Project

General

Profile

Activity

From 01/03/2017 to 02/01/2017

02/02/2017

11:55 PM Cleanup #960: Replace CmiEnforce(0) calls in verbs and net layer
So what do you want me to do here?
Replace CmiEnforce(0) with CmiAbort()?
Jaemin Choi
11:44 PM Bug #1390: AMPI_Alltoall crashes for short messages
We'll want to have MPI_Alltoall working for all message sizes in our 6.8.0 release. In the worst case, we could alway... Sam White
10:41 PM Feature #239: Charmrun wrapper around ibrun.symm for dual binary execution on host+mic with IMPI ...
Dual binary launching no longer relevant for KNL Sam White
10:38 PM Feature #190: Build target(s) for Intel MIC
KNL is self-hosted and doesn't require the complicated dual binary launch scheme. Since this is off the radar for now... Sam White
10:35 PM Bug #968: charm++ programs fail to run on BlueWaters due to craype-hugepages8M
Can this be closed? Sam White

02/01/2017

02:19 PM Feature #793 (Closed): Write default cuda kernelSelect so cuda build works out of the box
This bug is fixed/obviated by patch https://charm.cs.illinois.edu/gerrit/#/c/2063/ Michael Robson
02:18 PM Bug #792 (Closed): Enable cuda builds of charm to run non-cuda programs
This bug is fixed/obviated by patch https://charm.cs.illinois.edu/gerrit/#/c/2063/ Michael Robson
12:33 PM Bug #792: Enable cuda builds of charm to run non-cuda programs
This seems like a significant bug to me if I understand it correctly and if it is still pertinent: you can't build Ch... Sam White
01:58 PM Feature #315 (New): Reduction Starting messages - Performance optimal solution
Reassigning to Eric since he is already assigned the related issue. Sam White
01:57 PM Bug #598 (Closed): Unused tests in test/charm++
Sam White
01:55 PM Bug #805 (New): SDAG tracing counts time waiting for a when to be satisfied as overhead
Sam White
01:53 PM Bug #1005 (New): Some Projections views broken for AMPI with user-registered functions
Sam White
01:46 PM Cleanup #1065 (In Progress): Create a more efficient caching structure for location lookup
Change from std::map to std::unordered_map: https://charm.cs.illinois.edu/gerrit/#/c/1978/
Use tr1/unordered_map on ...
Sam White
12:56 PM Feature #105: Unify memory pool implementations where possible
Here's a related patch Nikhil originally wrote to make exceptions for large allocations (to not use the mempool when ... Sam White
12:44 PM Feature #105: Unify memory pool implementations where possible
CmiPool is now only used in the state-space search engine, so we don't really need to worry about that one anymore.
...
Sam White
12:40 PM Bug #234 (Rejected): net-linux-x86_64-*-smp-pgcc crashes in megatest
closing due to net- being deprecated and pgcc not generally being able to compile Charm (and the lack of requests fro... Sam White
12:36 PM Cleanup #537: Data races in handler registration and assignment to global index variables
Still hanging in gerrit, waiting on cleanup... Sam White
12:36 PM Feature #975: OFI Layer
Version 2 patch is here: https://charm.cs.illinois.edu/gerrit/#/c/1945/ Bilge Acun
12:35 PM Cleanup #536: Data Races in SMP PCQueue
Does the new lockless queue fix this? Sam White
12:33 PM Bug #756: CUDA build does not correctly find cuda location
Any update? Sam White
12:31 PM Feature #808 (Rejected): Implement a communication aware load balancer
This issue description is pretty vague to me, so closing for now. It can be reopened under a more descriptive listing... Sam White
12:30 PM Bug #815: Makefile for hybrid API is not using the system OPTS
Is this a real issue? If not close it please. Sam White
12:29 PM Feature #869 (Merged): Merge PICS framework to Charm
Marking this merged since there is nothing left from PICS that we decided we wanted to merge Sam White
12:28 PM Support #865: Update section 7.3 of Manual on Load Balancing Chare arrays
It would be good if you can do a quick read over and correction of this documentation since no else in the lab has yo... Sam White
12:25 PM Projections Feature #976: Tool for analyzing network message latencies
Can you document here where this tool is and how to use it? Sam White
12:04 PM Bug #571: pxshm shared queue lockless implementation is invalid
pxshm is becoming slightly more important with wider nodes, so bumping up the priority. Also putting under SMP becaus... Sam White
12:03 PM Bug #802: CUDA examples broken on Blue Waters
Is this issue up to date or did things get merged for it? Sam White
12:02 PM Cleanup #966 (Merged): cuda build fails unituitvely on blue waters when cudatoolkit isn't loaded
Jaemin Choi
12:01 PM Feature #885: extend physical node detection across partitions
Is this still desired? We should assign it to someone if so. Sam White
11:42 AM Feature #641: protect load balancer from variable cpu clock
It looks like this was fixed a while ago in the following two commits:
https://charm.cs.illinois.edu/gerrit/#/c/577/...
Sam White
11:38 AM Bug #1035 (Implemented): Idle PEs compete with comm thread for node queue lock
The new lockless queue (gerrit patch linked above) will address this issue, so re-assigning to Bilge as the owner of ... Sam White
11:24 AM Feature #1393 (Merged): Redesign of Hybrid API (GPU Manager) to support concurrent kernel execution
The original design of GPUManager had two data transfer streams and one kernel stream per GPUManager instance, which ... Jaemin Choi
09:59 AM Feature #1133: PMPI_ interface for AMPI
I started the implementation for weak symbols and non-MPI communication layers (#2 above) on a branch named 'ampi_wit... Sam White
09:29 AM Cleanup #1365 (Merged): Deduplicate AmpiRequest routines and their handling of NULL statuses
https://charm.cs.illinois.edu/gerrit/#/c/2163/ Sam White
08:32 PM Bug #1329: Hang in exit in TRAM test on gni-crayxc-smp
The bug does not happen in 'debug' build of Charm++ on Edison.
I am currently trying to find the exact location of...
Karthik Senthil

01/31/2017

05:53 PM Bug #1376 (Rejected): AMPI_Ireduce only creates a request at the root
Sam White
04:58 PM Cleanup #1058 (Merged): Make tracing warning about flushing logs more informative
Ronak Buch
03:06 PM Cleanup #1058 (Implemented): Make tracing warning about flushing logs more informative
Implemented at https://charm.cs.illinois.edu/gerrit/#/c/2177/. The documentation already contained the correct infor... Ronak Buch
03:42 PM Bug #1392 (Closed): Stampede test script fails during autobuild (verbs)
Testrun script used in autobuild uses charmrun that isn't allowed to run on a login node.
../../bin/testrun +p4 ...
Nitin Bhat
10:55 AM Feature #1351: For broadcasts, avoid copies for destination PEs on same node for [nokeep] messages
First, all features should have a target version of 6.8.0-beta1 or 6.8.1, not 6.8.0.
Second, even in smp mode it sho...
Jim Phillips
10:04 PM Support #1391 (Closed): Add an SMP/multicore build test to Jenkins
Decided at Core last week: before any commit is merged, it should have to pass the test suite on an SMP or multicore ... Sam White
06:47 PM Bug #1148: Define 'thisIndex' for Groups
It would be good to have this soon, before the 6.8.0 release. Sam White

01/30/2017

04:05 PM Bug #1390 (Merged): AMPI_Alltoall crashes for short messages
Running the megampi test for more number of ranks(instead of 4) crashes with a memory corruption. More investigation ... Karthik Senthil
10:29 AM Feature #858: improve efficiency of exclusive entry methods
In the parameter marshalling case, it seems like CmiReference(impl_msg) would suffice to balance out the deletion tha... Phil Miller
06:41 PM Feature #858: improve efficiency of exclusive entry methods
Some progress on at least the simple case here before the 6.8.0 release would be valuable. Sam White
10:18 AM Feature #1389 (Merged): AMPI ATAReq test/wait performs poorly
AMPI uses IATAReqs to track completion of a group of requests from MPI_Ialltoall* and MPI_Ineighbor_*. Currently we t... Sam White
07:02 PM Bug #1388 (Merged): charmrun doesn't compile on netlrts-win-x86_64
This commit broke charmrun compilation on net-win64 and netlrts-win-x86_64: https://charm.cs.illinois.edu/gerrit/#/c/... Sam White
06:40 PM Cleanup #1058: Make tracing warning about flushing logs more informative
This should be quick and would be valuable to have merged before the 6.8.0 release. Sam White
06:35 PM Feature #1105: AMPI support for large counts
We will merge this once the same changes for 64 bit messages are made to Charm. That is a separate issue. Sam White

01/29/2017

07:00 PM Support #1387 (New): Optimised algorithms for scatterv
Vipul Harsh
06:58 PM Feature #1248 (Implemented): Charm++ scatter(v) collective interface
Naive point to point implementation: https://charm.cs.illinois.edu/gerrit/#/c/2083/ Vipul Harsh

01/28/2017

10:11 AM Feature #1386 (Rejected): ckDestroy for Groups and NodeGroups
Jozsef Bakosi pointed out on the charm mailing list that we currently do not support deleting Groups. ckDestroy is on... Sam White

01/27/2017

09:56 PM Bug #1379: SDAG doesn't properly handle callbacks to [reductiontarget] methods with refnums
OK I didn't know that you couldn't do marshaling in callbacks although there are still two oddities about what you pr... Eric Mikida

01/26/2017

05:57 PM Bug #1379: SDAG doesn't properly handle callbacks to [reductiontarget] methods with refnums
I think we may be able to simplify things.
For a common parameter-marshalled method invocation, we can make sure t...
Phil Miller
05:46 PM Bug #1379: SDAG doesn't properly handle callbacks to [reductiontarget] methods with refnums
My thoughts after looking at this:
There are two ways in which reference numbers are used in SDAG which are "well-de...
Eric Mikida
01:02 PM Bug #1379: SDAG doesn't properly handle callbacks to [reductiontarget] methods with refnums
Semi-related is the issue of tuple reductions not being able to be delivered to [reductiontarget] entry methods: http... Sam White
12:58 PM Bug #1379: SDAG doesn't properly handle callbacks to [reductiontarget] methods with refnums
I think we should try to get this dealt with before 6.8.0 is released Phil Miller
01:39 PM Feature #1382: Add streaming floating point sum reducers
https://charm.cs.illinois.edu/gerrit/#/c/2159/ Sam White
01:39 PM Feature #1382 (In Progress): Add streaming floating point sum reducers
Sam White
01:29 PM Feature #1382 (Merged): Add streaming floating point sum reducers
We chose to make the default floating point sum reducer be not streaming when streaming reducers were added.
We can ...
Sam White
01:25 PM Feature #1322: PSM2 network layer
More on active messages:
https://github.com/01org/opa-psm2/blob/master/psm2_am.h
Jim Phillips
01:15 PM Feature #1322: PSM2 network layer
Presentation on OFI in MPICH:
https://www.openfabrics.org/images/eventpresos/workshops2015/DevWorkshop/Monday/mond...
Jim Phillips
01:07 PM Feature #1322: PSM2 network layer
Look for the first occurrence of psm2_am in the following:
https://github.com/01org/opa-psm2/blob/master/psm2.h
...
Jim Phillips
12:58 PM Feature #1113: smart-build.pl should detect supercomputers with specialized software environments...
Eric B - NB Phil Miller
12:57 PM Feature #1113: smart-build.pl should detect supercomputers with specialized software environments...
I think we should try to get this dealt with before 6.8.0 is released Phil Miller
12:54 PM Bug #1381: Crash in LrtsInitCpuTopo() on Quartz with verbs layer
This is the commit that I thought automated Mellanox vs. QLogic choice in verbs: https://charm.cs.illinois.edu/gerrit... Sam White
12:50 PM Bug #1381: Crash in LrtsInitCpuTopo() on Quartz with verbs layer
Building with --with-qlogic didn't help, I get the same error. Sam White
12:39 PM Bug #1381: Crash in LrtsInitCpuTopo() on Quartz with verbs layer
I only see this:
https://charm.cs.illinois.edu/gerrit/gitweb?p=charm.git;a=commitdiff;h=e3cfe89fffdd9349ababf5b13b01...
Jim Phillips
12:36 PM Bug #1381: Crash in LrtsInitCpuTopo() on Quartz with verbs layer
I added Bilge as a watcher, since I thought we had automated that build option? Sam White
12:35 PM Bug #1381: Crash in LrtsInitCpuTopo() on Quartz with verbs layer
You can fix this by adding --with-qlogic to the command line, although it will probably still be slower than MPI. Jim Phillips
12:03 PM Bug #1381 (Rejected): Crash in LrtsInitCpuTopo() on Quartz with verbs layer
Quartz is a new cluster with an Intel Omni-Path 100 Gb/s interconnect. If I build on verbs layer and run on more than... Sam White
12:44 PM Bug #1329: Hang in exit in TRAM test on gni-crayxc-smp
Core group would like to see some progress on this issue. Do you have access to Edison to reproduce this? Sam White
12:33 PM Feature #68 (In Progress): LRTS support for setting up a message to send and transmitting a GET h...
Is this solved by #1234? Eric Bohm
12:30 PM Feature #117: Projections traces with fewer files than 1-per-PE
This is a good point, and a change should definitely be made to plan ahead. I'll sketch out some possible designs as... Ronak Buch
12:28 PM Feature #109 (Merged): Test and merge section ID and manager work
Eric Bohm
10:32 AM Bug #1367 (Rejected): AMPI comm creation should use bulk construction
Distributed creation makes more sense and is already used. It's not distributed vs bulk that is the problem, it's tha... Sam White

01/25/2017

02:55 PM Bug #1294 (Merged): Darwin SMP failure in exit
Michael Robson
10:18 AM Bug #1294: Darwin SMP failure in exit
Reported to Apple as Bug #30186931. There doesn't seem to be any way in their system to CC others, so I'll update if/... Phil Miller
02:53 PM Bug #1379: SDAG doesn't properly handle callbacks to [reductiontarget] methods with refnums
Simple example program to reproduce this issue (hang) is attached. Written by Ed Hutter. Sam White
06:18 PM Feature #1378: 64-bit Charm message sizes
Previously merged, related patches:
Use size_t instead of int for all PUP interfaces: https://charm.cs.illinois.e...
Sam White

01/24/2017

04:46 PM Bug #1379 (Merged): SDAG doesn't properly handle callbacks to [reductiontarget] methods with refnums
If you want to do something like 'when rednResult[refnum](int iter, double sum) serial { }' where rednResult is a [re... Sam White
02:59 PM Feature #1378: 64-bit Charm message sizes
Also, note that AMPI has a patch in gerrit for MPI-3 large count support, which should work once this Charm message s... Sam White
02:59 PM Feature #1378 (In Progress): 64-bit Charm message sizes
Currently, PUP supports 64bit sizes, sizes of chares are 64bit, the GNI/Isomalloc mempool uses 64bit sizes, but Charm... Sam White
02:51 PM Feature #1164 (Merged): AMPI support for MPI-3's request-based RMA routines
Sam White
02:07 PM Bug #1360: AMPI megampi test fails on mpi-crayxc and darwin builds
Also seeing a failure on netlrts-darwin-x86_64 on Wit in megampi with +p1 +vp1. Doesn't reproduce on my Mac, but is ... Sam White
02:02 PM Feature #1366 (Merged): AMPI thread should not be awoken unless progress can be made
I will follow up with the similar changes for recv and probe. They each have their own complications.
Recv: https:...
Sam White
09:46 PM Feature #1366 (Implemented): AMPI thread should not be awoken unless progress can be made
Sam White

01/23/2017

03:34 PM Bug #1375: os-isomalloc failures during startup on SMP builds
After testing this more, it appears the above change broke '-memory isomalloc' on SMP mode (it was already broken for... Sam White
01:01 PM Bug #1375: os-isomalloc failures during startup on SMP builds
Fixed the hang during initialization: https://charm.cs.illinois.edu/gerrit/#/c/2151/
But now we get a failure afte...
Sam White
02:24 PM Feature #1366 (In Progress): AMPI thread should not be awoken unless progress can be made
Sam White
12:06 PM Cleanup #966 (Implemented): cuda build fails unituitvely on blue waters when cudatoolkit isn't lo...
Same thing happens on NCSA nano cluster when no CUDA toolkit directory exists.
Modified build script to bail out as ...
Jaemin Choi
09:48 PM Bug #1377 (Closed): linking megatest on Eos fails with undefined references to ceil and floor fro...
The trick for C++ 11 that works on Edison (Cray XC) fails to link C++ on OLCF Eos:... Jim Phillips
09:21 PM Bug #1376: AMPI_Ireduce only creates a request at the root
examples/ampi/creduce only waits on the result at the root. That should be changed as well. Sam White
09:20 PM Bug #1376 (Merged): AMPI_Ireduce only creates a request at the root
MPI_Ireduce and other "all-to-one" type nonblocking collectives currently only create and set the request on the root... Sam White

01/22/2017

04:16 PM Feature #1357 (In Progress): Add AMPI RMA routine tests to megampi
This isn't really a bug, update the commit message accordingly: https://charm.cs.illinois.edu/gerrit/#/c/2148/ Sam White
10:40 AM Bug #1331: Isomalloc hangs in startup for Clang non-SMP builds
I opened a separate issue for os-isomalloc failures on SMP builds here: https://charm.cs.illinois.edu/redmine/issues/... Sam White
10:39 AM Bug #1375 (Merged): os-isomalloc failures during startup on SMP builds
os-isomalloc fails during startup on the first access to a lock in cputopolgy.C. os-isomalloc works on Clang non-SMP ... Sam White

01/21/2017

10:42 PM Bug #1358 (Merged): AMPI Iget extensions are broken
Sam White

01/20/2017

05:41 PM Bug #1374 (Merged): Reduction clause doesn't work on the OpenMP integration with icc and clang
Reduction clause doesn't work on the OpenMP integration with icc and clang but works with GCC.
The reason why it ...
Seonmyeong Bak
04:46 PM Bug #1294 (Implemented): Darwin SMP failure in exit
We (Phil and I) isolated what we think might be a compiler bug that's causing this issue. See the attached files for ... Michael Robson
02:28 PM CharmDebug Bug #1373 (Merged): Unable to attach to an application it launched
Created launch configuration using the built in gui. When "start" is triggered, it was capable of launching the appli... Eric Bohm
02:23 PM Bug #1372 (Resolved): Using GCC-6 in support of Intel icc/icpc on Edison fails with no member "it...
Compatible versions of gcc and icc must be used together (gcc/4.9 and icc/15.9 work together). Sam White
09:15 AM Bug #1372: Using GCC-6 in support of Intel icc/icpc on Edison fails with no member "iterator_cate...
Using gcc/4.9.3 works. I'm not sure if gcc-5 works or not (or if that's available on Edison).
This suggests that t...
Sam White
09:01 AM Bug #1372 (Closed): Using GCC-6 in support of Intel icc/icpc on Edison fails with no member "iter...
Trying to build on Edison (which has an older gcc in /usr/bin) the advice
for C++11 code is to module load gcc (6.1...
Sam White
09:03 AM Feature #1366: AMPI thread should not be awoken unless progress can be made
Support in AMPI_Wait* routines: https://charm.cs.illinois.edu/gerrit/#/c/2152/
Similar changes can be made in bloc...
Sam White

01/19/2017

05:46 PM Feature #1371 (New): Within-node PUP API
When migrating a chare within the same logical node, we can optimize migration time by allowing the application to no... Sam White
05:40 PM Bug #1278: ReductionStarting overhead in SMP mode
In Core we discussed the possibility of passing CkArrayOptions through to the reduction managers just like what is al... Sam White
04:51 PM Feature #1352: CkArrayOptions callback for completion of chare array initialization
Reassign if urgent. Isomalloc hangs and other AMPI things are more important to me at the moment. This is only really... Sam White
11:49 AM Feature #1352: CkArrayOptions callback for completion of chare array initialization
AMPI's use case for this (intercomm creation) is one-at-a-time/element-by-element initial insertion. Perhaps we shoul... Sam White
04:47 PM Bug #1189 (Merged): AMPI resumeOnRecv should be a property of the thread, not each comm
Sam White
04:46 PM Bug #1358 (Implemented): AMPI Iget extensions are broken
Sam White
04:46 PM Bug #1240 (New): MPI_COMM_SELF handling issues in AMPI
Sam White
02:47 PM Bug #1367 (Rejected): AMPI comm creation should use bulk construction
Currently, AMPI subcommunicators and intercommunicators are created by ckNew with numInitial=0, then proxy[idx].inser... Sam White
11:27 AM Feature #1366 (Merged): AMPI thread should not be awoken unless progress can be made
AMPI's multiple completion routines currently block on any recv, while they could sleep until all or a specific reque... Sam White
11:33 PM Cleanup #1365 (Merged): Deduplicate AmpiRequest routines and their handling of NULL statuses
There are 2 parts to this:
1. A successful itest/test should have the same effect as a call to complete/wait, so tha...
Sam White

01/18/2017

05:16 PM Bug #1342 (Merged): Segfaults when using '-tracemode perfReport'
Ronak Buch
04:33 PM Bug #1342 (Implemented): Segfaults when using '-tracemode perfReport'
https://charm.cs.illinois.edu/gerrit/#/c/2136/1 Phil Miller
02:30 PM Cleanup #1363: Remove/deprecate dead machine layers
Not really a bug - their presence doesn't much hurt anything. Phil Miller
02:30 PM Cleanup #1363: Remove/deprecate dead machine layers
Format description to avoid ugly strike-through display Phil Miller
10:52 AM Bug #1080 (Merged): multicore projections tracing runs hang at startup on 129 pes
Phil Miller
08:58 AM Bug #1364: Review use of volatile variables in the runtime
The one in ckarrayreductionmgr was unused, so removed here: https://charm.cs.illinois.edu/gerrit/#/c/2132/ Sam White
08:36 AM Bug #1364 (New): Review use of volatile variables in the runtime
We've seen instances of volatile being used in places in the runtime where memory fences/barriers are actually needed... Sam White

01/17/2017

11:12 AM Bug #1331: Isomalloc hangs in startup for Clang non-SMP builds
Does rearranging that to not be a static constructor fix this crash? Phil Miller
09:03 AM Bug #1331: Isomalloc hangs in startup for Clang non-SMP builds
Yes, the failure is that a CmiNodeLock in conv-core/cputopology.C is NULL when it shouldn't be. Trying to figure out ... Sam White
09:00 AM Bug #1331: Isomalloc hangs in startup for Clang non-SMP builds
It looked like fairly similar failures across several of them to me. I suspect that the one we're seeing now ought to... Phil Miller
08:40 AM Bug #1331: Isomalloc hangs in startup for Clang non-SMP builds
Looking at autobuild from last night, os-isomalloc passed the tests isomalloc had been except netlrts-linux-smp, netl... Sam White
10:26 AM Bug #1337: Cpv Declarations of types with constructors may induce 'static initialization order fi...
Found another static variable of complex type in src/conv-core/cputopology.C: static CpuTopology cpuTopo;
Edit: nv...
Sam White
11:26 PM Bug #1155: AMPI's non-blocking collectives are not sequenced
Should be able to break this down into multiple parts:
1. Make the sequence number into the refnum instead of a memb...
Sam White
10:41 PM Feature #1158 (In Progress): AMPI scatter(v) performance is poor
We can wait until we have a Charm++ scatter operation to use in AMPI. Sam White
10:36 PM Feature #712 (Closed): Enable user-specified initial mappings of AMPI ranks to cores
No user of this, and the implementation is sitting on Ehsan's interop branch if need be. Sam White
10:33 PM Feature #1095: Improve AMPI error handling
The derived datatypes routines and RMA all need error handling. Sam White
09:45 PM Cleanup #1363 (Merged): Remove/deprecate dead machine layers
The @-sol, -cygwin, and net-@ layers are no longer needed. Probably the @-win32 and -cell@ layers too. We may want to... Sam White
09:29 PM Feature #23 (New): Optimization of MPI layer
It looks like none of the work mentioned above was ever merged... Sam White
09:25 PM Feature #13: CkCallback to section breaks
Is this code in a branch somewhere, waiting on the 64bit ID merge? Because 64bit IDs were merged months ago... If not... Sam White
09:19 PM Feature #460 (Closed): Reduction for stats collection
Stats reductions were added a while ago here: https://charm.cs.illinois.edu/gerrit/#/c/1116/ Sam White
09:18 PM Feature #639: method to distribute message receives across pes in node
Reassigning to PPL for now since Harshitha graduated... Sam White
09:12 PM Feature #19 (New): Extend TopoManager to work on an n-dimensional torus
Sam White

01/16/2017

05:00 PM Bug #1331: Isomalloc hangs in startup for Clang non-SMP builds
os-isomalloc seems to work everywhere isomalloc does, plus it works on clang non-SMP builds. The only caveat is that ... Sam White
03:41 PM Bug #1331: Isomalloc hangs in startup for Clang non-SMP builds
Great, I'll push the change to our tests to gerrit and start trying out os-isomalloc on a few different machines. Als... Sam White
03:16 PM Bug #1331: Isomalloc hangs in startup for Clang non-SMP builds
When I link against @-memory os-isomalloc@ instead of @-memory isomalloc@ (i.e. using the system malloc as the baseli... Phil Miller
09:35 AM Bug #1361 (Merged): multicore-linux-x86_64 should alias to multicore-linux64
If you try to build with linux-x86_64 instead of linux64 our build script aborts with the following error:... Sam White
09:24 AM Bug #1358 (In Progress): AMPI Iget extensions are broken
https://charm.cs.illinois.edu/gerrit/#/c/2125/ Sam White

01/15/2017

12:00 PM Bug #1360: AMPI megampi test fails on mpi-crayxc and darwin builds
For some reason I can't get an interactive job on Edison now, and I get the same exact error message when I try to su... Sam White
11:59 AM Bug #1360 (Merged): AMPI megampi test fails on mpi-crayxc and darwin builds
AMPI's megampi test passes on +p1 +vp1 and +p +vp2 but fails on +p2 +vp2. The same test passes all runs on Blue Waters. Sam White
11:42 AM Bug #1323 (In Progress): megatest multisection test failures
Marking 'In Progress' since we continue to see failures here, though some of the issues have been fixed. Sam White

01/14/2017

04:50 PM Bug #1149: Cray CC builds are broken
Blue Waters has cce/8.5.5 now (Edison still doesn't). I copied over gni-crayxc/conv-mach.sh to gni-crayxe/ and tried ... Sam White

01/13/2017

04:13 PM Bug #1359 (Merged): ODR violations among helper classes in LB strategies
Sam White
03:31 PM Bug #1359: ODR violations among helper classes in LB strategies
https://charm.cs.illinois.edu/gerrit/#/c/2124/ Phil Miller
03:31 PM Bug #1359: ODR violations among helper classes in LB strategies
Please link the the changes when setting status to _Implemented_, so that it's easy to cross-reference Phil Miller
03:28 PM Bug #1359 (Implemented): ODR violations among helper classes in LB strategies
Juan Galvez
01:01 PM Bug #1359: ODR violations among helper classes in LB strategies
I like your suggestion much better than mine. It applies equally well to all of the helper classes mentioned. Phil Miller
10:32 AM Bug #1359: ODR violations among helper classes in LB strategies
Another option is to just move the compare classes inside their load balancer class, which would probably be cleaner ... Juan Galvez
10:50 PM Bug #1359 (Merged): ODR violations among helper classes in LB strategies
... Phil Miller
07:36 PM Feature #1352: CkArrayOptions callback for completion of chare array initialization
I'm not sure of what you mean by 'all types of chare array creation' - I think we only need to be concerned with bulk... Phil Miller
07:07 PM Bug #1080 (Implemented): multicore projections tracing runs hang at startup on 129 pes
This ended up being caused by an array overflow overwriting a static guard, which then allowed a previously called fu... Ronak Buch

01/12/2017

03:53 PM Feature #1352: CkArrayOptions callback for completion of chare array initialization
One thing I am not sure of that you or someone might just know: Is there a single point of exit in the initialization... Sam White
02:36 PM Bug #1191: AMPI is missing MPI_Type_*_{attr,keyval} routines
No-op implementations merged. Sam White
02:35 PM Feature #969 (In Progress): AMPI support for collectives on inter-communicators
https://charm.cs.illinois.edu/gerrit/#/c/2084/ Sam White
02:34 PM Bug #1257 (Merged): AMPI missing support for PSCW synchronization of RMA
https://charm.cs.illinois.edu/gerrit/#/c/2016/ Sam White
02:33 PM Bug #1356 (Merged): AMPI_Accumulate is broken
Sam White
09:55 AM Bug #1358 (Merged): AMPI Iget extensions are broken
examples/ampi/onesided/ appears to be broken. Do 'make test TESTOPTS=++local' in that directory and debug. It may be ... Sam White
09:53 AM Feature #1357 (Merged): Add AMPI RMA routine tests to megampi
AMPI's RMA routines are not tested at all right now in nightly build or Jenkins.
Add simple tests for correctness to...
Sam White
11:23 PM Feature #952: Update AMPI's version of ROMIO
Need to implement generalized requests, datatypes, and anything else it uses that we don't currently support. Sam White
11:22 PM Feature #984: AMPI support for missing MPI-3 non-blocking collectives
The remainder of these are rarely used Sam White
11:21 PM Feature #1095: Improve AMPI error handling
THings are better now for 6.8.0, but more systematic changes can be made for future releases. Sam White
11:17 PM Bug #1104: AMPI instances may change if migrated while suspended
I think we have fixed all or almost all instances of this bug, but it is definitely something to keep an eye on since... Sam White

01/11/2017

05:04 PM Bug #1356 (Implemented): AMPI_Accumulate is broken
https://charm.cs.illinois.edu/gerrit/#/c/2115/ Sam White
02:40 PM Bug #1356: AMPI_Accumulate is broken
This was silly: I forgot that the ampi instance has a direct pointer to its ampiParent that it can just use rather th... Sam White
02:33 PM Bug #1356 (Merged): AMPI_Accumulate is broken
When I refactored AMPI's MPI_Op data structures to support non-commutative operations, I made the table to look up Op... Sam White
02:16 PM Bug #1331: Isomalloc hangs in startup for Clang non-SMP builds
TODO:
1. git bisect to see if this has ever worked
2. Run under valgrind with an eye on uninitialized variables.
Sam White
01:43 PM Bug #1324: Collision Detection library failures due to changes in demand creation
Collision detection library has no current owner. Will seek one in core meeting. Eric Bohm
01:43 PM Feature #1351: For broadcasts, avoid copies for destination PEs on same node for [nokeep] messages
If you do "grep -r "CMK_BROADCAST_USE" charm/src/* you will see that currently this support is turned off by this mac... Sam White
01:24 PM Bug #1337: Cpv Declarations of types with constructors may induce 'static initialization order fi...
Shifting to low priority as most of these are in code that is more likely to be eliminated than to become important. Eric Bohm
01:14 PM Bug #1327: AMPI derived datatype support for RMA routines
When I refactored AMPI's MPI_Op data structures to support non-commutative operations, I made the table to look up Op... Sam White
10:34 AM Bug #1294: Darwin SMP failure in exit
Have you tried alllocating an aligned buffer, passing that to pthread_attr_setstack() and pthread_create()?
What ver...
Sam White
10:14 AM Bug #1294 (In Progress): Darwin SMP failure in exit
I am only able to reproduce this error on wit and not on my personal laptop. Running Valgrind gives the following out... Michael Robson
07:57 PM Bug #1205: AMPI's -tlsglobals option is only supported by GCC
This missing support is known to Clang/LLVM since 2013. I added a comment about AMPI's desire for this support: https... Sam White
07:49 PM Bug #1350 (Merged): charmc's '-gen-cpm' option broken on bgclang builds
Sam White
07:49 PM Bug #1354 (Merged): CmiMemoryAtomics fail to build on bgclang
Follow up with making the workstealing queue use C11 atomics after the lockless queue patch is merged. Sam White

01/10/2017

04:07 PM Bug #1278: ReductionStarting overhead in SMP mode
Unconditionally defining GROUP_LEVEL_REDUCTION to 1 in ckreduction.h quiets this down, at the cost of some efficiency... Phil Miller
12:00 PM Feature #1352: CkArrayOptions callback for completion of chare array initialization
Could you get this implemented? Hand off to Ed, Karthik or such with your guidance if you think they can do it. Phil Miller
01:22 AM Bug #1354: CmiMemoryAtomics fail to build on bgclang
https://charm.cs.illinois.edu/gerrit/#/c/2111/3 Seonmyeong Bak
01:10 AM Bug #1354 (Merged): CmiMemoryAtomics fail to build on bgclang
CmiMemoryAtomics fail to build on bgclang because they use __atomic_fetch_and_add(sub) instead of __atomic_fetch_add(... Seonmyeong Bak
10:54 PM Bug #1350 (Implemented): charmc's '-gen-cpm' option broken on bgclang builds
Move the test for CMK_USING_BGCLANG after conv-config.sh is executed in charmc: https://charm.cs.illinois.edu/gerrit/... Sam White

01/09/2017

05:29 PM Feature #1353 (Merged): charmc hardcodes unversioned compiler names
charmc always compiles source files using `g++` (or `clang++` or similar). This means there is no easy way to test c... William Throwe
11:52 AM Bug #1350: charmc's '-gen-cpm' option broken on bgclang builds
The definition of CMK_USING_BGCLANG in the cc-clang.sh files is not visible to charmc, so this if-statement that adds... Sam White
10:34 PM Feature #1352 (Merged): CkArrayOptions callback for completion of chare array initialization
Add a callback to CkArrayOptions to be triggered when all of the initial objects are constructed in a chare array.
R...
Sam White
10:28 PM Feature #1351: For broadcasts, avoid copies for destination PEs on same node for [nokeep] messages
Related (or redundant): https://charm.cs.illinois.edu/redmine/issues/1184 Sam White
09:50 PM Feature #1351 (Closed): For broadcasts, avoid copies for destination PEs on same node for [nokeep...
For broadcasts with [nokeep] messages, RTS makes a copy for each PE on the same node. Copies can be avoided, if the ... Vipul Harsh

01/08/2017

05:02 PM Bug #1280 (Rejected): CkLoop gives bad results on > 64 pes per process
Phil Miller
04:31 PM Bug #1350 (Merged): charmc's '-gen-cpm' option broken on bgclang builds
This commit broke compilation using charmc's -gen-cpm option: https://charm.cs.illinois.edu/gerrit/#/c/2090/
See aut...
Sam White

01/07/2017

02:07 PM Cleanup #1346 (Implemented): Remove AMPI's MSGLOG support for sequential re-run of parallel execu...
https://charm.cs.illinois.edu/gerrit/#/c/1907/ Sam White
01:20 PM Bug #1347 (Merged): megampi fails on mpi-crayxc with inclusion of charm++.h
Sam White
12:39 PM Bug #1347 (Implemented): megampi fails on mpi-crayxc with inclusion of charm++.h
https://charm.cs.illinois.edu/gerrit/#/c/2106/ Sam White
12:00 PM Bug #1347 (Merged): megampi fails on mpi-crayxc with inclusion of charm++.h
All other AMPI examples/tests pass, and removing '#include "charm++.h"' from megampi fixes the failures in it.
It is...
Sam White

01/06/2017

11:49 AM Cleanup #1346: Remove AMPI's MSGLOG support for sequential re-run of parallel execution
Yeah, but BigSim + CODES are capable of doing that. That's exactly what Nikhil's and my paper at SC this year did, bu... Sam White
11:03 AM Bug #1345 (Merged): Workstealing task queue causes a race condition.
Seonmyeong Bak
09:13 PM Bug #1330 (Merged): llvm_openmp runtime doesn't build on BGQ or Darwin
Sam White

01/05/2017

02:19 PM Cleanup #1346: Remove AMPI's MSGLOG support for sequential re-run of parallel execution
Looks like that would have been very useful for things related to Bigsim performance predictions. Record execution of... Phil Miller
01:55 PM Cleanup #1346 (Implemented): Remove AMPI's MSGLOG support for sequential re-run of parallel execu...
Background on what this code is supposed to do: http://charm.cs.illinois.edu/manuals/html/ampi/manual.html#SECTION000... Sam White
08:26 AM Bug #1323: megatest multisection test failures
Nevermind, we're still getting a segfault from this test: https://charm.cs.illinois.edu/autobuild/cur/gni-crayxc.txt Sam White
10:53 PM Bug #1323: megatest multisection test failures
https://charm.cs.illinois.edu/gerrit/#/c/2095/
https://charm.cs.illinois.edu/gerrit/#/c/2096/
Should this be mark...
Sam White
10:51 PM Bug #1330: llvm_openmp runtime doesn't build on BGQ or Darwin
https://charm.cs.illinois.edu/gerrit/#/c/2090/ Sam White

01/04/2017

01:15 AM Bug #1345: Workstealing task queue causes a race condition.
Seonmyeong Bak wrote:
> Workstealing task queue incurs a race condition due to the following faults in the implement...
Seonmyeong Bak
10:52 PM Bug #1345: Workstealing task queue causes a race condition.
https://charm.cs.illinois.edu/gerrit/#/c/2101/ Sam White
10:51 PM Bug #1345 (In Progress): Workstealing task queue causes a race condition.
Sam White
06:10 PM Bug #1345 (Merged): Workstealing task queue causes a race condition.
Workstealing task queue incurs a race condition due to the following faults in the implementation.
inline static v...
Seonmyeong Bak
12:21 AM Feature #1298 (In Progress): Build MetisLB by default
https://charm.cs.illinois.edu/gerrit/#/c/1984/ Sam White
12:12 AM Bug #1327 (In Progress): AMPI derived datatype support for RMA routines
Sam White
12:11 AM Bug #1344 (Merged): tests/converse/commbench/pingpong-cachemiss: SMP builds crash due to mistaken...
Sam White
10:43 PM Documentation #1265 (In Progress): Document LLVM OpenMP runtime integration
Seonmyeong Bak
10:43 PM Documentation #1265 (Implemented): Document LLVM OpenMP runtime integration
Seonmyeong Bak

01/03/2017

02:45 PM Bug #1307 (Implemented): AMPI_Comm_free should free the ampi instance
The AMPI-level stuff is implemented, this is just blocked on the Charm++-level issue #1312 Sam White
10:24 AM Bug #1344 (Implemented): tests/converse/commbench/pingpong-cachemiss: SMP builds crash due to mis...
Fix in change 2098 Phil Miller
09:53 AM Bug #1344 (In Progress): tests/converse/commbench/pingpong-cachemiss: SMP builds crash due to mis...
... Phil Miller
09:17 AM Bug #1344: tests/converse/commbench/pingpong-cachemiss: SMP builds crash due to mistakenly shared...
Some more observations, with assertions, etc.
In startNextIter, the 'then' case of the if sees pva(nextNbr) == -1 ...
Phil Miller
08:13 AM Bug #1344 (Merged): tests/converse/commbench/pingpong-cachemiss: SMP builds crash due to mistaken...
From netlrts-linux-x86_64-smp-clang, run as
> ./charmrun +p4 ++ppn 2 ++debug-no-pause ./pgm...
Phil Miller
 

Also available in: Atom