Introducing the Hybrid Architecture for the DesignWare Interconnect Fabric for the AMBA 3 AXI protocol
The new Hybrid architecture for the DesignWare Interconnect Fabric for AMBA 3 AXI enables the combination of dedicated high performance channels with lower performance shared channels. By eliminating redundant logic and wires designers are able to realize lower area, power and reduced routing congestion. The new interconnect fabric architecture has been validated against the "AMBA 3 Assured" DesignWare Verification IP ensuring adherence to the AMBA 3 AXI Specification. Automated configuration and assembly with Synopsys' coreAssembler tool enables the IP to be rapidly deployed into the design environment.
Fig 1 - Traditional AMBA 3 AXI architecture: consists of individual dedicated buses that service specific master and slave devices
Up till now the standard AMBA 3 AXI interconnect fabric implementation as represented in Fig 1 targets the high performance SoC requirement by offering a multiple address, multiple data architecture which supports parallel traffic from independent masters and slaves. On a per master-slave link basis you have independent address, data and response channels, resulting in a full multiple address, multiple data architecture, capable of supporting even the most demanding system bus bandwidths requirements. While many component connections within a high performance SoC will require such a high bandwidth, there are also masters and slave connections which do not. For these system elements, a full multiple address multiple data architecture introduces a considerable amount of redundant logic which wastes die area, power and adds complexity to routing.
The new Hybrid architecture feature provides designers with a configurable, optimized architecture enabling redundancy to be eliminated, bringing about savings in area, power and routing congestion. The Hybrid architecture features allows low performance master slave links to be combined into single shared channels, eliminating the logic and wires associated with having dedicated channels for low performance masters and slaves. Using the hybrid architecture, high performance master-slave links are connected via dedicated channels, which ensures bandwidth requirements on these links can be met. Low performance master-slave links are connected via shared channels, resulting in reduced area, power and routing congestion within the interconnect while still meeting system performance requirements. The interconnects flexible configuration options enable the choice of dedicated or shared connection for each channel on each master-slave link within a single instance. The hybrid architecture provides a balance between area, performance and power which can be utilized in native AMBA 3 AXI based design and also minimizes the impact to the design when transitioning from an AMBA AHB based architecture to a pure AMBA 3 AXI based architecture.
For more information on the hybrid architecture and the other DesignWare Solutions for AMBA please visit: www.synopsys.com/amba
Fig 2 - Advanced hybrid architecture for the AMBA 3 AXI interconnect: provides a choice of connecting the master-to-slave link via a shared or dedicated bus within a single AMBA 3 AXI on-chip bus interconnect