You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.

257 lines
15 KiB

3 months ago
  1. Changelog
  2. ==============
  3. This changelog lists only the most important changes. Smaller (bug)fixes as well as non-mature features are not part of the changelog.
  4. The releases of major and minor versions contain an overview of changes since the last major/minor update.
  5. Version 1.6.x
  6. -------------
  7. ## Version 1.6.4 (20xx/xx)
  8. - Added support for model checking LTL properties in the sparse (and dd-to-sparse) engine. Requires building with Spot or an external LTL to deterministic automaton converter (using option `--ltl2datool`).
  9. - Added cmake options `STORM_USE_SPOT_SYSTEM` and `STORM_USE_SPOT_SHIPPED` to facilitate building Storm with [Spot](https://spot.lrde.epita.fr/).
  10. - Improved parsing of formulas in PRISM-style syntax.
  11. - Added export of schedulers that use memory (in particular optimizing schedulers for LTL properties)
  12. - Added support for PRISM models that use unbounded integer variables.
  13. - Added an export of check results to json. Use `--exportresult` in the command line interface.
  14. - Added computation of steady state probabilities for DTMC/CTMC in the sparse engine. Use `--steadystate` in the command line interface.
  15. - Implemented parsing and model building of Stochastic multiplayer games (SMGs) in the PRISM language. No model checking implemented (yet).
  16. - API: Simulation of prism-models
  17. - API: Model-builder takes a callback function to prevent extension of particular actions, prism-to-explicit mapping can be exported
  18. - API: Export of dice-formatted expressions
  19. - Prism-language/explicit builder: Allow action names in commands writing to global variables if these (clearly) do not conflict with assignments of synchronizing commads.
  20. - Prism-language: n-ary predicates are supported (e.g., ExactlyOneOf)
  21. - Added support for continuous integration with Github Actions.
  22. - `storm-pars`: Exploit monotonicity for computing extremal values and parameter space partitioning.
  23. ## Version 1.6.3 (2020/11)
  24. - Added support for multi-objective model checking of long-run average objectives including mixtures with other kinds of objectives.
  25. - Added support for generating optimal schedulers for globally formulae.
  26. - Simulator supports exact arithmetic.
  27. - Added switch `--no-simplify` to disable simplification of PRISM programs (which sometimes costs a bit of time on extremely large inputs).
  28. - Fixed issues with JANI inputs concerning .
  29. - transient variable expressions in properties,
  30. - constants in properties, and
  31. - integer variables with either only an upper or only a lower bound.
  32. - `storm-pomdp`: States can be labelled with values for observable predicates.
  33. - `storm-pomdp`: (Only API) Track state estimates.
  34. - `storm-pomdp`: (Only API) Reduce computation of state estimates to computation on unrolled MDP.
  35. ## Version 1.6.2 (2020/09)
  36. - Prism program simplification improved.
  37. - Revamped implementation of long-run-average algorithms, including scheduler export for LRA properties on Markov automata.
  38. - Support for step-bounded properties of the form ... [F[x,y] ... ] for DTMCs and MDPs (sparse engine).
  39. - Renamed portfolio engine to automatic
  40. - `storm-dft`: Fix for relevant events when using symmetry reduction.
  41. - `storm-pomdp`: Fix for --transformsimple and --transformbinary when used with until formulae.
  42. - `storm-pomdp`: POMDPs can be parametric as well.
  43. ## Version 1.6.0 (2020/06)
  44. - Changed default Dd library from `cudd` to `sylvan`. The Dd library can be changed back to `cudd` using the command line switch `--ddlib`.
  45. - Scheduler export: Properly handle models with end components. Added export in `.json` format.
  46. - CMake: Search for Gurobi prefers new versions.
  47. - CMake: We no longer ship xerces-c. If xerces-c is not found on the system, storm-gspn will not be able to parse xml-based GSPN formats.
  48. - CMake: Added option `STORM_LOAD_QVBS` to automatically download the quantitative verification benchmark set.
  49. - Eigen library: The source code of Eigen is no longer included but downloaded from an external repository instead. Incremented Eigen version to 3.3.7 which fixes a compilation issue with recent XCode versions.
  50. - Tests: Enabled tests for permissive schedulers.
  51. - `storm-counterexamples`: fix when computing multiple counterexamples in debug mode.
  52. - `storm-dft`: Renamed setting `--show-dft-stats` to `dft-statistics` and added approximation information to statistics.
  53. - `storm-pomdp`: Implemented approximation algorithms that explore (a discritization of) the belief MDP, allowing to compute safe lower- and upper bounds for a given property.
  54. - `storm-pomdp`: Implemented almost-sure reachability computations: graph-based, one-shot SAT-based, and iterative SAT-based.
  55. - `storm-pomdp': Various changes such that transformation to pMCs is now again supported (and improved).
  56. - Fixed several compiler warnings.
  57. Version 1.5.x
  58. -------------
  59. ## Version 1.5.1 (2020/03)
  60. - Jani models are now parsed using exact arithmetic.
  61. ## Version 1.5.0 (2020/03)
  62. - Added portfolio engine which picks a good engine (among other settings) based on features of the symbolic input.
  63. - Abort of Storm (via timeout or CTRL+C for example) is now gracefully handled. After an abort signal the program waits some seconds to output the result computed so far and terminates afterwards. A second signal immediately terminates the program.
  64. - Setting `--engine dd-to-sparse --bisimulation` now triggers extracting the sparse bisimulation quotient.
  65. - JIT model building is now invoked via `--engine jit` (instead of `--jit`).
  66. - DRN: support import of choice labelling.
  67. - Added option `--build:buildchoiceorig` to build a model (PRISM or JANI) with choice origins (which are exported with, e.g. `--exportscheduler`).
  68. - Implemented optimistic value iteration for sound computations and set it as new default in `--sound` mode.
  69. - Time bounded properties for Markov automata are now computed with relative precision. Use `--absolute` for the previous behavior.
  70. - Apply the maximum progress assumption while building a Markov automaton with one of the symbolic engines.
  71. - Added option `--build:nomaxprog` to disable applying the maximum progress assumption during model building (for Markov Automata).
  72. - Added hybrid engine for Markov Automata.
  73. - Improved performance of the Unif+ algorithm (used for time-bounded properties on Markov Automata).
  74. - Various performance improvements for model building with the sparse engine.
  75. - `storm-dft`: Symmetry reduction is now enabled by default and can be disabled via `--nosymmetryreduction`.
  76. - `storm-pomdp`: Only accept POMDPs that are canonical.
  77. - `storm-pomdp`: Prism language extended with observable expressions.
  78. - `storm-pomdp`: Various fixes that prevented usage.
  79. - Several bug fixes.
  80. Version 1.4.x
  81. -------------
  82. ### Version 1.4.1 (2019/12)
  83. - Implemented long run average (LRA) computation for DTMCs/CTMCs via value iteration and via gain/bias equations.
  84. - Added several LRA related settings in a new settings module. Note that `--minmax:lramethod` has been replaced by `--lra:nondetmethod`.
  85. ### Version 1.4.0 (2019/11)
  86. - Added support for multi-dimensional quantile queries.
  87. - Added support for multi-objective model checking under pure (deterministic) schedulers with bounded memory using `--purescheds`.
  88. - Allow to quickly check a benchmark from the [Quantitative Verification Benchmark Set](http://qcomp.org/benchmarks/) using the `--qvbs` option.
  89. - Added script `resources/examples/download_qvbs.sh` to download the QVBS.
  90. - If an option is unknown, Storm now suggests similar option names.
  91. - Flagged several options as 'advanced' to clean up the `--help`-message. Use `--help all` to display a complete list of options.
  92. - Support for parsing of exact time bounds for properties, e.g., `P=? [F=27 "goal"]`.
  93. - Export of optimal schedulers when checking MDPs with the sparse engine (experimental). Use `--exportscheduler <filename>`.
  94. - PRISM language: Support for the new `round` operator.
  95. - PRISM language: Improved error messages of the parser.
  96. - JANI: Allow bounded types for constants.
  97. - JANI: Support for non-trivial reward accumulations.
  98. - JANI: Fixed support for reward expressions over non-transient variables.
  99. - DRN: Added support for exact parsing and action-based rewards.
  100. - DRN: Support for placeholder variables which allows to parse recurring rational functions only once.
  101. - Fixed sparse bisimulation of MDPs (which failed if all non-absorbing states in the quotient are initial).
  102. - Support for export of MTBDDs from Storm.
  103. - Support for k-shortest path counterexamples (arguments `-cex --cextype shortestpath`)
  104. - New settings module `transformation` for Markov chain transformations. Use `--help transformation` to get a list of available transformations.
  105. - Support for eliminating chains of Non-Markovian states in MAs via `--eliminate-chains`.
  106. - Export to dot format allows for maximal line width in states (argument `--dot-maxwidth <width>`)
  107. - `storm-conv` can now apply transformations on a prism file.
  108. - `storm-pars`: Enabled building, bisimulation and analysis of symbolic models.
  109. - `storm-dft`: Support partial-order for state space generation.
  110. - `storm-dft`: Compute lower and upper bounds for number of BE failures via SMT.
  111. - `storm-dft`: Allow to set relevant events which are not set to Don't Care.
  112. - `storm-dft`: Support for constant failed BEs. Use flag `--uniquefailedbe` to create a unique constant failed BE.
  113. - `storm-dft`: Support for probabilistic BEs via PDEPs.
  114. - Fixed linking with Mathsat on macOS.
  115. - Fixed linking with IntelTBB for GCC.
  116. - Fixed compilation for macOS Mojave and higher.
  117. - Several bug fixes.
  118. Version 1.3.x
  119. -------------
  120. ## Version 1.3.0 (2018/12)
  121. - Slightly improved scheduler extraction
  122. - Environments are now part of the c++ API
  123. - Heavily extended JANI support, in particular:
  124. * arrays, functions, state-exit-rewards (all engines)
  125. * indexed assignments, complex reward expressions (sparse engine)
  126. * several jani-related bug fixes
  127. - New binary `storm-conv` that handles conversions between model files
  128. - New binary `storm-pomdp` that handles the translation of POMDPs to pMCs.
  129. - Maximal progress assumption is now applied while building Markov Automata (sparse engine).
  130. - Improved Unif+ implementation for Markov Automata, significantly reduced memory consumption.
  131. - Added support for expected time properties for discrete time models
  132. - Bug fix in the parser for DRN (MDPs and MAs might have been affected).
  133. - `storm-gspn`: Improved .pnpro parser
  134. - `storm-gspn`: Added support for single/infinite/k-server semantics for GSPNs given in the .pnpro format
  135. - `storm-gspn`: Added option to set a global capacity for all places
  136. - `storm-gspn`: Added option to include a set of standard properties when converting GSPNs to jani
  137. - `storm-pars`: Added possibility to compute the extremal value within a given region using parameter lifting
  138. - `storm-dft`: DFT translation to GSPN supports Don't Care propagation
  139. - `storm-dft`: Support DFT analysis via transformation from DFT to GSPN to JANI
  140. - `storm-dft`: Added SMT encoding for DFTs
  141. - `storm-dft`: Improved Galileo and JSON parser
  142. - Several bug fixes
  143. - Storm uses the `master14` branch of carl from now on
  144. ### Comparison with Version 1.2.0 (details see below)
  145. - Heavily extended JANI-support
  146. - New binary `storm-conv` that handles conversion between model files
  147. - New binary `storm-pomdp` that handles the translation of POMDPs to pMCs.
  148. - `storm-gspn` improved
  149. - Sound value iteration
  150. Version 1.2.x
  151. -------------
  152. ### Version 1.2.3 (2018/07)
  153. - Fix in version parsing
  154. ### Version 1.2.2 (2018/07)
  155. - Sound value iteration (SVI) for DTMCs and MDPs
  156. - Topological solver for linear equation systems and MinMax equation systems (enabled by default)
  157. - Added support for expected total rewards in the sparse engine
  158. - By default, iteration-based solvers are no longer aborted after a given number of steps.
  159. - Improved export for jani models
  160. - A fix in parsing jani properties
  161. - Several extensions to high-level counterexamples
  162. - `storm-parsers` extracted to reduce linking time
  163. - `storm-counterexamples` extracted to reduce linking time
  164. - `storm-dft`: improvements in Galileo parser
  165. - `storm-dft`: test cases for DFT analysis
  166. - Improved Storm installation
  167. - Several bug fixes
  168. ### Version 1.2.1 (2018/02)
  169. - Multi-dimensional reward bounded reachability properties for DTMCs.
  170. - `storm-dft`: transformation of DFTs to GSPNs
  171. - Several bug fixes
  172. ### Version 1.2.0 (2017/12)
  173. - C++ api changes: Building model takes `BuilderOptions` instead of extended list of Booleans, does not depend on settings anymore.
  174. - `storm-cli-utilities` now contains cli related stuff, instead of `storm-lib`
  175. - Symbolic (MT/BDD) bisimulation
  176. - Fixed issue related to variable names that can not be used in Exprtk.
  177. - DRN parser improved
  178. - LP-based MDP model checking
  179. - Sound (interval) value iteration
  180. - Support for Multi-objective multi-dimensional reward bounded reachability properties for MDPs.
  181. - RationalSearch method to solve equation systems exactly
  182. - WalkerChae method for solving linear equation systems with guaranteed convergence
  183. - Performance improvements for sparse model building
  184. - Performance improvements for conditional properties on MDPs
  185. - Automatically convert MA without probabilistic states into CTMC
  186. - Fixed implemention of Fox and Glynn' algorithm
  187. - `storm-pars`: support for welldefinedness constraints in mdps.
  188. - `storm-dft`: split DFT settings into IO settings and fault tree settings
  189. - `storm-dft`: removed obsolete explicit model builder for DFTs
  190. - Features for developers:
  191. * Solvers can now expose requirements
  192. * unbounded reachability and reachability rewards now correctly respect solver requirements
  193. * Environment variables (such as the solver precisions) can now be handled more flexible
  194. * changes to Matrix-Vector operation interfaces, in particular fixed some issues with the use Intel TBB
  195. Version 1.1.x
  196. -------------
  197. ### Version 1.1.0 (2017/8)
  198. - Support for long-run average rewards on MDPs and Markov automata using a value-iteration based approach.
  199. - Storm can now check MDPs and Markov Automata (i.e. MinMax equation systems) via Linear Programming.
  200. - Parametric model checking is now handled in a separated library/executable called `storm-pars`.
  201. - Wellformedness constraints on PMCs:
  202. * include constraints from rewards
  203. * are in smtlib2
  204. * fixed
  205. * computation of only constraints without doing model checking is now supported
  206. - Fix for nested formulae
  207. - JANI: Explicit engine supports custom model compositions.
  208. - Support for parsing/building models given in the explicit input format of IMCA.
  209. - Storm now overwrites files if asked to write files to a specific location.
  210. - Changes in build process to accommodate for changes in carl. Also, more robust against issues with carl.
  211. - `USE_POPCNT` removed in favor of `FORCE_POPCNT`. The popcnt instruction is used if available due to `march=native`, unless portable is set.
  212. Then, using `FORCE_POPCNT` enables the use of the SSE 4.2 instruction
  213. Version 1.0.x
  214. -------------
  215. ### Version 1.0.1 (2017/4)
  216. - Multi-objective model checking support now fully included
  217. - Several improvements in parameter lifting
  218. - Several improvements in JANI parsing
  219. - Properties can contain model variables
  220. - Support for rational numbers/functions in decision diagrams via sylvan
  221. - Elimination-based solvers (exact solution) for models stored as decision diagrams
  222. - Export of version and configuration to cmake
  223. - Improved building process
  224. ### Version 1.0.0 (2017/3)
  225. Start of this changelog