s14h
/
prism-accumulation
1
0
Fork 0
Code Releases Activity
Browse Source

NOTES. 

git-svn-id: https://www.prismmodelchecker.org/svn/prism/prism/trunk@2161 bbc10eb1-c90d-0410-af57-cb519fbb1720
master
Dave Parker 15 years ago
parent
bbaba8bddc
commit
45c1f3f367
4 changed files with 115 additions and 216 deletions
Whitespace
Show all changes Ignore whitespace when comparing lines Ignore changes in amount of whitespace Ignore changes in whitespace at EOL
Split View
Diff Options
Show Stats Download Patch File Download Diff File
  1. 171
      prism/NOTES
  2. 90
      prism/NOTES-PTAS
  3. 40
      prism/NOTES-SIM
  4. 30
      prism/TODO

171
prism/NOTES
View File

@ -1,27 +1,31 @@
Last stable version of PRISM on svn (3.3 plus some fixes) is rev 1405
(there is a copy of this, e.g. to get working old simulator, in ~/prism-old)
-----------------------------------------------------------------------
Stuff to remove when distributing early copies to people:
Remove for beta distributions:
------------------------------
* NOTES*
* TODO?
* examples/des
* examples/explicit
-----------------------------------------------------------------------
=======================================================
See also:
* NOTES-PTAS
* NOTES-SIM
* NOTES-ABSTR
TODO (before any release)
-------------------------
-----------------------------------------------------------------------
PTAs:
* GUI
* Tidy prism-examples/pta, incl. delete brp
* Clarify semantic/type checks (consistency with games/digital)
* Check guards/invariants for convexity (for now, neither can be non-convex, see below)
* Clarify time divergence issues
* non-zeno checks?
* Sort VarList (two types - before and after constants evaluated). Need for simulator too.
* Digital clocks: No time-bounded until yet
* Digital clocks: Translation done sep for each property (e.g. for cmax)?
* Bug fix: action alphabet (syntactic) for sync lost in PTA object construction
Filters, property semantics, etc.
=======================================================
TODO (before public release)
----------------------------
TODO:
Filters, property semantics, etc.
* Integer-valued props displayed as doubles - need to give types to StateValues?
- lots of other places too e.g. GUI results)
* Intervals for multiple initial states
@ -30,51 +34,106 @@ TODO:
* Restructure Result class? e.g. separate bracketed comment? (done I think)
- Errors (Exceptions) formatted properly?
Doc:
* Filters - no longer allowed minmax (or encode as range?)
* P>=p[...{filter}] means & not print (didn't print in 3.3 anyway, despite manual)
-----------------------------------------------------------------------
TODO:
- remove colons from @params
PTAs:
* Enforce well-formedness checks (i.e. guards/resets imply target invariants)
(as opposed to say supporting strong invariants,
where targets with false invariants cannot be entered)
- zone-based: check during reach using valids
(or just during FW reach?) (nb: need to split dpost to do check)
(see non-well-formed.nm/pctl for a test case)
(if can't do that, syntax check using sat)
- digital clocks: check invariants after transform/reach?
- fix adversary generation (de alfaro thesis)
- add adversary generation for other engines
- make action storage optional (when required e.g. for export) (especially for MCs)
- lp stuff - plugin support for eclipse/lpsolve (in prism-multi?)
- also look up lrs?
- finalise tra format with actions (needs to work with actions too)
- so also fix e.g. importtrans, prism-statra
DOC (before public release)
---------------------------
-----------------------------------------------------------------------
Abstraction/refinement:
TODO:
* exp reach for games
Other notes/ideas:
* refine opt: don't add ubsets of player1 choices
-----------------------------------------------------------------------
LTL...
bug:
Filters:
* Filters - no longer allowed minmax (or encode as range?)
* P>=p[...{filter}] means & not print (didn't print in 3.3 anyway, despite manual)
prism ~/prism-examples/dice/dice.pm \
PTAs:
* Where can clocks be used
- clock constraints (just (some) binops)
- in guards
- in invariants
- resets (normal updates, but to ints only)
- var decl.s
- generally must be convex (i.e. just conj of constraints) (can be relaxed a bit for digital clocks)
* Currently, modules cannot access non-local vars (and there are no globals)
(this is too restrictive, e.g. for message passing)
* Must have single initial state (is this true for digital clocks too?)
* Invariants must come straight after var decls
=======================================================
BUG (some time)
---------------
Simulator:
* approx mc of a property loses any current simulator path in gui.
is that ok? (seems to be buggy in 3.3.1 anyway)
* prism ~/prism-examples/dice/dice.pm \
-pctl '(s=0|s=1)=>P>=1 [ (P>=0.9 [ F<=1 s=3 ])=>(P>=0.5 [ F<=2 s=4 ]) ]'
time-bounds should be ok in inner pctl but are not
(time-bounds should be ok in inner pctl but are not)
TODO (some time)
----------------
PTAs:
* On-the-fly global reachability to allow (for A-R engine):
- access to other local and global vars
- system endsystem?
(then test on Arnd's BRP model + others)
* BRP example
* Translate non-convex guards to DNF and multiple transitions
* Investigate whether non-convex invariants can be supported (look at zone ops)
* Games: Optimise number of states in time-bounded PTAs
(extras added in old target states) using until in forwards reach?
* Fix: Creation of new names (adding extra _s) doesn't take into account prop file ot consts etc.
* Digital clocks: urgency?
* Digital clocks: optimisations?
* New version of valid2 (inv/g push outside)
* Combined complement + intersection (terminating early) for when dbm lists get big
* Are we ok to have non-diagonal c-closure algorithm?
(or is it ok if only use this on initial reach graph construction?)
Simulator:
* Investigate efficiency wrt old simulator
* Add (back) support for full loop detection? (embedded into Path*?)
* Add (back) early manual termination of sampling (thru expt stop?)
* Random initial state
* Variable overflow etc.
* Add support for "deadlock" and "init" (new EvaluationContext, model *and* state dependent)
* Seed issues (currently twice in one second = same seed)
* explicitbuildtest doesn't handle dupes in mdps (e.g. consensus)
* Explicit build doesn't handle multiple initial states
* GUI sim - add context menu to transition list with e.g. "show in model"
* Traviendo export?
* Code: Optimise/tidy Choices (ChoiceList/ChoiceSingleton/etc.)
Actions:
* Finalise tra format with actions (needs to work with MCs too)
- also fix e.g. importtrans, prism-statra
* make action storage optional (when required e.g. for export) (especially for MCs)
Adversary generation:
* Add adversary generation for other engines
bscc stuff
- regression testing on ltl stuff with 3 diff options
-----------------------------------------------------------------------
Abstraction/refinement:
* Exp reach for games
* Poss. refine opt: don't add ubsets of player1 choices
prism-multi
BSCC stuff:
* regression testing + perf. on ltl stuff with 3 diff options
DOC: no state rewards
Code tidy:
* remove colons from @params
DOC (some time)
---------------
Prism-multi:
* No state rewards

90
prism/NOTES-PTAS
View File

@ -1,90 +0,0 @@
TODO (before any release)
-------------------------
* GUI
* Tidy prism-examples/pta, incl. delete brp
* Clarify semantic/type checks (consistency with games/digital)
* Check guards/invariants for convexity (for now, neither can be non-convex, see below)
* Clarify time divergence issues
* non-zeno checks?
* Sort VarList (two types - before and after constants evaluated). Need for simulator too.
* Digital clocks: No time-bounded until yet
* Digital clocks: Translation done sep for each property (e.g. for cmax)?
* Bug fix: action alphabet (syntactic) for sync lost in PTA object construction
TODO (later)
------------
* Enforce well-formedness checks (i.e. guards/resets imply target invariants)
(as opposed to say supporting strong invariants,
where targets with false invariants cannot be entered)
- zone-based: check during reach using valids
(or just during FW reach?) (nb: need to split dpost to do check)
(see non-well-formed.nm/pctl for a test case)
(if can't do that, syntax check using sat)
- digital clocks: check invariants after transform/reach?
* On-the-fly global reachability to allow (for A-R engine):
- access to other local and global vars
- system endsystem?
(then test on Arnd's BRP model + others)
* BRP example
* Translate non-convex guards to DNF and multiple transitions
* Investigate whether non-convex invariants can be supported (look at zone ops)
Bugs
----
Maybe todo
----------
* Games: Optimise number of states in time-bounded PTAs
(extras added in old target states) using until in forwards reach?
* Fix: Creation of new names (adding extra _s) doesn't take into account prop file ot consts etc.
* Digital clocks: urgency?
* Digital clocks: optimisations?
Tidy
----
* Remove unneeded files from examples (prism-benchmark, rewards stuff)
Documentation
-------------
Where can clocks be used
* clock constraints (just (some) binops)
- in guards
- in invariants
* resets (normal updates, but to ints only)
* var decl.s
* generally must be convex (i.e. just conj of constraints) (can be relaxed a bit for digital clocks)
Currently, modules cannot access non-local vars (and there are no globals)
(this is too restrictive, e.g. for message passing)
Must have single initial state (is this true for digital clocks too?)
Invariants must come straight after var decls

40
prism/NOTES-SIM
View File

@ -1,40 +0,0 @@
NOW/NEXT
--------
TODO (before any release)
-------------------------
TODO (later)
------------
* investigate efficiency wrt old simulator
* add (back) support for full loop detection? (embedded into Path*?)
* add (back) early manual termination of sampling (thru expt stop?)
* random initial state
* variable overflow etc.
* add support for "deadlock" and "init" (new EvaluationContext, model *and* state dependent)
* seed issues (currently twice in one second = same seed)
* explicitbuildtest doesn't handle dupes in mdps (e.g. consensus)
* explicit build doesn't handle multiple initial states
* GUI sim - add context menu to transition list with e.g. "show in model"
* traviendo export?
CODE:
* Optimise/tidy Choices (ChoiceList/ChoiceSingleton/etc.)
BUGS (later)
------------
* approx mc of a property loses any current simulator path in gui.
is that ok? (seems to be buggy in 3.3.1 anyway)

30
prism/TODO
View File

@ -1,30 +0,0 @@
TODO:
new version of valid2 (inv/g push outside)
combined complement + intersection (terminating early) for when dbm lists get big
----------------------
BUGS:
----------------------
----------------------
./forwardreach examples/repudiation/originator.des examples/repudiation/recipient.des examples/repudiation/honest_deadline40.des '*:*:before' 'true' -min -opt -nopre
Final TPSG: 416 states, 1813 distribution sets, 2221 distributions, 3751 transitions, p1max/avg = 3/4.4, p2max/avg = 2/1.2
(avg > max)
----------------------
Questions:
is it safe to split (refine) multiple states concurrently in the way that we do?
do we want to check for duplicate sym states when creating new ones through refinement?
- is this safe?
are we ok to have non-diagonal c-closure algorithm?
or is it ok if only use this on initial reach graph construction?
Write Preview
Loading…
Cancel
Save