assaf shacham, keren bergman, luca p. carloni presented for hpcan session by: millad ghane nocs’07

On the Design of a Photonic Network-on-Chip

Assaf Shacham, Keren Bergman, Luca P. Carloni

Presented for HPCAN Session by: Millad Ghane

NOCS’07

Why NoC ?

Scaling Transistor Speed and Integrity Tighter Logic Power Dissipation

Increasing Number of Cores per Chip Bottleneck: Global Intrachip Communications▪ Bandwidth▪ Power

Performance-per-watt

Why Optic ?

Low Power Dissipation Ultra-high Throughput Minimal Latency End-to-end Transmission

No Repeating No Regeneration No Buffering

A silicon ring

resonator

Hybrid Architecture

Photonic Interconnection High Bandwidth Messages

Electronic Interconnection Low Bandwidth Messages Short Control Messages

Photonic Properties

Advantages Bit-rate Transparency▪ Not Switching by Every Bit of Data▪ Switching Once per Message

Low Loss in Optical Waveguide▪ Independence of Transmission Distance

Disadvantages No Storage Element E/O and O/E Conversions▪ Off-chip Lasers

Packet Life

D

D

SSending path-setup packet

Electronic Network

Optic Network

S

Packet Life

D

S D

SSending optical packet

Packet Life

D

S D

SSending path-teardown packet

Packet Life

D

S D

SSending path-blocked packet

path-teardown packet

Building Blocks

70 µm

70 µm

Photonic Switching Element

Electrical Router

Building Blocks Problems

No Injection/Ejection Port Not Deadlock-Free

Wide Turns

Approach

Injection

Torus Address

Ejection

Gateway Switch

Injection Switch

Ejection Switch

Torus Network Nodes

Access Points Address

Format

Deadlock Solution

Injection-Ejection Blocking [previous slide]

Intra-dimentional Blocking (Torus Network) Virtual Channel Flow Control▪ Circuit Switching

terminate-on-timeout packet

Simulation Parameters

POINTS Simulator Based on OMNET++

36-core CMP 6x6 planar

Chip size: 20 x 20 mm Uniform Traffic 3 Cases

Deadlock Message Size Optimization Increasing Path Diversity

2

Message Size Optimization

Long path-setup latency Nonoseconds

Super fast transmission

Overhead Ratio:

Message Size Optimization (cont.)

Message Duration: 50ns Message Size: 2KB Suitable for

DMA Trans

Increasing Path Diversity

PD=2 Less Hardware Less Overhead

Difference

Q & A ?

Thanks !