11 years agoMakefile: automate adding headers for new load balancers
Phil Miller [Tue, 14 Sep 2010 22:12:13 +0000 (17:12 -0500)]
Makefile: automate adding headers for new load balancers

11 years agopplmanual.sty should be deleted on a make clean
Abhinav S Bhatele [Tue, 14 Sep 2010 21:44:09 +0000 (16:44 -0500)]
pplmanual.sty should be deleted on a make clean

11 years agoMerge branch 'charm' of charmgit:charm into charm
Abhinav S Bhatele [Tue, 14 Sep 2010 21:38:21 +0000 (16:38 -0500)]
Merge branch 'charm' of charmgit:charm into charm

11 years agoManual: updated the steps to add a load balancer
Abhinav S Bhatele [Tue, 14 Sep 2010 21:37:10 +0000 (16:37 -0500)]
Manual: updated the steps to add a load balancer

11 years agocharmxi: remove archaic CVS header
Phil Miller [Tue, 14 Sep 2010 21:19:16 +0000 (16:19 -0500)]
charmxi: remove archaic CVS header

11 years agocharmxi -M: Stop generating stamp rules that can be implicit
Phil Miller [Tue, 14 Sep 2010 21:18:58 +0000 (16:18 -0500)]
charmxi -M: Stop generating stamp rules that can be implicit

11 years agoMakefile: Don't echo the echo
Phil Miller [Tue, 14 Sep 2010 21:06:04 +0000 (16:06 -0500)]
Makefile: Don't echo the echo

11 years agoAutomate ci file dependencies entirely
Phil Miller [Mon, 13 Sep 2010 16:33:11 +0000 (11:33 -0500)]
Automate ci file dependencies entirely

For the sake of avoiding circular dependencies in generating
Make.depends, teach the Makefile how to handle ci files
implicitly. Tell it to generate all decl/def.h files before generating
dependencies, so that the compiler doesn't fail on missing headers.

The problem of circular dependencies came to light from Abhinav trying
to add a new load balancer. The solution technique is adopted directly
from Ram's work on OpenAtom's build system, following common Make

11 years agoMerge branch 'charm' of charmgit:charm into charm
Filippo Gioachin [Fri, 10 Sep 2010 19:19:32 +0000 (14:19 -0500)]
Merge branch 'charm' of charmgit:charm into charm

11 years agoFixed ResetAllCRC for CMK_SEPARATE_SLOT case
Filippo Gioachin [Fri, 10 Sep 2010 19:19:24 +0000 (14:19 -0500)]
Fixed ResetAllCRC for CMK_SEPARATE_SLOT case

11 years agoMerge branch 'charm' of charmgit:charm into charm
Aaron Becker [Fri, 10 Sep 2010 15:35:23 +0000 (10:35 -0500)]
Merge branch 'charm' of charmgit:charm into charm

11 years agoNetFEM: Allow selective processing of input timesteps.
Aaron Becker [Fri, 10 Sep 2010 15:32:55 +0000 (10:32 -0500)]
NetFEM: Allow selective processing of input timesteps.

invoking "./NetFEM_To_Paraview low high stride" will process all steps
from low to high, strided by stride. You can also specify just 'low', or
just 'low' and 'high'. Default behavior is unaffected.

11 years agoWith a separate slot, you can hit a block not indexed. If so ignore it
Filippo Gioachin [Wed, 8 Sep 2010 23:56:36 +0000 (18:56 -0500)]
With a separate slot, you can hit a block not indexed. If so ignore it

11 years agoIncorrect calculation of memory needed for backup when separate slot is used
Filippo Gioachin [Wed, 8 Sep 2010 23:50:42 +0000 (18:50 -0500)]
Incorrect calculation of memory needed for backup when separate slot is used

11 years agoMerge branch 'charm' of /expand/home/gioachin/charm into charm
Filippo Gioachin [Wed, 8 Sep 2010 22:12:54 +0000 (17:12 -0500)]
Merge branch 'charm' of /expand/home/gioachin/charm into charm

11 years agoUnified memory reporting cross-object modification only when +memory_report is specified
Filippo Gioachin [Wed, 8 Sep 2010 21:40:14 +0000 (16:40 -0500)]
Unified memory reporting cross-object modification only when +memory_report is specified

11 years agoIn conditional mode, make the child receive locally sent messages
Filippo Gioachin [Tue, 7 Sep 2010 20:19:08 +0000 (15:19 -0500)]
In conditional mode, make the child receive locally sent messages

11 years agoadded an example of chare array broadcast
Aaron Becker [Tue, 7 Sep 2010 01:14:11 +0000 (20:14 -0500)]
added an example of chare array broadcast

Note: this example doesn't work if you don't specify a constructor for
the array because of some gremlin in charmxi.

11 years agoCharj : methods declared with the "entry" keyword are now
Minas Charalambides [Mon, 6 Sep 2010 20:15:44 +0000 (15:15 -0500)]
Charj : methods declared with the "entry" keyword are now
automatically made public as well (PUBLIC is being
added to the AST)

11 years agoinclude conv-config.h, without it, if controlpoints.h is included first, it never...
Gengbin Zheng [Fri, 3 Sep 2010 18:29:04 +0000 (13:29 -0500)]
include conv-config.h, without it, if controlpoints.h is included first, it never is activated since it is inside #ifdef CMK_WITH_CONTROLPOINT.

11 years agoOverload << operator for ckcomplex
Anshu Arya [Fri, 3 Sep 2010 17:51:28 +0000 (12:51 -0500)]
Overload << operator for ckcomplex

Can now output ckcomplex using conventional operator to simplify
output of complex numbers.

11 years agoIt appears to be good idea to pass OPTS_CPP (which contrains -D) to the preprocessing...
Gengbin Zheng [Fri, 3 Sep 2010 18:15:01 +0000 (13:15 -0500)]
It appears to be good idea to pass OPTS_CPP (which contrains -D) to the preprocessing phase of charmxi -E.

11 years agoMerge branch 'charm' of /expand/home/gioachin/charm into charm
Filippo Gioachin [Fri, 3 Sep 2010 16:21:33 +0000 (11:21 -0500)]
Merge branch 'charm' of /expand/home/gioachin/charm into charm

11 years agoAlways optimizing the charmdebug memory libraries
Filippo Gioachin [Fri, 3 Sep 2010 16:19:28 +0000 (11:19 -0500)]
Always optimizing the charmdebug memory libraries

11 years agofor SMP and getrusage timer mode, default getrusage does not return correct cpu time...
Gengbin Zheng [Wed, 1 Sep 2010 16:20:57 +0000 (11:20 -0500)]
for SMP and getrusage timer mode, default getrusage does not return correct cpu time since it accumulates across all threads. Disable cpu timer by using wallclock time instead to avoid confusion, however, keep it when the latest Linux kernel supports RUSAGE_THREAD mode.

11 years agoupdated configure for thre previous change to test RUSAGE_THREAD (which only is suppo...
Gengbin Zheng [Wed, 1 Sep 2010 16:17:23 +0000 (11:17 -0500)]
updated configure for thre previous change to test RUSAGE_THREAD (which only is supported after Linux 2.6.26)

11 years agotest RUSAGE_THREAD for getrusage.
Gengbin Zheng [Wed, 1 Sep 2010 13:57:40 +0000 (08:57 -0500)]
test RUSAGE_THREAD for getrusage.

11 years agogitignore ci.stamp files generated by charmc
Ramprasad Venkataraman [Tue, 31 Aug 2010 19:57:23 +0000 (14:57 -0500)]
gitignore ci.stamp files generated by charmc

11 years agoMerge branch 'ramv/x-array-delegation' into charm
Ramprasad Venkataraman [Tue, 31 Aug 2010 19:41:19 +0000 (14:41 -0500)]
Merge branch 'ramv/x-array-delegation' into charm

11 years agox-array sections: Update example code to use x-array section reductions too
Ramprasad Venkataraman [Mon, 30 Aug 2010 20:30:31 +0000 (15:30 -0500)]
x-array sections: Update example code to use x-array section reductions too

11 years agoCkMulticast: Support cross-array section reductions
Ramprasad Venkataraman [Mon, 30 Aug 2010 20:09:56 +0000 (15:09 -0500)]
CkMulticast: Support cross-array section reductions

Complete the first pass at delegation for cross-array sections
by enabling support for reductions along such sections.

Again, motive is to quickly bake the functionality into the library
without much heed to performance or optimization. Hence, reduction
support is implemented as reductions along each array section followed
by a final reduction step where all these messages are combined again.

- Tiny memleaks not of practical concern (see source comments).
- Only supports CkCallbacks as cross-array reduction clients.
    CkMulticast provides two mechanisms to specify
    reduction clients, one of which accepts a C function.
    This mechanism does not have cross-array reduction
    support built in yet. Again, should not be of concern
    in the near future.

11 years agox-array sections: Hello world example code to demo cross-array sections
Ramprasad Venkataraman [Fri, 27 Aug 2010 16:45:46 +0000 (11:45 -0500)]
x-array sections: Hello world example code to demo cross-array sections

11 years agoCkMulticast: Support multicasts to cross-array sections
Ramprasad Venkataraman [Fri, 27 Aug 2010 16:36:07 +0000 (11:36 -0500)]
CkMulticast: Support multicasts to cross-array sections

Quick implementation to add desired functionality with no efforts
on performance optimization. Treat a cross-array section simply as
a collection of individual array sections, and process each of these
(quote/unquote)sub-sections as usual.

Reduction support for x-array sections will follow in other commits,
but there are no checks/aborts in place now; so beware.

11 years agoCProxySection_ArrayBase: Fix bug caused by inconsistent new/deletes
Ramprasad Venkataraman [Thu, 26 Aug 2010 22:40:11 +0000 (17:40 -0500)]
CProxySection_ArrayBase: Fix bug caused by inconsistent new/deletes

Implementing cross-array-sections required new constructors for creating multi-array
sections. These constructors always used new[] even if the section spanned just one
chare array. The destructor on the other hand uses delete / delete[] depending on the
number of arrays in the section. Using these constructors to create a single-array
section causes segfaults in the destructor.

Quick fix so that new[] is used only if _nsid > 1 irrespective of which constructor
is invoked.

11 years agox-array sections: New method CProxySection_ArrayBase::ckGetNumSubSections().
Ramprasad Venkataraman [Wed, 25 Aug 2010 20:59:45 +0000 (15:59 -0500)]
x-array sections: New method CProxySection_ArrayBase::ckGetNumSubSections().

Make it possible to figure out if a section proxy targets a cross-array section, and if so,
the number of individual array sections that comprise such a cross-array section.

11 years agosome code docs for ckmulticast
Ramprasad Venkataraman [Tue, 24 Aug 2010 16:40:45 +0000 (11:40 -0500)]
some code docs for ckmulticast

11 years agoreindent and document array section util classes
Ramprasad Venkataraman [Mon, 23 Aug 2010 22:31:13 +0000 (17:31 -0500)]
reindent and document array section util classes

11 years agoChare Arrays: PUP elements in mapping class; needed for checkpoint
Abhinav S Bhatele [Mon, 30 Aug 2010 21:30:43 +0000 (16:30 -0500)]
Chare Arrays: PUP elements in mapping class; needed for checkpoint

11 years agoRevert "had forgotten to add elements to pup"
Abhinav S Bhatele [Mon, 30 Aug 2010 21:43:18 +0000 (16:43 -0500)]
Revert "had forgotten to add elements to pup"

This reverts commit ecb013a10292592e0a85403bb69cc161b576176b.

11 years agohad forgotten to add elements to pup
Abhinav S Bhatele [Mon, 30 Aug 2010 21:30:43 +0000 (16:30 -0500)]
had forgotten to add elements to pup

11 years ago-g removed from the wrong places in the Makefile
Abhinav Bhatele [Sun, 29 Aug 2010 05:03:08 +0000 (01:03 -0400)]
-g removed from the wrong places in the Makefile

11 years agoAdding a macro to supress a deprecation warning for ucontext routines on OSX 10.6...
Isaac Dooley [Thu, 26 Aug 2010 16:17:42 +0000 (11:17 -0500)]
Adding a macro to supress a deprecation warning for ucontext routines on OSX 10.6 (snow leopard).

The original error was this:
/usr/include/ucontext.h:42:2: error: #error ucontext routines are deprecated, and require _XOPEN_SOURCE to be defined

11 years agoMerge branch 'charm' of charmgit:charm into charm
Isaac Dooley [Wed, 25 Aug 2010 21:46:24 +0000 (16:46 -0500)]
Merge branch 'charm' of charmgit:charm into charm

11 years agoDisabling more code that causes extraneous build errors on net-darwin-x86/x86_64...
Isaac Dooley [Wed, 25 Aug 2010 21:45:57 +0000 (16:45 -0500)]
Disabling more code that causes extraneous build errors on net-darwin-x86/x86_64 on macs.

11 years agoMerge branch 'charm' of charmgit:charm into charm
Aaron Becker [Wed, 25 Aug 2010 21:32:45 +0000 (16:32 -0500)]
Merge branch 'charm' of charmgit:charm into charm

11 years agoCharj: make sure all sdag decls are in the right file and nowhere else
Aaron Becker [Wed, 25 Aug 2010 21:31:10 +0000 (16:31 -0500)]
Charj: make sure all sdag decls are in the right file and nowhere else

11 years agoDisabling code that causes extraneous build errors on net-darwin-x86/x86_64 on macs.
Isaac Dooley [Wed, 25 Aug 2010 20:34:14 +0000 (15:34 -0500)]
Disabling code that causes extraneous build errors on net-darwin-x86/x86_64 on macs.

The extraneous errors would occur in various files, but the build would still succeed. This commit eliminates many of the errors so that they don't polute the build output, falsely reporting errors in IDEs such as Eclipse (which hides the real errors).

11 years agoCharj: made SDAG grammar reflect actual SDAG implementation
Aaron Becker [Wed, 25 Aug 2010 16:20:59 +0000 (11:20 -0500)]
Charj: made SDAG grammar reflect actual SDAG implementation

11 years agoAdding guards around #include <malloc.h> calls for systems like OSX which doesn't...
Isaac Dooley [Wed, 25 Aug 2010 14:27:41 +0000 (09:27 -0500)]
Adding guards around #include <malloc.h> calls for systems like OSX which doesn't have malloc.h.

11 years agoCharj: do .ci codegen for sdag, infer atomic sections
Aaron Becker [Tue, 24 Aug 2010 23:12:36 +0000 (18:12 -0500)]
Charj: do .ci codegen for sdag, infer atomic sections

11 years agoCharj: fix generated projections tracing macro
Aaron Becker [Tue, 24 Aug 2010 21:09:37 +0000 (16:09 -0500)]
Charj: fix generated projections tracing macro

11 years agoCharj: add SDAG reference number support
Aaron Becker [Tue, 24 Aug 2010 21:05:33 +0000 (16:05 -0500)]
Charj: add SDAG reference number support

11 years agoCharj: fix sdag construct indent
Aaron Becker [Tue, 24 Aug 2010 21:04:53 +0000 (16:04 -0500)]
Charj: fix sdag construct indent

11 years agoUpdate NQueen problems about the bug in workstealing
Yanhua Yanhua [Tue, 24 Aug 2010 02:48:35 +0000 (21:48 -0500)]
Update NQueen problems about the bug in workstealing

11 years agoCharj: add macros for sdag pup/c'tor codegen
Aaron Becker [Mon, 23 Aug 2010 22:28:53 +0000 (17:28 -0500)]
Charj: add macros for sdag pup/c'tor codegen

11 years agofixed bugs in the default mapping of chare arrays
Abhinav S Bhatele [Mon, 23 Aug 2010 21:38:54 +0000 (16:38 -0500)]
fixed bugs in the default mapping of chare arrays

One fix is for making the blocked mapping more evenly balanced and not leaving
any processor with zero chares.

The second fix is for anytime insertions which did not work at all earlier.

11 years agoMerge branch 'charm' of charmgit:charm into charm
Aaron Becker [Mon, 23 Aug 2010 20:48:13 +0000 (15:48 -0500)]
Merge branch 'charm' of charmgit:charm into charm

11 years agoprepend string to timestamp files for compatibility with newer paraviews
Aaron Becker [Mon, 23 Aug 2010 20:34:42 +0000 (15:34 -0500)]
prepend string to timestamp files for compatibility with newer paraviews

11 years agorecord which methods contain sdag syntax in symbol table
Aaron Becker [Mon, 23 Aug 2010 20:32:01 +0000 (15:32 -0500)]
record which methods contain sdag syntax in symbol table

11 years agoAdd grammar for SDAG 'when' and 'overlap' statements
Aaron Becker [Mon, 23 Aug 2010 20:26:42 +0000 (15:26 -0500)]
Add grammar for SDAG 'when' and 'overlap' statements

11 years agoCharj: Forgot to check in these files.
Jonathan Lifflander [Mon, 23 Aug 2010 20:24:09 +0000 (15:24 -0500)]
Charj: Forgot to check in these files.

11 years agoCharj: Forgot to check in these files.
Jonathan Lifflander [Mon, 23 Aug 2010 20:24:09 +0000 (15:24 -0500)]
Charj: Forgot to check in these files.

11 years agouse charm++ functions instead of converse where possible
Aaron Becker [Mon, 23 Aug 2010 19:50:42 +0000 (14:50 -0500)]
use charm++ functions instead of converse where possible

11 years agoMerge branch 'charm' of charmgit:charm into charm
Yanhua Yanhua [Fri, 20 Aug 2010 18:06:20 +0000 (13:06 -0500)]
Merge branch 'charm' of charmgit:charm into charm

11 years agoupdate nqueen example
Yanhua Yanhua [Fri, 20 Aug 2010 18:05:57 +0000 (13:05 -0500)]
update nqueen example

11 years agoa typo in previous checkin.
Gengbin Zheng [Thu, 19 Aug 2010 16:59:18 +0000 (11:59 -0500)]
a typo in previous checkin.

11 years agoregenerate pup_f.f90 (update timestamp), and update Makefile.
Gengbin Zheng [Thu, 19 Aug 2010 16:56:21 +0000 (11:56 -0500)]
regenerate pup_f.f90 (update timestamp), and update Makefile.

11 years agofixed deprecated -static-libcxa
Gengbin Zheng [Thu, 19 Aug 2010 03:15:05 +0000 (22:15 -0500)]
fixed deprecated -static-libcxa

11 years agoCharj: Test 1D and 2D arrays.
Jonathan Lifflander [Thu, 19 Aug 2010 02:06:37 +0000 (21:06 -0500)]
Charj: Test 1D and 2D arrays.

11 years agoCharj: Modification to handle 2D arrays.
Jonathan Lifflander [Thu, 19 Aug 2010 02:05:44 +0000 (21:05 -0500)]
Charj: Modification to handle 2D arrays.

11 years agoCharj: Augment array accessor to work with domains.
Jonathan Lifflander [Thu, 19 Aug 2010 02:00:01 +0000 (21:00 -0500)]
Charj: Augment array accessor to work with domains.

11 years agoCharj: Test suite for backend C++ arrays.
Jonathan Lifflander [Thu, 19 Aug 2010 01:57:19 +0000 (20:57 -0500)]
Charj: Test suite for backend C++ arrays.

11 years agoCharj: New 2D multi-dimensional array support in C++.
Jonathan Lifflander [Thu, 19 Aug 2010 01:55:39 +0000 (20:55 -0500)]
Charj: New 2D multi-dimensional array support in C++.

11 years agoCharj: Fix SymbolResolver Array problem.
Jonathan Lifflander [Tue, 17 Aug 2010 21:04:55 +0000 (16:04 -0500)]
Charj: Fix SymbolResolver Array problem.

11 years agoCharj: Modification so List<TypeName> is used instead of List<String>.
Jonathan Lifflander [Tue, 17 Aug 2010 18:36:50 +0000 (13:36 -0500)]
Charj: Modification so List<TypeName> is used instead of List<String>.

Using a custom class (TypeName) for this will allow the type of
generics to be described.

11 years agoCharj: Instead of joining typenames by periods, use a list.
Jonathan Lifflander [Tue, 17 Aug 2010 16:52:17 +0000 (11:52 -0500)]
Charj: Instead of joining typenames by periods, use a list.

In the symbol resolver and definer, the qualified type name was being
joined by periods and then parsed by the Scope resolveType methods
to find the actual type. Now this is a List<String>.

11 years agoCharj: Removal of dead code.
Jonathan Lifflander [Tue, 17 Aug 2010 16:50:37 +0000 (11:50 -0500)]
Charj: Removal of dead code.

11 years agoCharj: Whitespace change.
Jonathan Lifflander [Tue, 17 Aug 2010 16:49:53 +0000 (11:49 -0500)]
Charj: Whitespace change.

11 years agoCharj: Bug fix with initializers.
Jonathan Lifflander [Tue, 17 Aug 2010 16:47:29 +0000 (11:47 -0500)]
Charj: Bug fix with initializers.

11 years agoCharj: Removal of old "genericTypeArgumentList" rule.
Jonathan Lifflander [Tue, 17 Aug 2010 06:04:39 +0000 (01:04 -0500)]
Charj: Removal of old "genericTypeArgumentList" rule.

11 years agoCharj: Correction to the tree walkers in the modifiers.
Jonathan Lifflander [Tue, 17 Aug 2010 06:03:32 +0000 (01:03 -0500)]
Charj: Correction to the tree walkers in the modifiers.

11 years agoCharj: Correction in finding the "basename"
Jonathan Lifflander [Tue, 17 Aug 2010 04:34:28 +0000 (23:34 -0500)]
Charj: Correction in finding the "basename"

charjc ./Test.cj was not working, so the code was modified so the "./"
is parsed out. charjc <full-path> still does not work.

11 years agoCharj: Small mistake fixed in the ArrayTest example program.
Jonathan Lifflander [Tue, 3 Aug 2010 19:39:41 +0000 (14:39 -0500)]
Charj: Small mistake fixed in the ArrayTest example program.

11 years agoset seed balancer default topology as torus_nd_5
Yanhua Yanhua [Wed, 18 Aug 2010 16:41:35 +0000 (11:41 -0500)]
set seed balancer default topology as torus_nd_5

11 years agofixed a bug in charmInit in BigSim, where CldRegisterInfoFn which means to
Gengbin Zheng [Tue, 17 Aug 2010 21:16:12 +0000 (16:16 -0500)]
fixed a bug in charmInit in BigSim, where CldRegisterInfoFn which means to
be Converse, for some reason is initialized in charm init, therefore has to
make sure it calls once on each Converse processor (not every target processor)

11 years agoset default topology for seed load balancer to torus_5d
Yanhua Yanhua [Tue, 17 Aug 2010 18:35:55 +0000 (13:35 -0500)]
set default topology for seed load balancer to torus_5d

11 years agoin workstealing, add steal work by 1
Yanhua Yanhua [Tue, 17 Aug 2010 17:16:11 +0000 (12:16 -0500)]
in workstealing, add steal work by 1

11 years agoCkMulticast Reductions: Set result message reference number to userFlag
Phil Miller [Thu, 12 Aug 2010 21:27:13 +0000 (16:27 -0500)]
CkMulticast Reductions: Set result message reference number to userFlag

11 years agoReductions: Set the reference number on result messages to the userFlag
Phil Miller [Thu, 12 Aug 2010 20:43:29 +0000 (15:43 -0500)]
Reductions: Set the reference number on result messages to the userFlag

Setting the reference number on the CkReductionMsg passed to the
callback to the reduction number was a dirty hack to allow Pritish to
get something predictable in the SDAG code Charisma generated. See
74e51ab49dc766f93bd3c297254ffb157294ae3a about that.

To expose this as reliable, controllable functionality, pick up the
caller-provided userFlag value (defaults to -1) and use that instead.

11 years agoAnother reducer function set for CCS requests from charmdebug
Filippo Gioachin [Tue, 10 Aug 2010 23:42:25 +0000 (18:42 -0500)]
Another reducer function set for CCS requests from charmdebug

11 years agoIf the program crashes during the delivery of a forced message, the sentinel message...
Filippo Gioachin [Tue, 10 Aug 2010 23:41:53 +0000 (18:41 -0500)]
If the program crashes during the delivery of a forced message, the sentinel message (-1) is in the queue. Ignore it.

11 years agoAdded new function to return a reply if no reply has already been sent, without abort...
Filippo Gioachin [Tue, 10 Aug 2010 20:13:35 +0000 (15:13 -0500)]
Added new function to return a reply if no reply has already been sent, without aborting the program if a reply has already been sent

11 years agoFixing case where traceBegin is called before first traceEnd.
Isaac Dooley [Sat, 7 Aug 2010 21:08:41 +0000 (16:08 -0500)]
Fixing case where traceBegin is called before first traceEnd.

11 years agoAdding preliminary support for traceBegin/traceEnd in control point tracing module.
Isaac Dooley [Sat, 7 Aug 2010 20:15:53 +0000 (15:15 -0500)]
Adding preliminary support for traceBegin/traceEnd in control point tracing module.

11 years agoAdd ACK message when one processor gets an idle request but does not have tasks to...
Yanhua Yanhua [Fri, 6 Aug 2010 04:15:23 +0000 (23:15 -0500)]
Add ACK message when one processor gets an idle request but does not have tasks to send back in work stealing. Also add a charmrun time option +stealonly1 to let work stealing only steal 1. By default it steals half of the whole tasks

11 years agomake the default topology for neighbor seed balancer as torus_nd_5, which performs...
Yanhua Yanhua [Thu, 5 Aug 2010 19:54:48 +0000 (14:54 -0500)]
make the default topology for neighbor seed balancer as torus_nd_5, which performs better than the old one torus2d

11 years agoRewording assumptions about priorities as discussed today in the group meeting.
Isaac Dooley [Thu, 5 Aug 2010 17:38:41 +0000 (12:38 -0500)]
Rewording assumptions about priorities as discussed today in the group meeting.

11 years agoMake the period time for work stealing smaller so the idle processor requests for...
Yanhua Yanhua [Thu, 5 Aug 2010 15:55:14 +0000 (10:55 -0500)]
Make the period time for work stealing smaller so the idle processor requests for tasks more often

11 years agoNetFEM: add prefix to timestep filenames for Paraview 3.x
Aaron Becker [Thu, 5 Aug 2010 00:07:01 +0000 (19:07 -0500)]
NetFEM: add prefix to timestep filenames for Paraview 3.x

11 years agogetifaddrs may return multiple eth IP addresses that confused skt_my_ip call.
Gengbin Zheng [Wed, 4 Aug 2010 04:02:36 +0000 (23:02 -0500)]
getifaddrs may return multiple eth IP addresses that confused skt_my_ip call.
Use the old scheme of gethostname if multiple eth exists.