Adaptive SoC Operations Using Policy-Based System Control
 

Adaptive SoC Operations Using Policy-Based System Control

on

  • 127 views

As the complexity of SoC operations grow, they too are more resembling networks. For example, the concept of distributed caches with coherency, recently introduced as an innovation in SoC interconnect ...

As the complexity of SoC operations grow, they too are more resembling networks. For example, the concept of distributed caches with coherency, recently introduced as an innovation in SoC interconnect technology, resembles the queues in a network device. But where are the equivalent QoS or PBR “policies” for the SoC that are present in networks and provide the key adaptive decision making components?
ChipStart’s SSM represents a control plane for SoCs that operates based on software policies. SSM is a key subsystem IP component that can be added to any SoC to provide the key missing components to enable adaptive SoC operations.

Statistics

Views

Total Views
127
Views on SlideShare
127
Embed Views
0

Actions

Likes
0
Downloads
0
Comments
0

0 Embeds 0

No embeds

Accessibility

Categories

Upload Details

Uploaded via as Microsoft Word

Usage Rights

© All Rights Reserved

Report content

Flagged as inappropriate Flag as inappropriate
Flag as inappropriate

Select your reason for flagging this presentation as inappropriate.

Cancel
  • Full Name Full Name Comment goes here.
    Are you sure you want to
    Your message goes here
    Processing…
Post Comment
Edit your comment

Adaptive SoC Operations Using Policy-Based System Control Adaptive SoC Operations Using Policy-Based System Control Document Transcript

  • Adaptive SoC Operations Using Policy Based System ControlNetwork devices often include “tap points” dispersed across the network’s flows that collect analyticsfor monitoring and adapting the network’s behavior according to the actual usage, priority, and type ofcontent passing through it. Policies such as Quality of Service (QoS), Policy Based Routing (PBR), andeven Call Admission Control (CAC) can then be applied based on the analytics that form the policies forthe adaptation.As the complexity of SoC operations grow, they too are more resembling networks. For example, theconcept of distributed caches with coherency, recently introduced as an innovation in SoC interconnecttechnology, resembles the queues in a network device. But where are the equivalentQoS or PBR“policies” for the SoC that are present in networks and provide the key adaptive decision makingcomponents?ChipStart’s SSM represents a control plane for SoCs that operates based on software policies. SSM is akey subsystem IP component that can be added to any SoC to provide the key missing components toenable adaptive SoC operations.The figure above represents a typical implementation of a multicore SoC which contains the SSMSubsystem IP. Software policies are loaded in the SSM Controller, which in turn converts those policiesinto commands. These commands are sent to the SSM MCB’s via the SSM bus for further conversion tosignals and messages to the corresponding IP Blocks. However, since SSM supports bidirectionalcommunications, the IP Blocks, via the SSM MCBs, can also feedback state data to the SSM Controller viathe SSM bus. This creates the infrastructure for adaptation.For example, each of the data plane caches associated with the IP blocks can be monitored for cachemisses by the SSM MCBs and reported to the SSM Controller. The SSM Controller then can send therolled up view of cache utilization as a global view analytic to the host processor. The host processorselects the appropriate SSM policy from a set of policies optimized for use cases, a decision that is madein conjunction with the application requirements, and loads the policy into the SSM Controller memoryfor execution. The SSM Controller can then work together with the memory scheduler to betteroptimize data block retrieval and distribution, driven by the SSM policy. The result, improved cacheutilization and increased system performance. Alternatively more complex polices can be loaded thatallow the SSM Controller itself to make decisions based on operations conditions. minimizing hostprocessor participation.
  • While the main benefit is more effective execution of the application, this can also lead to improvedpower management (turning on and off IP blocks when caches are empty for example) and morepredictable error recovery.Another alternative is to add intelligence to the SSM MCBs themselves, localizing the monitoring anddecision making, which is globally managed by the SSM Controller. This is especially effective when theIP Blocks transition to IP subsystems and hierarchical interconnect structures become a reality. By usingcontrol plane policy commandsto drive arbitrationdecisions for all the interconnects, data path controlglobally across the SoC and within the subsystems themselves can be tied efficiently to applicationbehavior. This effectively creates policy based routing. Congestion can also be detected which in turncan trigger flow control, using a profile of subsystem behavior, and communication back to the hostprocessor would enable the application to adapt as well.SoC architectures which compliment complex data plane interconnects with control plane subsystemswill scale more efficiently and with higher operations reliability. SSM is the industry’s first merchantsubsystem IP designed for adapting control planes on SoCs while abstracting specific devicepersonalization to software policies. SSM has also been designed such that overhead is minimized andreal estate and power consumption required are both nominal.