MSCF

#MSCF

MSCF is implemented using the C++ programming language,and easy-to-use interfaces are provided to parallelize processing algorithms.

1. To Compile

Note: this version is not a final release, and some components are still under testsing. The program has been tested on  Scientific Linux 6.7 operation system, compiled using g++ 4.9, OpenMPI 2.1, GDAL 1.9, and LibTIFF 4.0. The makefile (i.e., make_pcca) will compile a demonstration program, pCCA, which is able to manage raster data in parallel.

(1) Before compiling, make sure MPI, GDAL, and LibTIFF libraries have been installed.

(2) Open make_pcca and modify the lines that specify the locations of libraries.

(3) Type 'make -f make_pcca depend'.

(4) Type 'make -f make_pcca' to compile.

After successful compilation, an executable file named pCCA will be generated.

2. To Run

There are datasets, which include seven layers at 30m resolution of California in the US. Note that this data is only for testing if the pCCA program works as expected, not for demonstrating the performance.

2.1 Usage:

mpirun -np
um_proc\  pCCA  \workspace
um_urbanization-per-year\
um_iterations\ \MSCF-option(0/1)
um-row-subspaces
um-col-subspaces \task-farming(1/0)  \parallelIO(1/0) \with-writer(1/0) 

workspace: the directory where the input file is located and the output files will be written.

num_urbanization-per-year:the total numbers of nonurban cells convert to urban cells.

num_iterations:the numbers of iterations to run pCCA.

MSCF-option:  MSCF option, either 0 or 1. if 0, without using hypergraph partiioning; if 1, hypergraph partiioning is used.

num-row-subspaces: the number of sub-domains in the Y axis, for domain decomposition. If num-row-subspaces > 1 and num-col-subspaces = 1, the domain is decomposed as row-wise; if num-row-subspaces = 1 and num-col-subspaces > 1, the domain is decomposed as column-wise; if both > 1, the domain is decomposed as block-wise.

num-col-subspaces: the number of sub-domains in the X axis, for domain decomposition.

task-farming: load-balancing option, either 0 or 1. if 0, load-balancing; if 1, task farming.

parallelIO:I/O option, either 0 or 1. if 0, para_reading; if 1, centdel_reading.

with-writer: an option that specify whether a writer process will be used. If 0, no writer; if 1, use a writer.

2.2 Example:

mpirun -np 4 ./pCCA ./ 25000 50 1 5 5 0 0 0