win-x...
On a related note, the format of testing for mpi-win-smp seems to be running all tests with PPN = 1. Which means its... Eric Bohm
11:33 AM Bug #1955: tests/charm++/chkpt hangs for mpi-win-x86_64-smp
This bug is also seen on charm-6.8.2. Nitin Bhat
11:26 AM Projections Bug #459 (Merged): inline and local entry method can not be tracked back
Fixed in https://charm.cs.illinois.edu/gerrit/#/c/charm/+/3723/ Ronak Buch

08/07/2018

05:46 PM Feature #65 (Merged): MetaLB: Separate cost/benefit analysis for running LB strategy and performi...
The MetaBalancer patch has been merged. The random forest model will be updated to include more training runs. Chare ... Kavitha Chandrasekar
04:40 PM Bug #1959: examples/charm++/TRAM/randomAccessGroup crashes on mpi-win-x86_64-smp with debug optio...
examples/charm++/TRAM/randomAccessArray also fails in a similar manner. However, aggregateRandomAccessArray and aggre... Nitin Bhat
04:37 PM Bug #1959 (New): examples/charm++/TRAM/randomAccessGroup crashes on mpi-win-x86_64-smp with debug...
Charm build command: ./build LIBS mpi-win-x86_64 smp --enable-error-checking --without-romio --suffix=debug -j8 -g -O... Nitin Bhat
04:04 PM Feature #1952 (Merged): Pass CkNcpyBuffer instead of a separate CkNcpyAck object in the callback ...
Nitin Bhat
03:51 PM Projections Bug #1958 (New): "View" -> "Display Legend" is broken
The button to display a legend of entry method colors is broken and appears to do nothing right now. Sam White
03:49 PM Bug #1957 (Merged): Out of bounds std::vector accesses in NDMeshStreamer
Reproducible with the following one-liner:... Nitin Bhat
03:24 PM Bug #1956 (New): tests/charm++/sdag/migration and tests/charm++/sdag/anytimeMigration fail on mpi...
Charm build command: ./build LIBS mpi-win-x86_64 smp --enable-error-checking --without-romio --suffix=debug -j8 -g -O... Nitin Bhat
10:39 AM Bug #1947 (Merged): Charm pingpong hangs intermittently for mpi-win-x86_64-smp build
Nitin Bhat
08:50 AM Bug #1955 (New): tests/charm++/chkpt hangs for mpi-win-x86_64-smp
... Nitin Bhat

08/06/2018

01:56 PM Bug #1947 (Implemented): Charm pingpong hangs intermittently for mpi-win-x86_64-smp build
Fix: https://charm.cs.illinois.edu/gerrit/#/c/charm/+/4450/
The hang was because MPI was initialized in MPI_THREAD...
Nitin Bhat
01:37 PM Bug #1954 (Merged): Verbs Init code hangs when it tries connecting an inactive device
This issue was encountered when Yong Qin from Mellanox, tried running a charm++ application on their cluster which ha... Nitin Bhat

08/03/2018

04:38 PM Feature #1953: cmake-based build system
https://charm.cs.illinois.edu/gerrit/#/c/charm/+/4418/ Sam White

08/01/2018

05:07 PM Cleanup #1948: Rename Charm's graph.h to something less generic
RecBisectBfLB works okay after updating num_partitions to use CkNumPes(). It seems to balance load okay on a single n... Kavitha Chandrasekar
03:28 PM Feature #65 (Implemented): MetaLB: Separate cost/benefit analysis for running LB strategy and per...
Juan Galvez
03:28 PM Feature #177 (Implemented): objid_t: load balancing infrastructure should use objid_t
Juan Galvez
10:59 AM Bug #1395 (Implemented): AMPI_Get_elements returns incorrect results for struct type
https://charm.cs.illinois.edu/gerrit/#/c/charm/+/4341/ Sam White

07/31/2018

04:04 PM Bug #1947: Charm pingpong hangs intermittently for mpi-win-x86_64-smp build
Handing it off to Nitin. Evan Ramos
03:44 PM Bug #1947 (In Progress): Charm pingpong hangs intermittently for mpi-win-x86_64-smp build
I am now comparing @mpi-win-x86_64-smp@ to @mpi-linux-x86_64-smp@. My comparable launch line on Linux is @mpiexec -n ... Evan Ramos
02:14 PM Feature #1953 (In Progress): cmake-based build system
Our current home-grown build system consists of:
* the ./build shell script
* the ./smart-build.pl Perl script
* a...
Matthias Diener
11:23 AM Cleanup #1948: Rename Charm's graph.h to something less generic
Matthias Diener wrote:
> RecBisectBfLB does not seem to compile, and I think we should consider removing RecBisectBf...
Evan Ramos

07/30/2018

05:35 PM Bug #1947: Charm pingpong hangs intermittently for mpi-win-x86_64-smp build
If I change @warmupRun = true;@ to @warmupRun = false;@, then pingpong prints that 1D Arrays and 1D Threaded Arrays t... Evan Ramos
03:58 PM Bug #1947: Charm pingpong hangs intermittently for mpi-win-x86_64-smp build
Some verbose tracing messages from mpiexec:... Evan Ramos
03:40 PM Bug #1947: Charm pingpong hangs intermittently for mpi-win-x86_64-smp build
False alarm on the spinlock. It looks like that thread is the comm thread, and it is trying to pump messages:
<pre...
Evan Ramos
03:20 PM Bug #1947: Charm pingpong hangs intermittently for mpi-win-x86_64-smp build
Using my local machine, I managed to compile @mpi-win-x86_64-smp-gcc@ with Microsoft MPI 9.1 and MinGW-w64 GCC 7.3.0 ... Evan Ramos
12:20 PM Cleanup #1948: Rename Charm's graph.h to something less generic
RecBisectBfLB does not seem to compile, and I think we should consider removing RecBisectBfLB and the graph.h stuff. Matthias Diener
11:37 AM Feature #1952 (Implemented): Pass CkNcpyBuffer instead of a separate CkNcpyAck object in the call...
Gerrit: https://charm.cs.illinois.edu/gerrit/#/c/charm/+/4399/ Nitin Bhat
10:37 AM Cleanup #1897 (Merged): Cleanup AMPI DDT
Nitin Bhat

07/27/2018

04:41 PM Bug #1097 (Implemented): Attributes are not duplicated in MPI_Comm_dup
Fix to the various mpich attr tests: https://charm.cs.illinois.edu/gerrit/#/c/charm/+/4413/ Sam White
02:48 PM Cleanup #1897: Cleanup AMPI DDT
Error codes and constants:
https://charm.cs.illinois.edu/gerrit/#/c/charm/+/4408/
https://charm.cs.illinois.edu/ger...
Matthias Diener
08:56 PM Bug #937 (Merged): [local] entry methods don't set tracing event dependencies
Note that the '+tracenested' runtime option is needed to generate correct traces. Sam White
08:53 PM Bug #1951 (Merged): Isomalloc failure on netlrts-darwin-x86_64
Sam White

07/26/2018

03:35 PM Bug #1947: Charm pingpong hangs intermittently for mpi-win-x86_64-smp build
@mpi-win-x86_64-smp@ issues seem to be nondeterministic. In today's autobuild http://charm.cs.illinois.edu/autobuild/... Evan Ramos
12:15 PM Bug #1947: Charm pingpong hangs intermittently for mpi-win-x86_64-smp build
I can't manually reproduce this issue on Dexterity, even when copying the build lines from the autobuild log verbatim. Evan Ramos
02:49 PM Bug #1951 (Implemented): Isomalloc failure on netlrts-darwin-x86_64
https://charm.cs.illinois.edu/gerrit/4401 Evan Ramos
01:54 PM Bug #1951: Isomalloc failure on netlrts-darwin-x86_64
... Evan Ramos
01:13 PM Bug #1951: Isomalloc failure on netlrts-darwin-x86_64
I did this:... Evan Ramos
12:40 PM Bug #1951: Isomalloc failure on netlrts-darwin-x86_64
It gets stranger: I compiled with ASan and in that case execution successfully continues past the troublesome malloc ... Evan Ramos
11:07 AM Bug #1951: Isomalloc failure on netlrts-darwin-x86_64
@./jacobi.iso 1 1 1 1@ is sufficient to reproduce the crash. Evan Ramos
09:06 PM Bug #1951: Isomalloc failure on netlrts-darwin-x86_64
This is a weird one. It happens at the first call to meta_malloc, but a small number of meta_reallocs occur successfu... Evan Ramos
02:08 PM Feature #1952 (Merged): Pass CkNcpyBuffer instead of a separate CkNcpyAck object in the callback ...
Nitin Bhat

07/25/2018

11:55 AM Bug #1951 (Merged): Isomalloc failure on netlrts-darwin-x86_64
This only happens on a non-SMP darwin build with '-memory isomalloc', running './charmrun +p2 ./jacobi.iso 2 2 2 40 +... Sam White
11:40 AM Bug #1950: Isomalloc produces misleading memory usage numbers
In the above output, you can see that without Isomalloc, the memory usage remains relatively constant throughout the ... Sam White
11:37 AM Bug #1950 (Merged): Isomalloc produces misleading memory usage numbers
Without Isomalloc (using PUP instead):... Sam White

07/24/2018

03:23 PM Bug #1097 (In Progress): Attributes are not duplicated in MPI_Comm_dup
Sam White

07/20/2018

04:29 PM Feature #1428: AMPI TLS privatization support for IBM POWER
Sam White wrote:
> Only slightly related: our QuickThreads package lacks support for POWER, but it looks like a more...
Evan Ramos
12:55 PM Bug #1949 (New): Ensure that 'End of Program' message is printed consistently for every charm pro...
Build: verbs-linux-x86_64-smp
Machine: Golub
Iteratively run the hello example or any other simple program with m...
Nitin Bhat
11:05 AM Bug #1375 (Merged): os-isomalloc failures during startup on SMP builds
Sam White
07:40 AM Cleanup #1948: Rename Charm's graph.h to something less generic
Quinoa discussion: https://github.com/quinoacomputing/quinoa/pull/255#discussion_r203850796 Sam White

07/19/2018

04:46 PM Cleanup #1948: Rename Charm's graph.h to something less generic
Seems to be used only by RecBisectBfLB. Matthias Diener
04:36 PM Cleanup #1948 (Merged): Rename Charm's graph.h to something less generic
src/util/graph.h ends up in Charm's include/ directory, meaning that if an application tries to #include "graph.h" it... Sam White
02:41 PM Feature #1446 (Implemented): AMPI support for generalized requests
https://charm.cs.illinois.edu/gerrit/#/c/charm/+/4380/ Sam White
11:56 AM Feature #1446 (In Progress): AMPI support for generalized requests
Sam White
 

Also available in: Atom