An Initiative towards Open Network-on-Chip Benchmarks

1
An Initiative towards Open Network-on-Chip
Benchmarks

• C. Grecu, A. Ivanov (Univ. of British Columbia),
  P. Pande (Washington State Univ., US), A.
  Jantsch (Royal Institute of Technology, SE), E.
  Salminen (Tampere Univ. of Technology, FIN), U.
  Ogras, R. Marculescu (Carnegie Mellon University,
  US)

VTS 2007
2
Outline

• Who NoC WG
• Why Objectives and Benefits
• What Overview of NoC Benchmarks
• Performance Benchmarks
• Test and DFT Benchmarks
• Fault-Tolerance/Reliability Benchmarks
• When and Where
• Agenda (2007)
• Summary

3
NoC Challenges and Opportunities

• Engineering Knowledge and Practice
• Worldwide, industrial and academic
• Quality
• Increased product quality and performance at
  lower cost
• Evaluation
• Comparison and tradeoff
• Automation
• Standards, CAD, support and maintenance
• Adoption
• New products, new applications
• Deployment
• Manufacturing and sales

4
Who

• Academic participants (to date)
• C. Grecu, A. Ivanov (UBC - Canada)
• R. Marculescu, U. Ogras (CMU - USA)
• A. Jantsch (KTH - Sweden)
• P. Pande (WSU - USA)
• E. Salminen, A. Kulmala (TUT - Finland)
• Industrial Participation (to date)
• Sonics
• Nokia
• Toshiba
• Administrative/Logistical
• OCP-IP

5
Where and When

• Globally
• Bi-weekly teleconference
• Occasional face-to-face meetings
• E.g., DATE

6
NoC Benchmarks Benefits

• Goals of NoC WG are specifically to increase and
  accelerate
• Worldwide understanding of NoC paradigm in
  industry and academe
• Engineering and scientific development
  disseminated through technical literature
• General advancement of knowledge and state of the
  art and practice
• Quality of NoC solutions
• Enabling higher performance/low-cost products
• Evaluation comparison of NoC solutions
• Enabling fair and objective competition (academic
  and industrial)
• Methodology development automation (CAD) for
  NoC solutions
• Enabling reproducibility and porting from
  platform to platform, company to company,
• Open standards development
• Adoption
• Fostering management confidence, increase
  know-how and availability of highly qualified
  personnel
• Deployment
• Assessing and designing for manufacturability,
  quality and reliability of NoC-based products

7
NoC Benchmarks General Objectives

• Repository of standard NoC Information
• Representation Format Design Data
• Representation Format
• Description rules and requirements"
• Simulation Platforms/Parameters
• Metrics and Measurement Methodology
• Design Data
• Design specifications (black box)
• IP cores (nature, number)
• Design specifications (white box)
• Topology (NoC communication fabric)
• Data Traffic Models/Data
• Synthetic, application specific
• Interface information
• Above to be captured through Specifications/Deliv
  erables Document (in progress)

8
Challenge(s)

• Achieving goals under constraints
• IP protection/sensitivities
• Hardware
• Protocols
• Traffic
• Application-specific
• Relevance
• Abstraction
• Required complexity

9
Performance Benchmarks

• Benchmark Programs
• Programs or models mimicking real applications
  jointly exercising communication architecture
  (NoC) interfaced to the processing (computation)
  platform/elements as well as the system design
  methodology
• Useful for assessing effectiveness of particular
  NoC for given application or application domain
• Micro-Benchmarks
• Abstracted model aimed at exercising only a
  specific aspect of an NoC
• E.g., routing algorithms

10
Benchmark Programs

• Programs written in C, SystemC, VHDL .
• Information provided
• Functionality
• Application model
• Mapping and scheduling of application tasks to
  PEs
• Set of models for target PEs
• Usage
• Directions to connect PEs to NoC
• Instructions for configuration and compilation,
  and execution
• Topology and Mapping (Optional)
• Size and topology of NoC
• Structure and number of routers
• Resource (PE) binding
• Benchmark programs aim at aggregate performance
  of NoC

11
Performance Benchmarks How

• Propose to use Communication-Centric Application
  Modeling
• Based on Communication Task Graphs (CTGs)
• Finite State Machines (FSMs) used to model PEs in
  real applications
• Communication tasks
• Computation tasks
• Allows for effectively hiding proprietary/sensitiv
  e information of specific (real) applications

12
Communication-Centric Application Model A
picture
(25,48)
(46,13)
1
2
3
9
(25,22)
(16,25)
(40,14)

• application model
• annotated CTG

(18,40)
7
8
11
(16,25)
(22,36)
(40,14)
(16,25)
(25,22)
4
5
6
10
(bandwidth, latency)

• application mapping
• tasks onto PEs

• computation architecture
• computation parameters

PEn
PE3
PE2
PE1
NoC communication medium

• communication architecture
• communication parameters and models

13
Micro-Benchmarks (?Bs)

• ?Bs focus on single aspect/parameter of NoC-based
  architecture
• Packets
• Delay, latency, bandwidth, jitter, power
  consumption
• Routing, switching, buffering, flow control
• Transactions
• Packetization, end to end flow control, streaming
• Protocol and interface block evaluation
• Traffic Environment
• Temporal and spatial traffic distributions
• Congestion, arbitration, buffering, flow control
  mechanisms
• QoS
• best effort traffic
• Guaranteed services
• Scalability (network size topology)

14
Performance Benchmarks

• Additional Features
• Measurement point and methodology specification
• Interfaces and Sockets
• Flexibility (built-in through specification
  format)
• Core-centric and interconnect agnostic
• Allow Network Interfaces to deliver standard
  signals to NoC fabric
• E.g., OCP, AXI ,

15
NoC Test Benchmarks

• General Objective (Why)
• To measure efficiency of test methods and DFT
  architectures on NoCs-based systems
• e.g.,
• TAM design
• Test wrapper design
• Test scheduling

16
Test Benchmarks - Requirements

• Coverage Metrics Definition
• Fault models and set
• Low-level (stuck-at, open/shorts, crosstalk)
• High-level (mis-routing, data corruption, packet
  loss)
• Test modes supported (off-line, on-line)
• Test Type Definition
• Structural
• Fabrication correctness
• Functional Parametric
• Interaction between cores (computation) and data
  transport infrastructure (communication)
• Captured as integral part of NoC Test Benchmarks
• Enables Test Methodology Efficiency Evaluation
• How Apply to set of exemplary NoC-based systems
• What figure of merit parameters, e.g., test
  time, power, coverage, silicon area, etc.

17
NoC Test Benchmarks

• Input Standard, reproducible NoC Test Circuits
• Superset of ITC02 SoC Test Benchmarks
• Format and circuits
• Augmented/altered
• NoC fabric(s)
• Size, etc.
• Necessary information of NoC fabric(s)
• Connectivity
• Topology
• Components switches, routers, buffers
• Test-related data
• Number of scan-chains per router
• Number and size of buffers
• Number of tests sessions (per router, channel)
• Per test session number of test patterns, power
  dissipation, etc.

18
Test Benchmarks NoC Fabric Description

• XML-based
• Structured
• Portable
• Flexible
• Test info for
• Routers
• Channels (links)
• Inter-component connectivity routers cores
• TAM
• Versatile scenarios
• NoC reuse
• Dedicated
• Combination of above

19
NoC Benchmarks for On-line Test
Fault-Tolerance

• Objectives
• On-Line test efficiency
• Fault-tolerance qualities/capabilities
• Evaluated under the mission mode operating
  conditions of NoC
• Direct connection with Performance Benchmarks
• Required specifications
• NoC fabric (hardware) description
• Operating Conditions (traffic models, application
  data)
• Fault injection

www.ocpip.org/socket/whitepapers/
NoC-Benchmarks-WhitePaper-15.pdf
20
Benchmarks for On-Line Test Fault-Tolerance
Fault Injection Methodology

• Permanent and transient faults
• Fault models
• low-level (stuck-at, open, cross-talk, bit-flip)
• high-level (mis-routing, data corruption, packet
  loss)
• Assign
• permanent fault probabilities to NoC components
• transient fault probabilities to NoC data
• Account for defect clustering and burst errors

Pf (channel)
Pf (data) - message, packet, flit, bit -
21
Benchmarks for On-Line Test Fault-Tolerance

• On-line/Fault-Tolerance Quality Metrics
• Performance
• Detection (Error/Fault) coverage
• Error/Failure recovery
• Detection/Recovery time
• Cost/Performance Impact
• Si area
• Complexity
• QoS degradation
• Power dissipation

22
Agenda -- 2007

• Q1
• Public release of White Paper
• Recruitment of active industrial participants on
  WG
• Development of Specifications Document
• Q2
• Public disclosures at DATE NoC Workshop, VTS, NoC
  Symposium
• Completion of Specifications Document
• Initial development and implementation of
  benchmarks
• Performance
• Test and Fault Tolerance
• Q3
• Public release of Specifications Document
• Final development and implementation of
  Benchmarks (Release 1)
• Q4
• 1st release (beta) of benchmarks
• Initial evaluation and reports of benchmarks

23
Summary

• Help WANTED Welcome!
• Opinions/Contributions
• NoC Circuits
• Programs
• Traffic Data/Models
• Applications
• Special acknowledgements to OCP-IP

24
THANK YOU!