G. N. Srinivasa Prasanna

Learn More
Thk paper presents a novel approach to multiprocessor scheduling using the theory of optimal control. We start by assuming that every task can be executed in parallel on an arbitrary number of processors, and that a continuously variable number of processors can be assigned to the task at any time (by time-sharing). The optimal scheduling problem can then(More)
This paper considerably extends the multiprocessor scheduling techniques in [1], and applies it to matrix arithmetic compilation. In [1] we presented several new results in the theory of homogeneous multiprocessor scheduling. A directed acyclic graph (DAG) of tasks is to be scheduled. Tasks are assumed to be parallelizable - as more processors are applied(More)
—Financial transactions are specified in decimal arithmetic. Until the introduction of IEEE 754-2008, specialized software/ hardware routines were used to perform these transactions but it incurred a penalty on performance. In this paper, we show that if binary arithmetic is used to emulate decimal operations, then arbitrary error sequences can be generated(More)
Multimedia applications (also called multimedia systems) operate on datastreams, which are periodic sequences of data elements, called <i>datasets</i>. A large class of multimedia applications is described by the macro-dataflow graph model, with nodes representing parallelizable tasks, and arcs representing communication. This paper examines how such(More)