This paper addresses the problem of mapping a feedforward ANN onto a multiple bus system, MBS, with p processors and b buses so as to minimize the total execution time. We present an algorithm which assigns the nodes ...
详细信息
This paper addresses the problem of mapping a feedforward ANN onto a multiple bus system, MBS, with p processors and b buses so as to minimize the total execution time. We present an algorithm which assigns the nodes of a given computational layer (c-layer) to processors such that the computation lower bound inverted right perpendicular N-l/p inverted left perpendicular t(p)(l) and the communication lower bound inverted right perpendicular N-l/b inverted left perpendicular t(c) are achieved simultaneously, where N-l is the number of nodes in the mapped c-layer l and t(p)(l) and t(c) are the computation and communication times, respectively, associated with a node in the layer. When computation and communication are not overlapped, we show that the optimal number of processors needed is either 1 or p, depending on the ratio t(p)(l)/t(c). When computation and communication are overlapped, we show that the optimal number of processors needed is either 1 or (inverted right perpendicular t(p)(l)/t(c) inverted left perpendicular)b. We show that there is a unique arrangement of interlaces such that the total number of interlaces is minimum and the optimal time is reached. Finally, we compare the relative merits of the MBS simulating ANNs over the recently introduced checkerboarding scheme.
Karp and Miller's computation graphs have been widely studied because they are a useful abstraction of multiprocessor computation. A system has the determinacy property if whenever the same set of input streams ar...
详细信息
Karp and Miller's computation graphs have been widely studied because they are a useful abstraction of multiprocessor computation. A system has the determinacy property if whenever the same set of input streams are entered into the system, the resultant output is the same set of output streams. Karp and Miller showed that computation graphs have this property;Woo, Smith, and Agrawala showed that data flow schema also have the property of determinacy. We define a more general model than either of the foregoing, called a multiple component system model (MCSM). We prove that MCSM's have the determinacy property. We then define deadlocking and prove some additional properties of deadlocking MCSM's.
A generalization of Petri nets and vector addition systems, called GPN and MGPN, is introduced in this paper. Termination properties of this generalized formalism are investigated. Four subclasses of GPN’s are consid...
详细信息
A generalization of Petri nets and vector addition systems, called GPN and MGPN, is introduced in this paper. Termination properties of this generalized formalism are investigated. Four subclasses of GPN’s are considered. We distinguish forward-conflict-free, backward-conflict-free, forward-concurrent-free and backward-concurrent-free GPN’s. We also study the concept of strongly connected, strongly repetitive and conservative GPN’s. The main results obtained are (i) every strongly connected, strongly repetitive and forward- (or backward-) conflict-free GPN must be conservative, and (ii) every strongly connected, conservative and forward- (or backward-) concurrent-free GPN must be strongly repetitive. Since the class of forward-conflict-free GPN’s contains properly the structures of computation graphs of Karp and Miller and marked graphs, some results appearing in these studies can be obtained as corollaries. Specialized versions of our results for the case of Petri nets are also included.
Main memory and intermemory data transmission rate requirements of two-level memory multiprocessor systems are characterized. Operational software is modeled by computation graphs. For a fixed processing schedule, ind...
详细信息
暂无评论