University of Cambridge > Talks.cam > Computer Laboratory Computer Architecture Group Meeting > Spatial Computation

Spatial Computation

Add to your list(s) Download to your calendar using vCal

If you have a question about this talk, please contact Robert Mullins.

In order to improve energy efficiency, computer architects are increasingly relying on ‘custom computing’ by incorporating ASIC or reconfigurable hardware implementations of important applications or code regions into their designs. Spatial Computation is a new computational model that permits fully automated, direct compilation of code to hardware from an unmodified and unconstrained high level language like C. Unfortunately, spatial computation suffers from poor performance in the presence of complex control-flow, limiting its usefulness for general-purpose computation.

The first half of this talk is a tutorial on the spatial computation model, discussing the program intermediate representation currently used, as well as the execution model for SC followed by a discussion of reasons why this model currently offers poor performance on complex control-flow code.

The second half of the talk discusses a means of overcoming these performance issues by using a new intermediate representation designed for spatial computation, based on the Value-State Dependence Graph. This new IR exposes greater ILP by relaxing control-flow ordering constraints. An overview of the compiler infrastructure required for generating and validating this spatial representation is also presented.

This talk is part of the Computer Laboratory Computer Architecture Group Meeting series.

Tell a friend about this talk:

This talk is included in these lists:

Note that ex-directory lists are not shown.

 

© 2006-2020 Talks.cam, University of Cambridge. Contact Us | Help and Documentation | Privacy and Publicity