What to expect from GLARE installation?
First, GLARE is not a combinatorial library
platform, but an intermediate tool that can be used to improve the library design
process. The executable produced by the distribution can perform a library optimization
given that you provide the following input:
A text file with the list of reagents for each dimension (R groups) with their property of interest pre-calculated.
Typically, but not necessarily, a file per dimension is provided.
An input file that the user write to provide the information on how the different
dimensions are combined.
GLARE is run through a command line interface and the
output files are generated in form of text files with the extension
.glo (for GLARE output). In the platform independent releases
you can find an example that you can try following the information given on this page.
Platforms
GLARE is available on a variety of platforms. The compiled executable
is provided for few platforms such as Linux and Windows. Platform independent releases are also
provided for both Windows and Linux/UNIX (which differ by the endline special character). The compilation
setup uses the GNU automake tools.
Below, we provide instructions to test and compile GLARE.
To maximize the speed of execution, it is recommended to use a good compiler for your specific
platform.
on a variety of platforms including Windows Linux and UNIX. It has not been compiled on a MAC OS yet.
Compilation
In GLARE platform independent distributions, the GNU autotools
setup files should make the compilation a charm. Here is the usual compilation procedure:
$ gunzip glare-1.0b.source-linux.tar.gz
$ tar -xvf glare-1.0b.source-linux.tar
$ cd glare-1.0b.source-linux
$ ./configure
$ make
The executable is left in the sub-directory src and is normally
called glare.exe or glare. If you have a special compiler and you
want to pass special options to improve the speed of
the executable, you can specify the compiler by setting the option
CXX (c++) when invoking the configure command and the compiler
options can be set under CXXFLAGS options. For example:
$ ./configure CXX=xlC CXXFLAGS="-O4"
$ make
If you need to recompile GLARE, just type
$ make clean
$ make
Testing compilation
To test if the compilation successfully built the desired binary, you simply need to cd into the
main directory and type:
$ make check
The output for a successful result should be "TEST SUCCESSFULLY PASSED". This command requires
python to be available. If it is not available, then you can yourself
run the test and compare the result with a successful example:
$ src/glare.exe -i test/LIB03_definition
$ ...
------- ITERATION : 14 --------------
GOODNESS : 95%
NUMBER EVAL : 33489
CUMUL. EVAL. : 548518
KEPT IN STEP : 99.9 %
ACTUAL SIZE LIB03A : 651 x 631 x 1 = 410781
EFFECTIVENESS LIB03A : 34.9%
ACTUAL SIZE LIB03B : 76 x 631 x 1 = 47956
EFFECTIVENESS LIB03B : 59.4%
BRAVI EFFECT. : 36.2%
The last output iteration shown above is what should be obtained when
running the test. Small numerical differences are expected.
The speed of execution of
GLARE greatly depends on the compiler. For example, an executable
compiled on Windows/CYGWIN with g++ leads to an executable 20x slower that when it is
compiled with Microsof Visual C++. Timings for GLARE have also
been reported for a variety of libraries in the original published article. However, with
the distributed implementation, the algorithm should be about 2-3x slower due to added
user flexibility (the filtering rules were originally hard coded). The execution of the
compilation test should take between 0.5 to 2s if a good compiler is used and if run on
a decent computer (P3 1.5GHz or better).
|
Last site update: 26 January 2007
|
