Transcript A New Network Processor Architecture for High

A New Network Processor
Architecture for High-speed
Communication
Xiaoning Nie; Gazsi, L.; Engel, F.; Fettweis, G.
Signal Processing Systems, 1999. SiPS 99. 1999
IEEE Workshop on , 1999 , Page(s): 548 -557
What’s the Problem
• The requirement of high-speed communication
• The requirement of a flexible software/hardware
platform for protocol processing
• A Network Processor (NP) is then derived
Introduction
• The xDSL and cable modems provide 2 Mb/s for a
person.
• Programmable and configurable hardware design
need to be provided in system level requirement.
• A general purpose processor would waste portions
of its processing power due to no use of its
functionalities such as floating-point unit.
Seven Layers of OSI model
TELNET application
Application
Presentation
Session
Transport
Network
Data Link
Physical
TCP/IP
Ethernet
The Top-Level Architecture
• Host:
(1) completes webbrowsing or telnet sessions.
(2) implements by one or
several processors
• Adaption:
(1) prepares the information
for physical transmission
and delivers the applicationrelevant information to the
host layer.
(2) implements by one
processor.
• Media:
(1) responses for reliable physical
transmission and receiving of the
information.
(2) implements by hardwired manner
or DSPs.
Router/Switch
Host
Adaption
Media
Transport
Network
Physical
What tasks are for the NP
• Deals with protocols that always contain some
headers and the payloads.
• NP operations:
• The operations for data transfer (move, load, store, parse)
consumes a significant portion of the processing power.
• There can be very frequent interrupts especially if the
buffering has to be limited due to some network delay
constraints e.g. in video communication applications.
• There are many operations involved in the finite state
machines.
What tasks are for the NP
• For routing/switching efficient access to some lookup
tables is essential.
• Depending on the protocol, there can be many operations
that are not word or even byte-aligned.
Essential Characteristics for NP
• A fixed-point RISC based architecture
• No floating-point operations
• Data movement: register  port
memory  port
• Example:
move port1@ bit1 port2@ bit2
NP Architecture
• Separated memory buffer:
(1) communication interface data
(2) Software control data
• Zero-overhead, hardware-supported threads
Data Transfer
• Bit-operation:
(load) LOAD dest@ bit source@ bit bit-width
• Data transfer runs in one thread in parallel to the
other thread.
• Example:
(store) /*start block move and let the task #task run in
parallel*/
Repeat counter #task
STORE_BL port@ bit dest@ bit number
• Saving registers for the management of the DMA
controller.
Bit-manipulation Support
• Bit access, bit addressing
• Specifies bit-width, bit-offset
• Example:
Ry.Oy := Ry.Oy AND Rx.Ox @ Width
Comparison between ARM and
New NP
Loop Support by Compute &
Jump
Until R1=0,
otherwise always
jumps to L1
A “For Loop” Example
for (i=0;i<10;i++)
a[i]=11;
3 cycles
5 cycles
Until R3=0,
otherwise
always jump to
FOR
Conclusions
• Variable bit-width operations work efficiency, and
reduce power consumption.
• Compound instructions help to reduce execution
cycles of a program.
• Multi-thread hardwired hardware reduces register
usage due to the requirement of DMA controller.