acpl(1) [debian man page]

acpl(1) 						       Scotch user's manual							   acpl(1)

NAME

       acpl - compile a target architectures

SYNOPSIS

       acpl [options] [itfile] [otfile]

DESCRIPTION

       The  acpl  program  compiles a decomposition-defined target architecture file itfile of type 'deco 0' into a compiled decomposition-defined
       target architecture of type 'deco 1', stored in file otfile.

       Compiling a decomposition-defined architecture amounts to computing the distance matrix of  all	possible  subdomains,  from  the  distance
       matrix  of  all	terminal subdomains provided in the 'deco 0' format.  Since this computation is internally performed every time a 'deco 0'
       format is read, and is quadratic in time, pre-compiling the target architecture by means of acpl can save some time when repeatedly comput-
       ing mappings on a large decomposition-defined 'deco 0' target architecture.

       When the proper libraries have been included at compile time, acpl can directly handle compressed files, both as input and output. A stream
       is treated as compressed whenever its name is postfixed with a compressed file extension, such as in 'brol.tgt.bz2' or '-.gz'. The compres-
       sion formats which can be supported are the bzip2 format ('.bz2'), the gzip format ('.gz'), and the lzma format ('.lzma', on input only).

       Since  decomposition-defined target architecture files have a size which is quadratic in the number of target vertices, because of the dis-
       tance matrix structures, using compressed files to store them may save a lot of space, especially for compiled target architecture files.

OPTIONS

       -h     Display some help.

       -V     Display program version and copyright.

EXAMPLE

       Create a compiled cube-connected-cycle target architecture of dimension 4, and save it under the gzip(1) format to file 'ccc4c.tgt.gz'. The
       dash  '-'  standard file name is used so that the 'deco 0' target architecture description produced by amk_ccc(1) is read from the standard
       input, through the pipe.

	   $ amk_ccc 4 | acpl - ccc4c.tgt.gz

SEE ALSO

       amk_grf(1), amk_ccc(1), atst(1), dgmap(1), gmap(1), gmtst(1).

       Scotch user's manual.

AUTHOR

       Francois Pellegrini <francois.pellegrini@labri.fr>

								 February 14, 2011							   acpl(1)

amk_ccc(1)						       Scotch user's manual							amk_ccc(1)

NAME

       amk_ccc, amk_fft2, amk_hy, amk_m2, amk_p2 - create target architectures

SYNOPSIS

       amk_ccc [options] dim [tfile]

       amk_fft2 [options] dim [tfile]

       amk_hy [options] dim [tfile]

       amk_m2 [options] dimX [dimY] [tfile]

       amk_p2 [options] [wght0] [wght1] [tfile]

DESCRIPTION

       The amk_* programs create target architecture files for some common, regular topologies.

       amk_ccc	creates  a decomposition-defined cube-connected-cycle topology of dimension dim. The decomposition is performed first by bisection
       along the dimensions of the hypercube, then along the remaining cycle graphs.

       amk_fft2 creates a decomposition-defined fast-Fourier-transform topology of dimension dim. The  decomposition  is  performed  by  recursive
       bisection of the vertices, by descending dimension (that is, bit number in the labeling of the vertices).

       amk_hy  creates	a  decomposition-defined hypercube topology of dimension dim. The decomposition is performed by recursive bisection of the
       vertices, by descending dimension (that is, bit number in the labeling of the vertices). Save for experimentation purposes, this program is
       deprecated,  as the algorithmically-defined 'hcub' target architecture is a more convenient and efficient way to represent hypercube archi-
       tectures.

       amk_m2 creates a decomposition-defined 2D regular grid topology of dimensions dimX and dimY. The decomposition is  performed  by  recursive
       splitting  along  the  dimensions, either by cutting the longest one, or by one-way dissection, depending on the '-m' option flag. Save for
       experimentation purposes, this program is deprecated, as the algorithmically-defined 'mesh2D' and 'mesh3D' target architectures are a  more
       convenient and efficient way to represent 2D and 3D grid architectures.

       amk_p2 creates a weighted path graph topology comprising only two vertices of weights wght0 and wght1. This is just a helper program, which
       builds a 'wcmplt' algorithmically-defined complete graph with two vertices. It may be used to compute weighted bisections of a graph.

       When the proper libraries have been included at compile time, the amk_* programs can directly handle compressed files, both  as	input  and
       output.	A  stream  is  treated as compressed whenever its name is postfixed with a compressed file extension, such as in 'brol.tgt.bz2' or
       '-.gz'. The compression formats which can be supported are the bzip2 format  ('.bz2'),  the  gzip  format  ('.gz'),  and  the  lzma  format
       ('.lzma', on input only).

OPTIONS

       -mmeth For amk_m2 only. Perform either recursive dissection or one-way dissection, according to the given method flag:

	      n      perform nested dissection (default).

	      o      perform one-way dissection (cut across Y, then X).

       -h     Display some help.

       -V     Display program version and copyright.

EXAMPLE

       Create a cube-connected-cycle target architecture of dimension 4, and save it to file 'ccc4.tgt'.

	   $ amk_ccc 4 ccc4.tgt

       Run  gmap  to compute a bisection, into two parts of respective weights 3 and 5, of graph 'brol.grf' and save the resulting mapping to file
       'brol.map'. The dash '-' standard file name is used so that the target architecture description is read from the  standard  input,  through
       the pipe.

	   $ amk_p2 3 5 | gmap brol.grf - brol.map

SEE ALSO

       acpl(1), atst(1), amk_grf(1), dgmap(1), gmap(1).

       Scotch user's manual.

AUTHOR

       Francois Pellegrini <francois.pellegrini@labri.fr>

								 February 14, 2011							amk_ccc(1)