Project

General

Profile

Activity

From 08/08/2017 to 09/06/2017

09/06/2017

03:51 PM Feature #1393: Redesign of Hybrid API (GPU Manager) to support concurrent kernel execution
Includes 2 schemes: CUDA event-based & CUDA callback-based, but the default is CUDA event-based due to better perform... Jaemin Choi
02:47 PM Cleanup #567: Charm++ barnes-charm example cleanup
Rationalize this with the one in its own repository. Either there should only be one, or there should be a clear wel... Eric Bohm
01:32 PM Feature #1670 (Implemented): Migration of Taskqueue from Ck to Converse layer
Seonmyeong Bak
12:56 PM Feature #1670 (Merged): Migration of Taskqueue from Ck to Converse layer
The task queue is currently implemented in ck-core even though it doesn't use any constructs on ck-layer.
Taskqueue...
Seonmyeong Bak
01:25 PM Bug #1671: Verbs memory pool may leak pinned memory when message is deleted on a PE different fro...
Note that this is specific to a mempool in src/arch/verbs/machine-ibverbs.c, not the generic mempool that we have for... Sam White
01:12 PM Bug #1671 (In Progress): Verbs memory pool may leak pinned memory when message is deleted on a PE...
Receive a message on a node group
Section multicast a pointer to that message to the pe's local to that host.
Handl...
Eric Bohm
11:56 PM Feature #1666: Standalone OpenMP implementation based on Converse User-level threads
Basically, it works with a few lines of codes in converse and OpenMP runtime. Seonmyeong Bak

09/01/2017

02:07 PM Bug #1162: tracing runs segfault while writing logs
Still an issue? Sam White
02:04 PM Bug #1409: verbs crashes on Stampede KNL and Bridges
This is probably not a real issue, since MPI and OFI will be in 6.8.1 and OFI should be the preferred comm layer on b... Sam White
02:01 PM Bug #1661: Building charm with papi support is successful even when papi is not found (-lpapi)
Assigning to Ronak b/c tracing... Sam White

08/31/2017

05:49 PM Cleanup #621 (Implemented): Convert Converse from C to C++
Evan Ramos
05:49 PM Cleanup #622 (Implemented): Convert MPI Machine layer from C to C++
Evan Ramos
05:49 PM Cleanup #623 (Implemented): Convert Verbs machine layer from C to C++
Evan Ramos
05:49 PM Cleanup #624 (Implemented): Convert netlrts machine layer from C to C++
Evan Ramos
05:49 PM Cleanup #625 (Implemented): Convert uGNI machine layer from C to C++
Evan Ramos
05:49 PM Cleanup #626 (Implemented): Convert PAMI-LRTS machine layer from C to C++
Evan Ramos
05:49 PM Cleanup #627 (Implemented): Convert common LRTS code from C to C++
Evan Ramos
05:24 PM Feature #1098: Better CMake support
Regarding the output location of decl and def, what if I added something like `-output-dir <dir>`? Using the existing... Evan Ramos
03:43 PM Bug #1663 (Implemented): Fix CMK_PAGESIZE definitions
https://charm.cs.illinois.edu/gerrit/#/c/2978/ Sam White
03:34 PM Bug #1669 (Merged): Add machine layer files as Makefile dependencies for building the machine lay...
There are no existing Makefile dependencies for the pamilrts layer. Similarly, the gni layer doesn't have a few depen... Nitin Bhat
03:25 PM Bug #1441 (New): Lies at startup about "The comm. thread both sends and receives messages"
A pamilrts-smp build, which doesn't use a comm thread, also displays this message "The comm. thread both sends and re... Nitin Bhat
02:21 PM Bug #1664 (Implemented): Port Sameer's PAMI changes for POWER8 to PAMILRTS
Someone with access to one of the PAMI linux systems needs to test this: https://charm.cs.illinois.edu/gerrit/#/c/2977/ Sam White
01:47 PM Bug #1634: HDF5 issues in AMPI
h3. HDF5 serial tests working (all 62):
* testhdf5, cache, cache_api, cache_image, cache_tagging, lheap, ohdr, st...
Matthias Diener

08/30/2017

06:41 PM Bug #1259: AMPI missing built-in MPI_INFO_ENV
Yes Sam White
06:21 PM Bug #1259: AMPI missing built-in MPI_INFO_ENV
Should I submit another patch defining the MPI_Info objects to simplify AMPI_Migrate calls? Karthik Senthil
06:05 PM Bug #1259 (Implemented): AMPI missing built-in MPI_INFO_ENV
https://charm.cs.illinois.edu/gerrit/#/c/2968/ Sam White
06:10 PM Feature #1098 (In Progress): Better CMake support
https://charm.cs.illinois.edu/gerrit/#/c/2958/ Sam White
06:05 PM Bug #1268: AMPIF issues due to C++ main routine
^ That should work. Sam White
06:00 PM Feature #1389 (Implemented): AMPI ATAReq test/wait performs poorly
https://charm.cs.illinois.edu/gerrit/#/c/2943/ Sam White
02:52 PM Bug #1544: CMK_TIMER_USE_PPC64 inaccurate with variable clock speeds
I was trying to test this before I push it to Gerrit, but I've been unable to build a pami-linux-ppc64le on the machi... Ronak Buch
02:49 PM Feature #1434 (Merged): optimize degenerate CkLoop cases
Sam White
02:47 PM Feature #1394: Node-level message aggregation for CkMulticast
Any update on this? Sam White

08/29/2017

04:49 PM Feature #1667: Direct API for nocopy operations on sender-side and receiver-side
And similarly, that one-of-GET-or-PUT design can get us up and running on the other layers quickly, though, with an e... Phil Miller
04:48 PM Feature #1667: Direct API for nocopy operations on sender-side and receiver-side
If the generic layer gets done, possibly even with a hacky implementation that only really provides one of GET or PUT... Phil Miller
04:32 PM Feature #1667: Direct API for nocopy operations on sender-side and receiver-side
We may want to avoid calling this "direct" since there is already something else called CmiDirect. I think in general... Sam White
03:45 PM Feature #1667 (In Progress): Direct API for nocopy operations on sender-side and receiver-side
Initial commit: https://charm.cs.illinois.edu/gerrit/#/c/2967/ Nitin Bhat
10:15 AM Feature #1667 (Merged): Direct API for nocopy operations on sender-side and receiver-side
Since the receiver side API (https://charm.cs.illinois.edu/redmine/issues/1236) has restrictions which demand the use... Nitin Bhat
01:56 PM Bug #1668: Ensure that all libraries/modules will build as dynamic/shared objects (.so/.dylib vs .a)
Ok, I see why we don't merge the two directories - when linking under control of @charmc@, it addresses user requests... Phil Miller
12:47 PM Bug #1668 (Implemented): Ensure that all libraries/modules will build as dynamic/shared objects (...
remote: https://charm.cs.illinois.edu/gerrit/2962 Bug #1668 irecv Makefile: enable charmc's automatic generation of... Phil Miller
12:29 PM Bug #1668: Ensure that all libraries/modules will build as dynamic/shared objects (.so/.dylib vs .a)
Merging the two directories would be a separate issue - this is just about whether both static and dynamic objects ge... Phil Miller
12:27 PM Bug #1668: Ensure that all libraries/modules will build as dynamic/shared objects (.so/.dylib vs .a)
Can we merge lib/ and lib_so/ ? I think this separation is not necessary, as shared and static libraries have differe... Matthias Diener
12:25 PM Bug #1668: Ensure that all libraries/modules will build as dynamic/shared objects (.so/.dylib vs .a)
ROMIO is currently not able to be built dynamically. Bug #952 tracks an update for ROMIO which should make it easier ... Matthias Diener
12:23 PM Bug #1668: Ensure that all libraries/modules will build as dynamic/shared objects (.so/.dylib vs .a)
A further amendment to that statement - libqt.a is missing because it's only copied as
> @cp libqt.a ../lib/libckqt...
Phil Miller
12:19 PM Bug #1668: Ensure that all libraries/modules will build as dynamic/shared objects (.so/.dylib vs .a)
So, it turns out that libqt is wrong in the opposite direction - libqt.a gets built, but doesn't get copied to the li... Phil Miller
12:04 PM Bug #1668 (Merged): Ensure that all libraries/modules will build as dynamic/shared objects (.so/....
> @./build LIBS netlrts-darwin-x86_64 --build-shared --suffix=shared -j2 -g@... Phil Miller
12:28 PM Feature #952: Update AMPI's version of ROMIO
Note that the patch mentioned above (https://charm.cs.illinois.edu/gerrit/#/c/2940/) does not enable compilation as a... Matthias Diener
12:27 PM Feature #952: Update AMPI's version of ROMIO
Update our ROMIO to v1.2.6, which can be built as a shared object / dynamic library: https://charm.cs.illinois.edu/ge... Sam White

08/28/2017

04:27 PM Bug #1165: avoid -lm with Intel compiler
With some changes from a Charmworks intern, Ani, coming down the pike, this gets much easier to accommodate - we'll h... Phil Miller
02:35 PM Bug #1658 (Merged): Premature detection of Quiescence when TRAM is being used
Phil Miller

08/25/2017

02:16 PM Bug #1651 (Merged): AMPI Persistent send/recv requests are broken
Sam White
01:34 PM Feature #1666 (New): Standalone OpenMP implementation based on Converse User-level threads
The current OpenMP implementation only works with Charm++ and AMPI.
However, this OpenMP implementation basically c...
Seonmyeong Bak
01:01 PM Feature #1609: User-level thread implementation based on Boost context library
Copied reply from gerrit
This ULT implementation stores register values in stack. So, if stack for each ULT can be m...
Seonmyeong Bak

08/23/2017

05:03 PM Bug #1665 (Merged): DDT needs to reference count and free its type objects
Currently AMPI_Type_free is a no-op b/c DDT doesn't implement reference counting of its type objects. Sam White
01:13 PM Bug #1544: CMK_TIMER_USE_PPC64 inaccurate with variable clock speeds
Can you link to the change that fixes this? Jim Phillips
12:15 PM Bug #1545: Serialize std::vector with Custom Allocator
I've thought about this more over the last few months and have a few things to share:
1) I'm not exactly sure what...
Nils Deppe
07:52 PM Bug #1664 (Merged): Port Sameer's PAMI changes for POWER8 to PAMILRTS
src/arch/pami/machine.c contains the following:... Sam White
07:47 PM Bug #1663 (Merged): Fix CMK_PAGESIZE definitions
For some reason we have the pagesize as 8192 instead of 4096 on all kinds of Linux-based machine layers:... Sam White

08/22/2017

11:07 AM Support #1662 (Closed): Show error message about adding "--enable-tracing" when charm is built wi...
This bug will be handled as a part of https://charm.cs.illinois.edu/redmine/issues/1661 Nitin Bhat
11:05 AM Support #1662 (Closed): Show error message about adding "--enable-tracing" when charm is built wi...
Building charm++ with papi support, but without the --enable-tracing option (./build charm++ netlrts-linux-x86_64 pap... Nitin Bhat
11:00 AM Bug #1661: Building charm with papi support is successful even when papi is not found (-lpapi)
Also add an abort when building papi without --enable-tracing, and update the PAPI Projections documentation. Sam White
09:34 AM Bug #1661 (Merged): Building charm with papi support is successful even when papi is not found (-...
When I try to build charm with papi support like ... Nitin Bhat
03:19 AM Bug #1544 (Implemented): CMK_TIMER_USE_PPC64 inaccurate with variable clock speeds
Ronak Buch
03:14 AM Projections Feature #997 (Implemented): Color by user supplied parameter (e.g. timestep) in Communication ove...
Ronak Buch
03:14 AM Projections Feature #996 (Implemented): Color by user supplied parameter (e.g. timestep) in Time Profile
Ronak Buch
03:14 AM Projections Feature #995 (Implemented): Color by user supplied parameter (e.g. timestep) in non-timeline tools
Ronak Buch

08/21/2017

04:01 PM Bug #1660 (Rejected): AMPI tests /tests/migration and tests/megaampi fail on Stampede2 with MPI a...
The tests - tests/migration and tests/megampi fail when run on two processors on Stampede2 with the build mpi-linux-x... Nitin Bhat
03:24 PM Bug #1639 (Merged): AMPI MPI_IN_PLACE support is broken
Sam White

08/18/2017

09:23 AM Cleanup #1659: Cleanup DDT memory leaks / valgrind output
All 3 of these commits seem to make no difference in Valgrind output, though they make DDT more readable:
Use std:...
Sam White

08/17/2017

02:26 PM Cleanup #1659: Cleanup DDT memory leaks / valgrind output
Here's the valgrind output from PE 1 after migrating a single AMPI rank from PE 0 to PE 1. Sam White
02:03 PM Cleanup #1659: Cleanup DDT memory leaks / valgrind output
I'm using DDT after all dynamic memory allocation inside the CkDDT_DataType classes has been replaced with std::vecto... Sam White
02:01 PM Cleanup #1659: Cleanup DDT memory leaks / valgrind output
When we migrate only 1 rank from PE 0 to PE 1, only the valgrind output from PE 1 shows the DDT leaks. PE 0 is clean. Sam White
01:59 PM Cleanup #1659 (Merged): Cleanup DDT memory leaks / valgrind output
DDT causes a lot of noise in Valgrind output when migrating AMPI ranks. I haven't been able to find the source of its... Sam White

08/16/2017

12:49 PM Feature #975: OFI Layer
Other than pingpong, I haven't done any other synthetic performance tests.
I have tested it on NAMD, ChaNGa and Ope...
Nitin Bhat
12:39 PM Feature #975: OFI Layer
Do you have any synthetic performance tests besides ping-pong?
If you can make this compatible with the 6.8.0 head I...
Jim Phillips
10:02 AM Feature #975 (Implemented): OFI Layer
The current patch shows decent performance improvements over the MPI build on both Stampede2 and Bridges.
This pa...
Nitin Bhat

08/14/2017

11:43 AM Bug #1507 (Merged): ckio test failure on gni-crayxc
Thank you, done. Sorry for the trouble, and thank you for the fixes. Phil Miller
11:37 AM Bug #1507: ckio test failure on gni-crayxc
With the recent fix, I can no longer reproduce this bug. I recommend this issue be closed. Thomas Quinn
10:04 AM Support #222: Port miniFE to Charm++
Well, I think there was an unfixed issue with reproducibility with this. When running on multicore builds it seems to... Justin Szaday
09:59 AM Support #222 (Closed): Port miniFE to Charm++
This is now done, after having sat idle for a long time.
charmgit:benchmarks/mantevo/miniFE-2.0
Phil Miller

08/12/2017

04:49 PM Bug #1658: Premature detection of Quiescence when TRAM is being used
Confirmed that the patch fixes the test, even when modified to run itself in multiple iterations in a single job. Phil Miller
04:48 PM Bug #1658: Premature detection of Quiescence when TRAM is being used
https://charm.cs.illinois.edu/gerrit/2907 Phil Miller
04:27 PM Bug #1658 (Implemented): Premature detection of Quiescence when TRAM is being used
Confirmed that the attached test code fails consistently on netlrts-darwin-x86_64, with 2 PEs running on a single hos... Phil Miller
02:56 PM Bug #1658: Premature detection of Quiescence when TRAM is being used
The patch that fixes it is attached. I suggest, after due scrutiny and testing, we merge this, so that users have a b... Laxmikant "Sanjay" Kale
02:50 PM Bug #1658: Premature detection of Quiescence when TRAM is being used
A simple test program I wrote (after fixing the bug) demonstrates the problem consistently. Its a variation on hello,... Laxmikant "Sanjay" Kale
02:45 PM Bug #1658: Premature detection of Quiescence when TRAM is being used
The bug is (I am reasonably sure) due to faulty quiescence detection algorithm in qd.C. It employees 2 phases. Once t... Laxmikant "Sanjay" Kale
02:25 PM Bug #1658 (Merged): Premature detection of Quiescence when TRAM is being used
This was identified in Charm++ version of Quicksilver by Karthik and Nikhil, in an LLNL internship project. There are... Laxmikant "Sanjay" Kale
07:09 PM Documentation #1482 (Merged): Update Charm++ FAQ
Sam White
07:09 PM Documentation #1602 (Merged): FAQ: 5 . 0 . 3 Can I use TotalView?
Sam White
07:09 PM Documentation #1605 (Merged): FAQ: 6 . 0 . 8 Which C++ language features cause porting problems?
Sam White

08/11/2017

12:12 PM Documentation #1605 (Implemented): FAQ: 6 . 0 . 8 Which C++ language features cause porting prob...
https://charm.cs.illinois.edu/gerrit/2900 Phil Miller
11:15 AM Documentation #1602 (Implemented): FAQ: 5 . 0 . 3 Can I use TotalView?
https://charm.cs.illinois.edu/gerrit/2895 Phil Miller

08/10/2017

12:44 PM Bug #1634: HDF5 issues in AMPI
I have a patch to update romio to 1.2.6 (shipped with last version of mpich1) that compiles successfully with the cur... Matthias Diener
12:09 PM Bug #1634: HDF5 issues in AMPI
I'd like to know to know if it is building on AMPI yet, or if it requires any MPI-2 or MPI-3 features we don't have i... Sam White

08/09/2017

11:33 AM Feature #1655: Enable use of shm transport for regular messages in LRTS
Also, './build charm++ gni-crayxe xpmem' fails to build because it tries to build pxshm and xpmem both. The issue is ... Sam White
09:23 AM Feature #1657 (Merged): CMA support for nocopy sends using the Entry Method API across processes ...
It should be straightforward to implement this at least for the transfer of the nocopy payload: the small metadata me... Sam White
09:17 AM Bug #1634: HDF5 issues in AMPI
What's the status of updating ROMIO to get shared library support? Sam White

08/08/2017

05:09 PM Bug #1507: ckio test failure on gni-crayxc
The fix I pushed is a direct result of digging in to the crash here. I'm almost certain that this is fixed, but I am... Thomas Quinn
04:45 PM Bug #1507 (Feedback): ckio test failure on gni-crayxc
Tom, with the other fix that you recently pushed, could you test that this still reproduces, and potentially open a n... Phil Miller
04:29 PM Bug #1647 (Closed): ckNew(): CkReductionMgr not constructed on all PEs
Redmine #1652 is believed to be the root cause of this issue, and it's fix has been merged into 6.8.0. Sam White
04:29 PM Bug #1652 (Merged): CkArray::ckDestroy() does not delete CkMulticastMgr
Sam White
03:40 PM Bug #1652 (Implemented): CkArray::ckDestroy() does not delete CkMulticastMgr
We'll merge this into 6.8.0 Sam White
03:45 PM Documentation #1656 (Merged): Update manual entries on Load Balancing strategies
The section describing the built-in strategies should be more descriptive of the trade-offs in strategies and should ... Sam White
02:46 PM Documentation #1082 (Merged): Improve SMP mode documentation
Matthias Diener
12:43 PM Feature #1655 (Merged): Enable use of shm transport for regular messages in LRTS
Experimenting with different models has shown that CMA (Cross Memory Attach) is a good candidate for exploiting shm f... Sam White
 

Also available in: Atom