Project

General

Profile

Activity

From 03/29/2017 to 04/27/2017

Today

08:52 AM Bug #1538 (New): Support Shrink/Expand in verbs
Shrink/expand was never fully ported to verbs, but only to netlrts.
Necessary changes should be the same as netlrts,...
Bilge Acun
07:29 PM Bug #647: Make MeshStreamer classes [migratable] to support checkpoint/restart
I have completed adding PUP functions for all the involved classes in @VirtualRouter.h@ and for the @MeshStreamer@ ba... Karthik Senthil

04/26/2017

05:54 PM Bug #1493 (Implemented): Deleting an array also deletes all common elements from it's bound arrays
The fix was very straightforward, and is implemented here: https://charm.cs.illinois.edu/gerrit/2448
The only issu...
Eric Mikida
05:14 PM Bug #1493: Deleting an array also deletes all common elements from it's bound arrays
This bug looks to be due to the fact that ~CkMigratable() tries to access myRec in order to get the LBDB database for... Eric Mikida
05:06 PM Bug #1410: Tuple reducer leaks memory when using set/concat/custom reducers
Added documentation on reusing reduction msg memory in custom reducers: https://charm.cs.illinois.edu/gerrit/#/c/2446/ Sam White
04:22 PM Bug #1410 (Implemented): Tuple reducer leaks memory when using set/concat/custom reducers
https://charm.cs.illinois.edu/gerrit/#/c/2444/ Sam White
03:00 PM Bug #1410 (In Progress): Tuple reducer leaks memory when using set/concat/custom reducers
Sam White
01:41 PM Bug #1410: Tuple reducer leaks memory when using set/concat/custom reducers
Ah, I didn't think of that. We can indeed just check the returned pointer against the zeroth message passed into it, ... Sam White
12:05 PM Bug #1410: Tuple reducer leaks memory when using set/concat/custom reducers
Design 1 is also preferable (I think, if I understand correctly) because such a declaration would only have to appear... Phil Miller
11:54 AM Bug #1410: Tuple reducer leaks memory when using set/concat/custom reducers
It looks like the issue here is that all of the builtin reducers--except set and concat--reuse one of their input mes... Sam White
05:06 PM Feature #1398 (Implemented): Document addReducer's new option 'streamable'
https://charm.cs.illinois.edu/gerrit/#/c/2446/ Sam White
10:29 AM Bug #1507: ckio test failure on gni-crayxc
The interesting note from the case I looked into is that a sane array ID is returned into the proxy object, but the m... Phil Miller
10:27 AM Bug #1507: ckio test failure on gni-crayxc
I may have a simpler test case for this, or at least one that exhibits the same CmiAbort behavior. Phil Miller
09:02 AM Bug #1537 (Implemented): Support for Shrink/Expand in 6.8.0
https://charm.cs.illinois.edu/gerrit/#/c/2438/ Bilge Acun
09:55 PM Bug #1537: Support for Shrink/Expand in 6.8.0
Simply replacing _signal_ with _sigaction_ is not giving the same behavior for shrink/expand and the reason is the di... Bilge Acun
08:01 PM Bug #1537: Support for Shrink/Expand in 6.8.0
Bisecting shows that this commit that replaces signal with sigaction is the culprit: https://charm.cs.illinois.edu/ge... Bilge Acun
07:21 PM Bug #1537: Support for Shrink/Expand in 6.8.0
examples/charm++/shrink_expand is broken. Both shrink and expand operation hangs right after restart. Bilge Acun
11:15 PM Feature #1466 (Merged): Update list of available load balancing strategies in the manual
Sam White
11:13 PM Cleanup #1363 (Merged): Remove/deprecate dead machine layers
Sam White
11:12 PM Feature #1458 (Merged): Zero-copy send support for the MPI machine layer
Sam White
11:08 PM Bug #1275: DistributedLB: Objects not migrating after strategy runs
https://charm.cs.illinois.edu/gerrit/#/c/2437/ to fix hang when there are only 2 PEs. Kavitha Chandrasekar
10:31 PM Bug #1527 (Implemented): Unused Parameter Warning in Entry Methods with no Parameters
Fix: https://charm.cs.illinois.edu/gerrit/#/c/2436/1
This patch only targets removing warnings for entry methods t...
Nitin Bhat
07:22 PM Bug #1526 (Implemented): Unused Variable impl_obj in _call_ArchiveChare_CkMigrateMessage
Fix: https://charm.cs.illinois.edu/gerrit/#/c/2435/ Nitin Bhat

04/25/2017

05:42 PM Feature #1468: Enable pre-pinning memory for RDMA message sends
If we just provide a wrapper for malloc and keep track of its pinned-ness, it should be relatively straightforward li... Jaemin Choi
04:56 PM Feature #1468: Enable pre-pinning memory for RDMA message sends
Is this more complicated than providing the following?
A) Add a CkAlloc() routine that just calls CmiAlloc()
B) Mai...
Sam White
03:17 PM Feature #1468: Enable pre-pinning memory for RDMA message sends
Can we move this to 6.8.1? I don't think it will be finished anytime soon. Jaemin Choi
04:58 PM Bug #1537 (Implemented): Support for Shrink/Expand in 6.8.0
Test if it works. Bilge Acun
04:49 PM Bug #1530 (Implemented): Isomalloc on SMP mode always prints warning about +isomalloc_sync, even ...
https://charm.cs.illinois.edu/gerrit/#/c/2433/
The patch moves the warning to memory-isomalloc initialization.
Matthias Diener
04:46 PM Bug #1443: Serialization for std::unique_ptr Fails With Abstract Base Class
Alright, I'll try to get a patch in as quickly as possible for 6.9.0 then. Nils Deppe
04:43 PM Bug #1443: Serialization for std::unique_ptr Fails With Abstract Base Class
I think deferring to 6.9 would be preferable. Eric Mikida
04:39 PM Bug #1443: Serialization for std::unique_ptr Fails With Abstract Base Class
I just realized this would require some C++11 things, @std::enable_if@ and @std::is_base_of@. Rather than adding a te... Nils Deppe
04:28 PM Bug #1443: Serialization for std::unique_ptr Fails With Abstract Base Class
If you post a patch we can definitely take a look and see if we can get it into 6.8.0 Eric Mikida
04:35 PM Feature #1018: AMPI support for MPI-3's distributed graph interface
There is a noted regression in intercommunicators with this patch on gerrit. It's not too important to have before 6.8.0 Sam White
04:34 PM Support #865 (Rejected): Update section 7.3 of Manual on Load Balancing Chare arrays
Sam White
04:33 PM Feature #870: SDAG methods marked as [sync] should only return when run to completion
Re-assigning to Eric M since he has more experience with SDAG... It would be really nice to have this in 6.8.0 but I ... Sam White
04:12 PM Bug #1520: multicore-darwin-x86_64 megatest hangs when built with --enable-randomized-msgq --with...
The other related test failures with randomized queues are targeted to 6.8.1. Sam White
04:08 PM Bug #1227: Support template entry methods in generated code from TRAM [aggregate] entry method at...
Being deferred to 6.8.1. The current workaround is to use overloaded entry methods for each type, instead of a templa... Eric Mikida
03:59 PM Bug #833: mpi smp build is locked to one core per node by default
We're not immediately going to worry about the case where the user gives a bad pemap/commap argument set. Phil Miller
03:57 PM Bug #833: mpi smp build is locked to one core per node by default
Eric Bohm will test this on a machine that is known to have bad default mapping Sam White
03:59 PM Bug #931: avoid failing back to hostname ip lookup for physical node detection
Judgement is that this is relatively low impact. Deferring. Phil Miller
03:54 PM Feature #1237: Onesided sender side implementation for GNI layer
Implementing PUT calls from Sender to Receiver for non-aligned memory addresses as RDMA GET expects 4-byte aligned ad... Nitin Bhat
03:52 PM Bug #529: LrtsInitCpuTopo() operates asynchronously and overlaps with group constructors
Based on the preference for MPI machine layer on OmniPath systems for now, we're deferring this. Phil Miller
03:19 PM Bug #1529 (Implemented): Easy build-time option to elide LB support (cut tracing overhead, etc)
https://charm.cs.illinois.edu/gerrit/2431 Phil Miller
02:57 PM Bug #1529 (In Progress): Easy build-time option to elide LB support (cut tracing overhead, etc)
Phil Miller
03:18 PM Bug #1528 (Implemented): charmxi: Message type with no variable arrays can't be declared with { }
https://charm.cs.illinois.edu/gerrit/2432 Phil Miller
11:06 AM Feature #1449: AMPI support for MPI_Win_allocate_shared
Also need to support MPI_Win_query_shared Sam White
10:59 AM Feature #969: AMPI support for collectives on inter-communicators
(I)Barrier: https://charm.cs.illinois.edu/gerrit/#/c/2342/
(I)Scatter(v): https://charm.cs.illinois.edu/gerrit/#/c/2...
Sam White
10:57 PM Bug #901: Threads awoken by CthAwaken don't let Projections trace back to the event that woke them
With issue #901 resolved now, this issue has become a more visible sore of AMPI Projections usage. Note that AMPI thr... Sam White
10:10 PM Feature #1088 (Implemented): Trace MPI_ functions in AMPI
Matthias Diener
08:34 PM Cleanup #1535 (Implemented): Format/Indent all source code files that are a part of charmxi (xlat-i)
Used clang-format to format all .C and .h files in src/xlat-i/
Fix: https://charm.cs.illinois.edu/gerrit/#/c/2429/
Nitin Bhat
07:27 PM Feature #1536 (New): AMPI interface for mapping ranks to worker threads
AMPI can take advantage of the fact that ranks are 1D dense indices for MPI_COMM_WORLD and allow users to implement a... Sam White

04/24/2017

01:38 PM Feature #1088: Trace MPI_ functions in AMPI
WIP patch here: https://charm.cs.illinois.edu/gerrit/2426 Matthias Diener
01:38 PM Feature #1088 (In Progress): Trace MPI_ functions in AMPI
Matthias Diener
10:23 AM Cleanup #1535: Format/Indent all source code files that are a part of charmxi (xlat-i)
This patch include a clang-format file that can be used to do this automatically: https://charm.cs.illinois.edu/gerri... Sam White
10:14 AM Cleanup #1535 (Implemented): Format/Indent all source code files that are a part of charmxi (xlat-i)
The indentation for the source code files in src/xlat-i is quite messed up and non-uniform with different regions of ... Nitin Bhat
10:08 AM Support #1534 (New): Print out the job output, error and log for failed autobuild jobs launched d...
This was discussed in the Core meeting. It would be good to have the output, error and log to determine the failure r... Nitin Bhat
09:42 AM Bug #1533: State Space Search Examples Broken
Reviving that code is probably far down our list of priorities right now, we might as well have the fix for broken co... Sam White
02:50 AM Bug #1533 (New): State Space Search Examples Broken
The examples in examples/charm++/state_space_searchengine/ currently don't compile, due to "error: ‘class Main’ has n... Ronak Buch
12:06 AM Cleanup #1265: Document LLVM OpenMP runtime integration
https://charm.cs.illinois.edu/gerrit/#/c/2427/
Completed documentation and add examples to show how to build and u...
Seonmyeong Bak
11:57 PM Cleanup #1265 (Implemented): Document LLVM OpenMP runtime integration
Seonmyeong Bak

04/23/2017

02:01 PM Feature #1532 (New): Example for using Google benchmark library with Charm++
The attached code shows how to use Google's benchmark library with Charm++. It's nice because it automatically does e... Phil Miller
01:59 PM Bug #1531 (New): Main (scheduler) thread can suspend, but that confuses QD, where other thread su...
If I remove the @[threaded]@ tag from the attached code, it hangs at the first QD, while it doesn't with threaded in ... Phil Miller
01:55 PM Bug #1530 (Implemented): Isomalloc on SMP mode always prints warning about +isomalloc_sync, even ...
When running on an smp build, all programs print the message "@Warning> using Isomalloc in SMP mode, you may need to ... Phil Miller
01:49 PM Bug #1529 (Implemented): Easy build-time option to elide LB support (cut tracing overhead, etc)
Right now, to build without LB support, one has to edit the target build's conv-mach.h to change @#define CMK_LBDB_ON... Phil Miller
01:44 PM Bug #1528 (Implemented): charmxi: Message type with no variable arrays can't be declared with { }
The snippet... Phil Miller
10:22 AM Bug #1523: Verbs RDMA send fails on 0-byte sized message
I only see this error on Cab at LLNL, which has a QLogic Infiniband interconnect. On iForge at NCSA, which has a Mell... Sam White
10:21 AM Bug #1526: Unused Variable impl_obj in _call_ArchiveChare_CkMigrateMessage
Okay, that could be related then. Currently we have the migration constructors defined but are empty, e.g.... Nils Deppe
10:18 AM Bug #1526: Unused Variable impl_obj in _call_ArchiveChare_CkMigrateMessage
We recently made migration constructors unnecessary for chares that never actually migrate, so I think this is relate... Sam White
08:38 AM Bug #1526 (Implemented): Unused Variable impl_obj in _call_ArchiveChare_CkMigrateMessage
I get an unused variable warning in Charm++ v6.8.0-beta2 that was not present in v6.7.1. It appears in the generated ... Nils Deppe
10:19 AM Bug #1527: Unused Parameter Warning in Entry Methods with no Parameters
Ah okay, that explains it. I haven't used entry options (yet) so I wasn't able to provide more info. Thanks for looki... Nils Deppe
10:16 AM Bug #1527: Unused Parameter Warning in Entry Methods with no Parameters
It used to be that we didn't actually support CkEntryOptions on entry methods taking no arguments, but we recently ad... Sam White
09:27 AM Bug #1527 (Implemented): Unused Parameter Warning in Entry Methods with no Parameters
In v6.8.0-beta2 entry methods that do not have any parameters emit an unused parameter warning for @impl_e_opts@. The... Nils Deppe
09:00 AM Bug #1493: Deleting an array also deletes all common elements from it's bound arrays
This is the autobuild error seen on netlrts-linux-smp in examples/charm++/load_balancing/stencil3d Phil Miller
08:59 AM Bug #1493 (In Progress): Deleting an array also deletes all common elements from it's bound arrays
So, this results in a use-after-free error, as follows:... Phil Miller
07:12 PM Bug #1443: Serialization for std::unique_ptr Fails With Abstract Base Class
This issue is present in the v6.8.0 beta. Could it be resolved before v6.8.0 final? I can make a pull request with th... Nils Deppe

04/22/2017

10:54 AM Bug #1522: Verbs failure on small messages
This also happens on the iForge cluster at NCSA... Sam White
08:02 PM Bug #1522 (New): Verbs failure on small messages
If I run tests/charm++/pingpong on Cab @ LLNL with a payload size smaller than 9 bytes, it get the following:
<pre...
Sam White
10:30 AM Feature #1468: Enable pre-pinning memory for RDMA message sends
Phil suggested a complementary optimization on the GNI RDMA patch (https://charm.cs.illinois.edu/gerrit/#/c/1908/) wh... Sam White
10:27 AM Feature #68 (Closed): LRTS support for setting up a message to send and transmitting a GET handle
I think this issue is covered by the combination of the separate sender-side and receiver-side issues. Sam White
10:25 AM Feature #1236: Avoid receiver-side copy for large contiguous messages. API for charm and converse...
It would be good to know what is working and what isn't in this patch: https://charm.cs.illinois.edu/gerrit/#/c/1939/ Sam White
10:18 AM Feature #1237: Onesided sender side implementation for GNI layer
It looks like Nitin has been doing work on this too, so might need to coordinate. Sam White
10:14 AM Feature #195 (New): Modify old charm++/converse onesided/direct API and examples to new rdma API
Getting GNI and NetLRTS RDMA support working, plus getting real applications (NAMD, OpenAtom, ChaNGa) to use the API,... Sam White
07:06 AM Projections Feature #1525 (New): Bracketed User Events Track Numerical Values
It would be really useful to be able to pass a @vector<double>@ to the @traceUserBracketedEvent@ function to record v... Nils Deppe
10:39 PM Projections Feature #1524 (New): Time Profile With Bracketed User Events
Because of the structure of our code, we have only one entry method which then forwards to the correct member functio... Nils Deppe
08:03 PM Bug #1523 (New): Verbs RDMA send fails on 0-byte sized message
Obviously users shouldn't use RDMA sends for 0-byte sized arrays, but we shouldn't fail on this.
When I run AMPI o...
Sam White

04/21/2017

09:37 AM Bug #1267 (Merged): Integrate METIS graph partitioning library patched for global variable privat...
Phil Miller
09:36 AM Feature #1297 (Merged): Patch METIS to privatize global variables
Phil Miller
09:41 PM Feature #1521: CkIO file read support
Semi-related stackoverflow question: http://stackoverflow.com/questions/43272225/file-handling-support-in-charm Sam White
09:37 PM Feature #1521 (New): CkIO file read support
CkIO currently only supports writing files.
Branch rohan/ckioread: https://charm.cs.illinois.edu/gerrit/gitweb?p=c...
Sam White
07:22 PM Feature #1378: 64-bit Charm message sizes
Abort before trying to migrate a chare with size greater than a message can hold: https://charm.cs.illinois.edu/gerri... Sam White

04/20/2017

02:47 PM Bug #1510: Hang in tests/charm++/chkpt when using -tracemode perfReport
Trying this on linux now (instead of darwin) it works, so we should be safe to add "-tracemode perfReport" to netlrts... Sam White
01:08 PM Bug #931: avoid failing back to hostname ip lookup for physical node detection
That might be helpful? I haven't even begun to this one, and have two other 6.8 bugs that I've already started and am... Eric Mikida
10:41 AM Bug #931: avoid failing back to hostname ip lookup for physical node detection
Should the be re-assigned, perhaps to Juan? Phil Miller
12:34 PM Bug #1518 (Merged): net-linux-x86_64-ibverbs-smp-iccstatic build failure in machine-ibverbs.c
Jaemin Choi
11:34 AM Bug #1518: net-linux-x86_64-ibverbs-smp-iccstatic build failure in machine-ibverbs.c
Non-lrts layers do not include @util/machine-common-core.c@ and do not print information about the comm threads to th... Jaemin Choi
12:27 PM Bug #1520: multicore-darwin-x86_64 megatest hangs when built with --enable-randomized-msgq --with...
There are multiple other bugs in our test suite noted in issue #259. Sam White
11:21 AM Bug #1520: multicore-darwin-x86_64 megatest hangs when built with --enable-randomized-msgq --with...
Hangs in test 0 on multicore-linux64-iccstatic --no-build-shared --enable-randomized-msgq --with-prio-type=int --enab... Jim Phillips
11:16 AM Bug #1520 (New): multicore-darwin-x86_64 megatest hangs when built with --enable-randomized-msgq ...
multicore-darwin-x86_64 --no-build-shared --enable-randomized-msgq --with-prio-type=int --enable-error-checking -debu... Jim Phillips
10:37 AM Bug #1442: CkLoop fixed tree limits helper recruitment
I'm inclined to defer this as not release-critical, since CkLoop could do better, but it's not something that's gotte... Phil Miller
10:33 AM Bug #987 (Merged): Entry methods with no parameters can't be called with CkEntryOptions, can't ta...
Phil Miller
10:33 AM Bug #1519 (New): Entry methods with no parameters ignore priority from CkEntryOptions
We've fixed zero-argument entry methods to accept CkEntryOptions at the call site, and to respect group construction ... Phil Miller
10:28 AM Bug #901: Threads awoken by CthAwaken don't let Projections trace back to the event that woke them
This isn't a regression relative to 6.7, and I'm not immediately able to get back to this. Deferring for now. Phil Miller
10:01 AM Bug #1509: -tracemode summary always fails an assertion at exit
Looking at the second assertion failure, I'm not too surprised. The exit process goes from the middle of a running en... Phil Miller
09:32 AM Feature #1098: Better CMake support
Christoph and Jozsef encouraged us to take the file above and distribute it with charm. The only issue with Quinoa's ... Sam White

04/19/2017

03:31 PM Bug #1518: net-linux-x86_64-ibverbs-smp-iccstatic build failure in machine-ibverbs.c
I build both. If nothing else I can check for performance regressions. Jim Phillips
03:15 PM Bug #1518: net-linux-x86_64-ibverbs-smp-iccstatic build failure in machine-ibverbs.c
What is the motivation for using net-ibverbs-smp rather than verbs-smp? Sam White
03:07 PM Bug #1518: net-linux-x86_64-ibverbs-smp-iccstatic build failure in machine-ibverbs.c
Almost certainly caused by https://charm.cs.illinois.edu/gerrit/#/c/2360/ Jim Phillips
03:01 PM Bug #1518 (Merged): net-linux-x86_64-ibverbs-smp-iccstatic build failure in machine-ibverbs.c
... Jim Phillips
02:01 PM Bug #1511 (Merged): Darwin: build failure in sockRoutines.c on multicore-darwin-x86_64
Matthias Diener
09:18 AM Bug #1511 (Implemented): Darwin: build failure in sockRoutines.c on multicore-darwin-x86_64
https://charm.cs.illinois.edu/gerrit/#/c/2420/ Sam White
01:16 PM Feature #1098: Better CMake support
Nils Deppe said he was able to take Quinoa's cmake support for Charm and use that directly: https://github.com/quinoa... Sam White
01:12 PM Bug #854 (In Progress): RRMap and BlockLB broken for >1D chare arrays
Dong Hun Lee
10:11 AM Bug #854: RRMap and BlockLB broken for >1D chare arrays
RRMap using Nikhil's previous code.
https://charm.cs.illinois.edu/gerrit/#/c/2421/
Dong Hun Lee
12:52 PM Bug #1509: -tracemode summary always fails an assertion at exit
Okay, I'm able to reproduce it for runs with more than 1 PE without production. Ronak Buch
11:51 AM Bug #1509: -tracemode summary always fails an assertion at exit
You built with production, which disables CmiAsserts, which is the failure I saw. Sam White
11:48 AM Bug #1509: -tracemode summary always fails an assertion at exit
I couldn't reproduce this on the latest Charm (HEAD: ddc864b7e) on netlrts-linux-x86_64 (built with ./build charm++ n... Ronak Buch
11:38 AM Feature #1517 (New): Document PAPI Projections integration
This came up in the advanced tutorial session and there's nothing in the manual about how to use it or what exactly i... Sam White
10:45 AM Feature #1516 (New): Document the Statistics reducer
The Statistics reducer is not documented in the manual at all. Sam White
10:44 AM Feature #1515 (New): Document [inline] entry method optimization for const& args
This came up in the advanced tutorial session following the workshop. The new optimization is not in the manual, and ... Sam White
10:16 AM Bug #1514 (New): Throw a runtime error for registrations that occur after startup
This specifically comes up for templated entry methods that aren't explicitly instantiated in the ci file. If you for... Eric Mikida
09:41 AM Feature #1398: Document addReducer's new option 'streamable'
Advanced tutorial attendees asked about this... Sam White
09:26 AM Bug #1508 (Merged): AMPI incompatibility with shared library version of Charm++
Matthias Diener
09:19 AM Bug #1474 (Merged): mpi-win-x86_64 fails in collidethread example
Matthias Diener

04/18/2017

04:30 PM Support #1512 (New): Measure test coverage
It would be helpful to know the coverage that our tests provide for the core functionality.
One commonly available...
Eric Bohm
04:29 PM Bug #1510: Hang in tests/charm++/chkpt when using -tracemode perfReport
I think we can reasonably defer this, unless we think it's indicative of some broader deficiency that users will enco... Phil Miller
03:28 PM Bug #1511 (Merged): Darwin: build failure in sockRoutines.c on multicore-darwin-x86_64
Probably some header incompatibility:... Matthias Diener
02:46 PM Cleanup #1482: Update Charm++ FAQ
Charm++ tutorial webpage had this question that also can be added to Charm++ FAQ page:
"Is there syntax highlighti...
Bilge Acun
12:08 PM Feature #1481 (Merged): RDMA zero copy send implementation for multicore builds
Sam White
12:08 PM Feature #1504 (Merged): Enable automated testing of tracemodes
Sam White
11:50 AM Bug #1474 (Implemented): mpi-win-x86_64 fails in collidethread example
Implemented a simple fix ( https://charm.cs.illinois.edu/gerrit/#/c/2416/ ). On mpi-win64, AMPI and TCharm were not i... Matthias Diener

04/17/2017

02:30 PM Bug #1509: -tracemode summary always fails an assertion at exit
This seems to happen on *any* program using -tracemode summary?
This is examples/charm++/hello/1darray, which fail...
Sam White
02:10 PM Bug #1509 (New): -tracemode summary always fails an assertion at exit
If you do a make test OPTS="-tracemode summary" in tests/, it fails the first test:... Sam White
02:19 PM Bug #1510 (New): Hang in tests/charm++/chkpt when using -tracemode perfReport
Besides this one test, running make test OPTS="-tracemode perfReport" passes all tests.... Sam White
03:24 AM Cleanup #1265: Document LLVM OpenMP runtime integration
Sam White wrote:
> This should be easy, so bumping priority.
Some APIs and build,charmc and other changes will be...
Seonmyeong Bak
07:33 PM Bug #1508 (Implemented): AMPI incompatibility with shared library version of Charm++
Patch https://charm.cs.illinois.edu/gerrit/#/c/2409/ works around this issue. Matthias Diener

04/16/2017

04:27 PM Bug #1508: AMPI incompatibility with shared library version of Charm++
One way to work around this is by disabling ASLR, i.e. running ... Matthias Diener
04:11 PM Bug #1508: AMPI incompatibility with shared library version of Charm++
To reproduce:... Matthias Diener
03:59 PM Bug #1508 (Merged): AMPI incompatibility with shared library version of Charm++
AMPI crashes at application start-up when using the Charm++ libraries, i.e., when building charm with @./build ... -... Matthias Diener
04:01 PM Bug #1495 (Closed): PlasCom2/advect1d crashes on AMPI with ++local or multiple nodes
PlasCom2, as well as all other AMPI applications, crash during startup when using the shared library version of charm... Matthias Diener
12:27 AM Bug #1495 (In Progress): PlasCom2/advect1d crashes on AMPI with ++local or multiple nodes
Matthias Diener
12:26 AM Bug #1495: PlasCom2/advect1d crashes on AMPI with ++local or multiple nodes
A workaround is to use AMPI_Fallback_Main() instead of the function pointer passed to MPI_threadstart_t, i.e., using ... Matthias Diener
10:08 PM Bug #1495: PlasCom2/advect1d crashes on AMPI with ++local or multiple nodes
Also crashes on a single computer node on stampede1/golub when using ++local:... Matthias Diener
10:04 AM Bug #1507 (New): ckio test failure on gni-crayxc
The ckio test has failed the past 2+ days on autobuild on gni-crayxc, seemingly due to a race condition between array... Sam White
09:01 AM Feature #1504 (Implemented): Enable automated testing of tracemodes
https://charm.cs.illinois.edu/gerrit/2408
Second round fix, tested working.
Phil Miller
08:24 AM Feature #1504 (In Progress): Enable automated testing of tracemodes
Need to make the corresponding fix for bigsim compilation/linkage as well. Phil Miller
08:44 AM Feature #1070 (Closed): Migrate lagging 'net' builds to 'netlrts'
Phil Miller
08:43 AM Support #1496 (Closed): Add -Werror=vla to Jenkins per-commit builds
Done Phil Miller
01:02 AM Bug #1506: examples/hello/4darray breaks when doing sections
Also, this test was not correctly being run by Autobuild, because the makefile has the wrong path to config.mk. That ... Eric Mikida
01:00 AM Bug #1506 (New): examples/hello/4darray breaks when doing sections
Here's the end of the output for a fresh build of netlrts-x86_64 charm. It breaks when it enters phase 2 which is the... Eric Mikida

04/15/2017

04:42 PM Cleanup #1505 (New): Deprecate hapi_pool* in favor of hapi_host*
This is pending Jaemin's patch "#2275":https://charm.cs.illinois.edu/gerrit/#/c/2275/ and that patch that introduces ... Michael Robson
04:32 PM Bug #1487 (Implemented): Leaving -DGPU_MEMPOOl causes gpu manager to not build
https://charm.cs.illinois.edu/gerrit/#/c/2406/ Michael Robson
08:47 AM Feature #1504 (Merged): Enable automated testing of tracemodes
Phil Miller

04/14/2017

05:31 PM Bug #1268: AMPIF issues due to C++ main routine
I think we can provide a Fortran @PROGRAM@ entry point that we'd compile as a stand-alone object file, that just call... Phil Miller
02:26 PM Feature #1504: Enable automated testing of tracemodes
https://charm.cs.illinois.edu/gerrit/2403
Phil Miller
02:23 PM Feature #1504: Enable automated testing of tracemodes
Third option - ignore the tracemode when building a converse program, if that trace mode won't work with just Convers... Phil Miller
02:10 PM Feature #1504 (In Progress): Enable automated testing of tracemodes
Phil Miller
01:38 PM Feature #1504 (Merged): Enable automated testing of tracemodes
We don't currently test tracemode projections/summary/perfReport in Jenkins or autobuild.
To enable that, we need to...
Sam White
02:08 PM Bug #1488 (Implemented): GPU manager runs out of memory on talent
https://charm.cs.illinois.edu/gerrit/#/c/2401/ Michael Robson
02:08 PM Feature #1297: Patch METIS to privatize global variables
Thanks for the nm script. Patch here: https://charm.cs.illinois.edu/gerrit/#/c/2402/ Kavitha Chandrasekar
11:03 AM Feature #1297: Patch METIS to privatize global variables
... Phil Miller
11:44 AM Feature #1503 (New): Improve multi-GPU support
Currently we round-robin PE's to devices which can be a problem if we only create GPU managers on even ranks with an ... Michael Robson
11:10 AM Bug #1474 (In Progress): mpi-win-x86_64 fails in collidethread example
Sam White
11:09 AM Bug #1225 (Merged): Document TRAM [aggregate] entry method attribute
Sam White
11:07 AM Bug #1408: Improve visibility and usability of flushTraceLog()
bump Sam White
11:05 AM Support #1496: Add -Werror=vla to Jenkins per-commit builds
This should be quick and easy Sam White
10:59 AM Cleanup #1423 (Merged): Conditionalize PUP sizing on load balancers' needs
Phil Miller
10:59 AM Cleanup #1418 (Merged): Remove commlibInfo from _CkSectionInfoStruct
Phil Miller
10:57 AM Cleanup #1196 (In Progress): Document and exemplify chare class inheritance
Is that documentation sufficient?
https://charm.cs.illinois.edu/gerrit/#/c/2391/
Sam White
10:53 AM Bug #948 (Merged): CkExit() crashes called from main chare constructor with `-tracemode projections`
Sam White
10:50 AM Bug #700 (Merged): Calls to CkExit must be unique
Sam White
10:30 AM Bug #1475 (Merged): Define equality operators for proxies
Sam White

04/13/2017

06:25 PM Bug #1474: mpi-win-x86_64 fails in collidethread example
My previous comment was not correct, it turns control gets to main(...) just fine, and is also able to initialize AMP... Eric Mikida
03:09 PM Bug #1474: mpi-win-x86_64 fails in collidethread example
From what I understand, this bug came up after a change to demandCreation. One issue I have found so far is that AMPI... Eric Mikida
05:49 PM Cleanup #1082 (Implemented): Improve SMP mode documentation
https://charm.cs.illinois.edu/gerrit/#/c/2399/ Michael Robson
10:37 AM Cleanup #1082: Improve SMP mode documentation
This is sorely needed. Sam White
05:25 PM Bug #887 (In Progress): Investigate initialization of NullLB WRT thread safety
I am trying to replicate the error. For some reason the init method is always called on other PEs after the NullLB::i... Kavitha Chandrasekar
05:08 PM Cleanup #595: Cut long-running tests & examples out of 'make test'
The two commbench tests which take the longest are pingpong and flood. Each of these has entirely hard coded paramete... Eric Bohm
04:15 PM Cleanup #595: Cut long-running tests & examples out of 'make test'
We want megampi/pgm and Cjacobi3D/jacobi to run for more than a few seconds to test AMPI messaging/collectives/migrat... Sam White
04:04 PM Cleanup #595: Cut long-running tests & examples out of 'make test'
Here is an individual test time breakdown for make test netlrts-linux-x86_64 smp --with-production on intellect
http...
Eric Bohm
04:30 PM Feature #1468: Enable pre-pinning memory for RDMA message sends
Yes I think that's what we want is a CkAlloc. We can eventually provide a pre-pinned memory pool behind that, but at ... Sam White
04:26 PM Feature #1468 (In Progress): Enable pre-pinning memory for RDMA message sends
@CmiAlloc()@ calls @infi_CmiAlloc()@ underneath, which in turn calls @getInfiCmiChunk()@ where the memory is actually... Jaemin Choi
04:21 PM Feature #1492: Remove need for +LBCommOff
https://charm.cs.illinois.edu/gerrit/#/c/2396/ Kavitha Chandrasekar
03:44 PM Feature #1492 (In Progress): Remove need for +LBCommOff
I have added the ability for LB's to specify the commStatsOn, and have removed it from command line options. There ar... Kavitha Chandrasekar
04:11 PM Feature #1378: 64-bit Charm message sizes
Here's a first pass at the netlrts/multicore/smp build. This builds and passes the tests/examples, but no attempt has... Sam White
03:56 PM Bug #529: LrtsInitCpuTopo() operates asynchronously and overlaps with group constructors
This issue https://charm.cs.illinois.edu/redmine/issues/1381 might also be related to this. Bilge Acun
03:53 PM Feature #870: SDAG methods marked as [sync] should only return when run to completion
I think the problem is that we don't store the callback anywhere beyond the scope of the first generated function (he... Sam White
03:21 PM Feature #870: SDAG methods marked as [sync] should only return when run to completion
Two contradictory thoughts: (a) should we allow a a threaded method to be also an sdag method? Seems odd.. not quite ... Laxmikant "Sanjay" Kale
02:30 PM Feature #870: SDAG methods marked as [sync] should only return when run to completion
I am unable to reproduce bug. The example in the description works correctly. Vipul Harsh
03:25 PM Feature #1256: Document '-tracemode perfReport' in the manual
I didn't think perfReport even had a Projections view? Doesn't it just dump a text report at the end of the run? We d... Sam White
03:22 PM Feature #1256 (In Progress): Document '-tracemode perfReport' in the manual
I've been trying to fix the PICS view in Projections, but that's going to require a fairly large time investment, it ... Ronak Buch
10:37 AM Feature #1256: Document '-tracemode perfReport' in the manual
This should be quick/easy, so bumping priority. Sam White
03:21 PM Bug #1452 (Merged): verbs-linux-ppc64le xlC
Bilge Acun
03:15 PM Feature #195: Modify old charm++/converse onesided/direct API and examples to new rdma API
We have three different converse APIs that use or intend to use RDMA functionality. (CmiDirect, CmiOnesided, CmiRdma)... Nitin Bhat
03:02 PM Feature #1351 (Closed): For broadcasts, avoid copies for destination PEs on same node for [nokeep...
Unable to reproduce different messages. Seems like for nokeep messages, the RTS does send the same message to each re... Vipul Harsh
02:55 PM Cleanup #1432 (Implemented): document CkLoop caller function
Seonmyeong Bak
10:36 AM Cleanup #1432: document CkLoop caller function
This should be quick/easy, so bumping priority. Sam White
02:15 PM Cleanup #1363 (Implemented): Remove/deprecate dead machine layers
https://charm.cs.illinois.edu/gerrit/#/c/2393/ uses a #pragma and a launch time print to warn the user to use netlrt... Eric Bohm
02:02 PM Feature #1297: Patch METIS to privatize global variables
I looked at the object dump of the object files in the .bss section for global variables. It seems like all the globa... Kavitha Chandrasekar
01:58 PM Bug #987 (Implemented): Entry methods with no parameters can't be called with CkEntryOptions, can...
Fix: https://charm.cs.illinois.edu/gerrit/#/c/2392/
Charmxi now spits the code to set group dependency in the enve...
Nitin Bhat
01:45 PM Cleanup #1196: Document and exemplify chare class inheritance
I see that it exists but is commented out. Phil, do you have any idea about this? Vipul Harsh
01:44 PM Bug #1001: netlrts with replicas leaves zombie processes behind
Cannot reproduce the problem even with 6.7.0. Bilge Acun
01:34 PM Bug #1439 (Merged): net-linux-x86_64-ibverbs-smp-iccstatic with tracing or debug enabled segfault...
Bilge Acun
01:32 PM CharmDebug Bug #1485: CharmDebug in SMP mode does not work
bohm/charmdebugpatch branch still freezes with some ppn values, for example this one for tests/charm++/simplearrayhel... Bilge Acun
01:12 PM Feature #1466 (In Progress): Update list of available load balancing strategies in the manual
https://charm.cs.illinois.edu/gerrit/#/c/2390/ Kavitha Chandrasekar
10:35 AM Feature #1466: Update list of available load balancing strategies in the manual
This should be quick/easy, so bumping priority Sam White
12:52 PM Feature #1467 (Rejected): Avoid memory pinning overhead for RDMA sends within a process
Within process (charm node) rdma exchanges don't have memory pinning in the existing code. Memory Pinning happens onl... Nitin Bhat
11:02 AM Feature #1074 (Merged): Migrate net-linux-ppc to netlrts
Sam White
10:36 AM Cleanup #1265: Document LLVM OpenMP runtime integration
This should be easy, so bumping priority. Sam White
10:36 AM Feature #1398: Document addReducer's new option 'streamable'
This should be quick/easy, so bumping priority. Sam White
10:14 AM Cleanup #1482: Update Charm++ FAQ
Nitin Bhat
08:10 PM Bug #1500: Entry Methods Always Take lvalue References (feature/bug)
Okay, that's really good news. We would be happy to be involved in brainstorming, development and testing of any new ... Nils Deppe

04/12/2017

05:22 PM Bug #1410: Tuple reducer leaks memory when using set/concat/custom reducers
Juan noted in review of the above patch that CkMulticast is fine. The memory leak is the main concern left here (besi... Sam White
05:04 PM Bug #1441 (Merged): Lies at startup about "The comm. thread both sends and receives messages"
Sam White
05:02 PM Bug #1502 (New): Fix gfortran paths on Darwin
Anyone who wants to use gfortran on Darwin has to install their own compilers. Likely this is down with homebrew or m... Sam White
03:54 PM Bug #948 (Implemented): CkExit() crashes called from main chare constructor with `-tracemode proj...
https://charm.cs.illinois.edu/gerrit/2389 Phil Miller
03:35 PM Bug #948: CkExit() crashes called from main chare constructor with `-tracemode projections`
I've got most of a fix for this done. It will just short-circuit the exit-time code if initialization hasn't happened. Phil Miller
02:57 PM Bug #1500: Entry Methods Always Take lvalue References (feature/bug)
This is indeed the direction we're aiming to go, and want to move on in the not-too-distant future. There's room for ... Phil Miller
11:32 AM Bug #1500 (New): Entry Methods Always Take lvalue References (feature/bug)
I'm filing this bug/feature request is mainly to start a discussion of C++11 and template metaprogramming being used ... Nils Deppe
01:44 PM Bug #1312 (Merged): Deleting an array disables reclamation for all arrays bound to that location ...
Both patches have now been merged, so I think we're done here. Please re-open if that's not the case. Phil Miller
01:21 PM Bug #965: -swapglobals is broken for current gcc/ld
Pleasantly, the affected bits of code in ld are all part of libbfd, which popular distributions seem to link dynamica... Phil Miller
11:35 AM Bug #1501 (New): charmc fails to deal with rvalue references
This is probably related to issue 1500, but having an entry method take an rvalue reference causes the @charmc@ scrip... Nils Deppe
10:43 AM Feature #1187 (Merged): Automatic delegation of section work to CkMulticastMgr
Phil Miller
12:00 AM Feature #1428: AMPI TLS privatization support for IBM POWER
Only slightly related: our QuickThreads package lacks support for POWER, but it looks like a more recent version of Q... Sam White
11:56 PM Feature #1232: AMPI migration detection mechanisms
This is actually not too difficult to implement in AMPI: just use ckJustMigrated to check the new/old pe/node/whateve... Sam White
11:54 PM Bug #1268: AMPIF issues due to C++ main routine
This is actually going to be important for any AMPI Fortran code that we want to use tlsglobals with. Sam White
09:53 PM Bug #805 (Closed): SDAG tracing counts time waiting for a when to be satisfied as overhead
I can't seem to reproduce this bug, and the code looks pretty sound from what I can tell. The only way TRACE_EXECUTE_... Eric Mikida
09:51 PM Feature #1499: Remove the need to declare entry method parameters as "rdma"
This should maybe be a "next-gen"/ v7.0 feature, but I think the scope of this change is perhaps small enough to be d... Sam White
09:49 PM Feature #1499 (New): Remove the need to declare entry method parameters as "rdma"
Currently, in order to call an entry method via an RDMA send, the entry method being called has to be declared with i... Sam White
07:38 PM Bug #1225 (Implemented): Document TRAM [aggregate] entry method attribute
Implemented in https://charm.cs.illinois.edu/gerrit/#/c/2388/ Eric Mikida

04/11/2017

06:58 PM Bug #1495: PlasCom2/advect1d crashes on AMPI with ++local or multiple nodes
Also worth noting that is appeared to happen in IO. Fortran IO is known to be racy in SMP mode, but PlasCom2 doesn't/... Sam White
10:07 AM Bug #1495 (Closed): PlasCom2/advect1d crashes on AMPI with ++local or multiple nodes
... Matthias Diener
06:55 PM Feature #1497: Shared memory method to pass data between processes that share the same node
The main things to investigate here appear to be xpmem, pxshm, knem and cma. OpenMPI's "vader" shared-memory BTL is a... Sam White
04:53 PM Feature #1497 (New): Shared memory method to pass data between processes that share the same node
PXSHM exists for this purpose on the net layers. However it is not generally used in SMP mode to exchange data when ... Eric Bohm
06:19 PM Bug #1475 (Implemented): Define equality operators for proxies
Eric Mikida
06:19 PM Bug #1475: Define equality operators for proxies
Implemented in https://charm.cs.illinois.edu/gerrit/#/c/2387/ Eric Mikida
05:23 PM Bug #1470 (Closed): Investigate broken load balancers in mini-apps
Closing this issue since the mini-apps work okay. Following is the status:
1. amr mini-app in smp mode is fixed
2...
Kavitha Chandrasekar
02:02 PM Bug #1470: Investigate broken load balancers in mini-apps
No there no fixes for examples/test, only to external repos. Kavitha Chandrasekar
01:53 PM Bug #1470 (In Progress): Investigate broken load balancers in mini-apps
I don't think this needs to be targeted to 6.8.0 if the fixes are to mini-apps that are hosted in external repos. Are... Sam White
05:22 PM Bug #1498 (New): SDAG methods are not properly inherited by chare subclasses
chare A {
entry A();
entry void foo();
entry void bar() { when foo() {} };
};
chare B : A {
entry B()...
Eric Mikida
04:52 PM Support #1496 (Closed): Add -Werror=vla to Jenkins per-commit builds
This will help us catch MSVC breaking commits before they make it to mainline Sam White
04:27 PM Bug #1493 (Merged): Deleting an array also deletes all common elements from it's bound arrays
Phil Miller
01:06 AM Bug #1493 (Implemented): Deleting an array also deletes all common elements from it's bound arrays
Implemented a potential fix in https://charm.cs.illinois.edu/gerrit/2381. Eric Mikida
04:14 PM Feature #1297: Patch METIS to privatize global variables
I don't have a patch yet. I don't recall you mentioning that you did, I think you mentioned you might work on it. I c... Kavitha Chandrasekar
04:08 PM Feature #1297: Patch METIS to privatize global variables
Kavitha, did you have a patch for this? Do you recall if I said I did? Phil Miller
04:07 PM Feature #1187 (Closed): Automatic delegation of section work to CkMulticastMgr
Crashes fixed by this patch: https://charm.cs.illinois.edu/gerrit/#/c/2204/. Closing issue. Vipul Harsh
04:01 PM Feature #1187: Automatic delegation of section work to CkMulticastMgr
Please close this if it is done, or comment otherwise here. Sam White
04:04 PM Feature #34: Reduce Charm Message Send Overhead for Marshalled Messages
Does the RDMA send API address this adequately? Sam White
04:03 PM Feature #105 (In Progress): Unify memory pool implementations where possible
https://charm.cs.illinois.edu/gerrit/#/c/2306/ Sam White
04:02 PM Feature #1074: Migrate net-linux-ppc to netlrts
https://charm.cs.illinois.edu/gerrit/#/c/2337/ Phil Miller
04:02 PM Feature #1074: Migrate net-linux-ppc to netlrts
The above linked patch is for arm, not ppc Phil Miller
04:01 PM Feature #1072 (Merged): Migrate net-linux-arm7 to netlrts
Phil Miller
04:00 PM Feature #279: Windows: Display abort message in failure pop-up
Would be nice to have, but runs on Windows are currently mostly targeting production, and so shouldn't be seeing CkAb... Phil Miller
03:59 PM Feature #1298 (Merged): Build MetisLB by default
Sam White
02:03 PM Feature #1298 (Implemented): Build MetisLB by default
https://charm.cs.illinois.edu/gerrit/#/c/2319/ Kavitha Chandrasekar
03:58 PM Feature #789: Charmrun should test for SSH failures when node programs fail to launch
It'd be nice to have, but nothing is broken by missing this. Phil Miller
03:57 PM Feature #1382 (Merged): Add streaming floating point sum reducers
I don't think we need to add new nonstreaming built-in reducers for this. Users can implement them on their own if th... Sam White
03:56 PM Feature #1251: Document shrink/expand in the manual
Add this to the experimental features section. Eric Bohm
03:55 PM Feature #1481 (Implemented): RDMA zero copy send implementation for multicore builds
Sam White
03:54 PM Feature #262 (Rejected): MetaBalancer: add mechanism to inform user of AtSync call that resulted ...
The AMPI use case can be covered using ckJustMigrated: https://charm.cs.illinois.edu/gerrit/#/c/2200/ Sam White
03:48 PM Bug #1327: AMPI derived datatype support for RMA routines
The updated patch in gerrit still needs work, but no need to hurry before the release Sam White
03:46 PM Bug #78: AMPI failure with migration under Cray compiler due to tcmalloc bugs or incompatibility
Charm/AMPI do actually compile on the Cray Compiler now (CCE 8.5.4+), so I will try it out at some point before the r... Sam White
03:44 PM Bug #1439 (Implemented): net-linux-x86_64-ibverbs-smp-iccstatic with tracing or debug enabled seg...
Uploaded the patch here: https://charm.cs.illinois.edu/gerrit/#/c/2386 Bilge Acun
03:35 PM Bug #1312: Deleting an array disables reclamation for all arrays bound to that location manager
A fix to the actual functionality of reclaimRemote has been posted here: https://charm.cs.illinois.edu/gerrit/#/c/2385/ Eric Mikida
01:09 AM Bug #1312 (Implemented): Deleting an array disables reclamation for all arrays bound to that loca...
A potential fix for this is implemented in https://charm.cs.illinois.edu/gerrit/#/c/2380/, however it would appear th... Eric Mikida
03:09 PM Bug #1250: Tuple reductions cannot have [reductiontarget] callbacks
Progress so far on this bug can be found here:
https://charm.cs.illinois.edu/gerrit/gitweb?p=charm.git;a=commit;h=84...
Karthik Senthil
03:04 PM Bug #1250: Tuple reductions cannot have [reductiontarget] callbacks
We're not going to have this for 6.8.0 Sam White
02:10 PM CharmDebug Bug #1373 (Merged): Unable to attach to an application it launched
Bilge Acun
01:50 PM Bug #1473: verbs build hangs in tests/charm++/communication_overhead
This has been worked around for now. Sam White
12:48 PM Bug #1392 (Closed): Stampede test script fails during autobuild (verbs)
It was decided to not modify testrun (or charmrun) and keep it platform agnostic. The fix involved making changes to... Nitin Bhat
12:38 PM Bug #1452 (Implemented): verbs-linux-ppc64le xlC
https://charm.cs.illinois.edu/gerrit/#/c/2383 Bilge Acun

04/10/2017

05:29 PM Bug #700 (Implemented): Calls to CkExit must be unique
https://charm.cs.illinois.edu/gerrit/2377
That took long enough...
Phil Miller
04:49 PM Feature #1494 (New): Broadcast trees are not topology-aware
From Phil:
Looking at the implementation of our array->group->converse broadcasts, it seems that on systems with mea...
Sam White
04:31 PM Bug #1493 (Implemented): Deleting an array also deletes all common elements from it's bound arrays
When you call ckDestroy on an array proxy, it deletes all location records associated with it's elements. This result... Eric Mikida
04:20 PM Bug #1397 (Merged): Document that array creation must occur on PE0
Sam White
02:34 PM Bug #1421 (In Progress): Running leanmd with error checking enabled in Charm++ triggers assertion...
I am able to add a condition to check for destinations of a multicast message and remove them if they don't exist any... Kavitha Chandrasekar
01:49 PM Bug #1475: Define equality operators for proxies
For arrays/groups/nodegroups, this is trivial, but for singleton chares virtual IDs create an issue. A proxy that has... Eric Mikida
12:33 PM Cleanup #165: object location services: separate location caching from msg delivery and buffering
The current plan is to unify IDs by getting 64bit IDs to be used by all chare types (instead of just array elements).... Eric Mikida
12:29 PM Feature #179 (In Progress): objid_t: adapt callbacks to use objid_t
I'm currently working on getting the 64bit ID used for things other than chare arrays (singletons, groups, nodegroups... Eric Mikida
12:21 PM Cleanup #1423 (Implemented): Conditionalize PUP sizing on load balancers' needs
https://charm.cs.illinois.edu/gerrit/#/c/2375/ Sam White
12:04 PM Bug #1225: Document TRAM [aggregate] entry method attribute
That bit of documention can be reduced, since the header and module linkage is now automatic. The [aggregate] entry m... Phil Miller
11:59 AM Bug #1225 (In Progress): Document TRAM [aggregate] entry method attribute
This is already documented in the libraries manual, where all of TRAM is documented. Is that sufficient, or do we wan... Eric Mikida
11:12 AM Bug #994 (Merged): The Projections interface function 'traceUserSuppliedData()' is undocumented i...
Note that the commit message gets the bug number wrong. Too late now, though. Phil Miller

04/07/2017

04:12 PM Feature #1492 (In Progress): Remove need for +LBCommOff
The LB Database should not collect comm info unless a balancer that is running actually uses it. For PDES and fine-gr... Sam White
04:10 PM Cleanup #1491 (New): Update documentation of GPUManager
Update the documentation of GPUManager to match the redesign. Jaemin Choi
03:47 PM Cleanup #1490 (New): Automatically set NVCC path in Makefile of example program
We should create a script that is called inside each Makefile that finds out the path to NVCC and use that to compile... Jaemin Choi
03:45 PM Cleanup #1489 (New): Delete GPU dummy mempool
The dummy mempool inside GPUManager has not been used in a while, and should be removed. Jaemin Choi
11:21 AM Bug #1488 (Implemented): GPU manager runs out of memory on talent
... Dong Hun Lee
11:19 AM Bug #1487 (Implemented): Leaving -DGPU_MEMPOOl causes gpu manager to not build
... Dong Hun Lee
07:35 PM Bug #1486 (Merged): AMPI_{Test,Wait}{any,some} on Fortran need to adjust from 0-based indexing to...
Sam White

04/06/2017

05:24 PM CharmDebug Bug #1485: CharmDebug in SMP mode does not work
See the bohm/charmdebugpatch branch for a hacky workaround to that hang. Eric Bohm
10:13 AM CharmDebug Bug #1485: CharmDebug in SMP mode does not work
It simply freezes when you press "Start", nothing else is printed. Bilge Acun
08:43 AM CharmDebug Bug #1485 (New): CharmDebug in SMP mode does not work
It has been broken since at least Charm 6.6.0. Bilge Acun
03:15 PM Bug #1479 (Merged): Charm++ Fails to Compile on Arch Linux
Phil Miller
02:39 PM Bug #1479: Charm++ Fails to Compile on Arch Linux
Ah yea... I am an idiot, good catch! Okay, so then this ticket is resolved with your patch. Thanks! Nils Deppe
02:34 PM Bug #1479: Charm++ Fails to Compile on Arch Linux
Could you post the @./build@ command line you tried to run? I suspect it was something like
> ./build charm multic...
Phil Miller
02:23 PM Bug #1479: Charm++ Fails to Compile on Arch Linux
Thanks for the fix! I can confirm that the cpp part of the issue is resolved.
Here is the output for your other qu...
Nils Deppe
12:26 PM Bug #1486 (Implemented): AMPI_{Test,Wait}{any,some} on Fortran need to adjust from 0-based indexi...
https://charm.cs.illinois.edu/gerrit/2363 Phil Miller
12:13 PM Bug #1486 (Merged): AMPI_{Test,Wait}{any,some} on Fortran need to adjust from 0-based indexing to...
Phil Miller
11:01 AM Bug #1397 (Implemented): Document that array creation must occur on PE0
This hasn't been merged yet, and once it is it should be marked 'merged' since there was something pushed into the repo Sam White
10:59 AM Bug #1397 (Resolved): Document that array creation must occur on PE0
Dong Hun Lee
10:04 AM CharmDebug Bug #1373: Unable to attach to an application it launched
Fix to solve the second commit that breaks CharmDebug is implemented here: https://charm.cs.illinois.edu/gerrit/#/c/2362 Bilge Acun
08:45 AM Feature #1378: 64-bit Charm message sizes
I have a half-baked implementation on multicore builds. Some of the machine layers already support 64 bit message siz... Sam White
10:25 PM Support #1483 (Rejected): Welcome to Redmine
Sam White
10:04 PM Support #1483: Welcome to Redmine
Please delete, this was made in error. Freddie Sanger
10:03 PM Support #1483 (Rejected): Welcome to Redmine
For more info, see the "wiki":https://www.everipedia.com/Redmine/ Freddie Sanger
08:39 PM Cleanup #1482 (New): Update Charm++ FAQ
The “Charm++ FAQ” really needs to be updated. It is outdated in many ways: mentions of old/removed machine layers, no... Sam White
08:20 PM Bug #1409: verbs crashes on Stampede KNL
Verbs is one of the options on OmniPath (or at least most OmniPath systems). Which layer performs best (verbs, mpi, o... Sam White
07:37 PM Bug #1409: verbs crashes on Stampede KNL
I accidentally compiled and tested the netlrts layer, and that worked fine.
That brings up my question: are we suppo...
Jaemin Choi
07:13 PM Bug #1441 (Implemented): Lies at startup about "The comm. thread both sends and receives messages"
Patch pushed to gerrit [[https://charm.cs.illinois.edu/gerrit/#/c/2360/]] Jaemin Choi

04/05/2017

06:04 PM Feature #1481: RDMA zero copy send implementation for multicore builds
Implemented multicore RDMA support in netlrts here: https://charm.cs.illinois.edu/gerrit/#/c/2358/ Sam White
05:20 PM Feature #1481: RDMA zero copy send implementation for multicore builds
Just defining CMK_ONESIDED_IMPL to CMK_MULTICORE is not enough. When you link something that uses RDMA it fails with ... Sam White
05:09 PM Feature #1481: RDMA zero copy send implementation for multicore builds
Though it may make sense to separate the operability of the SMP implementation from the network-layer implementation,... Phil Miller
05:08 PM Feature #1481: RDMA zero copy send implementation for multicore builds
It suggests that the fix for this issue is very easy - just add ... Phil Miller
05:02 PM Feature #1481: RDMA zero copy send implementation for multicore builds
I assume what you mean by "generic SMP support" is the direct memcpy of the source buffer to a message on the recv'er... Sam White
03:29 PM Feature #1481: RDMA zero copy send implementation for multicore builds
Isn't this already covered by the generic SMP support? And even if it isn't, any implementation work here would just ... Phil Miller
01:00 PM Feature #1481 (Merged): RDMA zero copy send implementation for multicore builds
Currently, the RDMA send API uses the "generic" implementation in multicore builds. The generic implementation does t... Sam White
04:55 PM Bug #1397: Document that array creation must occur on PE0
Added documentation
https://charm.cs.illinois.edu/gerrit/#/c/2357/
Dong Hun Lee
03:22 PM CharmDebug Bug #1373 (In Progress): Unable to attach to an application it launched
There are more problems with CharmDebug, with this commit being fault: https://charm.cs.illinois.edu/gerrit/#/c/1873/... Bilge Acun
01:52 PM CharmDebug Bug #1373 (Merged): Unable to attach to an application it launched
Sam White
12:51 PM CharmDebug Bug #1373 (Implemented): Unable to attach to an application it launched
https://charm.cs.illinois.edu/gerrit/#/c/2356 Bilge Acun
12:16 PM CharmDebug Bug #1373: Unable to attach to an application it launched
I've replaced ReadonlyInfo's and ChareTypeInfo's sizes from int to long, but simply doing that hasn't solved the prob... Bilge Acun
12:06 PM CharmDebug Bug #1373: Unable to attach to an application it launched
Actually it should be as easy as grep'ing for "ReadonlyInfo" and "ChareInfo" inside charmdebug, changing the size fie... Sam White
11:46 AM CharmDebug Bug #1373: Unable to attach to an application it launched
git bisect showed this commit to be at fault: https://charm.cs.illinois.edu/gerrit/#/c/1903/
CharmDebug is probabl...
Sam White
09:57 AM Bug #704 (Closed): charmrun startup handles failed connection attempts very poorly
This seems to no longer be a noticeable problem, so closing. Mac OS users can typically just use @++local@, and thing... Phil Miller

04/04/2017

05:13 PM Feature #1480 (New): API to control whether a PE helps other threads that generate CkLoop/OpenMP/...
Jim brought up an issue that he doesn't want the PEs tasked with highly-critical PME work to participate in helping o... Phil Miller
05:04 PM Bug #677 (Closed): MPI Wrappers on BG/Q supersede include path's from command line, breaking AMPI...
I asked ANL staff about this way back when, and they opened a PMR with IBM. The outcome seems to be that the DEPRECAT... Phil Miller
04:01 PM Bug #1397: Document that array creation must occur on PE0
Here's the new API that can be used in the case of creation off PE 0: https://charm.cs.illinois.edu/gerrit/#/c/736/ Sam White
03:02 PM Bug #1397: Document that array creation must occur on PE0
Setting priority higher because these are easy to fix things that act as paper-cuts to users Phil Miller
03:58 PM Bug #1197 (Rejected): Charmrun fails to connect to nodes on Taub in netlrts build.
Since this issue occurs specifically on loading gcc/4.9 module, it needs for environment variables to be passed to th... Kavitha Chandrasekar
03:53 PM Bug #833: mpi smp build is locked to one core per node by default
Since we won't have hwloc integration for the release, I think it would be good to get this into 6.8.0 if it is ready? Sam White
03:52 PM Bug #1035: Idle PEs compete with comm thread for node queue lock
We are not planning on merging the lockless queue before the 6.8.0 release, since it is high risk this close to the r... Sam White
03:50 PM Bug #1002 (Merged): Changa's Final CkWaitQD() hangs after AtSync deletion-counting changes
Closing unless there are any objections... Sam White
03:43 PM Bug #1452: verbs-linux-ppc64le xlC
This should be quick/easy Sam White
03:43 PM Bug #522: static linking breaks on multicore builds with 'undefined reference to `get_myaddress''
It would seem that the thing to do about this would be to statically link everything with the exception of @libc@. Th... Phil Miller
03:43 PM Bug #999 (Merged): netlrts writeableDgrams is never reset to 0
The part of the above change that wasn't directly addressing this bug got reverted, but the fix for this is solidly i... Phil Miller
03:42 PM Feature #693 (New): add CcdCallBacksReset() to header file (or improve callback frequency)
Michael Robson
03:41 PM Feature #693 (In Progress): add CcdCallBacksReset() to header file (or improve callback frequency)
Michael Robson
03:41 PM Bug #159: Some CkCallback types are not valid across checkpoint/restart
Not seeming to affect any current applications, so deferring. Phil Miller
03:09 PM Bug #994 (Implemented): The Projections interface function 'traceUserSuppliedData()' is undocumen...
https://charm.cs.illinois.edu/gerrit/#/c/2354/ Ronak Buch
03:02 PM Bug #994: The Projections interface function 'traceUserSuppliedData()' is undocumented in the manual
Setting priority higher because these are easy to fix things that act as paper-cuts to users Phil Miller
03:04 PM Bug #1470: Investigate broken load balancers in mini-apps
Since the mini-apps would work with minimal changes, should we follow up on the email with suggestions? Kavitha Chandrasekar
03:02 PM Bug #1475: Define equality operators for proxies
Setting priority higher because these are easy to fix things that act as paper-cuts to users Phil Miller
03:02 PM Bug #1441: Lies at startup about "The comm. thread both sends and receives messages"
Setting priority higher because these are easy to fix things that act as paper-cuts to users Phil Miller
03:02 PM Bug #1408: Improve visibility and usability of flushTraceLog()
Setting priority higher because these are easy to fix things that act as paper-cuts to users Phil Miller
03:02 PM Bug #1225: Document TRAM [aggregate] entry method attribute
Setting priority higher because these are easy to fix things that act as paper-cuts to users Phil Miller
02:46 PM Bug #1479 (In Progress): Charm++ Fails to Compile on Arch Linux
https://charm.cs.illinois.edu/gerrit/2352
The @cpp@ part of this issue is addressed in the patch linked.
Phil Miller
01:13 PM Bug #1479: Charm++ Fails to Compile on Arch Linux
Hi Nils,
Thanks for your report. Could you say what shell Antergos uses as /bin/sh by default, or what you have in...
Phil Miller
01:35 PM Bug #1162: tracing runs segfault while writing logs
I thought I had added a log here before, but I guess not. I had tried to replicate this when the bug was originally ... Ronak Buch
01:10 PM Bug #1162: tracing runs segfault while writing logs
Has there been any investigation or follow-up on this? This could be pretty crippling for large-scale performance work. Phil Miller
01:25 PM Cleanup #473: Licensing of library code in Data Transfer library
www.magic-software.com is no longer a valid URL. Wayback machine doesn't have a copy due to robots.txt.
Magic So...
Eric Bohm
01:17 PM Bug #1392 (Resolved): Stampede test script fails during autobuild (verbs)
This would seem to have been dealt with, given that autobuild runs on Stampede pass. Please close if that's correct, ... Phil Miller
12:42 PM Feature #1039: reject pemap/commap with duplicate or too few cpus
Validation of the old command lines should be reconsidered when the binding substrate changes to hwloc. Eric Bohm
12:34 PM Feature #1386: ckDestroy for Groups
Retargetting this to a later version until someone makes a case for it being urgently required in 6.8.0. Eric Bohm

04/03/2017

06:31 PM Bug #965: -swapglobals is broken for current gcc/ld
Hacked ld in repo
@git clone charmgit:users/phil/binutils-gdb -b phil/ampi-swapglobals-hack@
Phil Miller
06:20 PM Bug #965 (In Progress): -swapglobals is broken for current gcc/ld
I have confirmed that with a modified ld to disable this optimization/conversion, and calling it by setting its path ... Phil Miller
05:30 PM Bug #965: -swapglobals is broken for current gcc/ld
The modified code to apply this conversion in modern ld is unconditional. Local references always get smashed. If we ... Phil Miller
01:39 PM Bug #965: -swapglobals is broken for current gcc/ld
Moving to PIE compilation/linking on the newer LD doesn't fix things. It still sees a 'local' reference to a global v... Phil Miller
01:09 PM Bug #965: -swapglobals is broken for current gcc/ld
And indeed, when compiled on Might, swapglobals appears to operate correctly. Phil Miller
01:06 PM Bug #965: -swapglobals is broken for current gcc/ld
Here's the patch that implemented that opitmization in GNU LD:
https://sourceware.org/ml/binutils/2012-08/msg00498.h...
Phil Miller
12:18 PM Bug #965: -swapglobals is broken for current gcc/ld
OK, so first issue on this is that the relocations in question are actually being optimized out at link time, so they... Phil Miller
11:22 AM Bug #965: -swapglobals is broken for current gcc/ld
I just tried out tests/ampi/jacobi3d on Courage, Ubuntu 14.04 with GCC 4.8.4 and ld 2.24, and it doesn't output the e... Phil Miller

04/02/2017

08:46 PM Bug #1479 (Merged): Charm++ Fails to Compile on Arch Linux
The default configuration of charm++ (at least v6.7.1) fails to build on the Arch Linux distro Antergos (I haven't tr... Nils Deppe

03/31/2017

04:22 PM Feature #1478 (New): Investigate use of pxshm in CmiAlloc
Currently, when the runtime is built with pxshm support, we use an extra copy into a posix shared memory buffer when ... Sam White
11:20 AM Bug #1477 (New): All Load Balancing Strategies should be CPU frequency (rate) aware
Under an assumption of full or near-full CPU load when applications and the RTS are running well, we've found that Tu... Phil Miller

03/30/2017

04:55 PM Cleanup #1476 (New): Fix Make.depends for libraries
Make.depends for libraries no longer works. It appears that when the main Make.depends was updated to use the correct... Eric Mikida
02:53 PM Bug #1475 (Merged): Define equality operators for proxies
We can overload the equality operator for proxies. This stems from John Mitchell's email to the charm mailing list. Sam White
 

Also available in: Atom