charm.git
10 years agofix path for charmc
Eric Bohm [Thu, 7 Jul 2011 17:56:16 +0000 (12:56 -0500)]
fix path for charmc

10 years agoAllReduce bug fixed, new benchmark added
Ehsan Totoni [Sun, 3 Jul 2011 17:34:19 +0000 (12:34 -0500)]
AllReduce bug fixed, new benchmark added

10 years agodocs: Documentation for using completion detection.
Jonathan Lifflander [Sat, 2 Jul 2011 21:21:05 +0000 (16:21 -0500)]
docs: Documentation for using completion detection.

10 years agoCompletion detector example.
Jonathan Lifflander [Sat, 2 Jul 2011 02:26:44 +0000 (19:26 -0700)]
Completion detector example.

10 years agoCkCallback: Clean up new array section constructor
Phil Miller [Thu, 30 Jun 2011 03:11:25 +0000 (20:11 -0700)]
CkCallback: Clean up new array section constructor

10 years agoRemove muddy bootprints
Phil Miller [Thu, 30 Jun 2011 03:09:05 +0000 (20:09 -0700)]
Remove muddy bootprints

10 years agoadd section callback
Xiang Ni [Thu, 30 Jun 2011 03:02:29 +0000 (22:02 -0500)]
add section callback

10 years agoDisabled the cpu affinity on BG/P and printed a warning instead of abortion if
Chao Mei [Tue, 28 Jun 2011 23:14:44 +0000 (18:14 -0500)]
Disabled the cpu affinity on BG/P and printed a warning instead of abortion if
the cpu affinity is set.

10 years agoOptimized the implementation for the broadcast in hypercube (which actually is
Chao Mei [Tue, 28 Jun 2011 22:59:19 +0000 (17:59 -0500)]
Optimized the implementation for the broadcast in hypercube (which actually is
a binomial tree implementation)

10 years agoallreduce chunk size changed
Ehsan Totoni [Tue, 28 Jun 2011 17:56:03 +0000 (12:56 -0500)]
allreduce chunk size changed

10 years agocharmxi: remove remaining CVS headers
Phil Miller [Mon, 27 Jun 2011 00:07:23 +0000 (17:07 -0700)]
charmxi: remove remaining CVS headers

10 years agocharmxi: Write syntax error messages to stderr, not stdout
Phil Miller [Mon, 27 Jun 2011 00:01:11 +0000 (17:01 -0700)]
charmxi: Write syntax error messages to stderr, not stdout

10 years agocharmxi: warn about superfluous CkMigrateMsg chare constructors
Phil Miller [Sun, 26 Jun 2011 23:59:30 +0000 (16:59 -0700)]
charmxi: warn about superfluous CkMigrateMsg chare constructors

10 years agocharmxi: Remove obsolete comment about ref params
Phil Miller [Sun, 26 Jun 2011 23:50:04 +0000 (16:50 -0700)]
charmxi: Remove obsolete comment about ref params

Entry methods taking reference parameters appear in a bunch of
different places, including some [local] entry methods. They're
clearly not defunct, and serve a useful purpose.

10 years agocharmxi: move standard headers below local, to avoid accidents
Phil Miller [Sun, 26 Jun 2011 03:59:57 +0000 (20:59 -0700)]
charmxi: move standard headers below local, to avoid accidents

10 years agocharmxi: remove pointless cast
Phil Miller [Sun, 26 Jun 2011 03:59:33 +0000 (20:59 -0700)]
charmxi: remove pointless cast

10 years agoControl Points: const-ify a string argument
Phil Miller [Sun, 26 Jun 2011 03:52:43 +0000 (20:52 -0700)]
Control Points: const-ify a string argument

10 years agofixed a bug in calling ckLocal for a local chare.
Gengbin Zheng [Fri, 24 Jun 2011 20:53:14 +0000 (15:53 -0500)]
fixed a bug in calling ckLocal for a local chare.

10 years agoallreduce example added to charm
Ehsan Totoni [Fri, 24 Jun 2011 16:25:18 +0000 (11:25 -0500)]
allreduce example added to charm

10 years agopipelined allreduce for large messages implemented, use -D_PIPELINED_ALLREDUCE_ to...
Ehsan Totoni [Fri, 24 Jun 2011 16:21:53 +0000 (11:21 -0500)]
pipelined allreduce for large messages implemented, use -D_PIPELINED_ALLREDUCE_ to enable

10 years agoMerge branch 'charm' of charmgit:charm into charm
Gengbin Zheng [Mon, 20 Jun 2011 19:34:11 +0000 (14:34 -0500)]
Merge branch 'charm' of charmgit:charm into charm

10 years agooptimization for stableLocation did not work due to miscalculated length of elements.
Gengbin Zheng [Mon, 20 Jun 2011 19:27:14 +0000 (14:27 -0500)]
optimization for stableLocation did not work due to miscalculated length of elements.

10 years agoswitched to memory-os for MPI-smp
Gengbin Zheng [Fri, 17 Jun 2011 05:20:06 +0000 (00:20 -0500)]
switched to memory-os for MPI-smp

10 years agofixed a bug in wrong calculation of MPI_DoubleInt for example, where the padding...
Gengbin Zheng [Thu, 16 Jun 2011 21:55:04 +0000 (16:55 -0500)]
fixed a bug in wrong calculation of MPI_DoubleInt for example, where the padding may change the actually space.

10 years agobash=>sh
Gengbin Zheng [Wed, 15 Jun 2011 20:49:14 +0000 (15:49 -0500)]
bash=>sh

10 years agoadded the memory usage report for BGP according to the wiki provided by ANL
Chao Mei [Wed, 15 Jun 2011 03:22:18 +0000 (22:22 -0500)]
added the memory usage report for BGP according to the wiki provided by ANL

10 years agoremoved the unnecessary cpv initialization as it is never used, and caused compilatio...
Chao Mei [Wed, 15 Jun 2011 02:56:56 +0000 (21:56 -0500)]
removed the unnecessary cpv initialization as it is never used, and caused compilation error in SMP mode

10 years agoMerge branch 'charm' of charmgit:charm into charm
Esteban Meneses [Tue, 14 Jun 2011 21:27:48 +0000 (16:27 -0500)]
Merge branch 'charm' of charmgit:charm into charm

10 years agoChanges to insert a section about Futures in the manual.
Esteban Meneses [Tue, 14 Jun 2011 21:27:29 +0000 (16:27 -0500)]
Changes to insert a section about Futures in the manual.

10 years agoMerge branch 'charm' of charmgit:charm into charm
YanhuaSun [Tue, 14 Jun 2011 20:15:29 +0000 (15:15 -0500)]
Merge branch 'charm' of charmgit:charm into charm

10 years agoadded post-mortem analysis, it is valid with +sumDetail option
YanhuaSun [Tue, 14 Jun 2011 20:15:18 +0000 (15:15 -0500)]
added post-mortem analysis, it is valid with +sumDetail option

10 years agoAdding long integer strategy for queueing.
Esteban Meneses [Mon, 13 Jun 2011 20:37:32 +0000 (15:37 -0500)]
Adding long integer strategy for queueing.

10 years agoAdding a couple of queueing types.
Esteban Meneses [Mon, 13 Jun 2011 16:46:28 +0000 (11:46 -0500)]
Adding a couple of queueing types.

10 years agofix the listings problem that latex2html has.
Eric Bohm [Fri, 10 Jun 2011 22:53:19 +0000 (17:53 -0500)]
fix the listings problem that latex2html has.
Spellcheck fix MSA entry.

10 years agoadd to nightly build test
Eric Bohm [Fri, 27 May 2011 18:53:06 +0000 (13:53 -0500)]
add to nightly build test

10 years agockfutures: now visible via charm++.h instead of charm.h
Ramprasad Venkataraman [Fri, 10 Jun 2011 21:47:17 +0000 (16:47 -0500)]
ckfutures: now visible via charm++.h instead of charm.h

charm.h is (was?) supposed to be include-able from C code. Although the
practical relevance of such a capability is anyone's guess. However, CkFutures
provides API that uses classes and was causing compilation issues in some
example programs. The declaration/definitions were also smeared across multiple
files obfuscating intents on where the API should be visible.

Fix by cleaning up ckfuture declaration locations and includes. This should fix
autobuild failures.

10 years agoAdd fwd declaration to make header self-sufficient
Ramprasad Venkataraman [Fri, 10 Jun 2011 16:11:10 +0000 (11:11 -0500)]
Add fwd declaration to make header self-sufficient

10 years agofile name change capability in bigsim read library
Ehsan Totoni [Fri, 10 Jun 2011 16:37:01 +0000 (11:37 -0500)]
file name change capability  in bigsim read library

10 years agofixed a bug in CkExit() with projections which has end_execute without matching begin...
Gengbin Zheng [Thu, 9 Jun 2011 21:26:44 +0000 (16:26 -0500)]
fixed a bug in CkExit() with projections which has end_execute without matching begin_execute.

10 years agoMerge branch 'ramv/cleanup-arridx-hierarchy' into charm
Ramprasad Venkataraman [Thu, 9 Jun 2011 19:25:16 +0000 (14:25 -0500)]
Merge branch 'ramv/cleanup-arridx-hierarchy' into charm

Clarify array index class hierarchy:
  CkArrayIndexBase       : union-friendly base class. holds all index data
  CkArrayIndex           : child class with all functionality. used everywhere in runtime code
  CkArrayIndexND (N=1..6): user visible grandchildren. empty types with just a constructor.

The ND variants are no longer strictly needed. charm and client code could
simply start using CkArrayIndex as the only index type. This will noticeably
simplify generated code in the decl/def etc.  On the other hand, ND classes in
the generated code may provide some type-based error checking at compile time.
For eg, it will not be possible to call an entry method with the wrong index
type etc.

10 years agockarrayindex: Move to new header
Ramprasad Venkataraman [Wed, 8 Jun 2011 21:16:13 +0000 (16:16 -0500)]
ckarrayindex: Move to new header

10 years agoupdated
Gengbin Zheng [Wed, 8 Jun 2011 15:59:33 +0000 (10:59 -0500)]
updated

10 years agoarray index: All indices (1D..6D) now use base class init()
Ramprasad Venkataraman [Tue, 7 Jun 2011 17:01:19 +0000 (12:01 -0500)]
array index: All indices (1D..6D) now use base class init()

10 years agoarray index: Provide protected init methods for array indices
Ramprasad Venkataraman [Tue, 7 Jun 2011 17:00:18 +0000 (12:00 -0500)]
array index: Provide protected init methods for array indices

10 years agoarray index: Introduce a union-friendly base class for ckarrayindex
Ramprasad Venkataraman [Thu, 2 Jun 2011 22:05:37 +0000 (17:05 -0500)]
array index: Introduce a union-friendly base class for ckarrayindex

There are only three entities in the charm codebase that put an array
index within a union:
    - envelope
    - callbacks
    - ckobjid

Create a base class that is used just by these three entities. CkArrayIndex is
now a safe and usable index class as it sports constructors, initialization
etc.  The remainder of the codebase and charm applications can and should use
the CkArrayIndex class.

Just at locations where the index objects in unions are handled, we provide a
somewhat "risky" cast operation from base to derived. This should be less
insane than casting across unrelated classes, which was the root of the
perennial problems with the original CkArrayIndex, IndexMax and IndexStruct
setup. However, we should actively explore mechanisms whereby even such casts
can be avoided without too much time/space costs.

10 years agockarrayindex: Delete child class copy assignment operators
Ramprasad Venkataraman [Wed, 1 Jun 2011 22:21:37 +0000 (17:21 -0500)]
ckarrayindex: Delete child class copy assignment operators

Since the index object is now as good as POD, allow the compiler
to generate its default copy assignment operator.

10 years agoUpdate docs to note that CkArrayIndexMax should be deprecated and removed
Ramprasad Venkataraman [Wed, 1 Jun 2011 20:50:33 +0000 (15:50 -0500)]
Update docs to note that CkArrayIndexMax should be deprecated and removed

10 years agoReplace all occurrences of CkArrayIndexMax with CkArrayIndex
Ramprasad Venkataraman [Wed, 1 Jun 2011 19:53:14 +0000 (14:53 -0500)]
Replace all occurrences of CkArrayIndexMax with CkArrayIndex

Leave a typedef behind so that applications that use CkArrayIndexMax can still
take their time to stop using it

10 years agockarrayindex: pretty indent and code docs
Ramprasad Venkataraman [Wed, 1 Jun 2011 18:15:44 +0000 (13:15 -0500)]
ckarrayindex: pretty indent and code docs

10 years agockarrayindex: Return combined count only for dense (upto 3D) arrays
Ramprasad Venkataraman [Wed, 1 Jun 2011 17:58:30 +0000 (12:58 -0500)]
ckarrayindex: Return combined count only for dense (upto 3D) arrays

Code for the creation of dense multi-dimensional chare arrays uses an array
index object to store the dimensions of the array.

The index object sports a funny method introduced to support such use. Ensure
that it returns a count on the number of array elements only for arrays upto 3D
(which are created dense by default).

10 years agockarrayindex: Always use operator== for comparisons
Ramprasad Venkataraman [Wed, 1 Jun 2011 17:42:01 +0000 (12:42 -0500)]
ckarrayindex: Always use operator== for comparisons

Also consolidate member function definitions to single location

10 years agockarrayindex: Delete copyFrom method thats no longer needed
Ramprasad Venkataraman [Wed, 1 Jun 2011 15:39:44 +0000 (10:39 -0500)]
ckarrayindex: Delete copyFrom method thats no longer needed

10 years agockarrayindexmax: Delete class and make the name synonymous with base class
Ramprasad Venkataraman [Tue, 31 May 2011 21:25:19 +0000 (16:25 -0500)]
ckarrayindexmax: Delete class and make the name synonymous with base class

CkArrayIndexMax existed because charm permitted array index objects of
different sizes and the Max object could hold the largest of them.  However,
recently charm has moved to support only a single-sized array index object and
hence the Max class has become vestigial.

Delete the class, and typedef CkArrayIndexMax to be synonymous with CkArrayIndex.
Code should eventually be cleaned to replace all references to Max, but the
typedef exists until then.

10 years agockarrayindexmax: Move trivial print methods to base class
Ramprasad Venkataraman [Tue, 31 May 2011 20:37:46 +0000 (15:37 -0500)]
ckarrayindexmax: Move trivial print methods to base class

10 years agockarrayindexmax: Move method getProjectionID() to base class
Ramprasad Venkataraman [Tue, 31 May 2011 20:12:45 +0000 (15:12 -0500)]
ckarrayindexmax: Move method getProjectionID() to base class

10 years agockarrayindexmax: Defer to base class pup
Ramprasad Venkataraman [Tue, 31 May 2011 19:46:45 +0000 (14:46 -0500)]
ckarrayindexmax: Defer to base class pup

10 years agockarrayindexmax: Move methods to base class
Ramprasad Venkataraman [Tue, 31 May 2011 18:51:15 +0000 (13:51 -0500)]
ckarrayindexmax: Move methods to base class

copyFrom() and operator== can belong in the base class

10 years agoremoved --dirty option as it is not available in git on Jaguar
Chao Mei [Wed, 8 Jun 2011 05:33:28 +0000 (00:33 -0500)]
removed --dirty option as it is not available in git on Jaguar

10 years agoMerge branch 'charm' of charmgit:charm into charm
Gengbin Zheng [Tue, 7 Jun 2011 17:04:35 +0000 (12:04 -0500)]
Merge branch 'charm' of charmgit:charm into charm

10 years agofixed for VC++
Gengbin Zheng [Tue, 7 Jun 2011 17:04:11 +0000 (12:04 -0500)]
fixed for VC++

10 years agoFixed a bug in passing argv for machine-specific initialization procedures.
Chao Mei [Tue, 7 Jun 2011 16:54:59 +0000 (11:54 -0500)]
Fixed a bug in passing argv for machine-specific initialization procedures.
This bug is exposed on MPI layer on "respect" which has MPI 1.2 version
installed. But on newer machines which has newer MPI version, say 2.1, then
this bug does not show up. So I think this bug is really a compatibility
issue with older MPI version.

10 years agoadded --always to --describe so that it works for branch checkout like in autobuild.
Gengbin Zheng [Tue, 7 Jun 2011 16:46:47 +0000 (11:46 -0500)]
added --always to --describe so that it works for branch checkout like in autobuild.

10 years agoremoved optimized-multicast macro definition as it is limited according to the report...
Chao Mei [Tue, 7 Jun 2011 14:50:12 +0000 (09:50 -0500)]
removed optimized-multicast macro definition as it is limited according to the report by Sameer Kumar

10 years agofixed the compiling errors reported by gnu compiler on Intrepid
Chao Mei [Tue, 7 Jun 2011 14:34:50 +0000 (09:34 -0500)]
fixed the compiling errors reported by gnu compiler on Intrepid

10 years agofixed a bug on multiple processors because variable is not initialized
YanhuaSun [Mon, 6 Jun 2011 04:53:58 +0000 (23:53 -0500)]
fixed a bug on multiple processors because variable is not initialized

10 years agoTSP problem fixed
YanhuaSun [Mon, 6 Jun 2011 02:21:17 +0000 (21:21 -0500)]
TSP problem fixed

10 years agoError check for inputs and some minor bugs
nikhil [Wed, 1 Jun 2011 19:39:41 +0000 (14:39 -0500)]
Error check for inputs and some minor bugs

10 years agoFix translate function
nikhil [Wed, 1 Jun 2011 19:27:50 +0000 (14:27 -0500)]
Fix translate function

10 years agofixed link time error caused by using_printCS in converse program.
Gengbin Zheng [Wed, 1 Jun 2011 06:03:51 +0000 (01:03 -0500)]
fixed link time error caused by using_printCS in converse program.

10 years agockarrayindex: Remove all occurrences of asMax()
Ramprasad Venkataraman [Tue, 31 May 2011 22:17:58 +0000 (17:17 -0500)]
ckarrayindex: Remove all occurrences of asMax()

10 years agodisable optimized multicast in bluegenep as it not necessarily gives best performance...
Chao Mei [Tue, 31 May 2011 20:31:05 +0000 (15:31 -0500)]
disable optimized multicast in bluegenep as it not necessarily gives best performance according to Sameer Kumar, and not work in all input cases

10 years agoMerge branch 'ramv/exorcise-arrayindex-bug' into charm
Ramprasad Venkataraman [Tue, 31 May 2011 16:00:45 +0000 (11:00 -0500)]
Merge branch 'ramv/exorcise-arrayindex-bug' into charm

Eliminate CkArrayIndexStruct and consolidate array index data
structures into one class hierarchy. The base class CkArrayIndex
now participates in unions and has shed its constructors.

10 years agockarrayindex: Add instantiation warning to code docs
Ramprasad Venkataraman [Tue, 31 May 2011 15:39:58 +0000 (10:39 -0500)]
ckarrayindex: Add instantiation warning to code docs

10 years agoRemove all traces of CkArrayIndexStruct. Avoid dangerous pointer casts
Ramprasad Venkataraman [Tue, 26 Apr 2011 18:05:01 +0000 (13:05 -0500)]
Remove all traces of CkArrayIndexStruct. Avoid dangerous pointer casts

10 years agockobjid: Use CkArrayIndex instead of CkArrayIndexStruct
Ramprasad Venkataraman [Tue, 26 Apr 2011 17:56:12 +0000 (12:56 -0500)]
ckobjid: Use CkArrayIndex instead of CkArrayIndexStruct

Needs further testing from fault tolerance subgroup

10 years agocharmdebug: Use CkArrayIndex instead of CkArrayIndexStruct
Ramprasad Venkataraman [Tue, 26 Apr 2011 17:26:36 +0000 (12:26 -0500)]
charmdebug: Use CkArrayIndex instead of CkArrayIndexStruct

10 years agoenvelope: For array indices store a base class (CkArrayIndex) object and not CkArrayI...
Ramprasad Venkataraman [Tue, 26 Apr 2011 17:12:48 +0000 (12:12 -0500)]
envelope: For array indices store a base class (CkArrayIndex) object and not CkArrayIndexStruct

This is a big step in the effort to get rid of CkArrayIndexStruct and all the dangerous
pointer casts that accompany it

10 years agockcallback: Use a CkArrayIndex base class object instead of CkArrayIndexStruct
Ramprasad Venkataraman [Fri, 22 Apr 2011 20:15:21 +0000 (15:15 -0500)]
ckcallback: Use a CkArrayIndex base class object instead of CkArrayIndexStruct

CkCallback has a union of data based on the different callback types. One of
these if for callbacks that go to array elements. This required storing an
array index in the union with the resulting stipulation that the stored index
object should not have any default constructors.  A CkArrayIndexStruct class
has existed in the codebase expressly for such situations. Because this class
has no relation to the hierarchy of 'normal' array index classes, dangerous
pointer casts have been the norm.

Because of such risky casts, this snippet of code in the callback has been
plagued with heisenbugs over many years, as compilers get stricter and exhibit
greater personality variations across brands and architectures. OpenAtom has
exposed the bugs a couple of times in the last couple of years.

To fix this issue for good (hopefully), the objective is to not cast across
unrelated class hierarchies, but rather make the base class amenable to
participation in a union.  This should help get rid of classes that existed
purely for such silly hacks to enable charm functionality.

10 years agockarrayindex: Remove all base class constructors
Ramprasad Venkataraman [Fri, 22 Apr 2011 20:00:01 +0000 (15:00 -0500)]
ckarrayindex: Remove all base class constructors

Putting an object in a union requires that the class not have any constructors.
Removing all base class constructors will mean that the array index base class
can now participate in a union; thus removing all purpose for the existence of
a CkArrayIndexStruct.

10 years agoadded a declearation
Yanhua Sun [Sun, 29 May 2011 10:48:53 +0000 (02:48 -0800)]
added a declearation

10 years agoweird mistake. two lines are wrongly deleted. added them now
Yanhua Sun [Sun, 29 May 2011 10:41:42 +0000 (02:41 -0800)]
weird mistake. two lines are wrongly deleted. added them now

10 years agoadd a function which is called before program exit to output statistic
Yanhua Sun [Sun, 29 May 2011 10:25:48 +0000 (02:25 -0800)]
add a function which is called before program exit to output statistic
information

10 years agoMulticast component to ckgraph
Nikhil Jain [Fri, 27 May 2011 22:02:28 +0000 (17:02 -0500)]
Multicast component to ckgraph

10 years agoEdits to xlc compiler options.
Sameer Kumar [Fri, 27 May 2011 17:42:52 +0000 (12:42 -0500)]
Edits to xlc compiler options.

10 years agoPOSE: Match printf/scanf formats to arguments to silence warnings
Phil Miller [Fri, 27 May 2011 17:42:16 +0000 (12:42 -0500)]
POSE: Match printf/scanf formats to arguments to silence warnings

10 years agoreset group dep field when reset envelope for recycling.
Gengbin Zheng [Fri, 27 May 2011 17:20:23 +0000 (12:20 -0500)]
reset group dep field when reset envelope for recycling.

10 years agoCharacterize smp bug by clarifying Eric's test to create single chare array
Ramprasad Venkataraman [Fri, 27 May 2011 16:44:13 +0000 (11:44 -0500)]
Characterize smp bug by clarifying Eric's test to create single chare array

The hang bug in smp mode happens on lab machines even with a single chare array.
Hence create just one in the test

10 years agonew test to demonstrate a race condition
Eric Bohm [Fri, 27 May 2011 03:10:45 +0000 (22:10 -0500)]
new test to demonstrate a race condition

10 years agoTests megacon: Match int sizes to pointer parameters to quiet warnings
Phil Miller [Thu, 26 May 2011 23:23:55 +0000 (18:23 -0500)]
Tests megacon: Match int sizes to pointer parameters to quiet warnings

10 years agoTests: Silence a type conversion warning
Phil Miller [Thu, 26 May 2011 23:19:41 +0000 (18:19 -0500)]
Tests: Silence a type conversion warning

10 years agoPOSE stats: De-duplicate printing code
Phil Miller [Thu, 26 May 2011 23:19:13 +0000 (18:19 -0500)]
POSE stats: De-duplicate printing code

10 years agoEnable Charm++/Converse build identification from program output
Phil Miller [Wed, 25 May 2011 23:07:02 +0000 (18:07 -0500)]
Enable Charm++/Converse build identification from program output

When making anything in the base build directory, check whether any
code has been changed or commited, and generate a version string
accordingly. Embed this string in a standalone object linked into
converse. Print the string at startup.

10 years agoMerge remote branch 'origin/machine-common-codes' into charm
Phil Miller [Thu, 26 May 2011 20:16:38 +0000 (15:16 -0500)]
Merge remote branch 'origin/machine-common-codes' into charm

10 years ago1. make the dynamic cap parameter configurable by reading from command line
Chao Mei [Wed, 25 May 2011 22:18:59 +0000 (17:18 -0500)]
1. make the dynamic cap parameter configurable by reading from command line

2. simplified the send op for both non-SMP and SMP that they share the same
   code of calling MPI_Isend function

10 years agosmall change to typed reductions
Pritish Jetley [Wed, 25 May 2011 20:23:30 +0000 (15:23 -0500)]
small change to typed reductions

10 years agoremoved commented section in 3.16; it seemed to be causing a blank code box to be...
Pritish Jetley [Wed, 25 May 2011 19:09:09 +0000 (14:09 -0500)]
removed commented section in 3.16; it seemed to be causing a blank code box to be rendered.

10 years agochanges to sections 3.14,3.15 and 3.16 of Charm++ manual
Pritish Jetley [Wed, 25 May 2011 19:03:33 +0000 (14:03 -0500)]
changes to sections 3.14,3.15 and 3.16 of Charm++ manual

10 years agodisable MPI_POST_RECV in default
Chao Mei [Mon, 23 May 2011 21:40:45 +0000 (16:40 -0500)]
disable MPI_POST_RECV in default

10 years agoadded printing out the description of the error number when creating the sts file
Chao Mei [Mon, 23 May 2011 21:31:37 +0000 (16:31 -0500)]
added printing out the description of the error number when creating the sts file

10 years agomake the parsing of #pes and #ppns more accurate as numbers should follow +p or
Chao Mei [Mon, 23 May 2011 19:35:28 +0000 (14:35 -0500)]
make the parsing of #pes and #ppns more accurate as numbers should follow +p or
ppn