Go for Decimals

Tags: decimal, floating point, fixed precision, IEEE 754R

Joshua Carpeggiani Software Engineer, ANZ joshua.carpeggiani@anz.com

Let me set the scene, it is February 1991, a missile has been launched with a military base in its sights. An anti air defence system kicks into gear ready to blow the enemy ballistic out of the sky, saving those living in the base.
The system turns on scans the sky for an incoming missile, looking at the part of the sky that sent it the trigger to activate.: It looks and looks, and looks, and looks.; A problem occurs; a floating point rounding error caused the internal clock to be off by a third of a second.; Under some circumstances, such inaccuracy might be unnoticeable but when attempting to scan the sky for an object moving five times the speed of sound the result was less than desirable.; The system can’t find anything. It turns off.; The missile hits the base moments later.; Less than 24 hours later a software update was deployed to fix the bug that caused the issue.; Melbourne university; Part time Software engineer; Summer internship; Weird story to how I got going; i met my now colleague in a fuit and veg shop and we started talking about software engineering; She mentioned that some jolly old fellow named marcelo was trying to get an summer internship program going on decimals and go; A project he started on .; A couple of weeks go by, some emails have been exchanged and the project finally gets the go-ahead.; I remember I was on the side of the road trying to hitchhike from melbourne to the gold coast; and I’d been stuck for a couple of hours at this point and was checking my emails to keep my sanity.; After this I frantically started going through “A tour of go” whilst on my journey

Floating point numbers

.image code/first_example.png 258 _

Floating point numbers

.image code/first_example.png 258 _

.image code/first_ans1.png _ 300

Floating point numbers

.image code/second.png 400 _

Floating point numbers

.image code/second.png 400 _
.image code/second_ans.png _ 500: 41 seconds going at 5 times the speed of sound is long enough to travel 71 kms

Not all numbering systems are created equal

Binary

Binary floating point numbers

Binary

Why?

Another Decimal library

http://www.github.com/anz-bank/decimal

Project Goal: Implement a 64 & 128 bit decimal floating point library
Conform as closely to IEEE 754-2008 standard as practical

Scientific notation

Another Decimal library

Applause

The End?

Use decimals for everything?

The End?

Use decimals for everything?
Why doesn’t everything use this?

The End?

Use decimals for everything?
Why doesn’t everything use this?
Are decimals even that efficient?

: IBM research; out of a survey in 2007 of enterprise databases;: 55% of numeric data was decimal

Go for Decimals

Joshua Carpeggiani Software Engineer, ANZ joshua.carpeggiani@anz.com

Floating point numbers

Floating point numbers

Floating point numbers

Floating point numbers

Not all numbering systems are created equal

Binary

Binary

Binary

Binary

Binary

Binary

Binary

Binary

Binary floating point numbers

Binary

Binary

Binary

Binary

Why?

Another Decimal library

Scientific notation

Scientific notation

Scientific notation

Scientific notation

Scientific notation

Scientific notation

Another Decimal library

Another Decimal library

Another Decimal library

Another Decimal library

Applause

Applause

The End?

The End?

The End?

The End?

Decimal decoding

Decimal decoding

Decimal decoding

Decimal decoding

Decimal decoding

Decimal decoding

Decimal decoding

Decimal decoding

Binary decoding

Binary decoding

Binary decoding

Decimal vs Floating point

Software vs Hardware

Binary floating point Add