Presentation
Rethinking the Usage of Self-Contained Reusable Components
DescriptionReusing well designed and verified components is an essential part of any complex chip design. It allows on-time delivery. Reusable components are usually self-contained: majority of parts needed to make the reused component functional is internal to the component. This results in majority of work focused on hooking up the component interface, avoiding the need to instantiate many external components and hooking them up manually. Sub-optimality is introduced in this process when some of the internal components can be shared with other instances of the same component or other components. With Moore's law not applicable anymore, new innovations are needed to achieve what Moore's law provided with respect to area and power optimization. At IBM, one of these methods is to rethink the designing of reusable components and how they are instantiated in the design. In this paper, we present a technique of 1) breaking common components into functionality specific and shared components, and 2) Automatically weaving these components together. This resulted in significant power and area savings on current IBM Power and Z processor chips.
Event Type
Engineering Track Poster
TimeTuesday, July 11th5:00pm - 6:00pm PDT
LocationLevel 2 Exhibit Hall
Back-End Design
Embedded Systems
Front-End Design
IP
RISC-V