Stallmann, Brglez, Ghosh: Data Evaluation Scripts
These scripts are in two categories.
The first category takes raw data from directories containing the
experimental subjects and summarizes the data. The raw data is in files
created by the execution scripts, files named TR?? that record
actual crossing numbers for each subject treated by treatment
?? or files named TIMING_TR?? that record time
and number of iterations for the treatment in question.
Each of these scripts has a single argument, which is a string called
class-name denoting the
set of subdirectories from which to gather data. In general this string
will contain globbing information, such as GB_*_rnd0_0?, and
needs to be put in quotes.
- sbg_confidence_intervals -
script to produce 95% confidence intervals for data
in TR?? files in specified class directories; number of subjects can
be set along with relevant t-table data
Usage: sbg_confidence_intervals class-name
- sbg_itr_summaries - script to
generate summaries of iteration data from
TIMING_TR?? files in specified class directories
Usage: sbg_itr_summaries class-name
- sbg_summaries to generate summaries
from TR?? files in specified class
directories (is careful to use the last 64 -- or whatever
the appropriate class size -- data entries; there may be 65
in case of an isomorphic class, the first being biased
because of the way it was generated.
Usage: sbg_summaries class-name
- sbg_summary2ci to generate confidence
intervals from existing
SUMMARY files; uses confidence.awk to process the
data (can set the number of subjects).
Usage: sbg_summary2ci class-name
- sbg_timing_summaries - script
to generate summaries from TIMING_TR?? files
in specified class directories
Usage: sbg_timing_summaries class-name
The second type of script generates tabular data appropriate for use in a
spreadsheet program. We used DeltaGraph, but others would work equally
well.
Most of these scripts
require two treatments as input -- these are given on the command line in
the form tr??, or dot
in the special case of the dot treatment. Most also
allow an additional argument, whose presence causes the minimum rather
than the average to be used when looking at crossing number statistics
for a class. The optional argument also causes either _scr or
_min, as
appropriate, to be added as a suffix to the output file name. When random
classes are involved, the corresponding iso classes are used when the
optional argument is present. The other scripts all require that
sbg_summaries has already been executed for the relevant classes.
- decomposition.awk - an awk
program for creating the DeltaGraph input for
the decomposition diagram representation of a particular bigraph
class (for use with csh script
sbg_decomposition)
- edges_[bhrx] - files giving the number of edges for various sizes
of butterflies, hypercubes, grids, and vlsi graphs, respectively
(edges_x also gives the sizes used for all random graph classes, except
those specifically based on communication graphs).
- quotient.awk - transforms two fields
to one as by taking their quotient (x if undefined, 1 in the special
case of 0/0).
- sbg_comm_ratios - creates a data
file suitable for DeltaGraph, plotting
ratios of crossing numbers between two treatments for
butterflies, hypercubes, and grids
- sbg_decomposition - creates
DeltaGraph input for the decomposition diagrams
representing a single size class (uses
decomposition.awk)
Usage: sbg_decomposition class [min]
- sbg_graph_ratios - creates a data
file suitable for DeltaGraph, plotting
ratios of crossing numbers between two treatments for
random graphs of various density factors.
- sbg_tree_ratios - creates a data
file suitable for DeltaGraph, plotting
ratios of crossing numbers between two treatments for
trees of various dimensions.
- sbg_vlsi_ratios - creates a data
file suitable for DeltaGraph, plotting
ratios of crossing numbers between two treatments for
all VLSI graphs.
Matthias Stallmann
Last modified: Mon Apr 9 13:53:19 2001