EE249 Project: Partitioning Algorithms & Modeling Methodologies For HW/SW Partitioning in Metropolis

EE249 Project:
Partitioning Algorithms &

Modeling Methodologies
for HW/SW Partitioning
in Metropolis
Mentor:
John Moondanos,
GSRC Visiting Fellow, UC Berkeley
&
Strategic CAD Labs Intel Corp.
DUSD(Labs)
Problem Statement
 HW/SW partitioning deals with assigning parts of a system
description to heterogeneous implementation units
 Key task in system level design due to the downstream cost & performance
consequences of the initial partitioning choices
 Multi-faceted
 Processor & flash on the same die or not?
 Functionality Partitioning into chips
 Hardware vs. Software functionality Implementation
 Which functions to which type of silicon?
 Difficult to model & analyze with conventional RTL Tools
 Of course, the focus is on how it affects power consumption
2
Goal of the Project
Briefly Review the literature to capture the state of the art
Develop technologies and methodologies for solving the

partitioning problem within the Metropolis environment
 Methodologies will focus more on the system modeling
methodology that is better suited for the capabilities of Metropolis
 Technologies will focus more on the algorithms that must be used
to accomplish the partitioning using the capabilities of the
Metropolis environment.
3
Suggestion for Design Driver for this Project
For the hardware: The PXA800F cell phone processor from

Intel
 Some publicly available introductory material on the PXA800F is
available in the “backup material section”
 Modeling of the Xscale can happen with the GnuPro simulator
For the Software: We have available Statistical Models for

typical applications that run on the PXA800F
4
Backup Material
Overview of the PXA800F cellular phone Processor
References
5
The Intel® PXA800F Cellular Processor
Full GSM/GPRS Class solution
 High-performance/Low-power Intel® XScale ™

technology core, providing class-leading
headroom for rich data applications
 Intel® Micro Signal Architecture
 Intel® On-Chip Flash Memory

 GSM/GPRS Communications Stack, RTOS and
applications code for a single-chip mobile solution
6
The Intel® XScale ™ in the PXA800F
 High-performance, power-efficient processor supports data-intensive applications
 Processor core operates at an adjustable clock frequency from 104 to 312 MHz
 Instruction cache and Data cache memories
 4 MB integrated Intel On-Chip Flash memory
 512 KB integrated SRAM

 Memory controller supports synchronous Flash mode, page mode Flash, SRAM,
DRAM, and variable latency
 DMA controller
 Clock units-GSM slow clocking, GSM frame timing, watchdog, RTC
 Supports a wide range of standard interfaces-SIM, UART, USB, I2C*, SPI, SSP,
Digital Audio Interface, MultiMediaCard, Secure Digital Card, Sony Memory Stick,
Dallas* 1-Wire* Interface, keypad, PWM D/A, JTAG
 Interfaces for Bluetooth, IrDA, GPS and digital camera peripherals
 LCD Controller for up to 120 x 240 display 16-bit color or gray scale
7
Intel Micro Signal Architecture in the PXA800F
 Performs GSM/GPRS baseband signal processing
 Modified Harvard architecture, dual-MAC, deep pipeline, 104 MHz

execution clock
 Instruction cache and 64 KB dual-banked data SRAM
 512 KB integrated Intel On-Chip Flash for field-upgradable signal
processing firmware
 Includes microprocessor instructions such as bit manipulation
 Includes cipher and Viterbi accelerators
 Multiple sleep modes and integrated power management minimize power

consumption
 Interface support-digital I/Q, voice codec, auxiliary serial port for mixed-
signal analog baseband, I2S audio codec interface, RF synthesizer serial
control interface, JTAG
8
The Memory Subsystem
The Intel® XScale ™

 Instruction and Data Cache
 4MB of Flash & 512KB of SRAM always at 104MHz
 Memory Controller managing accesses to external SRAM
The MSA
 Integrated 64KB SRAM for microcontroller like instructions
• Special instructions for maximizing GSM/GPRS performance
 512KB of flash for program store
9
PXA800F Block Diagram Smart Battery I/F
UARTs for
Bluetooth, IRDA
Synch Serial Port
GSM Sim card I/F
External Power
Management I/F
10
PXA800F Block Diagram Pulse Width
Memory Stick Modulator for
buzzer
Secure Card I/F
Programmable
Clock
Encrypt/Decrypt
Timing Control Unit
GSM data offloading
For basestation
MSA 11
timing
PXA800F Block Diagram
DSP Synchronous
Serial Ports
interfacing with RF,
speech
High Speed Logger

For debug
Viterbi error
Full Bandwidth (Hi-
decoding offloading
Fi) digital audio I/F
MSA 12
PXA800F Block Diagram
Peripheral
Bus 1
Peripheral
Bus 2
13
References
On the hardware-software partitioning problem: System

modeling and partitioning techniques Marisa López-Vallejo,
Juan Carlos López July 2003 ACM Transactions on Design
Automation of Electronic Systems (TODAES), Volume 8
Issue 3
A hardware/software partitioner using a dynamically
determined granularity Jörg Henkel, Rolf Ernst June
1997 Proceedings of the 34th annual conference on
Design automation conference
14
References
Issues in partitioning & design space eploration for

codesign: Dynamic hardware/software partitioning: a first
approach Greg Stitt, Roman Lysecky, Frank Vahid June
2003 Proceedings of the 40th conference on Design
automation
Hardware/software partitioning of software binaries Greg
Stitt, Frank Vahid November 2002 Proceedings of the 2002
IEEE/ACM international conference on Computer-aided
design
15

EE249 Project: Partitioning Algorithms & Modeling Methodologies For HW/SW Partitioning in Metropolis

Uploaded by

Copyright:

Available Formats

EE249 Project: Partitioning Algorithms & Modeling Methodologies For HW/SW Partitioning in Metropolis

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

EE249 Project: Partitioning Algorithms & Modeling Methodologies For HW/SW Partitioning in Metropolis

Uploaded by

Copyright:

Available Formats

EE249 Project:

Partitioning Algorithms &

 Difficult to model & analyze with conventional RTL Tools

 Of course, the focus is on how it affects power consumption

Briefly Review the literature to capture the state of the art

Develop technologies and methodologies for solving the

For the hardware: The PXA800F cell phone processor from

For the Software: We have available Statistical Models for

Overview of the PXA800F cellular phone Processor

Full GSM/GPRS Class solution

 High-performance/Low-power Intel® XScale ™

 Intel® On-Chip Flash Memory

 512 KB integrated SRAM

 Clock units-GSM slow clocking, GSM frame timing, watchdog, RTC

 Modified Harvard architecture, dual-MAC, deep pipeline, 104 MHz

 Includes cipher and Viterbi accelerators

 Multiple sleep modes and integrated power management minimize power

The Intel® XScale ™

Synch Serial Port

GSM Sim card I/F

Secure Card I/F

High Speed Logger

On the hardware-software partitioning problem: System

Issues in partitioning & design space eploration for

You might also like