1. Denis Caromel1 Troisieme partie Cours EJC 2003, AUSSOIS, www.inria.fr/oasis/ProActive Denis Caromel OASIS Team INRIA -- CNRS - I3S -- Univ. of Nice Sophia-Antipolis EJCP, Lundi 2 juin 2003 ASP: Asynchronous Sequential Processes An Imperative and Parallel Object Calculus

2. Denis Caromel2 The ASP calculus: Asynchronous Sequential Processes Together with Ludovic Henrio, and Bernard Serpette Objectives: Formally study the ProActive model Investigate various strategies for asynchronous calls Prove some equivalence between Sequential and Parallel programs Demonstrate the deterministic nature of sub-sets of the model

3. Denis Caromel3 ASP principles and Sequential Calculus Imperative sigma-calculus a la Abadi-Cardelli Sequential Language: Example: a point object A call to a point:

4. Denis Caromel4 Principles of Reduction

5. Denis Caromel5 The Heap -- Store Reduced objects: all fields are reduced to a location Heap Reference: Locations Heap Sequential Configuration Heap Well Formed properties

6. Denis Caromel6 Reduction Context From A. D. Gordon, P. D. Hankin, and S. B. Lassen, Compilation Equivalence of Imperative Objects, FST+TCS’97, LNCS Reduction context = an expression with a hole : provides a deterministic order of evaluation: Target object before parameters Fields of objects: left to right Assignment: left before right Abbreviation of substitution

7. Denis Caromel7 Sequential Transition System

8. Denis Caromel8 Sequential TS Properties From Gordon-Hankin-Lassen: Sequential reduction preserves well-formdness: Sequential reduction is deterministic (modulo renaming):

9. Denis Caromel9 Parallel calculus Language: Intuitive semantics: Object activation: copy together with a deep copy of its reachable objects Asynchronous calls: a Request Queue per Activity M is a list of method labels (names) used to specify the request to be served A list of already computed futures per activity At any time, a full or partial reply may be send back (not eager)

10. Denis Caromel10 Parallel Binary Tree

11. Denis Caromel11 Example of a parallel configuration

12. Denis Caromel12 Structure of Activities (3) Parallel configurations: name of activity [ a current expression being evaluated its own heap a reference towards the Active Object in the heap (root) a list of request to execute in FIFO order: a list of futures already computed (future Id --> obj in heap) A current future: Id associated to the current service

13. Denis Caromel13 3 additions to the Sequential Calculus In the store: Remote Objects (proxy) and Futures (proxy for futures) Reduction contexts: to reduce within an Active expression

14. Denis Caromel14 Deep (Parallel) Copy Basic ideas: upon an active object OA(alpha), or a future Fut(f beta->alpha ) the parallel copy stops copying, but pass the OA or Fut as is It means we pass futures as parameters of communications Transition System: Deep Copy (more operational def. exists)

15. Denis Caromel15 Parallel Transition System

16. Denis Caromel16 Properties The PTS is correct with respect to the well formed properties:

17. Denis Caromel17 Store partitioning

18. Denis Caromel18 A confluence property (1)

19. Denis Caromel19 A confluence property (2)

20. Denis Caromel20 A confluence property (3)

21. Denis Caromel21 A confluence property (4)

22. Denis Caromel22 A confluence property (5)

23. Denis Caromel23 Deterministic Object Networks (1)

24. Denis Caromel24 Deterministic Object Networks (2)

25. Denis Caromel25 Tree Topology Determinism

26. Denis Caromel26 Conclusion A library:ProActive 100% Java Parallelism, distribution, synchronization (CSCW), and mobility Reuse -- seamless Polymorphism with existing class types Asynchrony -- Wait-by-necessity An interactive tool towards Metacomputing: IC2D A calculus:ASP: Asynchronous Sequential Processes Capture the semantics, and demonstrates the independence of activities Results of confluence and determinism Mobility to be added ProActive vs. RMI alone : - 30% of code www.inria.fr/oasis/ProActive