Introduction to BME  (Informatics)


Prabhas Chongstitvatana

The lecture is divided into two parts. The first part concerns global trends and future of computing.  The second part discusses the fundamental of digital technology, processors.  We learn how to program in machine code (just a bit).  Mobile devices become ubiquitous hence we discuss the trade off and the future of information on everyday life.

Assignment

1) Write an essay on "Globally scientist pursues exascale computing, what should Thailand do?"  Read exascale computing initiative report.  Find out about Thai Nation Grid Computing of the past.  Then expand your opinion.  Please write no more than 5 pages.

2) Practice a bit of machine code programming.  Running the second example of doing multiplication.  Run it in the CPU 1001 simulator.  Next, describe a better multiplication program.  It is based on the idea of doing multiplication by doubling the multiplicand and halfing the multiplier.  If you can do it, write a machine code and test it in the simulator.  Assuming there is "divide by 2" and "doubling" instructions available.

Submit your report in the box in front of my office at Engineering Building 4, floor 18, room 13 by Monday 8th Dec. 2014, before 4pm.

Lecture

The fastest computer system for scientific use. 

top500.org 
What these machines look like.
The current champion can perform 33 Petraflops.  (33 x 10^15 floating point operation per second).  (as November 2014)
Exascale machine's goal is to achieve 1000x of current champion.
The race to build the fastest machine is an international "arm-race".  It is also a tool to energise advanced scientific research.

The Opportunities and Challenges of Exascale Computing, (pdf 2 Mb) US Department of Energy, Fall 2010
Kogge, P. "The tops in flops", IEEE spectrum, vol.48, issue 2, Feb 2011, pp.48-54,  local copy
Thai National Grid Computing

Current technology towards exascale (ppt)

Present everyday computing

MacBook Air 2014 has an intel i5 4250u processor.
Have a look at its specification
Notable:  Power Dissipation (TDP) and Clock Frequency

Four years ago desktop machine has Core 2 duo  power 65w
Two years ago notebook has Core 2 duo power 35w
Current notebook has i5  power 17w
Next year notebook will use Core M  power 4.5w

Mobile devices
Processors in mobile phones mostly are ARM (a name of a family of processor architecture)
Current high-end one: Snapdragon 800
They run 1GHz - 2.5GHz  but to conserve battery, they mostly stay idle.
intel enters the race by introducing very low power Atom family processor.
Here is the one used in a popular mobile phones today.  z2520  look at its specification.

CPU and GPU
CPU = central processing unit
GPU = graphic processing unit

GPU is a specialise processor. It is especially good at crunching numbers. GPU has a lot of these computing units. Typical PC machine has 40 of them.  Supercomputer in top500 has a large number of high-end GPUs (each has more than 200 computing units). Sometimes more than 50% of the processors are GPU. GPU is very efficient in terms of power per computation.

picture of NVIDIA k1 processor
picutre of Core M  more than half of the resource is devoted to Graphic computing unit

I did not have time to touch on Internet of Things. IoT will be deploy in a much larger scale than mobile phones. 

Basic of digital technology: microprocessor

http://www.cp.eng.chula.ac.th/~piak/project/cpu1001.htm
How it works
Write machine code
Run the code in the simulator

last update 28 Nov 2014