SDI Seminar

Speaker: Professor Brian Smith, Cornell University

Video as a High Level Data Type

Date: November 7, 1996


Developing programs that process multimedia data is time consuming, tedious, and error prone. Problems include the gap between the semantic level of designing algorithms and writing code, the constant tension between high performance and general purpose implementations, and the plethora of multimedia data formats and compression standards.

To address some of these problems, we developed a language for multimedia data processing called Rivl (pronounced "rival"). Rivl contains primitives to manpulate audio, images, and video as first class data objects. The Rivl run-time environment optimizes the execution of Rivl programs, providing many of the performance benefits of special-purpose code while retaining the benefits of programming at a high semantic level. This talk will describe the Rivl language, the optimizations used in its run-time system, and our future plans for the language.