Charm++

   Copyright (C) 1989-2013 Regents of the University of Illinois

INTRODUCTION

Charmexp is a fork of Charm++. It contains a branch named charmexp with a number of changes against charm-6.6.1 branch including fixup for GNU Autotools and backtrace demangling. This software includes code developed by the Parallel Programming Laboratory in the Department of Computer Science at the University of Illinois at Urbana-Champaign.

Charm++ is a message-passing parallel language and runtime system. It is implemented as a set of libraries for C++, is efficient, and is portable to a wide variety of parallel machines. Source code is provided, and non-commercial use is free.

GETTING THE LATEST CHARM SOURCE

You can use anonymous Git access to obtain the latest Charm++ source code, as follows:

git clone http://charm.cs.illinois.edu/gerrit/charm

In order to get charmexp branch do the following:

cd charm
git remote add github https://github.com/shvorin/charmexp
git fetch github
git checkout charmexp

PICKING A VERSION

First, you need to decide which version of charm++ to use. The "build" script in charm source directory takes several command line options to compile Charm++. The command line syntax is:

build [options ...] [--basedir=dir] [--libdir=dir] [--incdir=dir] [charmc-options ...]

For detailed help messages, pass -h or --help to the build script, i.e. ./build --help

REQUIRED:

specifies the parts of Charm++ to compile. The most often used is "charm++", which will compile the key Charm++ executables and runtime libraries. Other common targets are "AMPI" and "FEM". defines the CPU, OS and Communication layer of the machines. See "How to choose a " below for details.

OPTIONAL:

defines more detailed information of the compilations, including compilers, features to support, etc. See "How to choose " below. [--libdir=dir] specify additional lib paths for building Charm++. [--incdir=dir] specify additional include paths for building Charm++. [--basedir=dir] a shortcut to specify additional include and lib paths for building Charm++, the include path is dir/include and lib path is dir/lib.

Running build script, a directory of the name of combination of version and options like "---..." will be created and the build script will compile Charm++ under this directory.

For example, on an ordinary Linux PC:

./build charm++ net-linux-x86_64

will build charm++ in the directory: net-linux-x86_64/. The communication defaults to UDP packets and the compiler to gcc.

For a more complex example, on a Scyld workstation with the Intel C++ compiler, where you want the communication to happen over TCP sockets:

./build charm++ net-linux scyld icc tcp

will build charm++ in the directory: net-linux-scyld-tcp-icc/.

You can specify multiple options, however you can use at most one compiler option. The sequence of the options are not important given in build script, only one directory name will be generated, following the rules:

1. compiler option will be at the end;
2. other options are sorted alphabetically.

**** How to choose a ****

Here is the table for choosing correct version. The default setting of compiler in Charm version is gcc/g++. However, one can use to specify other compilers. See the detailed explanation of the below. (Note: this isn't a complete list. Run ./build for a complete listing)

Charm version OS Communication Default Compiler

---

net-linux-x86_64 Opteron Linux UDP GNU compiler net-darwin-x86_64 MacOS X UDP GNU C++ compiler net-linux Linux (32 bit) UDP/Myrinet GNU compiler net-sol Solaris UDP GNU compiler net-win32 Win32 UDP MS Visual C++ net-cygwin Win32/cygwin UDP GNU compiler mpi-sp IBM A/IX MPI A/IX xlC Compiler net-linux-ia64 IA64 Linux UDP/Myrinet GNU compiler mpi-linux PC Linux MPI GNU compiler mpi-darwin-x86_64 MacOS X MPI GNU C++ compiler mpi-linux-ia64 IA64 Linux MPI GNU compiler mpi-linux-x86_64 Opteron Linux MPI GNU compiler

To choose , your choice is determined by three options:

1.) The way a parallel program written in Charm++ will communicate:

"net-" Charm++ communicates using the regular TCP/IP stack

(UDP packets), which works everywhere but is fairly slow. Use this option for networks of workstations, clusters, or single-machine development and testing.

"gni-", "bluegene[lp]-", "pami-bluegeneq-" Charm++

communicates using direct calls to the machine's communication primitives.

"mpi-" Charm++ communicates using MPI calls.  Use this for

machines with a good MPI implementation (such as the Origin 2000).

"multicore-" Charm++ communicates using shared memory within a
single node

"sim-" and "uth-" are not actively maintained.  These are

single-processor versions: "uth-" simulates processors as user-level threads; "sim-" switches between processors and counts communications.

2.) Your operating system:

"linux"       Linux
"win{32,64}"  MS Windows with MS Visual C++ compiler (32/64-bit, resp.)
"cygwin"      MS Windows with Cygnus' Cygwin Unix layer
"darwin"      Apple Mac OS X
"sol"         Solaris
"aix"         IBM A/IX

3.) Some operating systems have other architecture options, such as:

"-x86"     For Solaris and Mac OS X, target x86 hardware (instead of
           SPARC or PPC).
"-ppc"     POWER/PowerPC
"-mips64"  MIPS, such as for SiCortex systems
    "-ia64"    Use Itanium(tm) IA-64 instructions (instead of x86).
    "-x86_64"  Use AMD64/EM64T 64-bit x86 instructions (instead of 32 bit).
"-cell"    Sony/Toshiba/IBM Cell PPE (e.g. Playstation 3,
           Mercury blades, Roadrunner)

Your Charm++ version is made by concatenating the options, e.g.:

"net-linux-x86_64" Charm++ for a network of 64-bit Linux workstations, compiled using g++.

"gni-crayxc" Charm++ for Cray XC30 systems using the system's compiler.

**** How to choose ****

above defines the most important OS, CPU and Communication of your machine, and most of time, it use the GNU gcc as default compiler. To use different compiler or demand additional special feature support, you need to choose from the following list:

• gcc3 - GNU GCC/G++ version 3

• acc - HP aC++ compiler

• cc - For Sun WorkShop C++ compilers;

• cc64 - For 64 bits Sun WorkShop C++ or IBM xlC compilers;

• cxx - DIGITAL C++ compiler;

• kcc - KAI C++ compiler;

• pgcc - Portland Group's C++ compiler;

• icc - Intel C/C++ compiler for Linux IA32

• ecc - Intel C/C++ compiler for Linux IA64

• mpcc - SUN Solaris C++ compiler for MPI.

• scyld - support Beowulf Scyld based on bproc;

• clustermatic - for Clustermatic Beowulf cluster based on bproc;

• gm - support MyriCom's Myrinet GM library;

• vmi - support NCSA's VMI library;

• tcp - for net- version, default communication is via UDP. Using option tcp will switch to TCP. TCP version of CHarm++ is usually slower than UDP, but it is more reliable.

• smp - Enable direct SMP support. An "smp" version communicates using shared memory within a machine; but normal message passing across machines. Because of locking, "smp" may slightly impact non-SMP performance. Try your application to decide if enabling smp mode improves performance.

• bigsim - compile Charm++ as running on the BigSim emulator.

• help - show supported options for a version. For example, for net-linux, running: > ./build charm++ net-linux help will give: supported options: gcc3 gm icc kcc pgcc scyld smp bluegene tcp

BUILDING THE SOURCE

If you have downloaded a binary version of Charm++, you can skip this step-- Charm++ should already be compiled. For win32/win64 systems, see README.win; for Cygwin version, see README.cygwin; for net- version, see README.net.

Once you have decided on a version, unpack Charm++, cd into charm, and run

 > ./build <target> <version> <opts>

is one of "charm++" The basic Charm++ language. "AMPI" An implementation of MPI on top of Charm++ "FEM" A Finite-Element framework on top of Charm++ "Tau" TAU's performance profiling/tracing

is described above

are build-time options (such as the compiler or "smp"), or command line options passed to the charmc compile script. Common compile time options such as -g, -O, -Ipath, -Lpath, -llib are accepted.

For example, on a Linux machine, you would run > ./build charm++ net-linux-x86_64 -O

This will construct a net-linux-x86_64 directory, link over all the Charm++ source code into net-linux-x86_64/tmp, build the entire Charm++ runtime system in net-linux-x86_64/tmp, and link example programs into net-linux-x86_64/examples.

Charm++ can be compiled with several optional features enabled or disabled. These include runtime error checking, tracing, interactive debugging, deterministic record-replay, and more. They can be controlled by passing flags of the form --enable-featurename or --disable-featurename to the build command: > ./build charm++ net-linux-x86_64 --disable-tracing

The list of optional features available is shown in the output of > ./build --help

Production optimizations: Pass the configure option --with-production to ./build to turn on optimizations in Charm++/Converse. This disables most of the run-time checking performed by Converse and Charm++ runtime. This option should be used only after the program has been debugged. Also, this option disables Converse/Charm++ tracing mechanisms such as projections and summary.

When Charm++ is built successfully, the diretory structure under the target directory will look like:

net-linux-x86_64/ | --- bin/ # all executables | --- doc/ # documentations | --- include/ # header files | --- lib/ # libraries | --- lib_so/ # dynamic libraries | --- examples/ # all example programs | --- tests/ # all test programs | --- tmp/ # Charm++ build directory

BUILDING A PROGRAM

To make a sample program, cd into examples/charm++/queens/. This program solves the N-queens problem-- find how many ways there are to arrange N queens on an NxN chess board such that none may attack another.

To build the program, type make. You should get an executable named "pgm".

RUNNING A PROGRAM

Following the previous example, to run the program on two processors, type

 > ./charmrun ./pgm 12 6 +p2

This should run for a few seconds, and print out: There are 14200 Solutions to 12 queens. Finish time=4.030000

Charmrun is used to provide a uniform interface to run charm programs. On some platforms, charmrun is just a shell script which calls the platform-specific start program, such as mpirun on mpi versions.

For net- version, charmrun is an executable which invokes rsh or ssh to start node programs on remote machines. You should set up a ~/.nodelist that enumerates all the machines you want to run jobs on, otherwise it will create a default ~/.nodelist for you that contains only localhost. Here is a typical .nodelist file:

group main ++shell /bin/ssh host

The default remote shell program is rsh, but you can define different remote shell you like to start remote processes in the ++shell option. You should also make sure that you can rsh or ssh to these machines without password authentication. Just type following command to verify: > rsh date If this gives you current date immediately, your running environment with this node has been setup correctly.

Now, for test running purpose, net- version charmrun comes with an easy-to-use "++local" options. No remote shell invocation is needed in this case. It starts node programs right on your local machine. This could be useful if you just want to run program on only one machine, for example, your laptop. This can save you all the hassle of setting up rsh/ssh or charmd daemons. To use this option, just type:

 > ./charmrun ++local ./pgm 12 100 +p2

However, for best performance, you should launch one node program per processor.

Build Charm++ in Dynamic libraries

In order to compile Charm++ into dynamic libraries, one need to specify "-build-shared" option as one of the Charm compiler script "charmc" at link time. For example, to compile Charm++ under net-linux/tmp, run

make charm++ OPTS='-O -build-shared'

Charm++'s dynamic libraries are compiled into lib_so/ directory. Typically, they are with ".so" suffix.

Note, "-build-shared" option is automatically turned on when building Charm++ using "build" script. So you don't need to pass "-build-shared" option to "build".

One can compile a Charm++ applicaiton linking against Charm++ dynamic libraries, run charmc with "-charm-shared" as one of the link options. For example:

charmc -o pgm pgm.o -charm-shared

You can then run the program as usual. Note, linking against Charm++ dynamic libraries produces much smaller size binaries and takes much less linking time.

FOR MORE INFORMATION

The Charm++ web page, with documentation, more programs, and the latest version of Charm++, is at http://charm.cs.illinois.edu/

The Charm++ mailing list, for questions, comments, suggestions, improvements, or bug reports is charm@cs.illinois.edu

AUTHORS

Charm++ was created and is maintained by the Parallel Programming Lab, in the Computer Science department at the University of Illinois at Urbana-Champaign. Our managing professor is Dr. L.V. Kale; students have included (in rough time order) Wennie Shu, Kevin Nomura, Wayne Fenton, Balkrishna Ramkumar, Vikram Saletore, Amitabh B. Sinha, Manish Gupta, Attila Gursoy, Balkrishna Ramkumar, Amitabh B. Sinha, Nimish Shah, Sanjeev Krishnan, Jayant DeSouza, Parthasarathy Ramachandran, Jeff Wright, Michael Lang, Jackie Wang, Fang Hu, Michael Denardo, Joshua Yelon, Narain Jagathesan, Zehra Sura, Krishnan Varadarajan, Sameer Paranjpye, Milind Bhandarkar, Robert Brunner and Jayant Desouza. Current developers include Terry Wilmarth, Gengbin Zheng, Orion Lawlor, Karthik Mahesh, and Neelam Saboo.

Charmexp fork was started by Artem Shvorin art@shvorin.net.