Project

General

Profile

Activity

From 03/31/2013 to 04/29/2013

04/29/2013

10:41 AM Feature #60: Exploit phase boundaries in iterative apps
But we should reach that conclusion after we understand the shape and size of the task. I still think the discussions... Ramprasad Venkataraman
10:27 AM Feature #60: Exploit phase boundaries in iterative apps
I don't think there's any realistic chance of that happening. We're starting from no design, grand ambitions, and num... Phil Miller
10:23 AM Feature #60: Exploit phase boundaries in iterative apps
Ideally, we should put this in place within the next month. Please convene a meeting this week to discuss guiding exa... Ramprasad Venkataraman

04/28/2013

04:10 PM Feature #29: Reduction Starting messages
Now that SC is past, we should get this moving again. Jim encountering reductionStarting as a performance issue in NA... Phil Miller

04/27/2013

07:36 PM Revision 17fdbbb7 (charm): Add Mesh Streamer to the Charm++ libraries manual.
Lukasz Wesolowski

04/26/2013

02:11 PM Revision b60c531c (charm): leanmd2 example code update
current status: whole code translates to Charm correctly
Charm code compiles without any complaint
howeve...
Bilge Acun
02:10 PM Revision 246c714d (charm): small bug fix, enable dereferencing an argument in an entry method call
Bilge Acun
01:37 PM Revision 7739592c (charm): fixing the Ref.cj functional test example
Bilge Acun
01:33 PM Revision c25d09b3 (charm): Adding the missing pup methods for message types
some things might still be missing.. Bilge Acun
01:17 PM Revision 319c7fd5 (charm): remove extra pointer dereference
when a variable object_ref_type is passed to an entry message
it was dereferenced unnecessarily
Bilge Acun

04/25/2013

02:19 PM Bug #104: Bluegene Q xlc smp won't link binaries when compiled with -g: "reference to `__tls_get_...
FWIW, I get the undefined tls errors only if I build charm and the program with -g -O0.
If I build charm with --with...
Ramprasad Venkataraman
11:12 AM Cleanup #188 (Closed): Abolish CmiBool
Ramprasad Venkataraman
11:10 AM Cleanup #192 (Merged): Remove CmiBool typedefs
Ramprasad Venkataraman
11:09 AM Support #191 (Closed): announce deprecation of CmiBool
Ramprasad Venkataraman

04/24/2013

04:30 PM Revision b24a70db (charm): build #188: Completely clobber all scaffolding that existed to setup C...
Cmi{Bool,False,True} are simply defined to be {bool,false,true}.
This should stay in place until v6.6 so that we can ...
Ramprasad Venkataraman
03:54 PM Feature #190 (Closed): Build target(s) for Intel MIC
We should incorporate the work done to build on MIC with the main codebase. If it's easily backported, it should go o... Phil Miller
03:11 PM Revision 2e954803 (charm): doc #188: Replace Cmi{Bool,False,True} with {bool,false,true}
Ramprasad Venkataraman
03:08 PM Revision 88038de2 (charm): cmiobjid #188: Replace Cmi{Bool,False,True} with {bool,false,true}
Ramprasad Venkataraman
03:06 PM Revision 2b594caa (charm): xlat #188: Replace Cmi{Bool,False,True} with {bool,false,true}
Ramprasad Venkataraman
03:05 PM Revision 0f048224 (charm): libs #188: Replace Cmi{Bool,False,True} with {bool,false,true}
Ramprasad Venkataraman
02:55 PM Revision a7238df3 (charm): ampi #188: Replace Cmi{Bool,False,True} with {bool,false,true}
Ramprasad Venkataraman
02:51 PM Revision 4f6213da (charm): bigsim #188: Replace Cmi{Bool,False,True} with {bool,false,true}
Ramprasad Venkataraman
02:46 PM Cleanup #188: Abolish CmiBool
There's a bit more mop-up work to be done:
* AMPI, mblock, packlib, and the bigsim emulator need the same search/rep...
Phil Miller
02:06 PM Cleanup #188: Abolish CmiBool
In review, the changes to CkCallback, usesAtSync, and PUP are fine to go ahead. Since CmiBool is already universally ... Phil Miller
02:01 PM Cleanup #188: Abolish CmiBool
I just realize now, after your comment, that CmiBool is only defined for C++ code and not for C. That simplifies a lo... Phil Miller
01:53 PM Cleanup #188: Abolish CmiBool
Autobuild indicates that c++ compilers recognize bool on all our test platforms. Hence CmiBool is simply typdefed to ... Ramprasad Venkataraman
10:21 AM Cleanup #188: Abolish CmiBool
Started reviewing @ramv/nuke-cmibool@
* I'm hesitant to change CkCallback right away, since those constructors are...
Phil Miller
09:53 AM Cleanup #188 (Closed): Abolish CmiBool
Replace it with standard C or C++ @bool@ wherever safe/possible Phil Miller
02:39 PM Revision ccb7b407 (charm): ck-ldb #188: replace Cmi{Bool,False,True} with {bool,false,true}
Ramprasad Venkataraman
02:39 PM Revision 29af3eab (charm): comlib #188: more replace Cmi{Bool,False,True} with {bool,false,true}
Ramprasad Venkataraman
02:39 PM Revision ee2cc165 (charm): #188: fix pointless casts to bool
Ramprasad Venkataraman
02:39 PM Revision fe8b3578 (charm): comlib #188: replace Cmi{Bool,False,True} with {bool,false,true}
Ramprasad Venkataraman
02:39 PM Revision 36d3ce7b (charm): ck-perf #188: replace Cmi{Bool,False,True} with {bool,false,true}
Ramprasad Venkataraman
02:39 PM Revision defc541d (charm): util #188: replace Cmi{Bool,False,True} with {bool,false,true}
Ramprasad Venkataraman
02:39 PM Revision 586678c6 (charm): pup #188: replace Cmi{Bool,False,True} with {bool,false,true}
Ramprasad Venkataraman
02:39 PM Revision d0b31aeb (charm): ck-core #188: replace Cmi{Bool,False,True} with {bool,false,true}
Ramprasad Venkataraman
01:55 PM Revision 30a2551a (charm): support for Charj Array usage in sdag statements
1- Array.h needs to be included in the ci file
2- "CharjArray::" tag needs to be added to the Array's
in the sdag ...
Bilge Acun
10:20 AM Bug #189 (Closed): CkBitVector's comparison operators lie
The basic @operator==@ and @operator<@ simply @return false;@ rather than actually computing anything. If they're in ... Phil Miller

04/23/2013

04:31 PM Revision af5da75d (charm): CentralLB: don't leak stats count message (fixes #187)
(cherry picked from commit 41119526be0019304aa8ba1e4803cb6d1770e35a) Phil Miller
04:30 PM Revision 41119526 (charm): CentralLB: don't leak stats count message (fixes #187)
Phil Miller
04:29 PM Bug #187 (Closed): CentralLB::ReceiveCounts leaks reduction message
Quoth Tom Quinn:
> I've discovered what I think is a memory leak in the CentralLB:...
Phil Miller
04:17 PM Bug #182 (Closed): New path to zlib on ALCF BG/Q machines
Tested working and cherry-picked to the stable branch. Thanks! Phil Miller
04:16 PM Revision b6eb8c6e (charm): zlib on mpi-bgq
(cherry picked from commit 7d27753bfeea2e8a955138822deb27966bf33f49) Yanhua Sun
04:16 PM Revision d9f98d21 (charm): fix the ZLIB path problem on BGQ
(cherry picked from commit df206d1897961b913c312ced89da23aee7841c32) Yanhua Sun
02:49 PM Revision ca58fbc2 (charm): working version on BGP
Xiang Ni

04/22/2013

04:19 PM Bug #147 (Rejected): Crash in ChaNGa while writing disk checkpoint
Tom says that it hasn't happened in about 20 runs after patching up the migration constructors and PUP routines. I th... Phil Miller
03:42 PM Revision e34f1ca6 (charm): Typo fix and reformat comment
Phil Miller
03:33 PM Revision 7a7140bc (charm): envelope::pup(): Treat groupinit and group messages separately
Phil Miller
02:13 PM Revision 2f40993b (charm): MetaBalancer with Temp control integrated
Harshitha Menon
02:05 PM Revision 7f5fca28 (charm): Disambiguate sqrt() call for VC++
Phil Miller
02:04 PM Revision 73180492 (charm): Install new header in include/
Phil Miller
12:29 PM Revision 944bf73e (charm): win32: provide a junky replacement for rand48 for various examples
Phil Miller
12:29 PM Revision d4bb4bce (charm): msa: convert CMK_OPTIMIZE to CMK_ERROR_CHECKING (issue #135)
Phil Miller
12:29 PM Revision 467d4c0d (charm): ampi,tcharm: Issue #135 - replace usage of archaic CMK_OPTIMIZE
Phil Miller
12:29 PM Revision 869e50c2 (charm): build: replace CMK_OPTIMIZE with --with-production (issue #135) and -O...
Phil Miller
12:01 AM Revision b404f294 (charm): init
Xiang Ni

04/21/2013

07:51 PM Revision dac324bf (charm): Adding Jonathan's fix to race condition on restart.
Esteban Meneses

04/20/2013

12:17 AM Feature #185: provide backup implementations of std::unordered containers
For eg: one use case is in CkMulticastMgr::setup(). We've replaced the original O(N) array with a std::map (where N =... Ramprasad Venkataraman
10:06 PM Revision 5cc9366c (charm): more fucking hacks
Jonathan Lifflander
09:33 PM Revision 597648ee (charm): hack
Jonathan Lifflander
08:27 PM Revision 23d3b471 (charm): Merge branch 'charm' of charmgit:charm into charm
Jonathan Lifflander

04/19/2013

04:43 PM Revision 2ad2ac83 (charm): Fixing silly printout.
Esteban Meneses
04:05 PM Revision 571cfb76 (charm): Fixing a minor bug.
Esteban Meneses
03:02 PM Revision fde6a855 (charm): Adding load-balancing independent support for checkpoint.
Esteban Meneses
01:26 PM Bug #104: Bluegene Q xlc smp won't link binaries when compiled with -g: "reference to `__tls_get_...
Any updates on this? Or any info of a workaround?
Does this mean we cannot compile smp builds with -g on bgq? That e...
Ramprasad Venkataraman
01:02 PM Revision 1670a852 (charm): a variety of bug fixes, new leanmd code
CkArgMsg changed to Ref<CkArgMsg> in main constructor Bilge Acun
10:58 AM Bug #182: New path to zlib on ALCF BG/Q machines
Same changes are also applied to mpi-bgq too now. Yanhua Sun
11:13 PM Revision f302e7cc (charm): ckmulticast: nuke internal class IndexPos
Ramprasad Venkataraman
11:13 PM Revision 796696fc (charm): ckmulticast: Fix O(numPE) data structure and loop in tree setup
Ramprasad Venkataraman
11:13 PM Revision 986c3571 (charm): ckmulticast: Some more streamlining of the ::setup() method
Avoid unnecessary struct construction / copies etc Ramprasad Venkataraman
10:02 PM Revision f4104ae4 (charm): Fixing minor bug.
Esteban Meneses
10:02 PM Revision 6153d6e3 (charm): Cleaning up checkpoint code in message-logging.
Esteban Meneses
09:51 PM Revision 0417fac2 (charm): Fixing minor bug.
Esteban Meneses
09:40 PM Revision f29a46df (charm): Cleaning up checkpoint code in message-logging.
Esteban Meneses

04/18/2013

03:34 PM Revision 7d27753b (charm): zlib on mpi-bgq
Yanhua Sun
03:29 PM Bug #182: New path to zlib on ALCF BG/Q machines
Does the same change need to be made for the mpi-bluegeneq builds? Phil Miller
03:29 PM Bug #182 (Resolved): New path to zlib on ALCF BG/Q machines
Yanhua Sun
03:28 PM Bug #182: New path to zlib on ALCF BG/Q machines
I added the correct path to cc-xlc.sh, also overwrite the _INC and _LIB variable to use new path. It is checked in.... Yanhua Sun
03:07 PM Revision 6a740898 (charm): Merge branch 'charm' of charmgit:charm into charm
Yanhua Sun
03:07 PM Revision df206d18 (charm): fix the ZLIB path problem on BGQ
Yanhua Sun
12:39 PM Revision 8b6cd501 (charm): arridx: fix operator< to do something sane
Ramprasad Venkataraman
12:39 PM Revision 0c027008 (charm): arridx: add const qualifier to print()
Ramprasad Venkataraman
10:24 AM Bug #186: ckout and ckerr should derive from std::ostream
When I first toyed with this idea in 2010, Gengbin sent me an email wondering why I couldn't just use std::cout and s... Ramprasad Venkataraman
10:20 AM Bug #186 (In Progress): ckout and ckerr should derive from std::ostream
A basic implementation using periodically flushed std::obufstreams is on ramv/ckostream.
There are some issues that ...
Ramprasad Venkataraman
10:08 AM Bug #186 (In Progress): ckout and ckerr should derive from std::ostream
One of the fundamental benefits of stream-based output is that any class can simply provide a stream inserter that se... Ramprasad Venkataraman
10:15 AM Revision 53119dfe (charm): Replacing endl with std::endl (#186)
Lukasz Wesolowski
10:15 AM Revision cafd423c (charm): ckout, ckerr and cklog are now instances of ck::obufstream (#186)
Originally of type CkStream. Now they inherit from std::ostream. Ramprasad Venkataraman
10:14 AM Revision f4c6e726 (charm): define macro to switch between any std::ostream in hello world code (#...
Ramprasad Venkataraman
10:13 AM Revision 6bd0eafc (charm): charm++ hello world example using std::ostreams, ie ck::out (#186)
Ramprasad Venkataraman
10:13 AM Revision f4229c39 (charm): ck::{out,err,log} provide std::ostream objects to charm++ programs (#186)
The existing ckout and ckerr objects are instances of home-grown classes that
bear no relation to std::ostream. This ...
Ramprasad Venkataraman
10:12 AM Revision 31d8b2fb (charm): ck::obufstream - a buffered ostream in charm that inherits from std::o...
Ramprasad Venkataraman
02:23 AM Revision 282cbb79 (charm): mlogft: hack the central lb not to send large O(P) message
Jonathan Lifflander
02:19 AM Revision 7d05c6a4 (charm): mlogft: hack the central lb not to send large O(P) message
Jonathan Lifflander

04/17/2013

04:06 PM Bug #158: Document and check restrictions on passing non-array callbacks to CkStartCheckpoint
I am not completely knowledgeable about why CkChareID carries a ptr around. However, our current perception of objid... Ramprasad Venkataraman
03:52 PM Feature #185 (Rejected): provide backup implementations of std::unordered containers
There are several locations in the charm runtime where hashtables are used, or are sorely needed. We do have a homegr... Ramprasad Venkataraman
02:49 PM Revision b2143760 (charm): [ckcache] initialize 'isReady' to false
Pritish Jetley
02:26 PM Revision fc378eca (charm): [coordination library] turn off verbosity
Pritish Jetley
02:08 PM Revision 0fb4d70d (charm): [ckcache] inheriting templates need this->foo() to be able to find foo...
Pritish Jetley
11:32 AM Cleanup #165 (In Progress): object location services: separate location caching from msg delivery...
Work in progress. The objective can be restated as the complete annihilation of CkLocRec and all its child classes. T... Ramprasad Venkataraman
11:03 AM Revision 7479be2f (charm): envelope: update comments to reflect split between array and arrayinit...
Ramprasad Venkataraman
11:03 AM Revision eec51c6b (charm): envelope: Fix assert to check against sentinel value from enum
Ramprasad Venkataraman
11:03 AM Revision fb58d788 (charm): envelope: Replace getsetArrayMgr() with separate get and set methods
Ramprasad Venkataraman
10:59 AM Bug #182: New path to zlib on ALCF BG/Q machines
An additional cc-gcc is not needed. The entries from conv-mach.sh are used in default case. If they need to be overwr... Nikhil Jain
12:14 AM Bug #182: New path to zlib on ALCF BG/Q machines
I modified conv-mach.sh to use current zlib path, this works. Projections generates .gz files.
However the point i...
Yanhua Sun
10:55 AM Revision 3e634302 (charm): [coordination library] use custom reducer instead of concat, because C...
Pritish Jetley
10:16 PM Revision 4ddfc055 (charm): [coordination] call base pup for CkSmpCoordinator
Pritish Jetley

04/16/2013

05:28 PM Revision 26dc96fa (charm): Provide a crude comparison operator< to permit CkArrayIndex in compari...
Ramprasad Venkataraman
11:43 AM Bug #182: New path to zlib on ALCF BG/Q machines
Yanhua, please test that the new libraries and updated code properly generate compressed projections logs without cra... Phil Miller

04/15/2013

09:53 AM Revision d85d0ccd (charm): ckarrayid: Define a comparison operator (<)
Ramprasad Venkataraman
11:20 PM Bug #183 (Closed): Additions to commbench cause segfaults on many platforms
Phil Miller

04/14/2013

03:05 PM Bug #184 (Closed): PAMI LRTS performance using ppn 64 is worse than that with PAMI
When I compared NAMD Apoa1 performance on 64 nodes, i found that PAMI LRTS is about 30% slower than PAMI using ppn 6... Yanhua Sun
02:58 PM Bug #81: Manytomany on PAMI SMP hangs without Async
I have not heard anything back from Sameer Yanhua Sun
08:15 PM Bug #81: Manytomany on PAMI SMP hangs without Async
Any word from Sameer about progress on this? Phil Miller
10:49 AM Revision e6c8bee5 (charm): megatest: increase numElements with numPEs in migration test
Ramprasad Venkataraman
10:49 AM Revision 74130086 (charm): msgqtest: make clean should remove ci.stamp files
Ramprasad Venkataraman
08:20 PM Bug #182: New path to zlib on ALCF BG/Q machines
I am not sure if the cause of crash on BGQ was the same as #71. Irrespective of that, the current paths on BGQ don't ... Nikhil Jain
08:16 PM Bug #182: New path to zlib on ALCF BG/Q machines
Are you referring to #71? I fixed that a month ago. Phil Miller
08:14 PM Bug #182: New path to zlib on ALCF BG/Q machines
This may be a noop since we do not use ZLIB implicitly on BGQ in our paths due to correctness issues. Nikhil Jain

04/13/2013

06:56 PM Bug #118: Segfault in examples/fem/simple2D under newer ICC
I just tried to reproduce this on Stampede with various versions of icc. None of them showed the crash, or related va... Phil Miller
03:11 PM Feature #170: objid_t: Basic definitions and use in envelope
I just realized a quirk in how we've been thinking of expressing and tracking indices for bound arrays (those sharing... Phil Miller
11:43 AM Bug #183 (Resolved): Additions to commbench cause segfaults on many platforms
Will close when autobuild passes these tomorrow. Phil Miller
11:41 AM Bug #183: Additions to commbench cause segfaults on many platforms
It is caused by cpv variable not initialized. I checked in a fix and tested on net smp. It worked. Yanhua Sun
11:37 AM Revision 318d7a9b (charm): fix the bug in flood test for smp case
Yanhua Sun
09:59 PM Revision 0bbab37e (charm): detect failure after the first checkpoint,put vector into array first ...
Xiang Ni

04/12/2013

05:34 PM Revision d206ca82 (charm): minor fix for trace projection summary
Yanhua Sun
04:42 PM Bug #148 (Closed): missing mlogfttest target
Phil Miller
04:13 PM Cleanup #74 (In Progress): Replace any remaining initcall declarations with initproc or initnode ...
Please update comments, error messages, documentation, and so forth, to reflect the change. Some bits from @git grep ... Phil Miller
04:02 PM Revision 2655e2a9 (charm): examples/charm++/leanmd: Fix mismatched new[]/delete
(cherry picked from commit 7c9cf21f2f70746c9732e7b9f87d4481b4f25db6) Phil Miller
03:17 PM Cleanup #135: Eliminate tests against archaic macro CMK_OPTIMIZE
... Phil Miller
02:18 PM Cleanup #135: Eliminate tests against archaic macro CMK_OPTIMIZE
Everyone should address instances of CMK_OPTIMIZE in the portions of the codebase they are responsible for, according... Phil Miller
02:10 PM Bug #161: SMP: Adapt to thread oversubscription, and provide flags to override
What produces a warning about @+CmiNoProcForComThread@ now should automatically configure the runtime appropriately, ... Phil Miller
12:31 PM Feature #170 (In Progress): objid_t: Basic definitions and use in envelope
Ramprasad Venkataraman
12:19 PM Feature #170: objid_t: Basic definitions and use in envelope
envelope setArrayMgr() only caters to the ForArrayEltMsg case. It is mostly used in scenarios that do not currently a... Ramprasad Venkataraman
11:42 AM Feature #170: objid_t: Basic definitions and use in envelope
Audit and fix all locations where ForIDedObjMsg needs to have a presence. git grep ForArrayEltMsg should bring these ... Ramprasad Venkataraman
11:30 AM Feature #170: objid_t: Basic definitions and use in envelope
Synchronize enums in charm.h with those defined in charmdebug. (MsgInfo.java)
Ramprasad Venkataraman
12:29 PM Bug #183 (Closed): Additions to commbench cause segfaults on many platforms
Mostly SMP builds, but some single-threaded builds as well. cf. autobuild. Phil Miller
10:21 AM Support #96: 6.5.0 on Titan
After a lot of email to help desk, they confirmed that on Titan they do not reset home directory as before (on Jaguar... Yanhua Sun
08:38 AM Bug #182: New path to zlib on ALCF BG/Q machines
Bump urgency and due date, because the old path will be broken on Monday. Phil Miller
08:36 AM Bug #182 (Closed): New path to zlib on ALCF BG/Q machines
From ALCF Support:
> ***Supported Version of ZLIB Available***
> ZLIB, a general purpose compression library, is pa...
Phil Miller

04/11/2013

06:08 PM Revision d2bdc973 (charm): Revert "loc-cache: rename at() to find() because it doesn't do any bou...
This reverts commit c77b93a834948e32e1710de1be5fec35251389d3.
The method in question is in some middle-ground betwee...
Ramprasad Venkataraman
06:08 PM Revision 9626d855 (charm): loc-cache: Rename test files etc
Ramprasad Venkataraman
06:08 PM Revision b85b9929 (charm): loc-cache: Add queue_demotable and cache_lru unit tests
Ramprasad Venkataraman
06:08 PM Revision 7a0321d3 (charm): build: Include lru cache and demotable queue headers in build
Ramprasad Venkataraman
06:08 PM Revision e6428d22 (charm): cache_lru: Add find() methods to lookup keys without exceptions
The method signatures do not conform to the std::map interface.
Instead of returning an iterator, they return a refer...
Ramprasad Venkataraman
06:08 PM Revision 988d115f (charm): cache_lru: Provide a reserve() method
Ramprasad Venkataraman
06:08 PM Revision 21cb7e91 (charm): loc-cache: remove mention of cache size from API.
Perhaps size-limited caches should have their limits specified elsewhere.
This culls the API down to essentials. We c...
Ramprasad Venkataraman
06:08 PM Revision 98b7feee (charm): cache_lru: Fix bug in erase()
Now passes unit tests Ramprasad Venkataraman
06:08 PM Revision 25516363 (charm): loc-cache: rename at() to find() because it doesn't do any bounds chec...
Ramprasad Venkataraman
06:08 PM Revision 5423a7f7 (charm): loc-cache: Modify to use cache_lru instead of unordered_map
However, code in the loc-cache that uses cache_t so far works well
with both cache_lru and with std::unordered_map. B...
Ramprasad Venkataraman
06:08 PM Revision 43799e9d (charm): cache_lru: A size-constrained dictionary with LRU retirement policies
API still under flux, but passes unit tests. Partial progress towards #113 Ramprasad Venkataraman
06:08 PM Revision a5992424 (charm): loc-cache: Do not provide non-const iterators to local elem indices
Ramprasad Venkataraman
06:08 PM Revision 47334121 (charm): loc-cache: Unify update() into one method. Incoherent thoughts on upda...
My opinions on how the onPE list of local elements and the (potentially) shared
cache of remote element locations sho...
Ramprasad Venkataraman
06:08 PM Revision 4cd4de00 (charm): loc-cache: more updates to cache update()
Ramprasad Venkataraman
06:08 PM Revision 17f9d469 (charm): loc-cache: erase() recognizes that element can have entry in both loca...
Ramprasad Venkataraman
06:08 PM Revision b9a40d13 (charm): loc-cache: will now also maintains a loc registry for the home domain ...
The home domain will be the same as the domain over which remote location caches
are shared This can be a PE, the who...
Ramprasad Venkataraman
06:08 PM Revision ad331d8e (charm): loc-cache: rename factory class
Ramprasad Venkataraman
06:08 PM Revision b7b05323 (charm): demotable_queue: Implement class queue_demotable_fifo
A queue that supports FIFO with efficient demotion of random elements to the
rear of the queue. Rudimentary function...
Ramprasad Venkataraman
06:08 PM Revision d1c22c1d (charm): demotable_queue: rename template arg in class queue_demotable_fifo
Ramprasad Venkataraman
06:08 PM Revision 9c54aaf0 (charm): demotable_queue: add promote(), erase(), reserve() and clear()
Ramprasad Venkataraman
06:08 PM Revision 32680c67 (charm): loc-cache: add headers to makefile list
Ramprasad Venkataraman
06:08 PM Revision bc5fb02d (charm): loc-cache: Eschew local/remote adjectives in favor of onPe, offPe etc
Also remove public methods:
numRemoteElems() (no reason to query how large the cache is)
numLocalElems() (pointless ...
Ramprasad Venkataraman
06:08 PM Revision 83acc682 (charm): std::hash specializations for group and array IDs
Ramprasad Venkataraman
06:08 PM Revision c65c019d (charm): loc-cache: Rudimentary unit test
Ramprasad Venkataraman
06:08 PM Revision 5fec57cf (charm): temp: operator< for CkArrayIndex to permit map/set in cache
Ramprasad Venkataraman
06:08 PM Revision 9f09c6ff (charm): loc-cache: Remote obj cache container is now supplied at construction
This should permit a shared remote obj cache that is instantiated by the client code Ramprasad Venkataraman
06:08 PM Revision 52d031f4 (charm): loc-cache: A singleton cache factory manages the obj location caches o...
- factory instantiates one or more cache instances depending on some policy.
For eg, one cache instance for each nu...
Ramprasad Venkataraman
06:08 PM Revision e5032448 (charm): loc-cache: when updating location, check both local and remote lists
Ramprasad Venkataraman
06:08 PM Revision c6a0bf12 (charm): arrayindex: a nutjob std::hash specialization for array indices
Ramprasad Venkataraman
06:04 PM Revision d7434bd6 (charm): loc-cache: A prototype container that caches object locations. #165
CkLocMgr does not just act as a location cache. Instead it acts
as a message delivery infrastructure. It also buffers...
Ramprasad Venkataraman
04:58 PM Revision de721779 (charm): send multiple msgs for flooding
Yanhua Sun
04:16 PM Revision 4b88abe4 (charm): add flood test to get maximum bandwidth
Yanhua Sun
09:03 AM Revision 74c9676f (charm): fix charj modifier lists
entry reductiontarget etc. was emitted wrong in message declarations Bilge Acun
10:37 PM Revision 70891087 (charm): message declaration bug fix
removed unnecessary modifiers in variable declaration inside of messages
and seperate them with new lines
Bilge Acun
09:34 PM Revision 87eaeed4 (charm): automatic analysis before exit
Yanhua Sun

04/10/2013

01:58 PM Bug #181 (Closed): performance problem in net-lrts smp build
charm net-lrts branch smp build is too slow, which can be observed in /tests/charm++/pingpong. The time difference ca... Bilge Acun
01:46 PM Support #94 (Closed): 6.5.0 on ANL BGQ systems (Vesta, Cetus/Mira)
Confirmed access to these locations. Nikhil Jain
11:52 AM Feature #180 (In Progress): object location services: improve implementation and protocol
We have had an ongoing discussion about improving the design, implementation and protocol for our distributed object ... Ramprasad Venkataraman
11:13 PM Revision 8326e669 (charm): fix error in previous commit
Xiang Ni
11:01 PM Bug #123 (Closed): Cray compiler won't link converse (non-Charm++) binaries due to missing typeinfo
And it tests out clean on craycc. All of tests and examples build. Phil Miller
09:48 PM Bug #123 (Resolved): Cray compiler won't link converse (non-Charm++) binaries due to missing type...
More in-depth surgery on header files to not expose @charm++.h@ and all the junk it defines all over the place done i... Phil Miller
10:55 PM Revision 9fe6df83 (charm): small change in the previous commit
Xiang Ni
10:43 PM Revision 15af6b4d (charm): refine failure generator
Xiang Ni
09:45 PM Revision 4a9c5ff4 (charm): pose/Makefile: improve safety of make -j
Phil Miller
09:44 PM Revision 7e297b0a (charm): Refs #123: Refactor headers to reduce exposure of types in charm++.h
Phil Miller

04/09/2013

02:52 PM Bug #123 (Closed): Cray compiler won't link converse (non-Charm++) binaries due to missing typeinfo
A little more on compatibility with commit:244f4506f896e0c020011768f8d9ff, and now all of @tests@ and most @examples@... Phil Miller
02:42 PM Feature #179 (In Progress): objid_t: adapt callbacks to use objid_t
Evaluate the implications of objid_t (basic implementations in #170) on callbacks and improve callback design / imple... Ramprasad Venkataraman
02:29 PM Feature #178 (New): objid_t: fault tolerance infrastructure should use objid_t
Once the basic objid_t scaffolding is available (#170), the impacts on the FT infrastructure and protocols should be ... Ramprasad Venkataraman
02:26 PM Feature #177 (In Progress): objid_t: load balancing infrastructure should use objid_t
Once the basic objid_t scaffolding is available (#170), the impacts on the LB infrastructure and strategies should be... Ramprasad Venkataraman
02:21 PM Feature #176 (In Progress): objid_t: tracing infrastructure should use objid_t
Once the basic objid_t scaffolding is available (#170), the impacts on the tracing infrastructure and projections log... Ramprasad Venkataraman
02:16 PM Feature #175 (Merged): objid_t: location cache should be keyed by objid_t
A LocMgr keyed by objid_t seems amenable to most situations when an inversion function is known / can be generated. W... Ramprasad Venkataraman
02:13 PM Feature #174 (Merged): objid_t: Distributed lookup protocol
This first requires design discussions to delineate the subtasks Ramprasad Venkataraman
02:12 PM Feature #173 (New): objid_t: Support user-defined conversion functions
-- at runtime
-- at compile-time (with accompanying generated code optimizations)
-- design discussion for how to d...
Ramprasad Venkataraman
02:11 PM Feature #172 (New): objid_t: Exploit compile-time declared size bounds
-- charmxi declaration syntax
-- code generation for conversion functions
-- change generated code to directly call...
Ramprasad Venkataraman
02:08 PM Feature #171 (Merged): objid_t: Specify array size at construction time via CkArrayOptions despit...
Map onto bulk constructor metadata described in #170 Ramprasad Venkataraman
02:05 PM Feature #170 (Merged): objid_t: Basic definitions and use in envelope
- objid_t definition (intial header file, typedefs, bit masks, basic functions etc)
- add a new extradata struct tha...
Ramprasad Venkataraman
01:17 PM Revision 244f4506 (charm): Don't export a genericly-named symbol that appears in other libraries ...
Phil Miller
10:02 AM Revision 353e15d0 (charm): split logpool from projections so that it can be used by controlpoints...
Yanhua Sun
09:27 AM Revision 833dfdca (charm): make every enrty method reductiontarget
Bilge Acun

04/08/2013

05:29 PM Support #97 (In Progress): 6.5.0 on Hopper
Please add builds that support @hugepages@ and @persistent@ in addition to what's there now. Those will require that ... Phil Miller
04:54 PM Support #97: 6.5.0 on Hopper
What's the path for me to check access to and eventually announce? Phil Miller
02:25 PM Support #97 (Resolved): 6.5.0 on Hopper
Michael Robson
11:09 AM Support #97: 6.5.0 on Hopper
Just the cce/debug targets, and then let's be done with this. Phil Miller
01:24 AM Support #97: 6.5.0 on Hopper
I re-built charm/LIBS on Hopper with -g/debug and it worked. It built megatest no problem. I even made a fresh build ... Michael Robson
07:17 PM Support #97: 6.5.0 on Hopper
It appears that everything but the Cray debug build (which is going to be resolved under #167), the Cray SMP builds (... Michael Robson
04:56 PM Support #98: 6.5.0 on NCSA systems (JYC, Blue Waters)
At what path can the builds be found? Phil Miller
04:53 PM Support #100 (Closed): 6.5.0 on Trestles at SDSC
Phil Miller
04:50 PM Support #100: 6.5.0 on Trestles at SDSC
They are accessible. Esteban Meneses
04:52 PM Support #99 (Resolved): 6.5.0 on Kraken
Please get someone else who has current access to Kraken to confirm that they can get to your builds and close this i... Phil Miller
04:46 PM Support #99: 6.5.0 on Kraken
All builds are successful now.
2) (OK) ./build AMPI mpi-crayxe -j8 --disable-error-checking -optimize
5) (OK)...
Esteban Meneses
04:42 PM Revision c2862238 (charm): [ckcache] added pup() methods for checkpointing.
Pritish Jetley
04:41 PM Revision c96e1c43 (charm): [smp coordination library] added pup method for checkpointing.
Pritish Jetley
04:09 PM Bug #123 (Resolved): Cray compiler won't link converse (non-Charm++) binaries due to missing type...
Fixed by not exposing Charm++-level classes in lower level headers. See commit:405ed93563561ed38536b0dcf6bfb687f4ff2d50. Phil Miller
03:56 PM Revision 405ed935 (charm): Make several headers more standalone and less order dependent
Phil Miller
03:18 PM Bug #141 (Closed): Update manual to reflect current Completion Detection API
Phil Miller
03:06 PM Cleanup #127 (Closed): Quantify code size/divergence impact of message logging FT
Phil Miller
03:02 PM Revision 33161067 (charm): Split out options appended to the version argument on the command line...
treat "net-linux-smp-icc" as "net-linux smp icc". Jim Phillips
02:24 PM Support #96: 6.5.0 on Titan
From Yanhua:
I sent an email to ask. It seems no current directory satisfies our purpose, it is either erased or o...
Michael Robson
12:00 PM Support #96: 6.5.0 on Titan
Yanhua: Any progress on identifying a stable location on Titan at which builds of Charm++ can be offered to users? Ha... Phil Miller
02:00 PM Revision eb5e5524 (charm): [ckcache] fixed conflicts with charm branch
Merge branch 'charm' into jetley/smpcache
Conflicts:
src/libs/ck-libs/cache/CkCache.C
src/libs/ck-libs/cache/CkCac...
Pritish Jetley
12:06 PM Feature #169 (Closed): ensure charm builds with C11/C++11 flags
Ensuring forward compatiblity with the latest available compiler support for the new standards might be useful as use... Ramprasad Venkataraman
12:00 PM Bug #168 (Rejected): make -use-new-std portable
When the user passes -use-new-std to charmc, it is required to detect the type of the compilation unit (C or C++ sour... Ramprasad Venkataraman
11:09 AM Revision 1025f27a (charm): added some extra functionality needed by changa
Pritish Jetley
11:08 AM Bug #167: Compiling Charm with debugging symbols crashes Cray Compiler 8.1.4 (but not 8.1.3 or 8....
According to Michael, this is *only* an issue on 8.1.4 Phil Miller

04/07/2013

04:39 PM Revision 76189f4e (charm): added support for multiple extern declarations in one line
previously every declaration has to be declared in a new line Bilge Acun
03:17 PM Revision a9ff7189 (charm): Merge branch 'charj-ref' of charmgit:charm into charj-ref
Bilge Acun
02:32 PM Revision 400e6c9a (charm): artificial notify failure
Xiang Ni

04/06/2013

01:05 PM Revision ee50bcde (charm): examples/charm++: Add a bunch of missing TESTOPTS
Phil Miller
12:48 PM Revision 6466e9c1 (charm): examples/charm++/leanmd: shorten test run by about 10x
Phil Miller
12:25 PM Revision e7c1d41a (charm): Rename files from .cc to .C to appease MS VC++
Phil Miller

04/05/2013

06:27 PM Bug #143: Parallel (make -j) build of AMPI libraries and client code is racy
This one is from ampicc/ampiCC/ampicxx symlinking our mpi.h and then deleting it when each compilation is done. In a ... Phil Miller
06:24 PM Bug #167 (Upstream): Compiling Charm with debugging symbols crashes Cray Compiler 8.1.4 (but not ...
This is reproducible on Blue Waters as well, with the current default Cray compiler module version, 8.1.4. It does no... Phil Miller
06:04 PM Bug #167 (Closed): Compiling Charm with debugging symbols crashes Cray Compiler 8.1.4 (but not 8....
I attempted to compile Charm on NERSC's Hopper using CC 4.1.40 and received the following output:... Michael Robson
04:03 PM Feature #166 (Closed): Graph performance components from autobuild
Parse output of nightly build and graph performance of microbenchmarks (ping pong) over time on web page. Eric Bohm
03:48 PM Support #160 (Closed): SDAG documentation of message API and CkSetRefNum
Eric Bohm
03:37 PM Bug #133 (In Progress): examples/ampi/CJacobi3D does not compile and hangs with icc on Taub
I'll check this against ICC 12, and see if we need to release-note a lack of support for ICC 11 Phil Miller
03:35 PM Bug #132: Pingpong hangs in smp version on Taub
Work on rationalizing the +p ++ppn argument lists and validating them. Eric Bohm
02:56 PM Bug #132: Pingpong hangs in smp version on Taub

Did you try with space between +ppn 1 instead of +ppn1 (space after +ppn).
Abhishek Gupta
02:51 PM Bug #78 (Upstream): AMPI failure with migration under Cray compiler due to tcmalloc bugs or incom...
What's the resolution? Will ampicc make sure that @-hsystem_alloc@ gets passed until this is fixed? Or are we just do... Phil Miller
02:41 PM Bug #78 (Resolved): AMPI failure with migration under Cray compiler due to tcmalloc bugs or incom...
Nikhil Jain
02:50 PM Bug #122: examples/ampi/Cjacobi3D leads to invalid memory access in isomalloc's pup
I assert valgrind is mistaken - it is reporting invalid access for a few bytes in middle of an allocated block. The b... Nikhil Jain
02:41 PM Bug #122: examples/ampi/Cjacobi3D leads to invalid memory access in isomalloc's pup
As in, valgrind doesn't report the same error for you? Or you assert that valgrind is mistaken? Phil Miller
02:39 PM Bug #122 (Resolved): examples/ampi/Cjacobi3D leads to invalid memory access in isomalloc's pup
Having done some tests, I found nothing suspicious - may be valgrind's weirdness. Nikhil Jain
02:44 PM Support #126 (In Progress): Document process launching arguments with aim of cross-machine ration...

https://wiki.engr.illinois.edu/display/ppl/Process+Launch+Parameters
Abhishek Gupta
02:27 PM Revision f514ca0f (charm): envelope.h: remove two asserts that are guaranteed to be true
Lukasz Wesolowski
02:22 PM Cleanup #165 (In Progress): object location services: separate location caching from msg delivery...
CkLocMgr acts as a cache that stores location data and also buffers undelivered messages, implements message delivery... Ramprasad Venkataraman
02:12 PM Feature #108: Fixed-size object IDs
Phil and I had an initial discussion of high level tasks. This will result in a list of subtasks that will be added t... Ramprasad Venkataraman
01:14 PM Support #98: 6.5.0 on NCSA systems (JYC, Blue Waters)
Hacked the script to express suitable options on gemini.
Note: there is still some kind race condition which can m...
Eric Bohm
01:06 PM Bug #163: Sending SIGINT to charmrun under SMP build run with ++local causes double-free
Hrm, the same symptom is also reproducible as follows... Phil Miller
12:45 PM Bug #163: Sending SIGINT to charmrun under SMP build run with ++local causes double-free
This seems to only occur when passing @++local@ to @charmrun@, and not without. That suggests that perhaps the issue ... Phil Miller
12:43 PM Bug #163: Sending SIGINT to charmrun under SMP build run with ++local causes double-free
For reference, this can be reproduced on @net-linux-x86_64-smp@ by running ... Phil Miller
12:31 PM Bug #163: Sending SIGINT to charmrun under SMP build run with ++local causes double-free
So, it looks like @CmiDestoryLocks()@ (sic) gets called twice on the signal-initiated-exit path in @src/arch/net/mach... Phil Miller
11:52 AM Bug #163 (Merged): Sending SIGINT to charmrun under SMP build run with ++local causes double-free
Launch a Charm++ program built with SMP using @charmrun ++local@ and stop its execution with Ctrl-C (SIGINT), and gli... Phil Miller
12:50 PM Revision 92cbe4bd (charm): resolve some memory leak issue, inject failure to certain pe for bette...
Xiang Ni
12:37 PM Cleanup #164: Repeated typo: Destory -> Destroy
Fixed the instances in persistent in commit:eed4862. Leaving @net/machine.c@ and @arch/util/machine-smp.c@ until #163... Phil Miller
12:32 PM Cleanup #164 (Merged): Repeated typo: Destory -> Destroy
The word 'destroy' is misspelled as 'destory' in several places in the codebase. I'm fixing some of them, but leaving... Phil Miller
12:36 PM Revision eed48623 (charm): Partial typo fix of 'destory'; references #164
Phil Miller
12:04 PM Bug #146 (Closed): xcastredn test won't build on Windows due to missing 'usleep' and ambiguous 'f...
Ramprasad Venkataraman
02:25 AM Support #97: 6.5.0 on Hopper
What's the Cray compiler issue in question? I believe I've successfully built Charm++ with the Cray compiler in the r... Phil Miller
07:39 PM Support #97: 6.5.0 on Hopper
All versions (debug, production, trace) of the three compilers (pgi, icc, gcc) built in both smp and non-smp have suc... Michael Robson
12:50 AM Revision 7bb5bc06 (charm): a fix in the previous commit
Xiang Ni
12:45 AM Revision 896d012c (charm): inject soft and hard failures according to exponential and weibull dis...
Xiang Ni

04/04/2013

06:33 PM Bug #147 (Feedback): Crash in ChaNGa while writing disk checkpoint
Phil Miller
06:28 PM Support #160 (Resolved): SDAG documentation of message API and CkSetRefNum
Applied in changeset charm|commit:7fbaac1721584e0da3272534a9ba49ac829853b4. Phil Miller
06:26 PM Bug #145 (Closed): xcastredn test Makefile generates malformed dependency file on OSX
Ramprasad Venkataraman
06:26 PM Revision 7fbaac17 (charm): Fix #160: Document SDAG refnum usage on messages
Phil Miller
06:25 PM Bug #144 (Closed): xcastredn test crashes on multicore-linux64
Ramprasad Venkataraman
12:43 PM Bug #162 (Closed): Segfault in examples/charm++/leanmd with pgcc
The test of @examples/charm++/leanmd@ has failed for a while on pgcc builds. It turns out that leanmd had mismatched ... Phil Miller
12:25 PM Bug #161: SMP: Adapt to thread oversubscription, and provide flags to override
If we're only going to warn on the thread to core oversubscription condition, then we should probably just automatica... Phil Miller
10:22 AM Bug #161 (Closed): SMP: Adapt to thread oversubscription, and provide flags to override
The flag indicates the primary effect only in SMP builds. In multicore builds, for instance, when there are no com th... Ramprasad Venkataraman
10:14 AM Revision 96bec571 (charm): xcastredn: remove a vestigial floor call
Ramprasad Venkataraman
09:52 AM Support #94: 6.5.0 on ANL BGQ systems (Vesta, Cetus/Mira)
Test update Phil Miller
09:50 AM Bug #146 (In Progress): xcastredn test won't build on Windows due to missing 'usleep' and ambiguo...
Please test stuff like this directly before saying it's done, so that we don't keep bouncing on it. The build failed ... Phil Miller
09:42 AM Feature #112: object location services: Share array element location cache above PE level
This seems like a pretty obvious engineering change, as opposed to any deep research endeavor - any particular reason... Phil Miller
09:40 AM Revision 7c9cf21f (charm): examples/charm++/leanmd: Fix mismatched new[]/delete
Phil Miller
08:07 PM Revision ba4f968c (charm): add communication statistic summary
Yanhua Sun

04/03/2013

05:24 PM Bug #144 (Resolved): xcastredn test crashes on multicore-linux64
Fixed a global variable that was not protected with Cpv. Hence, in shared memory settings this resulted in multiple m... Ramprasad Venkataraman
03:46 PM Bug #144 (In Progress): xcastredn test crashes on multicore-linux64
At first glance, this looks like a deeper issue with how converse broadcasts interact with ckmulticast reductions on ... Ramprasad Venkataraman
11:35 AM Bug #144: xcastredn test crashes on multicore-linux64
Small autobuild breakages should not be left to fester for several days running, since it hinders visibility into oth... Phil Miller
05:15 PM Revision 743f50fa (charm): xcastredn: Make global variable into Cpv to separate each PE's copy
This should fix redmine issue#144. multicore builds exposed the issue
by having the test fail in strange ways.
The t...
Ramprasad Venkataraman
04:40 PM Revision c7eadafd (charm): ckmulticast: make abort msg more informative
Ramprasad Venkataraman
04:36 PM Support #160 (Closed): SDAG documentation of message API and CkSetRefNum
CkSetRefNum and its utility in using SDAG with messages, instead of marshalled methods, is not documented in the manu... Eric Bohm
02:46 PM Bug #145: xcastredn test Makefile generates malformed dependency file on OSX
Oh yeah, and we can close then when we see the relevant tests pass tomorrow. Phil Miller
02:45 PM Bug #145: xcastredn test Makefile generates malformed dependency file on OSX
I'm not a fan of re-using the variable @$(CXX)@, which happens to point to @charmc@, to explicitly mean @charmc@. Ple... Phil Miller
02:42 PM Bug #145 (Resolved): xcastredn test Makefile generates malformed dependency file on OSX
ci file dependency generation now via charmc -M instead of sed hackery Ramprasad Venkataraman
11:35 AM Bug #145: xcastredn test Makefile generates malformed dependency file on OSX
Small autobuild breakages should not be left to fester for several days running, since it hinders visibility into oth... Phil Miller
02:43 PM Bug #146 (Resolved): xcastredn test won't build on Windows due to missing 'usleep' and ambiguous ...
Ramprasad Venkataraman
11:35 AM Bug #146: xcastredn test won't build on Windows due to missing 'usleep' and ambiguous 'floor'
Small autobuild breakages should not be left to fester for several days running, since it hinders visibility into oth... Phil Miller
02:40 PM Revision 5735dcd8 (charm): make clean should remove .ci.stamp file
Ramprasad Venkataraman
02:40 PM Revision bb6a3642 (charm): xcastredn: fix dependency gen, default build flags and args
Ramprasad Venkataraman
02:40 PM Revision 33d6261d (charm): xcastredn: dont sleep filler methods. actually do flops
Ramprasad Venkataraman
02:27 PM Support #94 (Resolved): 6.5.0 on ANL BGQ systems (Vesta, Cetus/Mira)
#104 prevents us building the 5 @xlc * (smp, async-smp) * debug@ targets, but the the remaining 43/48 targets are bui... Phil Miller
09:55 AM Feature #108: Fixed-size object IDs
Wiki page here:
https://wiki.engr.illinois.edu/x/Vbo1DQ
Ramprasad Venkataraman

04/02/2013

05:02 PM Support #94: 6.5.0 on ANL BGQ systems (Vesta, Cetus/Mira)
Builds to be found in @~phil/charm-6.5@ Phil Miller
04:31 PM Revision c054f923 (charm): Fixing a bug in a recently added function to resize user size of message.
Adding another helper function that takes as parameter the number of bytes by
which to shrink the message.
NDMeshStr...
Lukasz Wesolowski
04:24 PM Bug #147: Crash in ChaNGa while writing disk checkpoint
This may be a bug in ChaNGa, that its @Main::restart()@ is not @[threaded]@, but calls @CkCallbackResumeThread()@. Phil Miller
01:38 PM Bug #147: Crash in ChaNGa while writing disk checkpoint
Tom says:
> The build configuration is "gemini_gni-crayxe-hugepages-smp".
Attaching the logs he sent.
Phil Miller
04:22 PM Bug #141 (Resolved): Update manual to reflect current Completion Detection API
Phil Miller
04:21 PM Revision 53eb8d04 (charm): docs/charm++: Update completion detection to reflect added doneProduci...
Phil Miller
04:14 PM Feature #27: Make CkIO safe to use in real application code with proper flush and close calls
Has work resumed on this now that the release is out? Eric Bohm
03:03 PM Bug #159 (New): Some CkCallback types are not valid across checkpoint/restart
Per #158, many types of @CkCallback@ contain (possibly transitively, through other structures) raw pointers to object... Phil Miller
02:49 PM Bug #158 (Closed): Document and check restrictions on passing non-array callbacks to CkStartCheck...
In a bit of exploration in the ChaNGa source code following from #147, I discovered the following comment:
> /// For...
Phil Miller
12:58 PM Support #98 (In Progress): 6.5.0 on NCSA systems (JYC, Blue Waters)
was on vacation last week. Revising build script for bluewaters now. Eric Bohm
11:27 AM Revision 3998d4ad (charm): remove unnecesaary barrier which is already in machine-common-core
Bilge Acun
09:10 AM Revision cca3d57f (charm): Reference charmrun by relative path, too
Phil Miller
09:10 AM Revision f31bf315 (charm): examples/charm++: systematically ensure that all binaries are named by...
Phil Miller
07:53 AM Revision b8268171 (charm): update leanmad according to the new sytax
Charj is able to parse mostly the whole code now, however a
lot of charmc compilation errors exists
Bilge Acun
11:55 PM Revision 29a9940f (charm): split log entries from trace-projections
Yanhua Sun
09:32 PM Support #97 (In Progress): 6.5.0 on Hopper
Michael Robson
07:55 PM Bug #78: AMPI failure with migration under Cray compiler due to tcmalloc bugs or incompatibility
Per https://bluewaters.ncsa.illinois.edu/known-issues tcmalloc is known to be problematic on BW:
> TCMALLOC: Codes m...
Phil Miller
07:24 PM Revision 3a78c460 (charm): Cleanup #74: Updating uses of initcall to initnode
Ronak Buch
07:16 PM Cleanup #74 (Resolved): Replace any remaining initcall declarations with initproc or initnode as ...
Ronak Buch
07:07 PM Revision 9a595a90 (charm): Charj: Started to modify jacobi test to support new syntax
Ronak Buch

04/01/2013

06:44 PM Cleanup #74 (In Progress): Replace any remaining initcall declarations with initproc or initnode ...
Ronak Buch
06:19 PM Revision c1a8f753 (charm): Merge branch 'charm' of charmgit:charm into charm
Yanhua Sun
06:18 PM Revision 7d4537d7 (charm): add +write-analysis-file for trace projections to get some automatic
summary information Yanhua Sun
05:04 PM Revision 1ff399d1 (charm): Adding a function for modifying user data size in messages, which requ...
copying over priority and extradata fields from the end of the message.
Cleaning up some comments in envelope.h.
Lukasz Wesolowski
01:27 PM Revision b7a84683 (charm): bug fix in the previous commit
Xiang Ni
01:01 PM Revision 91673c4a (charm): only worker thread will write the projections log
Xiang Ni
12:08 PM Support #99: 6.5.0 on Kraken
Ping: this was supposed to get done last week. Please complete the missing builds and post the path at which users ca... Phil Miller
12:07 PM Support #97: 6.5.0 on Hopper
Ping: this was supposed to get done last week. Please do the builds and post the path to them so we can close this out. Phil Miller
12:07 PM Support #98: 6.5.0 on NCSA systems (JYC, Blue Waters)
Ping: this was supposed to get done last week. Please do the builds and post a path to them, or note that you can't s... Phil Miller
12:04 PM Support #100: 6.5.0 on Trestles at SDSC
Esteban, could you log in to Trestles and confirm that those paths are accessible to you? When that's done, please cl... Phil Miller
11:52 AM Support #100: 6.5.0 on Trestles at SDSC
module purge
module load pgi mvapich2
/build AMPI mpi-linux-x86_64 smp --suffix=pgcc-debug -j8 -g
/build AMPI mpi-...
Xiang Ni
10:28 AM Feature #140 (Closed): Architecture target for Cray XC in the stable branch.
Phil Miller
10:28 AM Bug #128 (Closed): Compilation with persistent on Cray fails
Phil Miller
09:26 AM Bug #148 (Resolved): missing mlogfttest target
% done doesn't really mean much - marking this Resolved, and will Close when tests pass tomorrow. Phil Miller
11:23 PM Bug #148 (Closed): missing mlogfttest target
On @mpi-linux-x86_64-mlogft@, autobuild experiences this:... Phil Miller
09:23 AM Revision f1590d5c (charm): Fixing Makefiles in test directory for FT versions.
Esteban Meneses
08:04 AM Bug #157 (Closed): Align -(no-)optimize flags with current release of Cray compilers
We provide a generic wrapper for various backend compilers in charmc, and a flag -optimize that's meant to provide go... Phil Miller
07:51 AM Bug #153 (Closed): Align -optimize flags with current release of IBM XL compilers
We provide a generic wrapper for various backend compilers in charmc, and a flag @-optimize@ that's meant to provide ... Phil Miller
07:50 AM Support #151 (New): Align -optimize flags with current release of GNU compilers
We provide a generic wrapper for various backend compilers in charmc, and a flag @-optimize@ that's meant to provide ... Phil Miller
07:48 AM Support #149: Choose backend options for -optimize on icc
Put this below a generalized audit task for other backend compilers. Phil Miller
07:41 AM Support #149: Choose backend options for -optimize on icc
The @-fast@ option to icc is nothing special in itself. It's a roll-up of other options, specifically documented in t... Phil Miller
11:38 PM Support #149 (New): Choose backend options for -optimize on icc
NERSC mailing list advised use of -fast -no-ipo as the preferred option for icc compilers on Edison. Phil did that an... Nikhil Jain
07:47 AM Support #150 (New): Align -optimize flags with current compiler releases
We provide a generic wrapper for various backend compilers in charmc, and a flag @-optimize@ that's meant to provide ... Phil Miller
07:10 AM Revision 72161566 (charm): Reference binary relative to ./ to appease Edison's aprun, redux
Phil Miller
11:19 PM Revision eb092905 (charm): examples/charm++: Only run bgtest on examples that actually have that ...
Phil Miller
11:07 PM Bug #147 (Rejected): Crash in ChaNGa while writing disk checkpoint
Tom Quinn reports:
> Sorry I'm slow getting back on this. I got it to read correctly once, but now I'm having tro...
Phil Miller
07:27 PM Bug #146 (Closed): xcastredn test won't build on Windows due to missing 'usleep' and ambiguous 'f...
... Phil Miller

03/31/2013

06:58 PM Bug #145 (Closed): xcastredn test Makefile generates malformed dependency file on OSX
... Phil Miller
06:56 PM Bug #144 (Closed): xcastredn test crashes on multicore-linux64
Output seems to be a bit out of order, possibly due to after-the-fact mixing of stdout and stderr, but the result of ... Phil Miller
05:56 PM Revision 88823570 (charm): bug fix: disable generate failure when not requested
Xiang Ni
05:45 PM Revision 4b95fe56 (charm): examples/charm++: Don't run NQueen example during testing, since Contr...
Phil Miller
05:40 PM Revision 8941c7f8 (charm): examples/charm++/matmul: And then don't link the missing BLAS library
Phil Miller
05:15 PM Revision 42249962 (charm): add auto analysis files
Yanhua Sun
05:01 PM Revision a94ead87 (charm): add fib example for control point
Yanhua Sun
04:30 PM Revision e354218b (charm): add examples to test performance analysis
Yanhua Sun
04:17 PM Revision dd43e413 (charm): Check in automatic performance analysis code, still in development phase
Yanhua Sun
11:52 AM Revision 7b637f96 (charm): examples/charm++/matmul: replace non-portable CBLAS DGEMM call with a ...
Phil Miller
11:06 AM Revision 497f15bc (charm): examples/charm++/completion/Makefile: don't force non-portable compila...
Phil Miller
10:58 AM Revision b0db7007 (charm): examples/charm++/NQueen/Makefile: test needs to depend on the binary t...
Phil Miller
12:48 AM Bug #143 (Merged): Parallel (make -j) build of AMPI libraries and client code is racy
Nikhil Jain
12:44 AM Bug #142 (Closed): Decide on default memory module
We use different memory modules as default based on the build - revisit why we use which one, and bring them up to da... Nikhil Jain
12:14 AM Revision 370935b1 (charm): Revert "Add appropriate flags for charmc -optimize when using Intel co...
It turns out the -fast flag to icc forces static linking, which is unacceptable
for various reasons (off the top of m...
Phil Miller
 

Also available in: Atom