Mini Project- Dual Processor Computation

Mini Project – Dual Processor Computation Author: University of Hertfordshire Date created: Date revised: 2009 Abstract The following resources come from the 2009/10 BEng in Digital Systems and Computer Engineering (course number 2ELE0065) from the University of Hertfordshire. All the mini projects are designed as level two modules of the undergraduate programmes. The objectives of this module are to demonstrate, within an embedded development environment: Processor – to – processor communication Multiple processors to perform one computation task using parallel processing This project requires the establishment of a communication protocol between two 68000-based microcomputer systems. Using ‘C’, students will write software to control all aspects of complex data transfer system, demonstrating knowledge of handshaking, transmission protocols, transmission overhead, bandwidth, memory addressing. Students will then demonstrate and analyse parallel processing of a mathematical problem using two processors. This project requires two students working as a team. In addition to the resources found below there are supporting documents which should be used in combination with this resource. Please see: Mini Projects - Introductory presentation. Mini Projects - E-Log. Mini Projects - Staff & Student Guide. Mini Projects - Standard Grading Criteria. Mini Projects - Reflection. You will also need the ‘Mini Project- Dual Processor Computation’ text document. © University of Hertfordshire 2009 This work is licensed under a Creative Commons Attribution 2.0 License .

Contents Lecture Overview ………………………………………………………………………….3 Day 1- Communication Between Two Processors …………………………………….4 The Problem Resources Asynchronous Communication Asynchronous Communication with Only Two Handshaking Signals Definition of Master/Slave The handshaking signals ACK and STROBE Preparation Day 2- Dual Processor Computation …………………………………………………..17 The Problem Resources Bandwidth Calculation Unit of transfer Asynchronous Communication with Only Two Handshaking Signals Handshake Cycle Time Bandwidth Measurement for this system An example calculation If one processor does the calculation Two processors sharing the calculation Time saving in first computation Repeated Computation – see improved time saving during 1s... Simpler notation: Normalise using tc Normalised values for example Method for timing each part of the calculation Preparation Background Reading ……………………………………………………………………42 Reflection …………………………………………………………………………………43 Remember ………………………………………………………………………………..45 Credits …………………………………………………………………………………….46

BH Dual Processor Computation Today’s Lecture The problem Breakdown to each day Resources List for each day Preparatory work Remind you for each day Reflection process Reflect on each day

Day 1 Communication between Two Processors

The Problem To establish Master-to-Slave asynchronous data transfer between two 68000 processors, using one 8-bit port and two additional control lines only. between two 68000 processors Master-to-Slave asynchronous using one 8-bit port two additional control only BH Dual Processor Computation

Resources I/O Expansion board Coloured wires BH Dual Processor Computation

Asynchronous Communication BH Dual Processor Computation Addr AS Data DTACK 1. AS signals valid address available on address bus – start of transfer 2. DTACK is asserted to signal that the data has been stored 4. ACK is negated to signal ready for next transfer 3. AS is negated to signal transfer complete Addr AS Data DTACK Master Slave

Asynchronous Communication with Only Two Handshaking Signals BH Dual Processor Computation Master Strobe Slave ACK I have data for you! Thanks, I’ve stored it! I might send some more Ok, I’m ready

Definition of Master/Slave Only the Master can initiate a data transfer The Slave can only respond BH Dual Processor Computation

The handshaking signals ACK and STROBE Master Controls STROBE Slave Controls ACK BH Dual Processor Computation

Preparation – Order of attack Plan your day (Friday) Break down the problem into steps and decide: How to start, what to do next… etc… BH Dual Processor Computation

Preparation – Resource map BH Dual Processor Computation Variable name Possible states Input or Output Allocated port pin

Preparation - Algorithms To transmit a single byte to the Slave To receive a single byte from the Master BH Dual Processor Computation

Modular Structure BH Dual Processor Computation This is only an example ! main function 1 function 2 function 3 function 4 function 5 function 6

Preparation – message format You must define a message format for data block transfer. Define a message header which contains: An address where the first byte will be stored in the slave’s memory (subsequent bytes will be stored contiguously) The number of bytes in the data block (max 256) BH Dual Processor Computation

Preparation Summary Create Resource Map: Define & allocate handshaking signals Define i/o port Write Algorithms: To transfer a single byte from Master to Slave Define message format BH Dual Processor Computation

Day 2 Dual Processor Computation

The Problem To establish the basic requirements for engaging a number of processors in solving a common problem. A simultaneous equation with obvious parallelism will be used to develop the necessary functions in ‘C’. A state-time diagram will be developed to summarise the efficiency of the proposed solution. BH Dual Processor Computation

Bandwidth Calculation This is the maximum number of bits that can be transferred in a second. Units: Mbits/second How much data is transferred each during one complete handshaking cycle? BH Dual Processor Computation

Unit of transfer In our system, how much data is transferred during one handshaking cycle? 1 byte BH Dual Processor Computation If we continuously transmit bytes, we can measure the time taken to transfer a byte…..

Handshake Cycle Time BH Dual Processor Computation Strobe ack time Unit transfer time 1 st cycle 2 nd cycle (µ s)

Bandwidth Measurement for this system Bandwidth = 8bits/ 20 µ s = 0.4Mbits/sec BH Dual Processor Computation

An example calculation Calculation of ‘z’ z = x / y x = (a * b) / e y = c + d BH Dual Processor Computation

If one processor does the calculation BH Dual Processor Computation x = a*b/e y = c+d z = x/y This is how long it would take a single processor to calculate z Processor 1 time

Two processors sharing the calculation BH Dual Processor Computation Processor 1 Processor 2 x = a*b/e y = c+d z = x/y time

Two processors sharing the calculation BH Dual Processor Computation Processor 1 x = a*b/e y = c+d z = x/y time to transfer data to processor 1 time Processor 2 tc

Two processors sharing the calculation BH Dual Processor Computation Processor 1 y = c+d z = x/y time Processor 2 tc x = a*b/e

Two processors sharing a calculation BH Dual Processor Computation time x = a*b/e y = c+d z = x/y Processor 1 x = a*b/e y = c+d z = x/y Processor 2 tc time saving over single processor

Two processors sharing a calculation BH Dual Processor Computation Processor 1 x = a*b/e y = c+d z = x/y time Processor 2 tc Processor 1 x = a*b/e y = c+d z = x/y Processor 2 tc time saving over single processor Single processor x = a*b/e y = c+d z = x/y

Time saving in first computation BH Dual Processor Computation Processor 1 tx y = c+d tz time Processor 2 tc Processor 1 tx y = c+d tz Processor 2 tc Single processor ty ty time saving over single processor tx ty tz

Repeated Computation – see improved time saving during 1 st repetition of calculation BH Dual Processor Computation tx tc tx tz tc tx tc ty tx tc tz tz tz ty ty ty End of 1 st calculation tx tz Single processor ty time saving over single processor P1 P2 P1 P2

Simpler notation: Normalise using tc To simplify the diagram, the calculation times can be described in units of tc . To do this, just divide each time by the time for tc eg: if tx takes 7 times as long as tc then it will become: tx/tc = 7 BH Dual Processor Computation

Normalised values for example Assume, just for the sake of argument, that: tc = 1 tx = 7 ty = 3 tz = 4 BH Dual Processor Computation

BH Dual Processor Computation Repeated Computation – see improved time saving during 1 st repetition of calculation tx tc tx tz tc tx tc ty tx tc tz tz tz ty ty ty tx tz Single processor ty P1 P2 12 12 14 8 P1 P2

Method for timing each part of the calculation Write a probe function – just an output pin which is asserted by the processor when it starts a calculation and negated at the end. This “probe” pin should be included in the resource map. Calls to the probe function should be added to the code at the beginning and the end of the calculation (the probe function itself will of course increase the measured time) BH Dual Processor Computation

BH Dual Processor Computation For this project, your level of preparation will be assessed. We will look at your logbook and the materials you have chosen to bring with you

Preparation - Algorithms Probe function Review receive/transmit single byte function and handshaking BH Dual Processor Computation

Preparation Summary Create Resource Map: Define & allocate handshaking signals Define i/o port Write Algorithm: Probe Review transmit/receive byte algorithms BH Dual Processor Computation

Background Reading From previous modules: 68000 architecture and C language programming Useful C program template Review related module lecture notes A general book on ‘C’ ( bring this to the lab too!!) BH Dual Processor Computation

Reflection Cycle BH Dual Processor Computation Reflection Observation Action

Reflection Would you break the problem down in the same way now? Did you spend enough time planning what you were going to do? What did you waste time on – what could you do to reduce the amount of time spent Do you think that you asked for too much help, or should you have asked for help sooner With hindsight, what preparation could you have done to help you make faster progress? Did you find the notes that you had made during preparation useful? – how could they be improved? Do you think that you could repeat the project in six months using only your logbook record as a reference. BH Dual Processor Computation

Remember Be methodical: Make progress in SMALL steps Document as you go (comments in code and diagrams, notes in logbook) Don’t go onto the next step until this one works FULLY and has been archived!! BH Dual Processor Computation

This resource was created by the University of Hertfordshire and released as an open educational resource through the Open Engineering Resources project of the HE Academy Engineering Subject Centre. The Open Engineering Resources project was funded by HEFCE and part of the JISC/HE Academy UKOER programme. © University of Hertfordshire 2009 This work is licensed under a Creative Commons Attribution 2.0 License . The name of the University of Hertfordshire, UH and the UH logo are the name and registered marks of the University of Hertfordshire. To the fullest extent permitted by law the University of Hertfordshire reserves all its rights in its name and marks which may not be used except with its written permission. The JISC logo is licensed under the terms of the Creative Commons Attribution-Non-Commercial-No Derivative Works 2.0 UK: England & Wales Licence. All reproductions must comply with the terms of that licence. The HEA logo is owned by the Higher Education Academy Limited may be freely distributed and copied for educational purposes only, provided that appropriate acknowledgement is given to the Higher Education Academy as the copyright holder and original publisher.

Mini Project- Dual Processor Computation

More Related Content

What's hot (20)

Viewers also liked (7)

Similar to Mini Project- Dual Processor Computation (20)

More from University of Hertfordshire, School of Electronic Communications and Electrical Engineering (20)

Recently uploaded (20)

Mini Project- Dual Processor Computation

Editor's Notes