R. Govindarajan, N.S.S. Narasimha Rao, E. R. Altman, and G. R. Gao. An enhanced co-scheduling method using reduced ms-state diagrams. In Proc. of the 12th Intl. Parallel Processing Symp., Orlando, FL, Mar. 1998. IEEE Computer Society. Merged with 9th Intl. Symp. on Parallel and Distributed Processing.  Home/Search   Document Details and Download   Summary   Related Articles   Check

.... an interesting approach that uses finite state automata for modeling complex resource constraints in instruction scheduling [15] Their approach was based on [13] and subsequently improved and applied to production compilers in [2] This approach was extended to software pipelining methods in [6, 7]. This paper focuses on the efficient construction of finite state automata for software pipelining methods. In the above methods [2, 6, 7, 13, 15] processor resources are modeled using a finite state automata (or state diagram) which is constructed from the resource usage table for each ....

....approach was based on [13] and subsequently improved and applied to production compilers in [2] This approach was extended to software pipelining methods in [6, 7] This paper focuses on the efficient construction of finite state automata for software pipelining methods. In the above methods [2, 6, 7, 13, 15], processor resources are modeled using a finite state automata (or state diagram) which is constructed from the resource usage table for each instruction (class) Each path in the state diagram represents a legal latency sequence which could be directly used by the scheduler for modeling resource ....

[Article contains additional citation context not shown here]

R. Govindarajan, N.S.S. Narasimha Rao, E. R. Altman, and G. R. Gao. An enhanced co-scheduling method using reduced ms-state diagrams. In Proc. of the 12th Intl. Parallel Processing Symp., Orlando, FL, Mar. 1998. IEEE Computer Society. Merged with 9th Intl. Symp. on Parallel and Distributed Processing.

.... of Modulo Schedule Pipelines A formal method which analyzes hardware pipelines operating under modulo scheduling (henceforth referred to as Modulo Scheduled Pipelines, or MS Pipelines) and establishes the maximum 3 number of initiations was developed by us for the co scheduling framework in [4, 6]. In MSpipelines, the resource usage is represented by a modified form of the reservation table, called the Cyclic Reservation Table (CRT) The CRT has exactly II columns. If the original reservation table had fewer than II columns, then blank columns are appended at the end to get II columns. ....

....8 ideas from hardware pipeline theory have been used to develop a Finite State Automaton (FSA) based approach to instruction scheduling [1, 10, 13, 14] However, these techniques are applied to instruction scheduling, while our method focuses on software pipelining. Several examples are given in [4, 6] to illustrate that neither classical pipeline theory nor the FSA based approaches can directly be used to analyze pipeline structures operating under modulo scheduling. In this paper we have developed the MS pipeline theory for architectures in which the processors do not share any resources. ....

[Article contains additional citation context not shown here]

R. Govindarajan, N.S.S. Narasimha Rao, E.R. Altman, and G.R. Gao. An enhanced Co-Scheduling method using reduced MS-State diagram. In Proc. of the 12th International Parallel Processing Symposium and 9th Symposium on Parallel and Distributed Processing, Orlando, FL, March 1998.

....the paths that lead to distinct offset sets In our example, removing state S 2 and the arcs that are connected to it, does not result in any loss of information. This motivates us to study the theory of reduced MS state diagrams which involve only paths corresponding to distinct offset sets. In [25] we proposed the construction of reduced MS state diagrams which involves distinct paths. We establish the necessary theory for reduced MS state diagrams in this paper. 5.2 Definitions We begin with the following definitions. Definition 5.1 A path S 0 p1 Gamma S 1 p2 Gamma S 2 Delta ....

R. Govindarajan, N.S.S. Narasimha Rao, E. R. Altman, and G. R. Gao. An enhanced Co-scheduling method using reduced MS-state diagrams. In Proc. of the Merged 12th Intl. Parallel Processing Symp. and 9th Intl. Symp. on Parallel and Distributed Processing, Orlando, FL, Apr. 1998.

.... an interesting approach that uses finite state automata for modeling complex resource constraints in instruction scheduling [14] Their approach was based on [12] and subsequently improved and applied to production compilers in [2] This approach was extended to software pipelining methods in [6, 7]. This paper focuses on the efficient construction of finite state automata for software pipelining methods. In the above methods [2, 6, 7, 12, 14] processor resources are modeled using a finite state automata (or state diagram) which is constructed from the resource usage table for each ....

....approach was based on [12] and subsequently improved and applied to production compilers in [2] This approach was extended to software pipelining methods in [6, 7] This paper focuses on the efficient construction of finite state automata for software pipelining methods. In the above methods [2, 6, 7, 12, 14], processor resources are modeled using a finite state automata (or state diagram) which is constructed from the resource usage table for each instruction (class) Each path in the state diagram represents a legal latency sequence which could be directly used by the scheduler for modeling resource ....

[Article contains additional citation context not shown here]

R. Govindarajan, N.S.S. Narasimha Rao, E. R. Altman, and G. R. Gao. An enhanced co-scheduling method using reduced ms-state diagrams. In Proc. of the 12th Intl. Parallel Processing Symp., Orlando, FL, Mar. 1998. IEEE Computer Society. Merged with 9th Intl. Symp. on Parallel and Distributed Processing.

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC