#### DMCA

## Allocation and scheduling for mpsocs via decomposition and no-good generation (2005)

### Cached

### Download Links

- [www.ijcai.org]
- [ijcai.org]
- [ijcai.org]
- [www-lia.deis.unibo.it]
- [ijcai.org]
- [ijcai.org]
- [lia.deis.unibo.it]
- DBLP

### Other Repositories/Bibliography

Venue: | In Procs. of the 11th Intern. Conference on Principles and Practice of Constraint Programming - CP 2005 |

Citations: | 27 - 13 self |

### Citations

698 |
Parallel Computer Architecture: A Hardware/Software Approach
- Culler, Singh, et al.
- 1998
(Show Context)
Citation Context ...hers in digital design automation have explored complete approaches for allocating and scheduling pre-characterized workloads on MPSoCs [1], instead of using on-line, dynamic (sub-optimal) schedulers =-=[5, 6]-=-. The multi-processor system we consider consists of a pre-defined number of distributed computation nodes, as depicted in Figure 1. All nodes are assumed to be homogeneous and made by a processing co... |

578 |
Synthesis and Optimization of Digital Circuits
- Micheli
- 1994
(Show Context)
Citation Context ...of such applications, and this explains why research efforts have been devoted to extending mapping and scheduling techniques to pipelined task graphs. An overview of these techniques is presented in =-=[13]-=-. IP formulations as well as heuristic algorithms have been traditionally employed. In [14] a retiming heuristic is used to implement pipelined scheduling, that optimizes the initiation interval, the ... |

377 | Partitioning procedures for solving mixed-variables programming problems - Benders - 2001 |

258 | Reconfigurable computing: a survey of systems and software,”
- Compton, Hauck
- 2002
(Show Context)
Citation Context ...hers in digital design automation have explored complete approaches for allocating and scheduling pre-characterized workloads on MPSoCs [1], instead of using on-line, dynamic (sub-optimal) schedulers =-=[5, 6]-=-. The multi-processor system we consider consists of a pre-defined number of distributed computation nodes, as depicted in Figure 1. All nodes are assumed to be homogeneous and made by a processing co... |

126 | The Future of Multiprocessor Systems-on-Chips. - Wolf - 2004 |

104 | A.: System Level Hardware/Software Partitioning Based on Simulated Annealing and Tabu Search, Design Automation for Embedded Systems,
- Eles, Peng, et al.
- 1997
(Show Context)
Citation Context ... reason, heuristic approaches are widely used. A comparative study of well-known heuristic search techniques (genetic algorithms, simulated annealing and tabu search) is reported in [18]. Eles et al. =-=[19]-=- compare the use of simulated annealing and tabu search for partitioning a graph into hardware and software parts while trying to reduce communication and synchronization between parts. More scalable ... |

98 | Algorithms for hybrid MILP/CP models for a class of optimization problems,” - Jain, Grossmann - 2001 |

95 |
Synthesis of Application-Specific Heterogeneous Multiprocessor Systems.
- Prakash, Parker
- 1992
(Show Context)
Citation Context ...-processor systems have been traditionally modelled as integer linear programming problems. An early example is represented by the SOS system, which used mixed integer linear programming (MILP) model =-=[8]-=-. SOS considers processor nodes with local memory, connected through direct point-to-point channels. The algorithm does not consider real-time constraints. Partitioning under timing constraints has be... |

71 | Logic-based benders decomposition.
- Hooker, Ottosson
- 2003
(Show Context)
Citation Context ...ssignment of tasks to machines, differently from our problem, where contributes to the objective function depend on pairs of assignments. In [3] and [7], Logic-Based BD (a variant of BD introduced in =-=[29]-=-, where the sub-problem should not necessarily be an IP problem) is used for Planning and Scheduling problems. Here different objective functions are considered: total cost minimization, makespan, tar... |

66 | Scheduling of Conditional Process Graphs for the Synthesis of Embedded Systems.
- Eles, Kuchcinski, et al.
- 1998
(Show Context)
Citation Context ...ation between parts. More scalable versions of these algorithms for large real-time systems are introduced in [20]. Many heuristic scheduling algorithms are variants and extensions of list scheduling =-=[21]-=-. Heuristic approaches provide no guarantees about the quality of the final solution. On the other hand, complete approaches which compute the optimum solution (possibly, with a high computational cos... |

50 |
Embedded System Synthesis by Timing Constraints Solving.
- Kuchcinski
- 1997
(Show Context)
Citation Context ...s can be obtained for small problems, while large problems require use of heuristics. The proposed framework is able to create pipelined implementations in order to increase the design throughput. In =-=[23]-=- the embedded system is represented by a set of finite domain constraints defining different requirements on process timing, system resources and interprocess communication. The assignment of processe... |

44 | A hybrid method for planning and scheduling. - Hooker - 2005 |

37 | Planning and Scheduling by Logic-Based Benders Decomposition,
- Hooker
- 2007
(Show Context)
Citation Context ...evice. These are the simplest kind of no-goods that can be added to the master problem since they state that the current solution must not be computed again. Even if they can be improved, as shown in =-=[7]-=-, we will show in the next section that they are very effective. 5 Experimental results To validate the strength of our approach, we now compare the results obtained using this model (Hybrid in the fo... |

37 | Branch-and-check: A hybrid framework integrating mixed integer programming and constraint logic programming
- Thorsteinsson
- 2001
(Show Context)
Citation Context ... is applied to an allocation and scheduling problem; the master problem (allocation) is based on CP and the sub-problem (scheduling) is solved using a real-time scheduler with fixed task priority. In =-=[28]-=- the branch and check framework is proposed using Benders Decomposition. This technique is applied to the problem of scheduling orders on dissimilar parallel machines, where a set of tasks, linked by ... |

34 | Architecture synthesis and partitioning of real-time systems: A comparison of three heuristic search strategies
- Axelsson
- 1997
(Show Context)
Citation Context ...nal cost. For this reason, heuristic approaches are widely used. A comparative study of well-known heuristic search techniques (genetic algorithms, simulated annealing and tabu search) is reported in =-=[18]-=-. Eles et al. [19] compare the use of simulated annealing and tabu search for partitioning a graph into hardware and software parts while trying to reduce communication and synchronization between par... |

29 | MILP based task mapping for heterogeneous multiprocessor systems
- Bender
- 1996
(Show Context)
Citation Context ...r timing constraints has been addressed in [9]. A MILP model that allows to determine a mapping optimizing a trade-off function between execution time, processor and communication cost is reported in =-=[10]-=-. Extensions of the IP formulation have also been used to account for memory allocation requirements, besides communication and computation ones. A hardware/software co-synthesis algorithm of distribu... |

29 | Improving Scalability of Task Allocation and Scheduling in Large Distributed Real-time Systems using Shared Buffers.
- Kodase, Wang, et al.
- 2003
(Show Context)
Citation Context ...graph into hardware and software parts while trying to reduce communication and synchronization between parts. More scalable versions of these algorithms for large real-time systems are introduced in =-=[20]-=-. Many heuristic scheduling algorithms are variants and extensions of list scheduling [21]. Heuristic approaches provide no guarantees about the quality of the final solution. On the other hand, compl... |

28 |
Constraint and Integer Programming: Toward a Unified Methodology.
- Milano
- 2004
(Show Context)
Citation Context ...ominates the other in terms of computational performance. The development of a hybrid CP-IP solver that captures the best features of both would appear to offer scope for improved overall performance =-=[24]-=-. However, the issue of communication between different modelling paradigms arises. One method is inherited from the Operations Research and is known as Benders Decomposition [2]: it is proved to conv... |

28 |
Logic circuit verification by Benders decomposition,
- Hooker, Yan
- 1995
(Show Context)
Citation Context ...is known as Benders Decomposition [2]: it is proved to converge producing the optimal solution. Benders Decomposition (BD) technique has been extensively used to solve a large variety of problems. In =-=[25]-=- BD is applied to a numeric algorithm in order to solve the problem of verifying logic circuits: results show that, for some kind of circuits, the technique is an order of magnitude faster w.r.t. othe... |

27 | Hardware/Software Co-Synthesis with Memory Hierarchies
- Li, Wolf
- 1999
(Show Context)
Citation Context ...n and computation ones. A hardware/software co-synthesis algorithm of distributed real-time systems that optimizes the memory hierarchy (caches) along with the rest of the architecture is reported in =-=[11]-=-. An integer IP CPs12 linear programming model is used in [12] to obtain an optimal distributed shared memory architecture minimizing the global cost to access shared data in the application, and the ... |

27 | Hybrid Benders Decomposition Algorithms in Constraint Logic Programming’. In:
- Eremin, Wallace
- 2001
(Show Context)
Citation Context ...thm in order to solve the problem of verifying logic circuits: results show that, for some kind of circuits, the technique is an order of magnitude faster w.r.t. other state of the art algorithms. In =-=[26]-=-, BD is embedded in the CP environment ECLiPSe and is shown that it can be useful in practice. There are a number of papers using Benders Decomposition in a CP setting. In [27] BD is applied to an all... |

23 | An optimal memory allocation for application-specific multiprocessor system-on-chip
- Meftali, Gharsalli, et al.
(Show Context)
Citation Context ...ithm of distributed real-time systems that optimizes the memory hierarchy (caches) along with the rest of the architecture is reported in [11]. An integer IP CPs12 linear programming model is used in =-=[12]-=- to obtain an optimal distributed shared memory architecture minimizing the global cost to access shared data in the application, and the memory cost. The above techniques lead to static allocations a... |

23 | Static scheduling of pipelined periodic tasks in distributed real-time systems
- Fohler, Ramamritham
- 1995
(Show Context)
Citation Context ...t optimizes the initiation interval, the number of pipeline stages and memory requirements of a particular design alternative. Pipelined execution of a set of periodic activities is also addressed in =-=[15]-=-, for the case where tasks have deadlines larger than their periods. Palazzari et al. [16], focus on scheduling to sustain the throughput of a given periodic task set and to serve aperiodic requests a... |

21 | Hardware-software partitioning and pipelined scheduling of transformative applications,”
- Chatha, Vemuri
- 2002
(Show Context)
Citation Context ... mapping and scheduling techniques to pipelined task graphs. An overview of these techniques is presented in [13]. IP formulations as well as heuristic algorithms have been traditionally employed. In =-=[14]-=- a retiming heuristic is used to implement pipelined scheduling, that optimizes the initiation interval, the number of pipeline stages and memory requirements of a particular design alternative. Pipel... |

20 | A constructive algorithm for memoryaware task assignment and scheduling.
- Szymanek, Kuchcinski
- 2001
(Show Context)
Citation Context ...d applied throughout the entire lifetime of the system. Constraint Programming (CP) is an alternative approach to Integer Programming (IP) for solving combinatorial optimization problems. The work in =-=[22]-=- is based on Constraint Logic Programming to represent system synthesis problem, and leverages a set of finite domain variables and constraints imposed on these variables. Optimal solutions can be obt... |

14 | Decomposition and learning for a hard real time task allocation problem
- Cambazard, Hladik, et al.
(Show Context)
Citation Context ... the art algorithms. In [26], BD is embedded in the CP environment ECLiPSe and is shown that it can be useful in practice. There are a number of papers using Benders Decomposition in a CP setting. In =-=[27]-=- BD is applied to an allocation and scheduling problem; the master problem (allocation) is based on CP and the sub-problem (scheduling) is solved using a real-time scheduler with fixed task priority. ... |

11 |
D.D.: A scheduling and pipelining algorithm for hardware/software systems
- Bakshi, Gajski
- 1997
(Show Context)
Citation Context ...to serve aperiodic requests associated with hard real-time constraints. Mapping of tasks to processors, pipelining of system specification and scheduling of each pipeline stage have been addressed in =-=[17]-=-, aiming at satisfying throughput constraints at minimal hardware cost. In general, even though IP is used as a convenient modelling formalism, there is consensus on the fact that pure IP formulations... |

7 |
Synthesis of pipelined systems for the contemporaneous execution of periodic and aperiodic tasks with hard real-time constraints
- Palazzari, Baldini, et al.
- 2004
(Show Context)
Citation Context ... of a particular design alternative. Pipelined execution of a set of periodic activities is also addressed in [15], for the case where tasks have deadlines larger than their periods. Palazzari et al. =-=[16]-=-, focus on scheduling to sustain the throughput of a given periodic task set and to serve aperiodic requests associated with hard real-time constraints. Mapping of tasks to processors, pipelining of s... |

6 |
System-Level Synthesis of Application-Specic Systems Using A* Search and Generalized ForceDirected Heuristics
- Lee, Potkonjak, et al.
- 1996
(Show Context)
Citation Context ...processor nodes with local memory, connected through direct point-to-point channels. The algorithm does not consider real-time constraints. Partitioning under timing constraints has been addressed in =-=[9]-=-. A MILP model that allows to determine a mapping optimizing a trade-off function between execution time, processor and communication cost is reported in [10]. Extensions of the IP formulation have al... |