KEMBAR78
Introduction to Data Oriented Design | PPT
Electronic Arts / DICE, profile picture
Uploaded byElectronic Arts / DICE
PPT, PDF53,914 views

Introduction to Data Oriented Design

Data-oriented design (DOD) focuses on how data is accessed and transformed, rather than how code is organized. This improves performance by minimizing cache misses and allowing better utilization of parallelism. The document provides an example comparing an object-oriented design (OOD) approach that stores related data together in objects, resulting in scattered memory access and many cache misses, versus a DOD approach that groups together data that is accessed together, resulting in fewer cache misses and faster performance.

In this document
Powered by AI

Introduces Data-Oriented Design (DOD), focusing on how data is read & written, and its importance.

Highlights the performance impact of data access, discussing memory latency & multithreading issues.

Describes challenges in Object-Oriented Design (OOD) like data isolation, cache misses, and optimization difficulties.

Details examples showing inefficiencies of OOD in code calls, emphasizing cache misses and performance issues.

Introduces Data-Oriented Design (DOD) concepts, focusing on output data and input needs to improve efficiency.

Presents a code example in DOD, showcasing its efficient data access and reduced memory issues.

Visual comparisons of data layouts in OOD and DOD, highlighting cache optimization implications.

Discusses the significance of optimizing for data access first, noting that many codes are memory-bound.

Focuses on real-game examples, emphasizing knowledge of data formats to optimize performance.

Compares old vs new culling systems, illustrating performance enhancements through DOD applications.

Summarizes the advantages of DOD: better performance, simpler code, and improved parallelizability.

Prompt for questions, providing links to additional resources on DOD and related topics.

Downloaded 807 times
Introduction to Data-Oriented Design
So what is this Data-Oriented Design?
It’s about on shifting focus to how data is read and written
Why should we care?
Performance
A read from memory takes ~600 cycles at 3.2 GHz
A read from memory takes 40 cycles at 300 MHz
Performance Disks (Blu-ray/DVD/HDD) Main Memory L2 Cache L1 Cache CPU / Registers Latency :( 600 cycles 40 cycles 1 – 2 cycles
Multithreading Cannot multithread without knowing how data is touched Adding locks always protects data not code Object Read? Write? Object update() Object Read? Write? Read? Write?
Offloading to co-unit If data is unknown hard/impossible to run on co-unit ? SPU/GPU/APU ?
Better design Data focus can lead to isolated, self-contained, interchangeable pieces of data and code This can make it easier to test data and code in isolation
Example - OOD class Bot { ... Vec3 m_position; ... float m_mod; ... float m_aimDirection; ... void updateAim(Vec3 target) { m_aimDirection = dot3(m_position, target) * m_mod; } }
Example - OOD class Bot { ... Vec3 m_position; ... float m_mod; ... float m_aimDirection; ... void updateAim(Vec3 target) { m_aimDirection = dot3(m_position, target) * m_mod; } } icache-miss
Example - OOD class Bot { ... Vec3 m_position; ... float m_mod; ... float m_aimDirection; ... void updateAim(Vec3 target) { m_aimDirection = dot3(m_position, target) * m_mod; } } icache-miss data-miss
Example - OOD class Bot { ... Vec3 m_position; ... float m_mod; ... float m_aimDirection; ... void updateAim(Vec3 target) { m_aimDirection = dot3(m_position, target) * m_mod; } } icache-miss data-miss Unused cached data
Example - OOD class Bot { ... Vec3 m_position; ... float m_mod; ... float m_aimDirection; ... void updateAim(Vec3 target) { m_aimDirection = dot3(m_position, target) * m_mod; } } icache-miss data-miss Unused cached data Very hard to optimize!
Example - OOD void updateAim(Vec3 target) { m_aimDirection = dot3(m_position, target) * m_mod; } Lets say we call this code 4 times (4 diffrent Bots)
Example - OOD void updateAim(Vec3 target) { m_aimDirection = dot3(m_position, target) * m_mod; } Lets say we call this code 4 times (4 diffrent Bots) iCache – 600
Example - OOD void updateAim(Vec3 target) { m_aimDirection = dot3(m_position, target) * m_mod; } Lets say we call this code 4 times (4 diffrent Bots) iCache – 600 m_position – 600
Example - OOD void updateAim(Vec3 target) { m_aimDirection = dot3(m_position, target) * m_mod; } Lets say we call this code 4 times (4 diffrent Bots) iCache – 600 m_position – 600 m_mod - 600
Example - OOD void updateAim(Vec3 target) { m_aimDirection = dot3(m_position, target) * m_mod; } Lets say we call this code 4 times (4 diffrent Bots) iCache – 600 m_position – 600 m_mod - 600 ~20 cycles
Example - OOD void updateAim(Vec3 target) { m_aimDirection = dot3(m_position, target) * m_mod; } Lets say we call this code 4 times (4 diffrent Bots) iCache – 600 m_position – 600 m_mod - 600 aimDir – 100 ~20 cycles
Example - OOD void updateAim(Vec3 target) { m_aimDirection = dot3(m_position, target) * m_mod; } Lets say we call this code 4 times (4 diffrent Bots) iCache – 600 m_position – 600 m_mod - 600 aimDir – 100 ~20 cycles iCache – 600 m_position – 600 m_mod - 600 aimDir – 100
Example - OOD void updateAim(Vec3 target) { m_aimDirection = dot3(m_position, target) * m_mod; } Lets say we call this code 4 times (4 diffrent Bots) iCache – 600 m_position – 600 m_mod - 600 aimDir – 100 ~20 cycles iCache – 600 m_position – 600 m_mod - 600 aimDir – 100 iCache – 600 m_position – 600 m_mod - 600 aimDir – 100
Example - OOD void updateAim(Vec3 target) { m_aimDirection = dot3(m_position, target) * m_mod; } Lets say we call this code 4 times (4 diffrent Bots) iCache – 600 m_position – 600 m_mod - 600 aimDir – 100 ~20 cycles iCache – 600 m_position – 600 m_mod - 600 aimDir – 100 iCache – 600 m_position – 600 m_mod - 600 aimDir – 100 iCache – 600 m_position – 600 m_mod - 600 aimDir – 100
Example - OOD void updateAim(Vec3 target) { m_aimDirection = dot3(m_position, target) * m_mod; } Lets say we call this code 4 times (4 diffrent Bots) iCache – 600 m_position – 600 m_mod - 600 aimDir – 100 ~20 cycles iCache – 600 m_position – 600 m_mod - 600 aimDir – 100 iCache – 600 m_position – 600 m_mod - 600 aimDir – 100 iCache – 600 m_position – 600 m_mod - 600 aimDir – 100 7680
Example - DOD
Example - DOD Design ”back to front” and focus on the output data
Example - DOD Design ”back to front” and focus on the output data Then add the minimal amount of data needed to do the transform to create the correct output
Example - DOD void updateAims(float* aimDir,const AimingData* aim, Vec3 target, uint count) { for (uint i = 0; i < count; ++i) { aimDir[i] = dot3(aim->positions[i],target) * aim->mod[i]; } }
Example - DOD void updateAims(float* aimDir,const AimingData* aim, Vec3 target, uint count) { for (uint i = 0; i < count; ++i) { aimDir[i] = dot3(aim->positions[i],target) * aim->mod[i]; } } What has changed?
Example - DOD void updateAims(float* aimDir,const AimingData* aim, Vec3 target, uint count) { for (uint i = 0; i < count; ++i) { aimDir[i] = dot3(aim->positions[i],target) * aim->mod[i]; } } Only read needed inputs What has changed?
Example - DOD void updateAims(float* aimDir,const AimingData* aim, Vec3 target, uint count) { for (uint i = 0; i < count; ++i) { aimDir[i] = dot3(aim->positions[i],target) * aim->mod[i]; } } Only read needed inputs Write to linear array What has changed?
Example - DOD void updateAims(float* aimDir,const AimingData* aim, Vec3 target, uint count) { for (uint i = 0; i < count; ++i) { aimDir[i] = dot3(aim->positions[i],target) * aim->mod[i]; } } Only read needed inputs Write to linear array Loop over all the data What has changed?
Example - DOD void updateAims(float* aimDir,const AimingData* aim, Vec3 target, uint count) { for (uint i = 0; i < count; ++i) { aimDir[i] = dot3(aim->positions[i],target) * aim->mod[i]; } } Only read needed inputs Write to linear array Loop over all the data Actual code unchanged What has changed?
Example - DOD void updateAims(float* aimDir,const AimingData* aim, Vec3 target, uint count) { for (uint i = 0; i < count; ++i) { aimDir[i] = dot3(aim->positions[i],target) * aim->mod[i]; } } Only read needed inputs Write to linear array Loop over all the data Actual code unchanged What has changed? Code separated
Example - DOD void updateAims(float* aimDir, const AimingData* aim, Vec3 target, uint count) { for (uint i = 0; i < count; ++i) { aimDir[i] = dot3(aim->positions[i], target) * aim->mod[i]; } }
Example - DOD void updateAims(float* aimDir, const AimingData* aim, Vec3 target, uint count) { for (uint i = 0; i < count; ++i) { aimDir[i] = dot3(aim->positions[i], target) * aim->mod[i]; } } iCache – 600
Example - DOD void updateAims(float* aimDir, const AimingData* aim, Vec3 target, uint count) { for (uint i = 0; i < count; ++i) { aimDir[i] = dot3(aim->positions[i], target) * aim->mod[i]; } } iCache – 600 positions – 600
Example - DOD void updateAims(float* aimDir, const AimingData* aim, Vec3 target, uint count) { for (uint i = 0; i < count; ++i) { aimDir[i] = dot3(aim->positions[i], target) * aim->mod[i]; } } iCache – 600 positions – 600 mod - 600
Example - DOD void updateAims(float* aimDir, const AimingData* aim, Vec3 target, uint count) { for (uint i = 0; i < count; ++i) { aimDir[i] = dot3(aim->positions[i], target) * aim->mod[i]; } } iCache – 600 positions – 600 mod - 600 ~20 cycles
Example - DOD void updateAims(float* aimDir, const AimingData* aim, Vec3 target, uint count) { for (uint i = 0; i < count; ++i) { aimDir[i] = dot3(aim->positions[i], target) * aim->mod[i]; } } iCache – 600 positions – 600 mod - 600 aimDir – 100 ~20 cycles
Example - DOD void updateAims(float* aimDir, const AimingData* aim, Vec3 target, uint count) { for (uint i = 0; i < count; ++i) { aimDir[i] = dot3(aim->positions[i], target) * aim->mod[i]; } } iCache – 600 positions – 600 mod - 600 aimDir – 100 ~20 cycles
Example - DOD void updateAims(float* aimDir, const AimingData* aim, Vec3 target, uint count) { for (uint i = 0; i < count; ++i) { aimDir[i] = dot3(aim->positions[i], target) * aim->mod[i]; } } iCache – 600 positions – 600 mod - 600 aimDir – 100 ~20 cycles
Example - DOD void updateAims(float* aimDir, const AimingData* aim, Vec3 target, uint count) { for (uint i = 0; i < count; ++i) { aimDir[i] = dot3(aim->positions[i], target) * aim->mod[i]; } } iCache – 600 positions – 600 mod - 600 aimDir – 100 ~20 cycles
Example - DOD void updateAims(float* aimDir, const AimingData* aim, Vec3 target, uint count) { for (uint i = 0; i < count; ++i) { aimDir[i] = dot3(aim->positions[i], target) * aim->mod[i]; } } iCache – 600 positions – 600 mod - 600 aimDir – 100 ~20 cycles 1980
Data layout OOD vs DOD pos0 mod0 aimDir0 pos0 Pos1 mod1 aimDir1 pos0 pos0 pos0 pos1 pos1 pos1 pos1 pos2 pos2 pos2 pos2 pos3 pos3 pos3 pos3 mod0 mod1 mod2 mod3 aimDir0 aimDir1 aimDir2 aimDir3 pos0 pos0 pos0
Data layout OOD vs DOD pos0 mod0 aimDir0 pos0 Pos1 mod1 aimDir1 pos0 pos0 pos0 pos1 pos1 pos1 pos1 pos2 pos2 pos2 pos2 pos3 pos3 pos3 pos3 mod0 mod1 mod2 mod3 aimDir0 aimDir1 aimDir2 aimDir3 pos0 pos0 pos0 Each color block is one 128 byte cache line
Data layout OOD vs DOD pos0 mod0 aimDir0 pos0 Pos1 mod1 aimDir1 pos0 pos0 pos0 pos1 pos1 pos1 pos1 pos2 pos2 pos2 pos2 pos3 pos3 pos3 pos3 mod0 mod1 mod2 mod3 aimDir0 aimDir1 aimDir2 aimDir3 pos0 pos0 pos0 Each color block is one 128 byte cache line
Data layout OOD vs DOD pos0 mod0 aimDir0 pos0 Pos1 mod1 aimDir1 pos0 pos0 pos0 pos1 pos1 pos1 pos1 pos2 pos2 pos2 pos2 pos3 pos3 pos3 pos3 mod0 mod1 mod2 mod3 aimDir0 aimDir1 aimDir2 aimDir3 pos0 pos0 pos0 Each color block is one 128 byte cache line
Data layout OOD vs DOD pos0 mod0 aimDir0 pos0 Pos1 mod1 aimDir1 pos0 pos0 pos0 pos1 pos1 pos1 pos1 pos2 pos2 pos2 pos2 pos3 pos3 pos3 pos3 mod0 mod1 mod2 mod3 aimDir0 aimDir1 aimDir2 aimDir3 pos0 pos0 pos0 Each color block is one 128 byte cache line
Data layout OOD vs DOD pos0 mod0 aimDir0 pos0 Pos1 mod1 aimDir1 pos0 pos0 pos0 pos1 pos1 pos1 pos1 pos2 pos2 pos2 pos2 pos3 pos3 pos3 pos3 mod0 mod1 mod2 mod3 aimDir0 aimDir1 aimDir2 aimDir3 pos0 pos0 pos0 Each color block is one 128 byte cache line
Data layout OOD vs DOD pos0 mod0 aimDir0 pos0 Pos1 mod1 aimDir1 pos0 pos0 pos0 pos1 pos1 pos1 pos1 pos2 pos2 pos2 pos2 pos3 pos3 pos3 pos3 mod0 mod1 mod2 mod3 aimDir0 aimDir1 aimDir2 aimDir3 pos0 pos0 pos0 Each color block is one 128 byte cache line
Data layout OOD vs DOD pos0 mod0 aimDir0 pos0 Pos1 mod1 aimDir1 pos0 pos0 pos0 pos1 pos1 pos1 pos1 pos2 pos2 pos2 pos2 pos3 pos3 pos3 pos3 mod0 mod1 mod2 mod3 aimDir0 aimDir1 aimDir2 aimDir3 pos0 pos0 pos0 Each color block is one 128 byte cache line
Data layout OOD vs DOD pos0 mod0 aimDir0 pos0 Pos1 mod1 aimDir1 pos0 pos0 pos0 pos1 pos1 pos1 pos1 pos2 pos2 pos2 pos2 pos3 pos3 pos3 pos3 mod0 mod1 mod2 mod3 aimDir0 aimDir1 aimDir2 aimDir3 pos0 pos0 pos0 Each color block is one 128 byte cache line
Data layout OOD vs DOD pos0 mod0 aimDir0 pos0 Pos1 mod1 aimDir1 pos0 pos0 pos0 pos1 pos1 pos1 pos1 pos2 pos2 pos2 pos2 pos3 pos3 pos3 pos3 mod0 mod1 mod2 mod3 aimDir0 aimDir1 aimDir2 aimDir3 pos0 pos0 pos0 Each color block is one 128 byte cache line
Its all about memory
Its all about memory Optimize for data first then code
Its all about memory Optimize for data first then code Most code is likely bound by memory access
Its all about memory Optimize for data first then code Most code is likely bound by memory access Not everything needs to be an object
Remember
Remember We are doing games, we know our data.
Remember We are doing games, we know our data. Pre-format. Source data and native data doesn’t need to be the same
Example: Area Triggers
Example: Area Triggers position position position position next position position position position next position position position position next Source data (Linked List)
Example: Area Triggers position position position position next position position position position next position position position position next Source data (Linked List) Native Data (Array) position position position position position position position position position position position position position position count
Example: Culling System
Example: Culling System Old System
Example: Culling System Old System New System (Linear arrays and brute force)
Example: Culling System Old System New System (Linear arrays and brute force) 3x faster, 1/5 code size, simpler
Data Oriented Design Delivers:
Better Performance
Often simpler code
More parallelizable code
Questions?
Links Data-Oriented Design (Or Why You Might Be Shooting Yourself in The Foot With OOP) http://gamesfromwithin.com/data-oriented-design Practical Examples in Data Oriented Design http://bitsquid.blogspot.com/2010/05/practical-examples-in-data-oriented.html The Latency Elephant http://seven-degrees-of-freedom.blogspot.com/2009/10/latency-elephant.html Pitfalls of Object Oriented Programming http://seven-degrees-of-freedom.blogspot.com/2009/12/pitfalls-of-object-oriented-programming.html Insomniac R&D http://www.insomniacgames.com/research_dev CellPerformance
Image credits Cat image: http://icanhascheezburger.com/2007/06/24/uninterested-cat photo by: Arinn capped and submitted by: Andy Playstation 3 and Playstation 2 Copyright to Sony Computer Entertainment Xbox 360 Image Copyright to Microsoft “ WTF” Code quality image: Copyright by Thom Holwerda http://www.osnews.com/comics

More Related Content

PPTX
Data oriented design and c++
byMike Acton
 
PPTX
A Step Towards Data Orientation
byElectronic Arts / DICE
 
PPSX
Dx11 performancereloaded
bymistercteam
 
PDF
Optimizing the graphics pipeline with compute
byWuBinbo
 
PPTX
Optimizing the Graphics Pipeline with Compute, GDC 2016
byGraham Wihlidal
 
PPSX
Oit And Indirect Illumination Using Dx11 Linked Lists
byHolger Gruen
 
PPTX
GDC Europe 2014: Unreal Engine 4 for Programmers - Lessons Learned & Things t...
byGerke Max Preussner
 
PDF
Unreal Engine Basics 01 - Game Framework
byNick Pruehs
 
Data oriented design and c++
byMike Acton
 
A Step Towards Data Orientation
byElectronic Arts / DICE
 
Dx11 performancereloaded
bymistercteam
 
Optimizing the graphics pipeline with compute
byWuBinbo
 
Optimizing the Graphics Pipeline with Compute, GDC 2016
byGraham Wihlidal
 
Oit And Indirect Illumination Using Dx11 Linked Lists
byHolger Gruen
 
GDC Europe 2014: Unreal Engine 4 for Programmers - Lessons Learned & Things t...
byGerke Max Preussner
 
Unreal Engine Basics 01 - Game Framework
byNick Pruehs
 

What's hot

PDF
Pitfalls of Object Oriented Programming by SONY
byAnaya Medias Swiss
 
PPTX
[IGC 2017] 펄어비스 민경인 - Mmorpg를 위한 voxel 기반 네비게이션 라이브러리 개발기
by강 민우
 
PDF
Windows Registered I/O (RIO) vs IOCP
bySeungmo Koo
 
PDF
NDC2017 언리얼엔진4 디버깅 101 - 게임 기획자, 프로그래머가 버그와 만났을 때 사용할 수 있는 지침들
by영욱 오
 
PDF
임태현, 게임 서버 디자인 가이드, NDC2013
bydevCAT Studio, NEXON
 
PPTX
Intrinsics: Low-level engine development with Burst - Unite Copenhagen 2019
byUnity Technologies
 
PDF
Deep Dive async/await in Unity with UniTask(EN)
byYoshifumi Kawai
 
PDF
NDC 2017 하재승 NEXON ZERO (넥슨 제로) 점검없이 실시간으로 코드 수정 및 게임 정보 수집하기
byJaeseung Ha
 
PDF
[KGC 2012]Boost.asio를 이용한 네트웍 프로그래밍
by흥배 최
 
PDF
Modern C++ 프로그래머를 위한 CPP11/14 핵심
by흥배 최
 
PDF
멀티스레드 렌더링 (Multithreaded rendering)
byBongseok Cho
 
PPSX
Holy smoke! Faster Particle Rendering using Direct Compute by Gareth Thomas
byAMD Developer Central
 
PPTX
DirectX 11 Rendering in Battlefield 3
byElectronic Arts / DICE
 
PPTX
Triangle Visibility buffer
byWolfgang Engel
 
PDF
NDC12_Lockless게임서버설계와구현
bynoerror
 
PDF
Compute shader
byQooJuice
 
PPTX
NDC 11 자이언트 서버의 비밀
by승명 양
 
PDF
스크린 스페이스 데칼에 대해 자세히 알아보자(워햄머 40,000: 스페이스 마린)
by포프 김
 
PDF
임태현, MMO 서버 개발 포스트 모템, NDC2012
bydevCAT Studio, NEXON
 
PDF
GDC2011 - Implementation and Application of the Real-Time Helper-Joint System
byJubok Kim
 
Pitfalls of Object Oriented Programming by SONY
byAnaya Medias Swiss
 
[IGC 2017] 펄어비스 민경인 - Mmorpg를 위한 voxel 기반 네비게이션 라이브러리 개발기
by강 민우
 
Windows Registered I/O (RIO) vs IOCP
bySeungmo Koo
 
NDC2017 언리얼엔진4 디버깅 101 - 게임 기획자, 프로그래머가 버그와 만났을 때 사용할 수 있는 지침들
by영욱 오
 
임태현, 게임 서버 디자인 가이드, NDC2013
bydevCAT Studio, NEXON
 
Intrinsics: Low-level engine development with Burst - Unite Copenhagen 2019
byUnity Technologies
 
Deep Dive async/await in Unity with UniTask(EN)
byYoshifumi Kawai
 
NDC 2017 하재승 NEXON ZERO (넥슨 제로) 점검없이 실시간으로 코드 수정 및 게임 정보 수집하기
byJaeseung Ha
 
[KGC 2012]Boost.asio를 이용한 네트웍 프로그래밍
by흥배 최
 
Modern C++ 프로그래머를 위한 CPP11/14 핵심
by흥배 최
 
멀티스레드 렌더링 (Multithreaded rendering)
byBongseok Cho
 
Holy smoke! Faster Particle Rendering using Direct Compute by Gareth Thomas
byAMD Developer Central
 
DirectX 11 Rendering in Battlefield 3
byElectronic Arts / DICE
 
Triangle Visibility buffer
byWolfgang Engel
 
NDC12_Lockless게임서버설계와구현
bynoerror
 
Compute shader
byQooJuice
 
NDC 11 자이언트 서버의 비밀
by승명 양
 
스크린 스페이스 데칼에 대해 자세히 알아보자(워햄머 40,000: 스페이스 마린)
by포프 김
 
임태현, MMO 서버 개발 포스트 모템, NDC2012
bydevCAT Studio, NEXON
 
GDC2011 - Implementation and Application of the Real-Time Helper-Joint System
byJubok Kim
 

Viewers also liked

PPTX
Technical direction
byMike Acton
 
PPTX
A Real-time Radiosity Architecture
byElectronic Arts / DICE
 
PPTX
Battlelog - Building scalable web sites with tight game integration
byElectronic Arts / DICE
 
PPTX
SPU-Based Deferred Shading in BATTLEFIELD 3 for Playstation 3
byElectronic Arts / DICE
 
PPTX
How data rules the world: Telemetry in Battlefield Heroes
byElectronic Arts / DICE
 
PPT
Stylized Rendering in Battlefield Heroes
byElectronic Arts / DICE
 
PPT
5 Major Challenges in Interactive Rendering
byElectronic Arts / DICE
 
PPT
Destruction Masking in Frostbite 2 using Volume Distance Fields
byElectronic Arts / DICE
 
PPTX
Scope Stack Allocation
byElectronic Arts / DICE
 
PPTX
Shadows & Decals: D3D10 Techniques in Frostbite (GDC'09)
byrepii
 
PPT
Bending the Graphics Pipeline
byElectronic Arts / DICE
 
PPTX
Building the Battlefield AI Experience
byElectronic Arts / DICE
 
PPT
Stable SSAO in Battlefield 3 with Selective Temporal Filtering
byElectronic Arts / DICE
 
PPTX
Future Directions for Compute-for-Graphics
byElectronic Arts / DICE
 
PPT
Level Design Challenges & Solutions - Mirror's Edge
byElectronic Arts / DICE
 
Technical direction
byMike Acton
 
A Real-time Radiosity Architecture
byElectronic Arts / DICE
 
Battlelog - Building scalable web sites with tight game integration
byElectronic Arts / DICE
 
SPU-Based Deferred Shading in BATTLEFIELD 3 for Playstation 3
byElectronic Arts / DICE
 
How data rules the world: Telemetry in Battlefield Heroes
byElectronic Arts / DICE
 
Stylized Rendering in Battlefield Heroes
byElectronic Arts / DICE
 
5 Major Challenges in Interactive Rendering
byElectronic Arts / DICE
 
Destruction Masking in Frostbite 2 using Volume Distance Fields
byElectronic Arts / DICE
 
Scope Stack Allocation
byElectronic Arts / DICE
 
Shadows & Decals: D3D10 Techniques in Frostbite (GDC'09)
byrepii
 
Bending the Graphics Pipeline
byElectronic Arts / DICE
 
Building the Battlefield AI Experience
byElectronic Arts / DICE
 
Stable SSAO in Battlefield 3 with Selective Temporal Filtering
byElectronic Arts / DICE
 
Future Directions for Compute-for-Graphics
byElectronic Arts / DICE
 
Level Design Challenges & Solutions - Mirror's Edge
byElectronic Arts / DICE
 

Similar to Introduction to Data Oriented Design

PPTX
Data oriented design
byMax Klyga
 
PPT
B Eng Final Year Project Presentation
byjesujoseph
 
PDF
Lab Log Summer 2016 - Sheng Li
bySheng Li
 
DOCX
ABB Project
byNick Bhardwaj
 
PPT
What does OOP stand for?
byColin Riley
 
PDF
High-Performance Physics Solver Design for Next Generation Consoles
bySlide_N
 
PDF
SOM
byLiu Zhen-Yu
 
PDF
Data-Oriented Design and C++ - Mike Acton Engine Director, Insomniac Games
bydarkasterion2009
 
PPT
Marek Suplata Projects
byguest14f12f
 
PDF
Parallelization and Optimization of Time-Domain Electronic Scattering Simulat...
byMichael Patterson
 
PPTX
Intro to data oriented design
byStoyan Nikolov
 
PPTX
Use Data-Oriented Design to write efficient code
byAlessio Coltellacci
 
PPT
Advanced Mobile Optimizations.ppt
byТранслируем.бел
 
PDF
A CGRA-based Approach for Accelerating Convolutional Neural Networks
byshtaxxx
 
PPTX
Intel’S Larrabee
byvipinpnair
 
PDF
ECS (Part 1/3) - Introduction to Data-Oriented Design
byPhuong Hoang Vu
 
PPTX
Parallel Futures of a Game Engine
byrepii
 
PDF
Using GPUs for parallel processing
byasm100
 
PDF
SPU gameplay
bySlide_N
 
PPTX
Lec05 buffers basic_examples
byTaras Zakharchenko
 
Data oriented design
byMax Klyga
 
B Eng Final Year Project Presentation
byjesujoseph
 
Lab Log Summer 2016 - Sheng Li
bySheng Li
 
ABB Project
byNick Bhardwaj
 
What does OOP stand for?
byColin Riley
 
High-Performance Physics Solver Design for Next Generation Consoles
bySlide_N
 
SOM
byLiu Zhen-Yu
 
Data-Oriented Design and C++ - Mike Acton Engine Director, Insomniac Games
bydarkasterion2009
 
Marek Suplata Projects
byguest14f12f
 
Parallelization and Optimization of Time-Domain Electronic Scattering Simulat...
byMichael Patterson
 
Intro to data oriented design
byStoyan Nikolov
 
Use Data-Oriented Design to write efficient code
byAlessio Coltellacci
 
Advanced Mobile Optimizations.ppt
byТранслируем.бел
 
A CGRA-based Approach for Accelerating Convolutional Neural Networks
byshtaxxx
 
Intel’S Larrabee
byvipinpnair
 
ECS (Part 1/3) - Introduction to Data-Oriented Design
byPhuong Hoang Vu
 
Parallel Futures of a Game Engine
byrepii
 
Using GPUs for parallel processing
byasm100
 
SPU gameplay
bySlide_N
 
Lec05 buffers basic_examples
byTaras Zakharchenko
 

More from Electronic Arts / DICE

PPTX
GDC2019 - SEED - Towards Deep Generative Models in Game Development
byElectronic Arts / DICE
 
PPT
SIGGRAPH 2010 - Style and Gameplay in the Mirror's Edge
byElectronic Arts / DICE
 
PDF
SEED - Halcyon Architecture
byElectronic Arts / DICE
 
PDF
Syysgraph 2018 - Modern Graphics Abstractions & Real-Time Ray Tracing
byElectronic Arts / DICE
 
PPTX
Khronos Munich 2018 - Halcyon and Vulkan
byElectronic Arts / DICE
 
PDF
CEDEC 2018 - Towards Effortless Photorealism Through Real-Time Raytracing
byElectronic Arts / DICE
 
PPTX
CEDEC 2018 - Functional Symbiosis of Art Direction and Proceduralism
byElectronic Arts / DICE
 
PPTX
SIGGRAPH 2018 - PICA PICA and NVIDIA Turing
byElectronic Arts / DICE
 
PPTX
SIGGRAPH 2018 - Full Rays Ahead! From Raster to Real-Time Raytracing
byElectronic Arts / DICE
 
PPTX
HPG 2018 - Game Ray Tracing: State-of-the-Art and Open Problems
byElectronic Arts / DICE
 
PDF
EPC 2018 - SEED - Exploring The Collaboration Between Proceduralism & Deep Le...
byElectronic Arts / DICE
 
PDF
DD18 - SEED - Raytracing in Hybrid Real-Time Rendering
byElectronic Arts / DICE
 
PDF
Creativity of Rules and Patterns: Designing Procedural Systems
byElectronic Arts / DICE
 
PPTX
Shiny Pixels and Beyond: Real-Time Raytracing at SEED
byElectronic Arts / DICE
 
PPTX
A Certain Slant of Light - Past, Present and Future Challenges of Global Illu...
byElectronic Arts / DICE
 
PPTX
Physically Based Sky, Atmosphere and Cloud Rendering in Frostbite
byElectronic Arts / DICE
 
PPTX
High Dynamic Range color grading and display in Frostbite
byElectronic Arts / DICE
 
PPTX
4K Checkerboard in Battlefield 1 and Mass Effect Andromeda
byElectronic Arts / DICE
 
PPTX
FrameGraph: Extensible Rendering Architecture in Frostbite
byElectronic Arts / DICE
 
PPTX
Lighting the City of Glass
byElectronic Arts / DICE
 
GDC2019 - SEED - Towards Deep Generative Models in Game Development
byElectronic Arts / DICE
 
SIGGRAPH 2010 - Style and Gameplay in the Mirror's Edge
byElectronic Arts / DICE
 
SEED - Halcyon Architecture
byElectronic Arts / DICE
 
Syysgraph 2018 - Modern Graphics Abstractions & Real-Time Ray Tracing
byElectronic Arts / DICE
 
Khronos Munich 2018 - Halcyon and Vulkan
byElectronic Arts / DICE
 
CEDEC 2018 - Towards Effortless Photorealism Through Real-Time Raytracing
byElectronic Arts / DICE
 
CEDEC 2018 - Functional Symbiosis of Art Direction and Proceduralism
byElectronic Arts / DICE
 
SIGGRAPH 2018 - PICA PICA and NVIDIA Turing
byElectronic Arts / DICE
 
SIGGRAPH 2018 - Full Rays Ahead! From Raster to Real-Time Raytracing
byElectronic Arts / DICE
 
HPG 2018 - Game Ray Tracing: State-of-the-Art and Open Problems
byElectronic Arts / DICE
 
EPC 2018 - SEED - Exploring The Collaboration Between Proceduralism & Deep Le...
byElectronic Arts / DICE
 
DD18 - SEED - Raytracing in Hybrid Real-Time Rendering
byElectronic Arts / DICE
 
Creativity of Rules and Patterns: Designing Procedural Systems
byElectronic Arts / DICE
 
Shiny Pixels and Beyond: Real-Time Raytracing at SEED
byElectronic Arts / DICE
 
A Certain Slant of Light - Past, Present and Future Challenges of Global Illu...
byElectronic Arts / DICE
 
Physically Based Sky, Atmosphere and Cloud Rendering in Frostbite
byElectronic Arts / DICE
 
High Dynamic Range color grading and display in Frostbite
byElectronic Arts / DICE
 
4K Checkerboard in Battlefield 1 and Mass Effect Andromeda
byElectronic Arts / DICE
 
FrameGraph: Extensible Rendering Architecture in Frostbite
byElectronic Arts / DICE
 
Lighting the City of Glass
byElectronic Arts / DICE
 

Introduction to Data Oriented Design

  • 1.
  • 2.
    So what isthis Data-Oriented Design?
  • 3.
    It’s about onshifting focus to how data is read and written
  • 4.
  • 5.
  • 6.
    A read frommemory takes ~600 cycles at 3.2 GHz
  • 7.
    A read frommemory takes 40 cycles at 300 MHz
  • 8.
    Performance Disks (Blu-ray/DVD/HDD)Main Memory L2 Cache L1 Cache CPU / Registers Latency :( 600 cycles 40 cycles 1 – 2 cycles
  • 9.
    Multithreading Cannot multithread without knowing how data is touched Adding locks always protects data not code Object Read? Write? Object update() Object Read? Write? Read? Write?
  • 10.
    Offloading to co-unitIf data is unknown hard/impossible to run on co-unit ? SPU/GPU/APU ?
  • 11.
    Better design Datafocus can lead to isolated, self-contained, interchangeable pieces of data and code This can make it easier to test data and code in isolation
  • 12.
    Example - OODclass Bot { ... Vec3 m_position; ... float m_mod; ... float m_aimDirection; ... void updateAim(Vec3 target) { m_aimDirection = dot3(m_position, target) * m_mod; } }
  • 13.
    Example - OODclass Bot { ... Vec3 m_position; ... float m_mod; ... float m_aimDirection; ... void updateAim(Vec3 target) { m_aimDirection = dot3(m_position, target) * m_mod; } } icache-miss
  • 14.
    Example - OODclass Bot { ... Vec3 m_position; ... float m_mod; ... float m_aimDirection; ... void updateAim(Vec3 target) { m_aimDirection = dot3(m_position, target) * m_mod; } } icache-miss data-miss
  • 15.
    Example - OODclass Bot { ... Vec3 m_position; ... float m_mod; ... float m_aimDirection; ... void updateAim(Vec3 target) { m_aimDirection = dot3(m_position, target) * m_mod; } } icache-miss data-miss Unused cached data
  • 16.
    Example - OODclass Bot { ... Vec3 m_position; ... float m_mod; ... float m_aimDirection; ... void updateAim(Vec3 target) { m_aimDirection = dot3(m_position, target) * m_mod; } } icache-miss data-miss Unused cached data Very hard to optimize!
  • 17.
    Example - OODvoid updateAim(Vec3 target) { m_aimDirection = dot3(m_position, target) * m_mod; } Lets say we call this code 4 times (4 diffrent Bots)
  • 18.
    Example - OODvoid updateAim(Vec3 target) { m_aimDirection = dot3(m_position, target) * m_mod; } Lets say we call this code 4 times (4 diffrent Bots) iCache – 600
  • 19.
    Example - OODvoid updateAim(Vec3 target) { m_aimDirection = dot3(m_position, target) * m_mod; } Lets say we call this code 4 times (4 diffrent Bots) iCache – 600 m_position – 600
  • 20.
    Example - OODvoid updateAim(Vec3 target) { m_aimDirection = dot3(m_position, target) * m_mod; } Lets say we call this code 4 times (4 diffrent Bots) iCache – 600 m_position – 600 m_mod - 600
  • 21.
    Example - OODvoid updateAim(Vec3 target) { m_aimDirection = dot3(m_position, target) * m_mod; } Lets say we call this code 4 times (4 diffrent Bots) iCache – 600 m_position – 600 m_mod - 600 ~20 cycles
  • 22.
    Example - OODvoid updateAim(Vec3 target) { m_aimDirection = dot3(m_position, target) * m_mod; } Lets say we call this code 4 times (4 diffrent Bots) iCache – 600 m_position – 600 m_mod - 600 aimDir – 100 ~20 cycles
  • 23.
    Example - OODvoid updateAim(Vec3 target) { m_aimDirection = dot3(m_position, target) * m_mod; } Lets say we call this code 4 times (4 diffrent Bots) iCache – 600 m_position – 600 m_mod - 600 aimDir – 100 ~20 cycles iCache – 600 m_position – 600 m_mod - 600 aimDir – 100
  • 24.
    Example - OODvoid updateAim(Vec3 target) { m_aimDirection = dot3(m_position, target) * m_mod; } Lets say we call this code 4 times (4 diffrent Bots) iCache – 600 m_position – 600 m_mod - 600 aimDir – 100 ~20 cycles iCache – 600 m_position – 600 m_mod - 600 aimDir – 100 iCache – 600 m_position – 600 m_mod - 600 aimDir – 100
  • 25.
    Example - OODvoid updateAim(Vec3 target) { m_aimDirection = dot3(m_position, target) * m_mod; } Lets say we call this code 4 times (4 diffrent Bots) iCache – 600 m_position – 600 m_mod - 600 aimDir – 100 ~20 cycles iCache – 600 m_position – 600 m_mod - 600 aimDir – 100 iCache – 600 m_position – 600 m_mod - 600 aimDir – 100 iCache – 600 m_position – 600 m_mod - 600 aimDir – 100
  • 26.
    Example - OODvoid updateAim(Vec3 target) { m_aimDirection = dot3(m_position, target) * m_mod; } Lets say we call this code 4 times (4 diffrent Bots) iCache – 600 m_position – 600 m_mod - 600 aimDir – 100 ~20 cycles iCache – 600 m_position – 600 m_mod - 600 aimDir – 100 iCache – 600 m_position – 600 m_mod - 600 aimDir – 100 iCache – 600 m_position – 600 m_mod - 600 aimDir – 100 7680
  • 27.
  • 28.
    Example - DODDesign ”back to front” and focus on the output data
  • 29.
    Example - DODDesign ”back to front” and focus on the output data Then add the minimal amount of data needed to do the transform to create the correct output
  • 30.
    Example - DODvoid updateAims(float* aimDir,const AimingData* aim, Vec3 target, uint count) { for (uint i = 0; i < count; ++i) { aimDir[i] = dot3(aim->positions[i],target) * aim->mod[i]; } }
  • 31.
    Example - DODvoid updateAims(float* aimDir,const AimingData* aim, Vec3 target, uint count) { for (uint i = 0; i < count; ++i) { aimDir[i] = dot3(aim->positions[i],target) * aim->mod[i]; } } What has changed?
  • 32.
    Example - DODvoid updateAims(float* aimDir,const AimingData* aim, Vec3 target, uint count) { for (uint i = 0; i < count; ++i) { aimDir[i] = dot3(aim->positions[i],target) * aim->mod[i]; } } Only read needed inputs What has changed?
  • 33.
    Example - DODvoid updateAims(float* aimDir,const AimingData* aim, Vec3 target, uint count) { for (uint i = 0; i < count; ++i) { aimDir[i] = dot3(aim->positions[i],target) * aim->mod[i]; } } Only read needed inputs Write to linear array What has changed?
  • 34.
    Example - DODvoid updateAims(float* aimDir,const AimingData* aim, Vec3 target, uint count) { for (uint i = 0; i < count; ++i) { aimDir[i] = dot3(aim->positions[i],target) * aim->mod[i]; } } Only read needed inputs Write to linear array Loop over all the data What has changed?
  • 35.
    Example - DODvoid updateAims(float* aimDir,const AimingData* aim, Vec3 target, uint count) { for (uint i = 0; i < count; ++i) { aimDir[i] = dot3(aim->positions[i],target) * aim->mod[i]; } } Only read needed inputs Write to linear array Loop over all the data Actual code unchanged What has changed?
  • 36.
    Example - DODvoid updateAims(float* aimDir,const AimingData* aim, Vec3 target, uint count) { for (uint i = 0; i < count; ++i) { aimDir[i] = dot3(aim->positions[i],target) * aim->mod[i]; } } Only read needed inputs Write to linear array Loop over all the data Actual code unchanged What has changed? Code separated
  • 37.
    Example - DODvoid updateAims(float* aimDir, const AimingData* aim, Vec3 target, uint count) { for (uint i = 0; i < count; ++i) { aimDir[i] = dot3(aim->positions[i], target) * aim->mod[i]; } }
  • 38.
    Example - DODvoid updateAims(float* aimDir, const AimingData* aim, Vec3 target, uint count) { for (uint i = 0; i < count; ++i) { aimDir[i] = dot3(aim->positions[i], target) * aim->mod[i]; } } iCache – 600
  • 39.
    Example - DODvoid updateAims(float* aimDir, const AimingData* aim, Vec3 target, uint count) { for (uint i = 0; i < count; ++i) { aimDir[i] = dot3(aim->positions[i], target) * aim->mod[i]; } } iCache – 600 positions – 600
  • 40.
    Example - DODvoid updateAims(float* aimDir, const AimingData* aim, Vec3 target, uint count) { for (uint i = 0; i < count; ++i) { aimDir[i] = dot3(aim->positions[i], target) * aim->mod[i]; } } iCache – 600 positions – 600 mod - 600
  • 41.
    Example - DODvoid updateAims(float* aimDir, const AimingData* aim, Vec3 target, uint count) { for (uint i = 0; i < count; ++i) { aimDir[i] = dot3(aim->positions[i], target) * aim->mod[i]; } } iCache – 600 positions – 600 mod - 600 ~20 cycles
  • 42.
    Example - DODvoid updateAims(float* aimDir, const AimingData* aim, Vec3 target, uint count) { for (uint i = 0; i < count; ++i) { aimDir[i] = dot3(aim->positions[i], target) * aim->mod[i]; } } iCache – 600 positions – 600 mod - 600 aimDir – 100 ~20 cycles
  • 43.
    Example - DODvoid updateAims(float* aimDir, const AimingData* aim, Vec3 target, uint count) { for (uint i = 0; i < count; ++i) { aimDir[i] = dot3(aim->positions[i], target) * aim->mod[i]; } } iCache – 600 positions – 600 mod - 600 aimDir – 100 ~20 cycles
  • 44.
    Example - DODvoid updateAims(float* aimDir, const AimingData* aim, Vec3 target, uint count) { for (uint i = 0; i < count; ++i) { aimDir[i] = dot3(aim->positions[i], target) * aim->mod[i]; } } iCache – 600 positions – 600 mod - 600 aimDir – 100 ~20 cycles
  • 45.
    Example - DODvoid updateAims(float* aimDir, const AimingData* aim, Vec3 target, uint count) { for (uint i = 0; i < count; ++i) { aimDir[i] = dot3(aim->positions[i], target) * aim->mod[i]; } } iCache – 600 positions – 600 mod - 600 aimDir – 100 ~20 cycles
  • 46.
    Example - DODvoid updateAims(float* aimDir, const AimingData* aim, Vec3 target, uint count) { for (uint i = 0; i < count; ++i) { aimDir[i] = dot3(aim->positions[i], target) * aim->mod[i]; } } iCache – 600 positions – 600 mod - 600 aimDir – 100 ~20 cycles 1980
  • 47.
    Data layout OODvs DOD pos0 mod0 aimDir0 pos0 Pos1 mod1 aimDir1 pos0 pos0 pos0 pos1 pos1 pos1 pos1 pos2 pos2 pos2 pos2 pos3 pos3 pos3 pos3 mod0 mod1 mod2 mod3 aimDir0 aimDir1 aimDir2 aimDir3 pos0 pos0 pos0
  • 48.
    Data layout OODvs DOD pos0 mod0 aimDir0 pos0 Pos1 mod1 aimDir1 pos0 pos0 pos0 pos1 pos1 pos1 pos1 pos2 pos2 pos2 pos2 pos3 pos3 pos3 pos3 mod0 mod1 mod2 mod3 aimDir0 aimDir1 aimDir2 aimDir3 pos0 pos0 pos0 Each color block is one 128 byte cache line
  • 49.
    Data layout OODvs DOD pos0 mod0 aimDir0 pos0 Pos1 mod1 aimDir1 pos0 pos0 pos0 pos1 pos1 pos1 pos1 pos2 pos2 pos2 pos2 pos3 pos3 pos3 pos3 mod0 mod1 mod2 mod3 aimDir0 aimDir1 aimDir2 aimDir3 pos0 pos0 pos0 Each color block is one 128 byte cache line
  • 50.
    Data layout OODvs DOD pos0 mod0 aimDir0 pos0 Pos1 mod1 aimDir1 pos0 pos0 pos0 pos1 pos1 pos1 pos1 pos2 pos2 pos2 pos2 pos3 pos3 pos3 pos3 mod0 mod1 mod2 mod3 aimDir0 aimDir1 aimDir2 aimDir3 pos0 pos0 pos0 Each color block is one 128 byte cache line
  • 51.
    Data layout OODvs DOD pos0 mod0 aimDir0 pos0 Pos1 mod1 aimDir1 pos0 pos0 pos0 pos1 pos1 pos1 pos1 pos2 pos2 pos2 pos2 pos3 pos3 pos3 pos3 mod0 mod1 mod2 mod3 aimDir0 aimDir1 aimDir2 aimDir3 pos0 pos0 pos0 Each color block is one 128 byte cache line
  • 52.
    Data layout OODvs DOD pos0 mod0 aimDir0 pos0 Pos1 mod1 aimDir1 pos0 pos0 pos0 pos1 pos1 pos1 pos1 pos2 pos2 pos2 pos2 pos3 pos3 pos3 pos3 mod0 mod1 mod2 mod3 aimDir0 aimDir1 aimDir2 aimDir3 pos0 pos0 pos0 Each color block is one 128 byte cache line
  • 53.
    Data layout OODvs DOD pos0 mod0 aimDir0 pos0 Pos1 mod1 aimDir1 pos0 pos0 pos0 pos1 pos1 pos1 pos1 pos2 pos2 pos2 pos2 pos3 pos3 pos3 pos3 mod0 mod1 mod2 mod3 aimDir0 aimDir1 aimDir2 aimDir3 pos0 pos0 pos0 Each color block is one 128 byte cache line
  • 54.
    Data layout OODvs DOD pos0 mod0 aimDir0 pos0 Pos1 mod1 aimDir1 pos0 pos0 pos0 pos1 pos1 pos1 pos1 pos2 pos2 pos2 pos2 pos3 pos3 pos3 pos3 mod0 mod1 mod2 mod3 aimDir0 aimDir1 aimDir2 aimDir3 pos0 pos0 pos0 Each color block is one 128 byte cache line
  • 55.
    Data layout OODvs DOD pos0 mod0 aimDir0 pos0 Pos1 mod1 aimDir1 pos0 pos0 pos0 pos1 pos1 pos1 pos1 pos2 pos2 pos2 pos2 pos3 pos3 pos3 pos3 mod0 mod1 mod2 mod3 aimDir0 aimDir1 aimDir2 aimDir3 pos0 pos0 pos0 Each color block is one 128 byte cache line
  • 56.
    Data layout OODvs DOD pos0 mod0 aimDir0 pos0 Pos1 mod1 aimDir1 pos0 pos0 pos0 pos1 pos1 pos1 pos1 pos2 pos2 pos2 pos2 pos3 pos3 pos3 pos3 mod0 mod1 mod2 mod3 aimDir0 aimDir1 aimDir2 aimDir3 pos0 pos0 pos0 Each color block is one 128 byte cache line
  • 57.
  • 58.
    Its all aboutmemory Optimize for data first then code
  • 59.
    Its all aboutmemory Optimize for data first then code Most code is likely bound by memory access
  • 60.
    Its all aboutmemory Optimize for data first then code Most code is likely bound by memory access Not everything needs to be an object
  • 61.
  • 62.
    Remember We aredoing games, we know our data.
  • 63.
    Remember We aredoing games, we know our data. Pre-format. Source data and native data doesn’t need to be the same
  • 64.
  • 65.
    Example: Area Triggersposition position position position next position position position position next position position position position next Source data (Linked List)
  • 66.
    Example: Area Triggersposition position position position next position position position position next position position position position next Source data (Linked List) Native Data (Array) position position position position position position position position position position position position position position count
  • 67.
  • 68.
  • 69.
    Example: Culling SystemOld System New System (Linear arrays and brute force)
  • 70.
    Example: Culling SystemOld System New System (Linear arrays and brute force) 3x faster, 1/5 code size, simpler
  • 71.
  • 72.
  • 73.
  • 74.
  • 75.
  • 76.
    Links Data-Oriented Design(Or Why You Might Be Shooting Yourself in The Foot With OOP) http://gamesfromwithin.com/data-oriented-design Practical Examples in Data Oriented Design http://bitsquid.blogspot.com/2010/05/practical-examples-in-data-oriented.html The Latency Elephant http://seven-degrees-of-freedom.blogspot.com/2009/10/latency-elephant.html Pitfalls of Object Oriented Programming http://seven-degrees-of-freedom.blogspot.com/2009/12/pitfalls-of-object-oriented-programming.html Insomniac R&D http://www.insomniacgames.com/research_dev CellPerformance
  • 77.
    Image credits Catimage: http://icanhascheezburger.com/2007/06/24/uninterested-cat photo by: Arinn capped and submitted by: Andy Playstation 3 and Playstation 2 Copyright to Sony Computer Entertainment Xbox 360 Image Copyright to Microsoft “ WTF” Code quality image: Copyright by Thom Holwerda http://www.osnews.com/comics