MALOC (Minimal Abstraction Layer for Object-oriented C) is a small, portable, abstract C environment library for object-oriented C programming. MALOC is used as the foundation layer for a number of scientific applications, including MC, SG, and APBS. MALOC can be used as a small stand-alone abstraction environment for writing portable C programs which need access to resources which are typically architecture-dependent, such as INET sockets, timing routines, and so on. MALOC provides abstract datatypes, memory management routines, timing routines, machine epsilon, access to UNIX and INET sockets, MPI, and so on. All things that can vary from one architecture to another are abstracted out of an application code and placed in MALOC. To port the application code to a new architecture, only the small MALOC library needs to be ported (usually just "./configure ; make"). MALOC takes the pain of varying UNIX (and Win32) platforms with differing library and header layous completely out of the software development picture.
MALOC is a class library written in Clean OO C. "Clean" refers to the fact that the language is both legal C++ and legal ANSI/ISO/Standard C, and can be compiled with any standard C or C++ compiler. "OO" refers to the programming style employed -- object-oriented. An Clean OO C implementation consists of a set of "Objects" (Clean C structs) which are operated on by a collection of "methods" (Clean C subroutines) which always have a pointer to the Object as their first argument. This special argument is always written as "thee", analogous to the implicit "this" pointer in C++. (An Clean OO C implementation can be turned into a C++ implementation with a simple AWK/SED or Perl Script.) As a result of this Clean OO C implementation, MALOC can be used as a set of C++-like class libraries, it can be safely software engineered into other large software packages, and it can be built on just about any UNIX-like platform with either a C or a C++ compiler, including e.g. Linux, IRIX, and Win32. To use some of the graphics and parallel computing features, your platform must also have some form of standard INET sockets (WINSOCK will work). MALOC is easily buildable from source on any UNIX-like system, and uses a GNU autoconf build environment.
Detailed information about MALOC can be found in the User's Guide and Programmers's Guide.
While MALOC is itself a self-contained software package, it is one of several components of FETK (the Finite Element ToolKit). FETK consists of the following components written in Clean OO C:
MALOC is self-contained, and requires only an ANSI-C compiler on a UNIX or Win32 platform. PUNC, SG, and MC are also self-contained, but rely on MALOC having been previously installed on the platform. Additional features of MC are enabled if PUNC is available, but PUNC is not required to build MC. The MC eXtension libraries MCX are constructed on top of MALOC and MC, and in order install and use MCX one must first correctly configure and install both MALOC and MC. MCX is made up of a number of individual libraries developed by members of our group, or contributed by one of a number of colleagues. More information about FETK can be found on the FETK website:
MALOC is copyrighted, but is redistributable in source and binary form under the following license. The MALOC source and binary code can be downloaded from the following locations:
What is MALOC and what does MALOC stand for?
See the Overview.
How to I obtain a copy of the MALOC binaries and/or source code?
See the Obtaining MALOC section.
I managed to get a copy of "maloc-VERSION.[i386|src|tar].[rpm|gz]"; how do I now install MALOC?
See the Installation Instructions.
You gave me a "patch.gz" file to fix a bug in MALOC; how do I apply the patch?
To apply patches to upgrade MALOC to a new version, you first obtain the patch from me or my webpage as a single file with a name like "patch.gz". You apply the patch after you have unpacked the maloc-VERSION.tar.gz file as described in the installation instructions. To apply the patch, cd into the directory containing the root MALOC directory (called "maloc" after unpacking maloc-VERSION.tar.gz) and execute the "patch" program as follows (the patch program exists on most UNIX machines):
Patch files are simply the output from a recursive "diff" that are used to represent all differences between two directory trees. For example, to create a patch representing the changes from version 1.0 of MALOC (in directory maloc_1.0 for example) to version 1.1 of MALOC (e.g. in directory maloc_1.1), I would normally type the following:
I really don't know what I'm doing; how to I get more documentation for MALOC?
The User's Guide and the Programmers's Guide contain all of the MALOC documentation.
Why did you develop MALOC? Can't you just use ANSI-C?
Anyone who has ported C codes that use sockets and signals to various UNIX machines quickly realizes that there has to be a better way to do things. GNU Autoconf solves this problem, but also then writing a small C library which holds all of the autoconf-generated machine-dependent details internally simplifies software development a great deal. MALOC plays this role in MC as well as in the entire FETK project.
What is in all of these subdirectories? Where exactly is "MALOC"?
MALOC consists of several (class) libraries from which you will call routines to handle your application. You simply link your application to the MALOC static or shared library. The MC package automatically looks for the MALOC libraries as part of its Autoconf configuration process.
maloc
|
------------------------
/ | | | \
config doc examples src tools
The src directory has the additional subdirectory structure:
src
|
-------------------------
/ | | | | \
aaa_inc aaa_lib base psh vsh vsys
Within each library source directory is an additional subdirectory,
"maloc". The "maloc" subdirectory contains public headers for the library,
representing the library API; these headers will be installed in the
specified header install directory during the install procedure after
building MALOC.
The following is a brief description of each subdirectory of the package.
maloc - The entire MALOC package
maloc/config - GNU Autoconf scripts and non-unix config files
maloc/doc - MALOC documentation
maloc/examples - Complex examples and data files for using MALOC
maloc/src - MALOC source code (all source and headers)
maloc/src/aaa_inc - Header installation tools
maloc/src/aaa_lib - Library installation tools
maloc/src/*/maloc - The MALOC headers (API)
maloc/src/base - Source for M. Holst's BASE (MALOC foundation headers)
maloc/src/vsys - Source for M. Holst's VSYS (Virtual SYStem layer)
maloc/src/vsh - Source for M. Holst's VSH (Virtual command SHell)
maloc/src/psh - Source for M. Holst's PSH (Parallel vSH extension)
maloc/tools - Some binary tools for use with MALOC
Okay, I seem to have installed MALOC correctly; how do I actually use it now?
Using MALOC is pretty simple; it is a very object-oriented implementation, although it is written in C. It is actually written in an object-oriented way using a subset of ANSI/Standard C, sometimes referred to as Clean OO C. Clean C refers to the overlapping subset of ANSI/Standard C and C++, so you can compile the code as a legal C++ or ANSI/Standard C code. Using the code consists of constructing objects (represented by C structs) and manipulating these objects using appropriate methods (represented by C functions which follow a certain object-oriented prototype convention).
What is the class hierarchy? How are the various libraries related?
Detailed information on the class relationships can be found in the Programmers's Guide. The following directed graph shows the class library dependencies. (This tends to evolve as MALOC is developed.)
ANSI-C ===> base ==> vsys ==> vsh ==> psh
/\ /\ /\
|| || ||
UNIX/INET sockets || MPI
||
signals
Clean C is simply a subset of ISO C which is compatible with (the current draft) standard C++; i.e., Clean C is the intersection of the standard accepted definitions of C and C++. ISO C is currently the internationally accepted version of C; ISO C was previously called ANSI C, and the term Standard C (as opposed to Tradiational or K&R C) is also now used for ISO C. Clean C programs are ISO C programs that use only the new C++-style function prototypes, and avoid using names that are reserved words in C++ (such as class, public, etc). There are a few other differences between ISO C and C++ that one has to be careful of (such as struct/union scopes, typedef syntax, storage classes, const and cast usage, initializers, "sizeof" differences), but the prototypes and reserved words are the main restrictions. Clean C programs can then be compiled with any ISO C compiler or any standard C++ compiler on any architecture. In other words, Clean C programs are actually C++ programs, but they only use C++ features that can be handled with an ISO C compiler. Clean C is a very small, simple, fast, powerful, yet extremely portable language. The name "Clean C" for this C dialect seems to have been first coined in the book of Harbison and Steele, "C: A Reference Manual"; this excellent C reference has a thorough discussion of the various dialects of C.
What is Clean OO C?
The "OO" in "Clean OO C" refers to the programming style employed -- Object Oriented (it does not refer to syntax as did the word "Clean" above). An "Clean OO C" implementation consists of a set of "Objects" (Clean C structs) which are operated on by a collection of "methods" (Clean C subroutines) which always have a pointer to the Object as their first argument. This special argument is always written as "thee", analogous to the implicit "this" pointer in C++. (An Clean OO C implementation can be turned into a C++ implementation with a simple AWK/SED or Perl Script.) Much like C++, a careful Clean OO C implementation can be used as a set of C++-like class libraries, it can be safely software engineered into other large software packages, and it can be compiled on just about any UNIX-like platform with either a C or a C++ compiler, including e.g. Linux, IRIX, and Win32.
Why would someone program in Clean OO C?
Clean OO C seems to be a good language compromise: a very small language, but powerful enough to provide a great deal of abstraction. In addition, one can continue to use various integrated C++ development environments under Win32 (such as the Watcom and Microsoft products), making use of the symbolic debuggers and so on, since Clean OO C programs are still legal C++ programs. The book of A. Holub, "C+C++: Programming with Objects in C and C++" was a very useful guide for doing an object-oriented implementation in (ISO) C.
Avoiding use of the inline function in C++ is a difficult sacrifice to make in moving to Clean OO C (note that GCC actually provides the inline function, although most ISO C compilers do not). Inlining in Clean OO C can be emulated using macros. This can be made reasonably safe by implementing two versions of all macro'd routines -- a normal Clean OO C routine, and a macro version. Switching between the two can be as simple as setting or unsetting a macro in a makefile; this way if things start to go wrong, the macros can be turned into real functions with prototypes, and the code can be stepped through with a debugger. Another useful feature of C++, from a practical large-package software engineering point of view, is the "public/protected/private" keywords in class definitions. This functionality can be recovered to a small degree by careful use of header files and the "static" keyword in a Clean OO C implementation.
Why is MALOC written in Clean OO C rather than C++ or Objective-C?
A brief explanation can be found here.
I want to write software to extend MALOC; besides the Clean OO C framework, what is the "Coding Style"?
A fairly complete description of the coding style guidelines used in writing MALOC (and all of FETK ) can be found in the Programmers's Guide.
I wrote a cool MALOC extension; how do I get it included in the MALOC package?
To keep the core FETK libraries as small, efficient, stable, and bullet-proof as possible, I distribute extensions to any of MALOC, PUNC, MC, and/or SG in a package called MCX (MC eXtension Library). MCX is developed collaboratively by a number of collaborators. In addition, several application-specific software packages have been written such as GPDE (Geometric PDE) which use FETK to solve a number of different application problems. If you wish to contribute to MCX, and are willing to release you contributions under the GNU LGPL, then I would be delighted to include it in MCX.
If you need special modifications to FETK to support your particular contribution or your particular FETK -based application, let me know. If you provide me with a clear description of an interface to FETK that you need, then I will almost always consider making the modifications to FETK for you. In any event, the general guideline for such proposed changes is that they be fairly minimal to avoid breaking too many other things that rely on the existing structure of FETK , and general enough to be useful for other extensions (and in particular, not application-specific).
Available distribution formats
MALOC is distributed in both binary format (as a binary RPM file maloc-VERSION.i386.rpm for i386-based versions of Linux) and in source format (as a source RPM file maloc-VERSION.src.rpm and as a gzipped tar file "maloc-VERSION.tar.gz").
Installation using the binary RPM file
The following rpm command will install all of the MALOC headers and libraries into /usr/local/include and /usr/local/lib, and will install the MALOC documentation into /usr/share/doc/packages/maloc:
rpm -Uvh maloc-VERSION.i386.rpm
Installation and rebuilding from sources using the source RPM file
The following rpm command will unpack the source rpm file "maloc-VERSON.src.rpm" into the MALOC gzipped tar file containing the sources called "maloc-VERSION.tar.tar.gz" and into a small RPM spec file called "maloc-VERSON.spec":
rpm -Uvh maloc-VERSION.src.rpm
The sources can then be unpacked and built using the directions for
the gzipped tar file below.
Alternatively, the following rpm command will do these steps for you:
rpm -bp maloc-VERSION.spec
Rebuilding binary and source RPM files from the gzipped tar file
The MALOC sources contain the RPM spec file "maloc-VERSON.spec" in the root source directory; as a result, rebuilding the RPM files from sources can be done using the rpm command:
rpm -ta maloc-VERSION.tar.gz
The result will be the corresponding source and binary rpm files,
named "maloc-VERSON.src.rpm" and "maloc-VERSION.i386.rpm".
Normally, these files are written to /usr/src/redhat/SRPMS
and /usr/src/redhat/RPMS respectively, but you must be logged in
as root for these to work.
The destination directories can be overriden using arguments to the
rpm program (see the rpm manpage).
Installation and building from sources using the gzipped tar file
The following command will unpack MALOC into a number of subdirectories and files on any UNIX machine (and on any WinNT machine with the GNU-Win32 tools gzip and tar).
gzip -dc maloc-VERSION.tar.gz | tar xvf -
Building the package using the GNU "configure" shell script and "make"
The "configure" shell script in the "maloc" directory (the toplevel directory created when you unpacked the MALOC tar.gz file) will build the entire package. This is a standard GNU autoconf-generated configuration script. For a list of the possible configuration options, type:
./configure --help
You should be able to build MALOC by simply typing:
./configure
make
make install
However, it is often advantageous to keep the original source directory
pristine; the configure script can actually be run outside the source
tree, which will keep all files created by the build outside the source
tree. (This idea is related to the section below describing how to build
binaries for multiple architectures at the same time using the same source
tree, and requires that your version of make has the VPATH facility, such
as GNU make.) For example, I build MALOC in a separate directory from the
source tree as follows:
gzip -dc maloc-VERSION.tar.gz | tar xvf -
mkdir maloc_build
cd maloc_build
../maloc/configure
make
make install
Building binaries for multiple architectures in the same source directory
If you have a version of "make" that supports the VPATH facility (such as any recent version of GNU make), then you can build the package for multiple architectures in the same source directory (in fact, you can do the compiles at the same time without collisions). This is very useful if you have your home directory on an NFS volume that you share among multiple architectures, such as SGI, Linux, etc. To build MALOC for all the systems at the same time, you simply make an additional subdirectory in the main MALOC directory for each architecture, copy "configure" into it, "cd" into the subdirectory, and then install as above. For example, on a linux machine you would do the following:
mkdir linux
cp configure linux/.
cd linux
./configure
make
make install
If you mount the same NFS home directory on for example an OpenStep box,
you could at the same time do the following:
mkdir next
cp configure next/.
cd next
./configure
make
make install
Again, both builds can actually be done outside the source tree rather
than in a subdirectory of the source tree, as described in the previous
section.
Building shared libraries rather than static libraries
(MIKE: give an overview of libtool.)
Rebuilding the configure script and the Makefile.in files
If for some reason you actually need to rebuild the configure script or the Makefile.in files using the GNU autoconf suite, you should read the block of documentation at the top of the configure.in file. The commentary I put there explains exactly how the GNU autoconf suite must be used and in what order, and exactly what files are produced at each step of the process. A script called "bootstrap" which automates this process is located in the config subdirectory of the MALOC source tree.
Platform-specific information
Below is some platform-specific build/usage information for MALOC.
Things should work as described above.
Things should work as described above.
Things should just work, but you may have to set the CC environment variable as follows before typing ./configure:
export CC="/bin/cc"
or you might need to use:
export CC="/bin/cc -ObjC"
If you are on a 64-bit IRIX box such as an Onyx, Octane, or Origin, set the CC environment variable as follows before typing ./configure:
export CC="/bin/cc -64"
If you are on a 32-bit IRIX box such as an O2 or Indy,
set the CC environment variable as follows before typing ./configure:
export CC="/bin/cc -32"
Unless you have the Cygwin environment, you need to use one of the included project file collections for one of the commercially available ANSI C or C++ compilers for the Win32 environment.
What you end up with
Once the build completes via the "configure;make;make install" procedure above with no errors, the MALOC library (libmaloc.a and/or libmaloc.so) is installed into the specified installation directory. You can also build some useful tools that employ the MALOC library by cd-ing into the "tools" subdirectory and repeating the "configure;make;make install" procedure.
Using MALOC on a parallel computer
MALOC provides abstractions to both INET sockets and MPI for communication support in parallel computing software. Control of the overall structure is accomplished through MALOC-shell scripts you write, or through calls to the MALOC library. (Going through the MALOC-shell is much simpler, and you have access to all the communication possibilities through the MALOC-shell.) The use of the parallel features of MALOC is explained in more detail in the documentation that is included in the MC library.
Getting MALOC to find your installation of MPI
If your installation of MPI is located in an unusual directory, then the configuration script may have trouble finding the MPI library (libmpi.a) or the MPI header file (mpi.h). Again, the configure script prints out the state of affairs quite clearly as to whether it found the library and the header. If you have MPI and configure is not finding it, then here are several possible solutions, each of which usually works. They are listed in preferred order (i.e. you should try Solution 1 first, and if that doesn't work try Solution 2, and so on).
Have your system administrator install MPI in a proper system directory so that MALOC (and other AUTOCONF-based codes) can find it!
locate mpi.h
locate libmpi.a
locate libmpich.a
On my Redhat6.2 Linux box, the following output is produced:
bash:~% locate mpi.h
/usr/share/mpi/include/mpi.h
bash:~% locate libmpi.a
bash:~% locate libmpich.a
/usr/share/mpi/lib/libmpich.a
Therefore the MPI header is installed as:
/usr/share/mpi/include/mpi.h
and the MPI library is installed as:
/usr/share/mpi/lib/libmpich.a
export FETK_MPI_INCLUDE=/usr/share/mpi/include
export FETK_MPI_LIBRARY=/usr/share/mpi/lib
./configure --enable-mpi
make clean; make; make install
The configure script should now report that it successfully
found the library and header and thus enabled MPI, and then
MALOC should compile without error.
MALOC (Minimal Abstraction Layer for Object-oriented C) was conceived, designed, and implemented over several years by Michael Holst, beginning with an initial implementation in 1994. Various colleagues have contributed ideas and/or code to MALOC (see the credits list below).
MALOC (Minimal Abstraction Layer for Object-oriented C)
Copyright (C) 1994-2006
Michael Holst TELE: (858) 534-4899
Department of Mathematics FAX: (858) 534-5273
UC San Diego, AP&M 5739 EMAIL: mholst@cam.ucsd.edu
La Jolla, CA 92093 USA WEB: http://cam.ucsd.edu/~mholst
MALOC was designed to be a portable abstraction layer for use in the
development of MC (Manifold Code), an adaptive multilevel finite element
package also developed by
Michael Holst.
MALOC was developed almost
entirely on a home-grown 90Mhz Pentium PC running various flavors of
Linux and [Free|Net|Open]BSD, using primarily GNU, BSD, and other free
software development tools. Most of the development occurred during the
hours of 10pm to 2am on a daily basis for several years, under heavy
influence of Starbuck's coffee, with helpful advice provided by Mac and
Mochi (two cats knowledgable in socket programming and numerical analysis).
MALOC is currently released under the GNU LGPL (GNU Lesser General Public License). What this means is that you may redistribute it and/or modify it under the terms of the GNU LGPL as published by the Free Software Foundation; either version 2.1 of the license, or (at your option) any later verison. You should have received a copy of the GNU LGPL with this distribution of MALOC; a copy can be found here. If you did not receive a copy of the GNU LGPL, please write to me and also write to: The Free Software Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA.
MALOC was initially released under the GNU GPL (GNU General Public License). What this means is that like all GNU softare, MALOC is freely redistributable in source code form following the rules outlined in the text of the GNU GPL.
Credits
Below is a credits list for the people that have contributed to MALOC in one way or another. The fields below follow the credits file format used in the Linux kernel CREDITS file to allow for easy manipulation via shell scripts. The fields are as follows:
N: name of contributor
E: email address
W: web address
P: PGP key ID and fingerprint
D: description of primary contributions
S: snail-mail address
N: Michael Holst E: mholst@cam.ucsd.edu W: http://cam.ucsd.edu/~mholst P: 1024/0xB5212DCD D: maloc/* -- The package structure D: maloc/acconfig.h -- The platform abstraction information D: maloc/configure.in -- The GNU autoconf/automake structure D: maloc/config/* -- The GNU autoconf/automake shell scripts D: maloc/doc/* -- The package documentation D: maloc/examples/* -- The package examples D: maloc/src/aaa_inc/* -- Library header build structure D: maloc/src/aaa_lib/* -- Static and shared library build structure D: maloc/src/base/* -- M. Holst's MALOC Foundation headers D: maloc/src/vsys/* -- M. Holst's Virtual System abstraction layer D: maloc/src/vsh/* -- M. Holst's Virtual Shell abstraction layer D: maloc/src/psh/* -- M. Holst's Parallel virtual Shell D: maloc/tools/* -- Tools built on the libraries S: Department of Mathematics S: UC San Diego, AP&M 5739 S: La Jolla, CA 92093 USA N: Nathan Baker E: nbaker@wasabi.ucsd.edu D: maloc/src/psh/vcom.c -- Vcom Class (MPI abstraction) S: Department of Chemistry S: UC San Diego S: La Jolla, CA 92093 USA N: Stephen Bond E: sdbond@illinois.edu D: maloc/maloc.spec -- RPM support (for building src/binary RPMs) S: Department of Computer Science S: University of Illinois S: Urbana, IL 61801 USA
Contacting the Author
If you have questions or comments about MALOC, please feel free to contact me at mholst@cam.ucsd.edu.
This version of MALOC is distributed under the following guidelines:
This program is free software; you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation; either version 2 of the License, or (at your option) any later version.
This program is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details.
You should have received a copy of the GNU General Public License along with this program; if not, write to the Free Software Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
The LGPL (GNU Lesser General Public License) below is copyrighted by the Free Software Foundation. However, the instance of software that it refers to, my package in this case, is copyrighted by myself as the author of the package. Any additional software that I distribute with my software is copyrighted by the authors of those pieces of software (see the individual source files for author information). ---Michael Holst
GNU LESSER GENERAL PUBLIC LICENSE
Version 3, 29 June 2007
Copyright (C) 2007 Free Software Foundation, Inc.