Manual Reference Pages  - sift (n)

NAME

sift - high performance state space explorator and checker

Part of Tina Toolbox for analysis of Petri nets and Time Petri nets.

CONTENTS

Synopsis
Description
Options
Examples
See Also
Authors

SYNOPSIS

sift [-help] [-p]
[-R | -P | -V | -Q |
-D | -F | -F1 |
-W | -M | -H | -S | -E | -G]
[-equal | -incl | -hull]
[-f form | -dead] [-c n] [-t s] [-b n]
[-q | -k] [-NET | -NDR | -TPN | -PNML | -TTS]
[-df | -bf | -rf] [-tree] [-inh] [-tc] [-pr] [-dt] [-stats]
[infile] [outfile] [digestfile] [errorfile]

DESCRIPTION

sift builds various state space abstractions for extended Time Petri nets. It takes as input descriptions in textual form (.net, .pnml, .tpn formats) or graphical form (.ndr format of files produced by nd, .pnml with graphics), or a Time Transition System description in .tts format.

sift serves similar purposes as the tina tool, with less options but handles large states spaces much more efficiently. sift also allows to check reachability properties on the fly.

OPTIONS

-help Recalls options.
 

Operating mode options:
 

-R Builds the marking reachability graph of a Petri net (untimed, or with temporal information discarded). This option forgets time constraints, it sets flag -tc.

-V

Builds the Covering Step Graph of a Petri net, according to the technique of Vernadat/Azema/Michel. The construction preserves deadlocks. This option forgets time constraints and priorities, it sets flags -tc -pr.

-P Builds the partial marking graph of a Petri net, according to the persistent sets technique. The heuristic retained is to minimize local branching. This option forgets time constraints and priorities, it sets flags -tc and -pr.

-Q Builds the partial marking graph of a Petri net. Combine the persistent sets (-P) and covering steps (-V) methods. This option forgets time constraints and priorities, it sets flags -tc and -pr.

-D Builds the essential state graph of a Time Petri net, according to the technique of Popova. This graph preserves reachability and linear time temporal properties. Static intervals must be closed or unbounded and left-closed. If no temporal information is specified, or if all transitions bear interval [0,w[, then -D is silently replaced by -R.

-F | -F1 Builds the subgraph of the state graph obtained by firing integer delay transitions (unit delays if -F1) and discrete transitions. This graph strictly includes that obtained by -D and preserves the same properties. Static intervals must be closed or unbounded and left- closed. If no temporal information is specified, or if all transitions bear interval [0,w[, then -F/-F1 is silently replaced by -R.

-W Builds the linear state class graph of a Time Petri net, according to the technique of Berthomieu/Menasche [IFIP1983]. The linear state class graph preserves marking reachability properties and linear time temporal properties. If no temporal information is specified, or if all transitions bear interval [0,w[, then -W is silently replaced by -R. This option removes priorities, it sets flag -pr.

-M Option -M computes the set of markings of a TPN. Option -M is typically faster than -W and yields smaller state space abstractions, but it does not preserve firing sequences. ktz files created in mode -M only capture markings and the firable transitions represented as loop transitions on the markings they are firable from.

-H Option -H computes an overapproximation of the set of markings of a TPN. It is typically much faster than options -M and -W and produces a set of classes with exactly one firing domain associated with each marking. Compared to -W, all classes associated with the same marking are merged into a single class obained from the dbm-hull of their domains (the smallest dbm including all of them). As for mode -M, ktz files created in mode -H only capture markings and the firable transitions.

-S Builds the strong state class graph, according to the technique of Berthomieu/Vernadat [TACAS2003]. That graph preserves state reachability properties and linear time temporal properties; it is also compatible with priorities. If no temporal information is specified, or if all transitions bear interval [0,w[, then -S is silently replaced by -R.

-E Option -E computes the set of states of a TPN. Option -E is typically faster than -S and yields smaller state space abstractions, but it does not preserve firing sequences. ktz files created in mode -E only captured markings and the firable transitions, represented as loop transitions on the markings they are firable from.

-G Option -H computes an overapproximation of the set of states of a TPN. It is typically much faster than options -S and -E and produces a set of strong classes with exactly one clock domain associated with each marking. Compared to -S, all classes associated with the same marking are merged into a single class obained from the dbm-hull of their domains (the smallest dbm including all of them). As for mode -E, ktz files created in mode -G only capture markings and the firable transitions.

-equal Identify two state classes when they are equal (default). This flag has the effects of -W if no priorities are specified, or of -S otherwise; It is ignored if the net is untimed or if some operating mode is simultaneously set.

-incl Identify two state classes when one is included in the other. This flag has the effects of -M if no priorities are specified, or of -E otherwise; It is ignored if the net is untimed or if some operating mode is simultaneously set.

-hull Merge all state classes with same marking into their dbm hull. This flag has the effects of -H if no priorities are specified, or of -G otherwise; It is ignored if the net is untimed or if some operating mode is simultaneously set.

Exploration strategy flags (when meaningful):
 

-df Depth-first exploration order

-bf Breadth-first exploration order (default)

-rf Random exploration order

-inh Removes inhibitor arcs from the input net.

-pr Removes priority constraints from the input net.

-tc Removes time constraints from the input net.

-dt Forgets data processing when reading a tts description.

Storage strategy flags:
 

-tree Do not store states; assume any state computed is new.

Stopping conditions:
 

If some of these conditions fail and the output is ktz, then the ktz file captures a partial state space including the offending state and in which all non fully explored states bear property "sub".

-c n Stop if the number of state enumerated exceeds n; no effects if n=0.

-t n Stop if exploration lasts longer than n seconds; no effects if n=0.

-b n Stops if the marking of some place exceeds n; no effects if n=0. Sift may take advantage of this value for choosing the representation of states.

-f form Stop if some state does not obey property form; no effects if form=T. Formula form is any modality-free formula accepted by the selt model checker (check man selt for details). In addition, an atomic proposition is provided, "safe" or "L.safe", then asserts that the marking of each place does not exceed 1.

-dead Stop if some deadlock state is found; shorthand for -f "-L.dead".

Other flags:
 

-p Disables state space generation, just parses input net.

-stats Prints progress information on the fly.

Input format and source:
 

-NET | -NDR | -PNML | -TPN | -TTS
  Specifies the format of the input net. This flag is necessary when the input net is read on standard input, or read from a file that does not bear the expected extension. By default, the net is assumed in .net or

infile Where the net is read. The input format is determined by the file type, according to the table below. If absent or specified by "-", the net is read on standard input in the format specified by the input flag. If both an infile and some input flag are present, then the format defined by the input flag supersedes that determined by the infile extension.

file extension          input format
--------------------------------------------------------------
.net                    net format
.ndr                    ndr format
.tpn                    tpn format
.pnml                   pnml format
.tts                    tts format

Output format and destination:
 

-q | -ktz
  Specifies the format of the output. This flag is necessary when output is produced on standard output, or written in a file that does not bear the expected extension. -q means no output is printed, while -k means some output is produced in .ktz proprietary compressed format.

outfile
  Where results are written. The output format is determined by the file type, according to the table below. If absent or specified by "-", then results are written on standard output in the format specified by the output format flag. If both an outfile and some output flag are present, then the format defined by the output flag supersedes that determined by the outfile extension.

file extension          output format
--------------------------------------------------------------
.ktz                    tina binary ktz format
any other               textual output

Digest destination:
 

digestfile
 

In addition to (possibly) its results in ktz form in file outfile, sift prints a summary of results in file digestfile (default stdout).

Errors destination:
 

errorfile
  Where error messages are written. By default, errors are printed on standard error.

Default options:
 

Depend upon the features of the input net:

If no time constraint:     sift -R -c 0 -t 0 -b 0 -f T -q
Else, if no priorities:    sift -W -c 0 -t 0 -b 0 -f T -q
Otherwise:                 sift -S -c 0 -t 0 -b 0 -f T -q

EXAMPLES

sift -f "p1 \/ t4" < ifip.ndr
sift -PNML mynet.xml
sift -f safe abp.ndr abp.ktz -stats

SEE ALSO

nd(n), tina(n), tedd(n), plan(n), struct(n), ktzio(n), ndrio(n), selt(n), muse(n), pathto(n), play(n), formats(n)

AUTHORS

Bernard Berthomieu LAAS/CNRS, 2000-2012, Bernard.Berthomieu@laas.fr.


Tina Tools sift (n) Version 3.5.0
Generated by manServer 1.07 from src/sift.n using man macros.