Top Banner
Author : Jing Lin, Xiaola Lin, Liang Tang Publish Journal of parallel and Distributed Computing MAKING-A-STOP: A NEW BUFFERLESS ROUTING ALGORITHM FOR ON-CHIP NETWORK
14

Author : Jing Lin, Xiaola Lin, Liang Tang Publish Journal of parallel and Distributed Computing MAKING-A-STOP: A NEW BUFFERLESS ROUTING ALGORITHM FOR ON-CHIP.

Dec 31, 2015

Download

Documents

Ruth Gibbs
Welcome message from author
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Page 1: Author : Jing Lin, Xiaola Lin, Liang Tang Publish Journal of parallel and Distributed Computing MAKING-A-STOP: A NEW BUFFERLESS ROUTING ALGORITHM FOR ON-CHIP.

Author : Jing Lin, Xiaola Lin, Liang Tang

Publish Journal of parallel and Distributed Computing

MAKING-A-STOP: A NEW BUFFERLESS ROUTING ALGORITHM FOR ON-CHIP

NETWORK

Page 2: Author : Jing Lin, Xiaola Lin, Liang Tang Publish Journal of parallel and Distributed Computing MAKING-A-STOP: A NEW BUFFERLESS ROUTING ALGORITHM FOR ON-CHIP.

OUTLINE• Some background

• Virtual channel

• wormhole switching

• MAS

• Microarchitecture

• Evaluation

• Simulation results

Page 3: Author : Jing Lin, Xiaola Lin, Liang Tang Publish Journal of parallel and Distributed Computing MAKING-A-STOP: A NEW BUFFERLESS ROUTING ALGORITHM FOR ON-CHIP.

VIRTUAL CHANNEL• A channel designation which differs from that of the actual radio channel (or range of

frequencies) on which the signal travels

Page 4: Author : Jing Lin, Xiaola Lin, Liang Tang Publish Journal of parallel and Distributed Computing MAKING-A-STOP: A NEW BUFFERLESS ROUTING ALGORITHM FOR ON-CHIP.

WORMHOLE SWITCHING• Large network packets are broken into small pieces called flits (flow control digits). The

first flit, called the header flit holds information about this packet's route (namely the destination address) and sets up the routing behavior for all subsequent flits associated with the packet. The head flit is followed by zero or more body flits, containing the actual pay load of data. The final flit, called the tail flit, performs some book keeping to close the connection between the two nodes. One thing special about wormhole flow control is the implementation of virtual channels.

Page 5: Author : Jing Lin, Xiaola Lin, Liang Tang Publish Journal of parallel and Distributed Computing MAKING-A-STOP: A NEW BUFFERLESS ROUTING ALGORITHM FOR ON-CHIP.

PROBLEM

• buffer routing need more power and area and also adds the complexity of on-chip network design.

Page 6: Author : Jing Lin, Xiaola Lin, Liang Tang Publish Journal of parallel and Distributed Computing MAKING-A-STOP: A NEW BUFFERLESS ROUTING ALGORITHM FOR ON-CHIP.

MAKE A STOP• 藉由增設一動態 register

來給 flits stop 以取代 buffers 並 減少 deflected

Page 7: Author : Jing Lin, Xiaola Lin, Liang Tang Publish Journal of parallel and Distributed Computing MAKING-A-STOP: A NEW BUFFERLESS ROUTING ALGORITHM FOR ON-CHIP.
Page 8: Author : Jing Lin, Xiaola Lin, Liang Tang Publish Journal of parallel and Distributed Computing MAKING-A-STOP: A NEW BUFFERLESS ROUTING ALGORITHM FOR ON-CHIP.

• Arbitration unit

• register array

• switch crossbar

Page 9: Author : Jing Lin, Xiaola Lin, Liang Tang Publish Journal of parallel and Distributed Computing MAKING-A-STOP: A NEW BUFFERLESS ROUTING ALGORITHM FOR ON-CHIP.

• Router has five input ports and five output ports

• Warm-up period of 100,000 cycles

• 10,000 packets are injected per node

Page 10: Author : Jing Lin, Xiaola Lin, Liang Tang Publish Journal of parallel and Distributed Computing MAKING-A-STOP: A NEW BUFFERLESS ROUTING ALGORITHM FOR ON-CHIP.

TRAFFIC PATTERNS• OR traffic assumes each node uniformly injects packets to randomly distributed

destinations in the network.

• In TR traffic, the node ( i , j ) only communicates with node ( j , i ) .

• In HS traffic, four hotspot nodes are located at the center of the network. The 80% of traffic is sent to randomly distributed destinations in the network, while the remaining traffic is to the hotspot nodes.

Page 11: Author : Jing Lin, Xiaola Lin, Liang Tang Publish Journal of parallel and Distributed Computing MAKING-A-STOP: A NEW BUFFERLESS ROUTING ALGORITHM FOR ON-CHIP.

LATENCY • Packet latency is

calculated from the time

when the first flit of

a packet is generated,

to the time when its last

flit is ejected at the

destination, including

source queueing time.

Page 12: Author : Jing Lin, Xiaola Lin, Liang Tang Publish Journal of parallel and Distributed Computing MAKING-A-STOP: A NEW BUFFERLESS ROUTING ALGORITHM FOR ON-CHIP.

• the average hop

count per packet as

a function of injected

rate under different

synthetic traffic

patterns before

network saturates.

Page 13: Author : Jing Lin, Xiaola Lin, Liang Tang Publish Journal of parallel and Distributed Computing MAKING-A-STOP: A NEW BUFFERLESS ROUTING ALGORITHM FOR ON-CHIP.

• The power by flits

traversing through the

channels + the power

needed by the switches in the router.

Page 14: Author : Jing Lin, Xiaola Lin, Liang Tang Publish Journal of parallel and Distributed Computing MAKING-A-STOP: A NEW BUFFERLESS ROUTING ALGORITHM FOR ON-CHIP.