Project

General

Profile

Activity

From 04/26/2018 to 05/25/2018

Today

03:18 PM Support #1905 (Closed): pami* autobuilds failing since C -> C++ conversion
Fix for adding separate targets: https://charm.cs.illinois.edu/gerrit/#/c/charm/+/4198/
Additional fixes were made...
Nitin Bhat
02:00 PM Cleanup #1915: Make the build script clean/remove old builds
All 3 of those patches have been merged, so rebuilding an already built target should work smoothly now.. We could st... Sam White

05/24/2018

04:18 PM Feature #1731: Complete spack installation script
Patch for the manual is here: https://charm.cs.illinois.edu/gerrit/4229 Matthias Diener
12:42 PM Feature #1731: Complete spack installation script
Yeah it could be Sam White
12:40 PM Feature #1731: Complete spack installation script
By the way, should this be mentioned somewhere in the manual (e.g., Appendix A. Installing Charm++)? Matthias Diener
11:43 AM Feature #1731 (Merged): Complete spack installation script
The patch has been merged upstream. Matthias Diener
03:00 PM Feature #112: object location services: Share array element location cache above PE level
Another use case is supporting sends to chare array elements from [immediate] methods of node groups. That currently ... Sam White
02:42 PM Bug #1924 (New): Calls to chare array element entry methods can fail from [immediate] node group ...
If an [immediate] entry method executes on the comm thread, then sending a message to chare array element will abort ... Sam White
02:36 PM Bug #1922: Isomalloc fails with large memory footprints
The problem here appears to be that the application tried to allocate a buffer of size >= 256 MB, which is the max si... Sam White

05/23/2018

04:02 PM Feature #1731 (Implemented): Complete spack installation script
Matthias Diener
04:01 PM Feature #1731: Complete spack installation script
Pull request here: https://github.com/spack/spack/pull/8260
Matthias Diener
03:02 PM Feature #1731 (In Progress): Complete spack installation script
Ok, I'll do it. Matthias Diener
02:59 PM Feature #1731: Complete spack installation script
Yeah that'd be useful, can you do that? Edward Kim is I think no longer at Charmworks. Sam White
02:50 PM Feature #1731: Complete spack installation script
It would be nice to be able to build the git version of charm++ via spack (which is something spack supports: https:/... Matthias Diener
02:13 PM Feature #1731 (Merged): Complete spack installation script
Edward Kim got his pull request merged into the Spack repo: https://github.com/spack/spack/pull/7291 Sam White
03:33 PM Feature #1584 (In Progress): Allow exit codes in CkExit() / CkAbort()
Sam White
02:25 PM Cleanup #1923 (New): Remove BigSim support
Remove all BigSim-related code from Converse, Charm, AMPI, etc. after the 6.9.0 release Sam White
01:52 PM Support #1512 (In Progress): Measure test coverage
Eric Bohm

05/22/2018

05:07 PM Feature #1584: Allow exit codes in CkExit() / CkAbort()
https://charm.cs.illinois.edu/gerrit/#/c/charm/+/4219/ has a proof-of-concept implementation for CkExit() based on th... Matthias Diener
04:05 PM Feature #1584: Allow exit codes in CkExit() / CkAbort()
I don't believe changing all existing code would make it out of committee. Evan Ramos
04:03 PM Feature #1584: Allow exit codes in CkExit() / CkAbort()
Another option would be to always require an exit code for CkExit() (like exit() does), and to change all existing code. Matthias Diener
03:56 PM Feature #1584: Allow exit codes in CkExit() / CkAbort()
Matthias Diener wrote:
> We could do something like this: https://stackoverflow.com/questions/3046889/optional-param...
Evan Ramos
12:27 PM Feature #1584: Allow exit codes in CkExit() / CkAbort()
It would be nice to get this into 6.9.0 since Jozsef Bakosi recently requested it. Sam White
04:03 PM Bug #1922: Isomalloc fails with large memory footprints
From Yidong Xia and his postdoc at Idaho National Lab, who are currently running MiniMD on AMPI:... Sam White
11:54 AM Bug #1922 (New): Isomalloc fails with large memory footprints
Isomalloc has not been stress tested for its efficiency for large memory footprint applications.
What is the memory ...
Sam White
02:42 PM Cleanup #1915 (Implemented): Make the build script clean/remove old builds
https://charm.cs.illinois.edu/gerrit/4215
https://charm.cs.illinois.edu/gerrit/4216
https://charm.cs.illinois.edu/g...
Evan Ramos
01:30 PM Documentation #1845 (Merged): Documentation for the Zerocopy Direct API
Sam White
11:50 AM Feature #1921 (New): Make Isomalloc/mempool not use the pool for large allocations
Currently Isomalloc always uses the mempool by default, and the mempool only tries to use pooled memory.
For large a...
Sam White
11:47 AM Feature #1920 (New): Make mempool aware of Isomalloc's per-slot metadata
By default Isomalloc uses the mempool implementation in src/arch/util/mempool.{h,c}.
That mempool always allocates m...
Sam White
08:54 PM Bug #1268 (Merged): AMPIF issues due to C++ main routine
Sam White

05/21/2018

12:11 PM Documentation #1917 (Merged): Inadequate documentation about immediate messages (section 10.1.3.2)
Nitin Bhat
11:13 AM Documentation #1917 (Implemented): Inadequate documentation about immediate messages (section 10....
Fix: https://charm.cs.illinois.edu/gerrit/#/c/charm/+/4210/ Nitin Bhat

05/18/2018

04:52 PM Bug #1268 (Implemented): AMPIF issues due to C++ main routine
https://charm.cs.illinois.edu/gerrit/4203 Evan Ramos
01:16 PM Bug #1699: [inline] entry methods should use perfect forwarding in C++11
Nils identified an issue with the above patch for Spectre.
New patches from Evan are here:
https://charm.cs.illin...
Sam White
08:22 AM Feature #1919 (New): Zerocopy Direct API: Allow user to pass an arbitrary reference pointer into ...
This feature is useful for AMPI and was requested by Sam Nitin Bhat
08:17 PM Feature #1882 (Merged): Print slightly more detailed CPU topology info at startup by default
Sam White

05/17/2018

05:08 PM Bug #1268: AMPIF issues due to C++ main routine
I cobbled this together:
args.f90:...
Evan Ramos
03:16 PM Bug #1268: AMPIF issues due to C++ main routine
Yeah, I wouldn't be too surprised if there were issues with C++ codes if our main was always in Fortran.
As for pa...
Sam White
02:50 PM Bug #1268: AMPIF issues due to C++ main routine
One potential issue is that support for reading argv/argc from Fortran was only added in Fortran 2003, and converting... Evan Ramos
02:11 PM Bug #1268: AMPIF issues due to C++ main routine
That could also work. I think Phil's idea above was to compile Charm++/AMPI as a library without a main(), plus two s... Sam White
01:39 PM Bug #1268: AMPIF issues due to C++ main routine
I took a look at this and I'm not sure how to go about replacing the @main@ in ck-core/main.C with a Fortran @PROGRAM... Evan Ramos
03:54 PM Feature #1909 (Merged): Proper support for callbacks to [immediate] entry methods of node groups
Nitin Bhat
03:54 PM Documentation #1916 (Merged): Clarify in the manual that immediate entry methods are only support...
Nitin Bhat
09:39 AM Documentation #1916 (Implemented): Clarify in the manual that immediate entry methods are only su...
Fix: https://charm.cs.illinois.edu/gerrit/#/c/charm/+/4197/ Nitin Bhat
09:24 AM Documentation #1916 (Merged): Clarify in the manual that immediate entry methods are only support...
Nitin Bhat
03:54 PM Documentation #1911 (Merged): Better explanation of [immediate] entry methods
Nitin Bhat
11:27 AM Documentation #1911 (Implemented): Better explanation of [immediate] entry methods
Fix: https://charm.cs.illinois.edu/gerrit/#/c/charm/+/4199/ Nitin Bhat
10:46 AM Documentation #1911 (In Progress): Better explanation of [immediate] entry methods
In addition to this change, I also see another related section that seems incomplete as well:
10 . 1 . 3 . 2 Immed...
Nitin Bhat
09:59 AM Documentation #1911: Better explanation of [immediate] entry methods
I think the manual should say that [immediate] entry methods can run on the comm thread in SMP mode, and because of t... Sam White
01:54 PM Feature #1918 (New): Enable immediate method tracing when CMK_SMP_TRACE_COMMTHREAD is enabled
By default, it looks like immediate entry methods are not traced as they are marked with CK_EP_TRACEDISABLE by charmx... Nitin Bhat
11:25 AM Documentation #1917 (Merged): Inadequate documentation about immediate messages (section 10.1.3.2)
Currently the manual just says:
10 . 1 . 3 . 2 Immediate Messages
Immediate messages are special messages that sk...
Nitin Bhat

05/16/2018

06:20 PM Feature #1909 (Implemented): Proper support for callbacks to [immediate] entry methods of node gr...
Fix: https://charm.cs.illinois.edu/gerrit/#/c/charm/+/4196/ Nitin Bhat
12:12 PM Feature #1909 (In Progress): Proper support for callbacks to [immediate] entry methods of node gr...
Nitin Bhat
04:09 PM Bug #1913 (Merged): uFcontext threads segfault on ppc64le
Sam White
11:08 AM Bug #1913: uFcontext threads segfault on ppc64le
Make Boost uFcontext threads the default ULT implementation on all ppc64le targets: https://charm.cs.illinois.edu/ger... Sam White
11:07 AM Bug #1913 (Implemented): uFcontext threads segfault on ppc64le
Updating our Boost threads to Boost 1.67.0 fixes the issue: https://charm.cs.illinois.edu/gerrit/#/c/charm/+/4044/ Sam White
10:32 AM Bug #1913 (Merged): uFcontext threads segfault on ppc64le
Boost uFcontext threads offer the lowest overhead context switching of all our user-level thread packages, so we'd li... Sam White
04:08 PM Cleanup #1915: Make the build script clean/remove old builds
I agree with what you've said: our './build' should behave like 'make'. Sam White
03:31 PM Cleanup #1915: Make the build script clean/remove old builds
I'm not sure if removing old builds should necessarily be the default, but either way something sensible should be do... Eric Mikida
03:25 PM Cleanup #1915 (Implemented): Make the build script clean/remove old builds
If a user has a netlrts-linux-x86_64/ and a netlrts-linux-x86_64-smp/, then they do a "./build charm++ netlrts-linux-... Sam White
04:07 PM Feature #1866 (Merged): Update Boost ULTs to v1.67.0
Sam White
11:07 AM Feature #1866 (Implemented): Update Boost ULTs to v1.67.0
Sam White
04:07 PM Feature #1778 (Merged): Clean and portable way of generating a working libcharm.so for charmpy
Sam White
03:15 PM Feature #1914 (New): Make AtSync load balancing the default rather than periodic
Periodic LB is currently the default, even though no production applications use it.
The only other uses for it woul...
Sam White
03:11 PM Support #1905: pami* autobuilds failing since C -> C++ conversion
Make examples/ampi/pingpong/ run faster, especially on BGQ: https://charm.cs.illinois.edu/gerrit/#/c/charm/+/4195/ Sam White
03:04 PM Support #1905 (In Progress): pami* autobuilds failing since C -> C++ conversion
Sam White
03:04 PM Support #1905: pami* autobuilds failing since C -> C++ conversion
This may be due to examples/ampi/pingpong recently being added to 'make test' and the fact that AMPI pingpong is very... Sam White
02:59 PM Support #1905: pami* autobuilds failing since C -> C++ conversion
2 hours seems like a long time for the tests to run. Maybe #1872 describes the problem? Evan Ramos
02:56 PM Support #1905: pami* autobuilds failing since C -> C++ conversion
I found that autobuild began failing because of the addition of new tests which caused the total execution time to go... Nitin Bhat
02:50 PM Feature #1450 (Implemented): Clean up and add CUDA example programs
Sam White
02:49 PM Cleanup #1454 (Implemented): GPUManager API change
Sam White
02:49 PM Documentation #1491 (Implemented): Update documentation of GPUManager
Sam White
02:49 PM Feature #1456 (Implemented): Add more stream callbacks for use after HToD transfer and kernel exe...
Sam White
11:48 AM Feature #1875 (Rejected): Test whether hwloc can be used via a git submodule in charm++
I think we can close this report for now. As Evan said, if we use an embedded hwloc (as opposed to a system hwloc), s... Matthias Diener
10:34 AM Bug #1774: Thread migration fails on ppc64le builds
I think Isomalloc needs +isomalloc_sync on this system, but isomalloc_sync itself is hanging during startup. We need ... Sam White
08:38 AM Feature #1712 (Rejected): Avoid intermediate ctx to scheduler in case of ULTs
Sam White
08:06 PM Projections Bug #1910 (Merged): java.lang.NullPointerException in projections.analysis.Analysis.hasSumFiles
Ronak Buch
07:06 PM Documentation #1912 (Implemented): Consistent definition of "PE" throughout the manual
https://charm.cs.illinois.edu/gerrit/4188 Evan Ramos

05/15/2018

04:56 PM Bug #1633: NodeGroup Broadcasts creates many copies of the message for point to point sends
When running megatest as @./pgm +p2@, failure sometimes occurs before the first test begins, with the message
<pre...
Evan Ramos
03:40 PM Bug #1774: Thread migration fails on ppc64le builds
Programs with "-thread uFcontext" crash with a seg fault:... Nitin Bhat
02:21 PM Projections Bug #1910 (Implemented): java.lang.NullPointerException in projections.analysis.Analysis.hasSumFiles
https://charm.cs.illinois.edu/gerrit/#/c/charm/+/4176/
https://charm.cs.illinois.edu/gerrit/#/c/projections/+/4177/
Ronak Buch
01:40 PM Projections Bug #1910: java.lang.NullPointerException in projections.analysis.Analysis.hasSumFiles
The sts file is version 9. I deleted the space by hand, and projections works. Thomas Quinn
11:50 AM Projections Bug #1910 (In Progress): java.lang.NullPointerException in projections.analysis.Analysis.hasSumFiles
Oh, wait, sorry, I didn't see the space between "CmiUInt8" and ">". That does indeed seem erroneous and likely due to... Ronak Buch
11:48 AM Projections Bug #1910: java.lang.NullPointerException in projections.analysis.Analysis.hasSumFiles
The sts file is correct, but it should have failed earlier, in fact. What version is reported in the sts file (it sho... Ronak Buch
07:49 PM Projections Bug #1910: java.lang.NullPointerException in projections.analysis.Analysis.hasSumFiles
I think I found the problem: the sts file has a line:
CHARE 50 CkCacheManager<CmiUInt8 > -1
which projections can't...
Thomas Quinn
07:13 PM Projections Bug #1910 (Merged): java.lang.NullPointerException in projections.analysis.Analysis.hasSumFiles
I'm trying to analyze some projections traces from ChaNGa (linked with -tracemode projections), and trying to open an... Thomas Quinn
09:59 AM Documentation #1912 (Implemented): Consistent definition of "PE" throughout the manual
The term PE should be consistently defined as a thread/process with a scheduler that chares can live on.
The last...
Sam White
09:06 AM Documentation #1911 (Merged): Better explanation of [immediate] entry methods
The manual's entry on [immediate] entry methods doesn't say anything about why you'd use them or not use them and whe... Sam White
08:23 PM Feature #1420 (Merged): Lockless queue build option --enable-lockless-queue (off by default)
Merged the patch that adds the lockless queue but leaves it off by default. To enable it, build with '--enable-lockle... Sam White
07:07 PM Cleanup #1906 (Merged): Add {mpi,netlrts,multicore}-linux-ppc64le build targets
Sam White

05/14/2018

06:30 PM Feature #1876: Use IP multicast for faster broadcast and multicast on netlrts
It looks like the network protocol aspect of IP multicast is fairly simple. Sending packets is the same as with unica... Evan Ramos
05:48 PM Feature #1909 (Merged): Proper support for callbacks to [immediate] entry methods of node groups
Currently, we don't keep track of whether a node group method is [immediate] or not inside of our callback infrastruc... Sam White
05:46 PM Feature #1864 (Merged): Charm++ type traits for compile-time querying of types
Sam White
05:46 PM Feature #1863 (Merged): PUP support for templated abstract base classes
Sam White
01:04 PM Projections Bug #1886: Remove support for long-broken Topology Display tool
Followup patch to remove .topo file creation: https://charm.cs.illinois.edu/gerrit/#/c/charm/+/4171/ Matthias Diener
09:26 AM Bug #1774: Thread migration fails on ppc64le builds
If that doesn't work, then my next guess would be that Isomalloc is broken, possibly because we need '+isomalloc_sync... Sam White
09:12 AM Documentation #1908 (New): Document PUP::able and associated macros
This and the related functionality needs to be documented in the manual: https://charm.cs.illinois.edu/gerrit/#/c/cha... Sam White

05/12/2018

10:37 AM Feature #13: CkCallback to section breaks
Is this still a real issue? Sam White
10:19 AM Bug #1774: Thread migration fails on ppc64le builds
My current hypothesis is that context threads aren't working, so we should switch the default to use uFcontext thread... Sam White
10:17 AM Cleanup #1906 (Implemented): Add {mpi,netlrts,multicore}-linux-ppc64le build targets
Added mpi/multicore/netlrts aliases here: https://charm.cs.illinois.edu/gerrit/#/c/charm/+/4169/ Sam White
09:38 AM Support #1905: pami* autobuilds failing since C -> C++ conversion
It looks like verbs-linux-x86_64-smp failed in a bigsim test last night: http://charm.cs.illinois.edu/autobuild/cur/v... Sam White

05/11/2018

04:16 PM Feature #1844: SMP/non-SMP agnostic job launching arguments
After reading over our documentation it is better than I remembered at avoiding saying that PEs are cores/hyperthread... Sam White
03:40 PM Feature #1844: SMP/non-SMP agnostic job launching arguments
Where is "PE" described like a Core/PU? The only divergence I am aware of is how a PE is a worker thread in SMP mode ... Evan Ramos
07:07 PM Feature #1844: SMP/non-SMP agnostic job launching arguments
I think we might want to avoid "PE" since that term is overloaded in Charm already: a PE is sometimes described like ... Sam White
03:35 PM Cleanup #1907 (New): Replace CmiNumNodes et al with hwloc
With hwloc in place, there is no need to maintain our custom querying of hardware PU counts. Evan Ramos
01:10 PM Support #1905: pami* autobuilds failing since C -> C++ conversion
If you set a breakpoint on @exit()@, does it get hit, and if so what is the backtrace? Evan Ramos
09:24 AM Support #1905 (Closed): pami* autobuilds failing since C -> C++ conversion
All pami{lrts}-bluegeneq-* autobuilds have been failing since the C -> C++ conversion was done last week.
The last t...
Sam White
11:08 AM Cleanup #1906 (Merged): Add {mpi,netlrts,multicore}-linux-ppc64le build targets
I think Jim Philips said before that mpi-linux-ppc works perfectly fine on ppc64le, but for consistency/completeness ... Sam White
09:57 AM Bug #1904: Review CMK_PCQUEUE_LOCK
Also see comments on this patch: https://charm.cs.illinois.edu/gerrit/#/c/charm/+/4159/ Sam White

05/10/2018

06:47 PM Feature #1844: SMP/non-SMP agnostic job launching arguments
Bump Evan Ramos
06:46 PM Feature #1875: Test whether hwloc can be used via a git submodule in charm++
For the moment, we now include a hotfix for hwloc in order to resolve #1895, something which would have been impossib... Evan Ramos
06:43 PM Feature #789 (Merged): Charmrun should test for SSH failures when node programs fail to launch
Sam White
05:40 PM Feature #789 (Implemented): Charmrun should test for SSH failures when node programs fail to launch
++verbose is definitely enough to diagnose the described problem, but I've added this fact to the documentation in or... Evan Ramos
06:43 PM Feature #1882 (Implemented): Print slightly more detailed CPU topology info at startup by default
https://charm.cs.illinois.edu/gerrit/4164 Evan Ramos
05:58 PM Bug #1332: assumes remote shell is OpenSSH
Bump. Evan Ramos
05:11 PM Bug #522 (Closed): static linking breaks on multicore builds with 'undefined reference to `get_my...
Closing because the issue has not been reproduced. Evan Ramos
03:42 PM Cleanup #1898: C++ cleanup of code formerly compiled as C
Added C to C++ renaming tutorial. Evan Ramos
02:20 PM Bug #1904 (New): Review CMK_PCQUEUE_LOCK
From comments on https://charm.cs.illinois.edu/gerrit/418:
Eric Bohm:
> Change looks good, has this been tried on...
Evan Ramos
01:50 PM Bug #1902 (Merged): ckreduction.h {} xlC error: expected class member or base class name
Sam White
11:51 AM Bug #1902 (Implemented): ckreduction.h {} xlC error: expected class member or base class name
https://charm.cs.illinois.edu/gerrit/#/c/charm/+/4158/ Sam White
11:45 AM Bug #1902: ckreduction.h {} xlC error: expected class member or base class name
It looks like this fails due to an extra comma whenever CMK_ERROR_CHECKING is disabled:
+ reducerStruct(reducer...
Jim Phillips
11:43 AM Bug #1902 (Merged): ckreduction.h {} xlC error: expected class member or base class name
Build failure on ORNL Summit pami-linux-ppc64le-smp
../bin/charmc -optimize -production -I. -c -o DummyLB.o Du...
Jim Phillips
12:26 PM Bug #1903 (New): All load balancing framework should be off if we don't have any load balancers l...
Currently, if no load balancers are linked, NullLB is created and used. But certain features like periodic load balan... Eric Mikida
12:14 AM Feature #1865: Implement zero copy translation for move semantics and rvalue refs
Much obliged to you for posting such an incredible article! I discovered your site ideal for my necessities. It conta... Anonymous
09:42 PM Bug #1895 (Merged): Charm's hwloc can collide with OpenMPI's
Sam White
09:41 PM Bug #1881 (Merged): Make mpi-linux-x86_64 use ifort if mpicc wraps ICC
Sam White
09:40 PM Documentation #1883 (Merged): Update AMPI manual section on global variable privatization
Sam White

05/09/2018

03:29 PM Documentation #1883 (Implemented): Update AMPI manual section on global variable privatization
https://charm.cs.illinois.edu/gerrit/#/c/charm/+/4157/ Sam White
02:49 PM Feature #1420: Lockless queue build option --enable-lockless-queue (off by default)
Evan has already been working on this, and Seonmyeong is leaving PPL next week. Sam White
12:42 PM Projections Bug #1886: Remove support for long-broken Topology Display tool
Ronak Buch wrote:
> I think all of this will still work if things are run as Eric said. Using a different version of...
Ronak Buch
12:34 PM Projections Bug #1886 (Merged): Remove support for long-broken Topology Display tool
Sam White
12:37 PM Projections Bug #1901 (New): Enable viewing bracketed user events on a per-PE basis
When using Projections to visualize traces with AMPI and a high degrees of virtualization (>8 ranks/core), displaying... Sam White
11:58 AM Cleanup #1897 (In Progress): Cleanup AMPI DDT
Move simple routines and serialize() for primitive types into the header: https://charm.cs.illinois.edu/gerrit/#/c/ch... Sam White
11:12 AM Feature #871: Return data from [sync] SDAG methods
Reassigned to Eric M to review state of this task. Eric Bohm
11:07 AM Cleanup #1315: examples/charm++/jacobi*d are non-exemplary HPC code, using nested arrays
Assigning to Juan for eventual reassignment to interns or newer junior students as this seems like a worthy learning ... Eric Bohm
11:05 AM Projections Feature #1004: View showing messages grouped by size
Assigned to Ronak, but this seems like a good intern project. Eric Bohm
11:01 AM Cleanup #1898: C++ cleanup of code formerly compiled as C
Eric Bohm wrote:
> Assigning to Evan as I think he's already accomplished most of this task.
The work I've done i...
Evan Ramos
11:00 AM Cleanup #1898: C++ cleanup of code formerly compiled as C
We should probably break this down into a few different subtasks based on the categorical changes left. Some of the ... Eric Bohm
10:59 AM Cleanup #1898: C++ cleanup of code formerly compiled as C
Assigning to Evan as I think he's already accomplished most of this task. Eric Bohm
10:57 AM Charm-NG Feature #1144: Batched message delivery to objects for better cache behavior
Shifted to Charm-NG as this should be considered in the mix of what we do to revise scheduling. Eric Bohm
10:57 AM Feature #1876: Use IP multicast for faster broadcast and multicast on netlrts
I think we probably want to implement this with verbs first, rather than for netlrts, since we care more about verbs ... Sam White
10:42 AM Bug #1641 (Rejected): charmrun with nodelist option (++nodelist) fails on campus cluster
Eric Bohm
10:38 AM Bug #1740: Failure at LrtsInit with OFI build with gni provider on Edison
Is this still an issue? Or should this be closed? Eric Bohm
10:35 AM Bug #1816: Cray CCE/8.5 builds are broken
Seems like simplest answer here to have configure test for 8.5 and reject it with a (8.5 not support, use 8.6) message. Eric Bohm
10:22 AM Bug #1881 (Implemented): Make mpi-linux-x86_64 use ifort if mpicc wraps ICC
https://charm.cs.illinois.edu/gerrit/#/c/charm/+/4151/ Sam White

05/08/2018

06:22 PM Bug #1633 (In Progress): NodeGroup Broadcasts creates many copies of the message for point to poi...
The std::atomic prerequisite has been satisfied as of https://charm.cs.illinois.edu/gerrit/4108 but the existing impl... Evan Ramos
12:24 PM Bug #1633: NodeGroup Broadcasts creates many copies of the message for point to point sends
Enabling this for SMP builds causes megatest to hang at @test 0: initiated [groupring (milind)]@ when run as @./charm... Evan Ramos
03:19 PM Bug #1774: Thread migration fails on ppc64le builds
I see that it hangs for all ppc64le targets (both smp and nonsmp, all of verbs, pami and pamilrts).
I'm not very ...
Nitin Bhat
11:41 AM Bug #1899: AMPI jacobi.iso crashes in migration on gni-crayxe-persistent-smp autobuild
Stacktrace: ... Nitin Bhat
10:21 AM Bug #1899 (New): AMPI jacobi.iso crashes in migration on gni-crayxe-persistent-smp autobuild
... Nitin Bhat
11:21 AM Bug #1895 (Implemented): Charm's hwloc can collide with OpenMPI's
https://charm.cs.illinois.edu/gerrit/4145
I am also seeking to merge this fix upstream: https://github.com/open-mp...
Evan Ramos
09:26 PM Cleanup #1898: C++ cleanup of code formerly compiled as C
Also std::atomic rather than compiler builtins like __sync_* Sam White

05/07/2018

04:38 PM Feature #1184: SMP-safe CmiReference and CmiFree
Cross-referencing from Gerrit change:
Sam White wrote:
> Turning CmiReference and CmiFree into atomics does solve...
Evan Ramos
11:52 AM Feature #1778 (Implemented): Clean and portable way of generating a working libcharm.so for charmpy
Patch here: https://charm.cs.illinois.edu/gerrit/#/c/charm/+/4040/
Juan Galvez

05/06/2018

02:57 PM Feature #541 (Merged): SMP mesage passing must enforce memory ordering consistency
Sam White
01:07 PM Cleanup #1898 (New): C++ cleanup of code formerly compiled as C
Now that we compile machine and parts of Converse as C++, we can clean up some parts of the code using templates, STL... Sam White
12:46 PM Cleanup #536 (Merged): Data Races in SMP PCQueue
Sam White
12:45 PM Bug #1896 (Merged): Linker errors using Cray CCE 8.6.5
Sam White

05/05/2018

04:29 PM Bug #1774: Thread migration fails on ppc64le builds
Does AMPI work on Summit (rather than SummitDev)? We want it working for v6.9.0. Also does it work on verbs-linux-ppc... Sam White
12:18 PM Bug #1896 (Implemented): Linker errors using Cray CCE 8.6.5
https://charm.cs.illinois.edu/gerrit/#/c/charm/+/4134/ Sam White
10:14 AM Cleanup #1897 (In Progress): Cleanup AMPI DDT
Now that CkDDT confined to AMPI, we can remove the CkDDT constants and replace them directly with MPI_ constants.
Th...
Sam White

05/04/2018

04:50 PM Bug #1896: Linker errors using Cray CCE 8.6.5
... Evan Ramos
04:46 PM Bug #1896 (Merged): Linker errors using Cray CCE 8.6.5
Building on Cori. Our CmiAbort uses this, but Cray CCE doesn't seem to support it:... Sam White
10:05 AM Feature #1778 (In Progress): Clean and portable way of generating a working libcharm.so for charmpy
Patch here: https://charm.cs.illinois.edu/gerrit/#/c/charm/+/4040/
Possibly needs a few tweaks to simplify build l...
Juan Galvez

05/03/2018

02:54 PM Bug #1438 (Merged): off-node messages show created by receiving node comm thread
Ronak Buch
11:15 AM Bug #1895 (Merged): Charm's hwloc can collide with OpenMPI's
https://lists.cs.illinois.edu/lists/arc/charm/2018-05/msg00004.html Sam White

05/02/2018

06:49 PM Bug #1633 (Implemented): NodeGroup Broadcasts creates many copies of the message for point to poi...
https://charm.cs.illinois.edu/gerrit/4109 Evan Ramos
06:48 PM Feature #1184 (Implemented): SMP-safe CmiReference and CmiFree
https://charm.cs.illinois.edu/gerrit/4108
Implementing this appears to have been trivial. It required renaming a l...
Evan Ramos
06:32 PM Bug #1438 (Implemented): off-node messages show created by receiving node comm thread
https://charm.cs.illinois.edu/gerrit/#/c/projections/+/4105/ Ronak Buch
05:19 PM Bug #1364: Review use of volatile variables in the runtime
Convert volatile int commThdExit + a CmiNodeLock to std::atomic<int>: https://charm.cs.illinois.edu/gerrit/#/c/charm/... Sam White
05:17 PM Cleanup #539 (In Progress): Data race in ConverseExit
Converted the volatile int + CmiNodeLock to a std::atomic<int>, but this may well hurt performance. So we need to ben... Sam White
03:34 PM Feature #1834: Functionality to Broadcast large buffers using the nocopy API
Implementational Scheme for Nocopy API (1 source - n destinations)
1. When `CkNcpySource` is passed over a proxy...
Nitin Bhat
02:33 PM Bug #1887: Custom array indices segfault in CkVec inside of LB framework
No. Haven't heard from him. Eric Mikida
12:56 PM Feature #1860: Support HostBuffer shared memory allocation of one buffer per physical host at sam...
Just a stray idea from group meeting discussion, but we should keep in mind the idea of backing the hostBuffer with v... Michael Robson
12:48 PM Projections Bug #1886: Remove support for long-broken Topology Display tool
I think all of this will still work if things are run as Eric said. Using a different version of the j3d library and ... Ronak Buch
11:15 AM Projections Bug #1886: Remove support for long-broken Topology Display tool
The .topo files were only used for display and analysis in topology aware studies. AFAIK. When it worked, the featu... Eric Bohm
11:13 AM Bug #1894 (Rejected): AMPI zero copy patches are hanging on gni-crayxc
Nevermind, this is an issue with the commit that adds a free list for CkNcpy objects. The earlier patches all work.
...
Sam White
11:06 AM Bug #1894: AMPI zero copy patches are hanging on gni-crayxc
To reproduce:... Sam White
11:05 AM Bug #1894 (Rejected): AMPI zero copy patches are hanging on gni-crayxc
Testing this whole series of patches on Cori (Haswell), I've found that it hangs when calling rget() for an inter-pro... Sam White
08:57 AM Bug #1881: Make mpi-linux-x86_64 use ifort if mpicc wraps ICC
Also, on mpi-darwin-x86_64 we should link with the correct libc++/libstdc++ depending on if mpicxx is a wrapper for c... Sam White
12:25 AM Feature #1769: AMPI implementation of Buffered sends
One quick mpich-test that we can cross of is bsendalign. We currently return NULL for MPI_Buffer_detach, but we shoul... Sam White
12:21 AM Feature #1866: Update Boost ULTs to v1.67.0
It'd be good to merge this before 6.9.0-beta testing, but I don't think it's necessary Sam White
12:13 AM Feature #177: objid_t: load balancing infrastructure should use objid_t
This needs to be discussed in Core as to whether we should break the API that NAMD is suing or defer to a later release. Sam White
12:10 AM Documentation #1656: Update manual entries on Load Balancing strategies
Bump Sam White

05/01/2018

05:32 PM Feature #1420: Lockless queue build option --enable-lockless-queue (off by default)
I've updated the patch to use C++11 atomics. Evan Ramos
04:45 PM Bug #937 (In Progress): [local] entry methods don't set tracing event dependencies
This is one of two remaining bugs targeted at v6.9.0 Sam White
04:45 PM Bug #1438: off-node messages show created by receiving node comm thread
This is one of two remaining bugs targeted at v6.9.0 Sam White
04:43 PM Bug #1881: Make mpi-linux-x86_64 use ifort if mpicc wraps ICC
Not a release blocker Sam White
04:41 PM Bug #1892 (Merged): no rule to make target .../stdarg.h error on Windows
Sam White
02:17 PM Bug #1892 (In Progress): no rule to make target .../stdarg.h error on Windows
Please test https://charm.cs.illinois.edu/gerrit/4093 Evan Ramos
04:41 PM Bug #1858 (Merged): tests/ampi/privatization fails on mpi-linux-x86_64-smp autobuild
Sam White
02:29 PM Bug #1858: tests/ampi/privatization fails on mpi-linux-x86_64-smp autobuild
Sam White wrote:
> Can you try out an mpi-crayxc build too, to see if that should also set this? We might need all M...
Evan Ramos
03:36 PM Bug #1893 (New): cmitls sometimes cannot find __executable_start
Currently cmitls uses the symbol @__executable_start@ in order to find the .tbss segment's metadata.... Evan Ramos
02:33 PM Bug #1891 (Merged): PumpOneSidedRDMATransactions 'putOp' undeclared on Cray GNI
Sam White
01:55 PM Bug #1891 (Implemented): PumpOneSidedRDMATransactions 'putOp' undeclared on Cray GNI
Fix: https://charm.cs.illinois.edu/gerrit/#/c/charm/+/4092/ Nitin Bhat
12:30 PM Feature #921: Entry tag [inline] is unable to optimize away most of the overhead
Not really, we can do the testing on the beta version of the release as part of our usual release testing. Sam White
12:26 PM Feature #921 (Implemented): Entry tag [inline] is unable to optimize away most of the overhead
Sam White wrote:
> Yeah, I am unsure what is left to do here that isn't covered by Bug #1699. It might be good to ha...
Evan Ramos
09:14 AM Projections Bug #1886: Remove support for long-broken Topology Display tool
I don't think so. The only other thing that came to mind was BigSim, since it can simulate different network topologi... Sam White
09:07 AM Bug #1887: Custom array indices segfault in CkVec inside of LB framework
Any update from James on if this fixed the issue? Sam White

04/30/2018

05:00 PM Bug #1892: no rule to make target .../stdarg.h error on Windows
That error is everywhere in machine.d:
machine.o machine.d : machine.c \
/cygdrive/cProgram\ Files\ (x86)...
Jim Phillips
04:46 PM Bug #1892 (Merged): no rule to make target .../stdarg.h error on Windows
../bin/charmc -optimize -production -o ../lib/libconv-core.a convcore.o conv-conds.o conv-taskQ.o queueing.o msgm... Jim Phillips
04:59 PM Bug #1858: tests/ampi/privatization fails on mpi-linux-x86_64-smp autobuild
Can you try out an mpi-crayxc build too, to see if that should also set this? We might need all MPI builds to do it... Sam White
04:56 PM Bug #1858 (Implemented): tests/ampi/privatization fails on mpi-linux-x86_64-smp autobuild
-It looks like non-SMP mpi-linux-x86_64 already defines CMK_NOT_USE_TLS_THREAD to 1. I'm not sure why this did not ca... Evan Ramos
04:49 PM Bug #1858: tests/ampi/privatization fails on mpi-linux-x86_64-smp autobuild
https://charm.cs.illinois.edu/gerrit/4088 Evan Ramos
04:43 PM Bug #1858: tests/ampi/privatization fails on mpi-linux-x86_64-smp autobuild
I was able to resolve the issue with the following change:... Evan Ramos
10:24 AM Bug #1858: tests/ampi/privatization fails on mpi-linux-x86_64-smp autobuild
Semi-related: I was just looking at -tlsglobals support in charmc, and it looks like we are adding "-Wl,--allow-multi... Sam White
04:03 PM Bug #1891 (Merged): PumpOneSidedRDMATransactions 'putOp' undeclared on Cray GNI
On gemini_gni-crayxe-persistent-smp:
../bin/charmc -optimize -production -I. -o machine.o machine.c
In file in...
Jim Phillips
08:37 PM Bug #1867 (Merged): Disable partitions test on pamilrts-bluegeneq
Verified that partitions works for ppc64le builds on Summit. Nitin Bhat

04/29/2018

06:04 PM Cleanup #617: Rename windows commands to not reference Windows NT
Is this still an issue? If not, close it. Sam White
06:02 PM Bug #543 (Closed): charmrun under causalft should respect ++local
causalft and mlogft are not production features and have been broken since at least 64-bit IDs were merged for 6.8.0 Sam White
06:00 PM Support #513: Allow arbitrary ordering of options when building charm
A fix for at least part of the issue is here: https://charm.cs.illinois.edu/gerrit/#/c/charm/+/3997/ Sam White
05:59 PM Projections Feature #505: show communication threads in usage profile
Bump. Still no record of what happened here Sam White
05:58 PM Support #504 (Closed): Check for errors in values returned by library calls
With the C++ attribute to warn when a function's return value is unused, we've fixed a lot of standard library calls ... Sam White
05:57 PM Cleanup #475: Licensing of parallel random number generator from NCSA
As discussed in Core, we could get rid of this code entirely in favor of standard C++11 random number generators Sam White
05:55 PM Cleanup #306 (Closed): FEM & ParFUM libraries suffer from several poor coding practicies
I think these libraries are currently more likely to be deprecated/removed than to receive time spent on code cleanup. Sam White
05:52 PM Feature #239 (Closed): Charmrun wrapper around ibrun.symm for dual binary execution on host+mic w...
Sam White
05:45 PM Feature #23: Optimization of MPI layer
Small optimization to use MPI-3's MPI_Mprobe and MPI_Mrecv where possible: https://charm.cs.illinois.edu/gerrit/#/c/c... Sam White
11:22 AM Support #1725: Improve pup_stl testing
This patch needs to be updated with more STL containers and types: https://charm.cs.illinois.edu/gerrit/#/c/charm/+/3... Sam White

04/27/2018

04:58 PM Bug #1858: tests/ampi/privatization fails on mpi-linux-x86_64-smp autobuild
Switching the machine layer to use @MPI_THREAD_SERIALIZED@ instead proves the hypothesis incorrect, as it still crash... Evan Ramos
04:45 PM Bug #1858: tests/ampi/privatization fails on mpi-linux-x86_64-smp autobuild
> My current hypothesis is that the crash is related to the fact that the MPI machine layer uses the first thread as ... Sam White
04:34 PM Bug #1858: tests/ampi/privatization fails on mpi-linux-x86_64-smp autobuild
I tried the following hack:... Evan Ramos
04:04 PM Bug #1858: tests/ampi/privatization fails on mpi-linux-x86_64-smp autobuild
My current hypothesis is that the crash is related to the fact that the MPI machine layer uses the first thread as th... Evan Ramos
02:05 PM Bug #1699: [inline] entry methods should use perfect forwarding in C++11
Sam White wrote:
> This was merged last night and broke the Windows autobuilds:
Addressed in https://charm.cs.il...
Evan Ramos
10:50 AM Bug #1699 (Merged): [inline] entry methods should use perfect forwarding in C++11
This was merged last night and broke the Windows autobuilds: ... Sam White
12:32 PM Projections Bug #1886: Remove support for long-broken Topology Display tool
Are the .topo files used anywhere outside Projections? Matthias Diener
10:54 AM Projections Bug #1886: Remove support for long-broken Topology Display tool
Follow-up to remove .topo file generation still needs to be done, as noted in the gerrit patch. Sam White
10:54 AM Projections Bug #1886 (Implemented): Remove support for long-broken Topology Display tool
Sam White
10:55 AM Cleanup #1890 (Merged): Avoid defining VERSION in conv-autoconfig.h to avoid conflicts with appli...
Sam White
07:08 PM Cleanup #1890 (Implemented): Avoid defining VERSION in conv-autoconfig.h to avoid conflicts with ...
https://charm.cs.illinois.edu/gerrit/#/c/charm/+/4070/ Sam White
07:02 PM Cleanup #1890 (Merged): Avoid defining VERSION in conv-autoconfig.h to avoid conflicts with appli...
When compiling LAMMPS on AMPI I get this:... Sam White
09:20 AM Bug #1871 (Merged): get_put_pingpong segfaults on gni-crayxc-smp
Nitin Bhat

04/26/2018

05:22 PM Feature #1184: SMP-safe CmiReference and CmiFree
Should be possible to fix this with a use of std::atomic that is similar to the approach used in the pcqueue.h fix. ... Eric Bohm
04:43 PM Bug #1858: tests/ampi/privatization fails on mpi-linux-x86_64-smp autobuild
The failure occurs on the first @CthResume@ that shows up in any backtrace from @switchTLS@.
mpi-linux-x86_64-smp:...
Evan Ramos
03:46 PM Bug #1858: tests/ampi/privatization fails on mpi-linux-x86_64-smp autobuild
Interesting comment here:... Evan Ramos
12:07 PM Bug #1871 (Implemented): get_put_pingpong segfaults on gni-crayxc-smp
Fix: https://charm.cs.illinois.edu/gerrit/#/c/charm/+/4066/ Nitin Bhat
10:20 AM Projections Bug #1886: Remove support for long-broken Topology Display tool
Patch to remove it: https://charm.cs.illinois.edu/gerrit/#/c/projections/+/4065/ Matthias Diener
 

Also available in: Atom