UBM Tech - Game Network

My Shopping Cart

[ 0 ] items in cart

View Cart | Checkout

Game Developer Research

Research Reports

Gamasutra

Contractor Listings

GDC Vault

Individual Subscription

GDC Audio Recordings

App Developers Conference 2013

GDC Next 2013

GDC Europe 2013

	GDC 2013
	GDC Online 2012
	GDC Europe 2012
	GDC 2012
	GDC 2011
	GDC 10
	GDC 09
	GDC Austin 08
	GDC Mobile 08
	GDC 08
	GDC Austin 07
	GDC Mobile 07
	GDC 07
	GDC 06
	GDC 05
	GDC 04
	GDC 03
	GDC 01
	GDC 2000 & Before

Newest Item(s)

Why Now Is the Best Time Ever to Be a Game Developer

Ingress: Design Principles Behind Google's Massively Multiplayer Geo Game

Playing with 'Game'

Gathering Your Party with Project Eternity (GDC Next 10)

D4: Dawn of the Dreaming Director's Drama (GDC Next 10)

Using Plot Devices to Create Gameplay in Storyteller (GDC Next 10)

How I Learned to Stop Worrying and Love Making CounterSpy (GDC Next 10)

Luck and Skill in Games

Minimalist Game Design for Mobile Devices

Broken Age: Rethinking a Classic Genre for the Modern Era (GDC Next 10)

Storefront > GDC Vault Store - Audio Recordings > GDC 2010

View larger image

QTY:

R-Trees -- Adapting out-of-core techniques to modern memory architectures
Price	$3.95
Adjustment
Type
Stock	Unlimited
Status
Weight	0 lb, 0 oz
SKU	GDC10-10437
Statistics

Description

R-Trees -- Adapting out-of-core techniques to modern memory architectures
Speaker: Sebastian Sylvan (Principal Software Engineer, Rare Ltd.)
Date/Time: Saturday (March 13, 2010) 11:05am — 11:30am
Location (room): Room 132, North Hall
Track: Programming
Format: 25-minute Lecture
Experience Level: All

Session Description
Code running on modern in-order CPUs pay a hefty cost for cache misses. This leads to major performance problem for pointer-heavy data structures, such as spatial hierarchies. This problem is similar to the problems encountered by people doing out-of-core processing for the last several decades. They needed to minimize disk seeks, we need to minimize memory fetches. An R-tree is a BV-hierarchy based on large fixed-size blocks, commonly used in disk based spatial processing. This talk shows how to adapt them for in-memory use, reaping major benefits with respect to cache behaviour, as well as SIMD processing, and more!

Intended Audience
Anyone interested in spatial indexing structures for graphics, physics, gameplay or any other purpose, or anyone interested in cache-friendly data structures in general, as the concepts discussed are not entirely R-tree-specific. Some experience with writing performance code is preferable.

Please leave this field blank.