GDC 2009 – Josh Jones 

IGDA

There was as usual lots of good debating and discussion, and the workshops went very well. The IGDA is still gearing towards supporting local chapters and focusing on members, while handling international relations and forming new ventures such as the GGJ and the coders SIG. This was some of what I talked about at the last local chapter meeting (2^nd Tuesday in May) where everyone shared GDC experiences and won raffle prizes.  
 
 

Sessions 
 
Advanced Graphics with DirectX (Tutorial) 
 
This is a basic class but goes over all the ground work and is good for a refresher or starting out. This was an all-day course and they had everything from A to Z plus a couple other topics from other speakers at the end.  
 
 
API in DX10 has been redone 
-no fixed function 
-no more legacy code 
-non-kernel drivers (doesn't kill the OS) 
-Vista = Memory Management is in the OS 
 
Benchmark Mode 
-check performance 
-automated 
-debugging 
-easier for card makers 
 
Constant Buffers 
-HLSL, Vector, Color 
-close together 
-best kept as a smaller heap of buffers sorted by frequency of access 
-bind about 5 and share them 
 
Dynamic Buffer 
-updates 
-non GPU geometry (particles, translucency) 
 
Early 2 Optimize 
-no alpha test -> shader discard pixel 
-have two shaders 
1. write-z 
2. full shader 
 
Formats 
-Texture - bigger, slower 
-Render targets 
 
Geometry Shader 
-less is better 
-DX11 introduces tessellation 
 
High Batch Count 
-instancing 
 
Input Assembly 
-DXOptimize 
-Input compression 
 
Juggling with states 
-DX10 = immutable states 
-states at load time 
-dirty states 
 
Klears 
-use Z clear() 
-only color if needed 
 
Level of Detail 
-still important 
 
Multi GPU 
-optimize for frame pipeline 
-per GPU 
 
No Way Jose! 
-don't render skybox first (HUD last is ok) 
-don't forget frustum culling 
 
Output Streaming 
-multi-pass GS (Geometry Shaders) 
 
Parallelism 
-CPU | GPU - check with PIX - STAGING > DEFAULT 
 
Queries 
-get results when available only 
 
Resolving MSAA Buffers 
-resolve !free 
-avoid depth buffer 
-MSAA Render Target -> Resolve Operation -> Non-MSAA Back Buffer 
 
Shadow Mapping 
-DST 
-ATI: D16 
-NVIDIA: D32_S8 
-disable color 
-depth-only is much faster 
-no clip-discard 
-cut aniso 
-Gather() DX 10.1 
 
Transparency 
-ZPS, transparency & opacity 
-clip/discard early 
 
Updating Textures 
-do not create and destroy textures at runtime 
-create during load and use ring buffer 
 
Verifying Performance 
-use IHV tools 
-ATI: GPU perfshader 
-NVIDIA: NVPerfHUD 
 
Writing Fast Shaders 
-ALU: TEX HW ratio 
-Parallel Branching Interpolation 
 
Xtra Performance 
-larger triangle with unused space is better than two triangles 
-larger triangles in small complex areas instead of all small triangles 
 
Your call to action 
-Manage resources 
-Balancing 
-Texture Updates 
-Parallel (Multi GPU) 
-Instancing 
 
Z-Buffer Access 
-Bind Dynamic Buffer as texture 
 
cem@nvidia.com 
nicolas.thibieroz@amd.com 
 
 
Rain 
-scrolling texture, specialized shader 
Sun rays 
-raytrace march, single cascade shader 
MSAA deferred rendering - in/out pixel freq 
-lighting stages 
-DX10.1 or less adds 
G-Buffer optimizations 
-Shader A2C when hi-rez/hi MSAA 
-SV_POS 1-3 pit pattern 
-x+y = coverage 
-alpha -> alpha' 
-zoom 
Min-Max Shadow Map 
-originally presented by Dmitriev at GDC07 
 
DirectX 11 
-SV_Coverage - bit mask - System Value 
-available to shaders during stages 
-double precision -slower but now available 
-Gather(); - now has channel, offset (-32 - 31) 
-ddx_coarse - potential derivatives - same for all pixels 
-f32tof16 - f16 low 16 bits - for packing & compression - first bit high - reverse bits 
-Unordered Access Views - main new feature in DX11 
-Raw View - Tesselation - Displacement Maps - Hull Shader, Tesselator, Domain Shader 
-Structured Buffers - bind as SRV or UAV 
-Buffer Append/Consume 
-Atomic Operations - PS & CS - Interlock Add, etc 
-Compute Shader - GPGPU - post-processing, filtering, sorting, prep AI 
-1-per-frame - full execution, loops, etc. 
-Thread model - access to parallels - group & thread 
-TLS (Thread Local Storage) - 32 KB - sync before data work 
 
 
 
The Creation of a Kids' Virtual World - Real Data (Ridemakerz) 
 
This talk was okay- there was quite a lot of ego and advertising, but at the very end we did get some real numbers.  
 
 
Extend Experience 
Players in free play can move to online micropayments, member revenue, or retail purchase 
Players can become members by signing up directly, by online micropayments, or by retail purchase 
Members have discounts on online micropayments and retail purchases 
Make it Real didn't work -> Make it a Toy 
Still unsure about free play before registration 
Load swf's on demand 
Make waiting fun 
Real World Achievements - work on car with parent 
$1/mo per user revenue 
 
Revenue 
5-10 sponsors/partners 
Shop (micropayments for avatar changes) 5-8% 
E-commerce 1-5% 
Premium Member 1.5-3% 
4 mo beta - 5045 registered (one of them was me, haha) 
retail store coupons 60% 
parent email response 20% 
relogin 8% 
level up 5% 
 
 
 
Avoiding legal pitfalls in Virtual Worlds 
 
This talk by Mark Methenitis (a lawyer from Dallas and IGDA) was pretty good, but like any law talk mostly focused on past legal cases and worries about video game patents.  
 
 
worlds.com holds patents that are under scrutiny and may affect the virtual world space 
other early starters include Meridian and AlphaWorld - 1995 
copyright concerns 
-ensure EULA has a clause to cover it if you allow user-generated content 
-e.g. Marvel and City of Heroes (Marvel lost) 
-have a contact to handle copyright claims 
derivative works (mods, etc) 
-ensure EULA/ToS is clear on who owns what and who is responsible for what 
Illegal to have online gambling 
-do not mix game economy with real world currency 
Government plan for making money: 
1. Make Rules 
2. ??? -> Enforce 
3. Profit 
Governments are having trouble being able to tax online, so they are raising fines 
-expect to be required to have stickers and standard terms 
 
 
 
Discovering New Development Opportunities (Iwata keynote)  
 
This excellent keynote gave great insights into the workings of Nintendo and how they succeed and are driving and supporting the industry. Iwata managed to show off his successes, have a positive, uplifting keynote, and show support for Nintendo developers, amazingly all without being too pretentious or seemingly offending anyone in the room. It probably helped that he gave away free copies of Rhythm Heaven at the end, haha.  
 
 
Death Spiral 
Financial Pressure -> Less Time -> Poorer Quality -> Less Sales -> Financial Pressure -> and on and on... 
e.g. HAL Laboratories 
-stick to the "core concept" - distortion leads to delay 
-always comes from human observation 
-they like this activity - why? 
-Gardening -> Pikmin 
-Puppies -> Nintendogs 
-Excercise -> Wii Fit 
-20% non-gamers 
-50% increase in women gamers 
-Miyamoto's personal life is now under NDA haha 
Upward Spiral 
Ideas -> Prototype -> Small Teams -> Multiple Projects -> Trial and Error -> Mass Production 
Ideas are everywhere - always observe 
Explain ideas personally - no design doc 
Small teams - even 1 programmer 
Miyamoto seems like perfectionist - but is judging the trial and error of each small project 
-Kidnapping - randomly grab someone not on the team and have them play the game - over the shoulder observation 
New ideas that surprise you are good 
-Tsunku dance lessons demo -> Rhythm Heaven 
WiiWare is 90% 3rd Party 
New Wii support released 
-HD SD Ram 32G channels 
DSi demo 
-animation maker 
-eyetoy type game 
 
 
 
My Lessons Learned from Flagship Studios 
 
This was an excellent talk from an obviously humbled man who had a lot of important things to say. Mostly business lessons but good for everyone to know. Good speaker and kept audience attention well.  
 
 
Flagship Studios: What happened? 
July 2008 100+ employees laid off 
Flagshipped - lost IPs -> creditors - controversy/bad blood 
Made common mistakes 
-why didn't we listen? 
-just like VH1's behind the music, drugs, reunion, the whole deal 
-so much optimism/hype 
-inflated valuation 
 
Failure is not an option... but it is a Grim Reality 
-had no plan B 
-just to survive is success right now 
-scale it back 
Know your economic model 
-identify your first money in 
-don't reinvent the wheel 
-reeducation -> many steps -> $$$ 
-velvet rope - too much free area - not work in retail - should have gone with standard subscription 
Don't do more than One New Thing 
-Changing from developing something you're experienced with 
-Changing to a new dev team 
-Changing to a new genre 
Limit Your Partnerships 
-Each will have high expectations 
-Team cannot say no (producer needs to watch for requests which are slowing the project and cut them off at the source) 
Be careful with Multiple Territories 
-N languages = N builds (and less polish for each) 
-Each Region/Publisher wants "something special" 
Infrastructure is Harder than you think 
-Billing, Billing, Billing - learn from Jessica Mulligan 
-Partner with good Ops - don't try to be an online publisher 
-Customer Service Tools & billing are important features 
Don't Believe the Hype 
-Never stop doubting 
-Slow roll marketing 
-Sell! Your hype is always better than reality 
-You usually get offered more before launch 
If you're offered money, TAKE IT!!! 
-Always there when you don't need it (buy shares, services) 
-Unless you have 18 months of operating costs in the bank to burn (asked for audience raise of hands to see who had that- no one did) 
-2008 VC: Red 5, Slipgate, Trion, Perpetual, etc. etc. 
-25-30 mil -> stick in bank, delay launch, pay for delay penalties 
- < 100% of something is better than > 100% of 0 
-cash, equity okay 
-we didn't take it because we thought we could make it on our own 
See the difference between the forest and a tree 
-Had no outside point of view 
-Outsiders can make an impact 
-Better band than a business 
-get Board of Directors or Board of Advisors to be your outsiders 
We're All in this Together 
-Buy game to compete with others 
-Don't wish failure on others 
-Went down because didn't sell enough product 
-others will die through no fault of their own 
-support each other 
-small budget can still get big audience 
-500k copies (big for PC retail) - no subscription 
-if you do microtransactions - run your operation super light 
 
 
 
Solid Game Design (Kojima keynote) 
 
This keynote was okay, and everyone has a writeup about it since it was during the main conference and nothing else was scheduled for the timeslot. Kojima joked that the only reason he came was because they said if he did, they would offer him an award. He offered many good insights hidden among a giant history lesson. It seemed like it would have been much more effective on a Japanese audience.  
 
 
Making the Impossible Possible 
-No creation yet has been truly revolutionary 
-Barrier of Impossibility - make a game 
Possible: done before 
Impossible: not experienced 
Discard Preconceived Notions - go around the wall - set up a new mission 
Use tools (mario->mushroom - snake->ladder, box, etc.) 
Other routes, new ideas, new vision 
Barriers: Hardware Technology -> ground level 
Software Technology can lift you above that somewhat 
Game Design - new idea, tool - helps you get over the barrier 
Mission: Create a combat game for the MSX2 
-no sprites - limit - algorithm = chase down 
-no combat -> escape -> hide -> ! 
-Metal Gear came about just because the hardware could not handle bullets 
Design - ladder - sight range, change AI, "!" notify mechanism 
MG2:SS - Evasion minimap, 3 alert modes, hearing and vision for AI 
MGS sep, oct in US, 1998 - worldwide release - 3D area (ducts) - real time cutscenes - camera changes 
MGS2 no new hardware 
-go for immersion via game design 
-shadows, AI, environmental effects (rain, fog lens), new player actions 
-motion capture, puzzles, shooting 
When designing, map out areas you have already explored and look for holes 
-Metal Gear had not taken place outside much so MGS3 focused more on outside environments 
MGS3 2004 new engine, survival, heal, open natural environments, capture food 
MGS4 2008 Factions, Alliances - Japanese Review 
Each time he showed the intro cutscene he expected people to laugh 
Work together and support industry 
 
 
 
A Whirled Tour 
 

This incredible talk was chock full of useful information and practical data.  The speaker went very fast so I was scrambling to take what notes I could.   

 
 
Virtual World + Social Network + Casual Games Website 
Coins - earned, Bars - real world currency 
-economy was spiraling out of control until they introduced some items that can only be bought with bars 
-limit earned coins to 600/day 
-there is also Bling which content creators gain through sales- Bling can be converted to real world money 
Creators can develop games, even MMOs to sell 
Players can earn trophies 
Item and Level packs are more sellable 
Franchise merchandise also sells 
Creators can develop avatars 
Tech - Google Web Kit - Java - JavaScript - Flash - in-house Java server 
Nodes - Rooms - individual servers that handle specific sets of tasks 
LB Router - RPC - DOBJ (proprietary small message format) 
Server infrastructure/colo: clients <-> S3 <-> FC <-> Nodes, LBR1, LBR2, BNode <-> BC <-> DB 
IM is handled through 1-on-1 chats b/t nodes 
Flash - 15-20 Whirled avatars before dying 
Clients -> Node -> room data change rooms, change server nodes, pass around session data 
Nodes share what room/game/thing it has constantly - manual/routing 
Group chat = 1 Node hosts, sends to nodes with those channels 
Distributed locking mechanism - before publishing node routing info 
Clients have a many-to-many relationship with Nodes - Google Juice @Inject transient keyword 
Amazon S3 - download - Remix - (Editors - can easily grab an image editor in flash - instructions to easily make an avatar with your head) 
Creators that make games have access to an API that lets them award achievements (trophies), and prizes (pets, items) 
Games sell content packs - stream from S3 (own mp3's) - customize MOBs, AVs, even Whirled GUI 
Score ratings (keep a distribution, higher percent -> rating, easy to catch cheaters who go way above the bell curve) 
Game Coin Budget - games can award this many coins, based on scores, duration of game, humanity (isBot), etc. 
Protect from exploits - running unknown code on server(!) - Tamarin, Flash VM 
all user-created games are run on secured firewalled server - S3, 80 machines, quotas - fail check - scale database 
Alternate Virtual Reality Games DirectPlay ssh 47624  
samskivert.com/work/2009/gdc/WhirledTour.pdf 
Amazon ACC - TSUNG - EC2 - 35 I/O thread, DB thread, 2000 = 1 cpu - darkstar 
 
 
 
Cinematic Next-Generation Action NARUTO 
 
A great session with up front post-mortem level quality discussion. Many interesting techniques from real development.  
 
 
CyberConnect2 - formed in 1996 - Kyuushuu - 109-120 employees 
good corporate environment => good development 
Sports - hard license 
War - bad in Japan 
Movie - Japan strengths - Anime, Manga 
2006 - months, # of people 
R&D - 6, 10 - 1 designer, artists, maybe 2 programmers 
Proto - 6, 10 
Base - 6, 24 
Main - 6, 38 
High Vol - 9, 65 
Publisher progress - report every 6 months - document new effects and R&D 
organize team with people best suited for the phase 
Naruto Ultimate Ninja Storm Art 
Character 
-soft clean lively colors - bend head - normals - toon shading - emphasize chin 
-3 LoD - 2.4k, 1.4k, and 7k - flatten eyes 
-soft light shader (except areas with high chroma) - artist select outline shader 
-52 22 tree 67 facial polygons - bend bones for exaggerated expressions - stretch tris for anime blur - free bones 
Background 
-hand drawn touch - silhouettes - 3D Max - special lighting shield - different vantage points 
Effects 
-Layer - ZSort no ZBuffer - 8-loop hachi_maru animations 
-billboard fx inside simple geometry e.g. fireball sphere + billboard inside + animated flames 
Video 
-cinematics - script - match to game system, fx, mechanics, design 
-director check 
-licenser - check vs license 
-prototype - pose only, get camera angle, timing, etc 
-director check 
-motion - full animations - imagery fix - white stand-in enemy 
-director check 
-publisher - check, edit, fix, finalize 
 
 
 
Experimental Gameplay Session 
 
I came to the last half of this session since it was happening at the same time as the CC2 session. I got the list of all the games from a friendly attendee. There were as usual many awesome new ideas presented and many inspiring moments throughout the whole session. 
 
 
Here is the list and notes from the person I sat next to, Joe Mathes. 
 
* <http://iandallas.com/games/swan/> Unfinished Swan - 3D Exploration 
Game <http://iandallas.com/games/swan/>  
 
* Shadow Physics 
<http://www.avclub.com/articles/avc-at-gdc-09-day-four-wake-me-when-your-game-does,25806/> - You're a shadow not cast by anything interacting with the 
shadows cast actual 3d objects. The link is to an article that mentions it; 
it doesn't seem to have a page of its own anywhere 
* Miegakure <http://www.enquirerherald.com/371/story/603701.html> - 
the 4D platformer that broke everyone's brains. As above, link is to an 
article 
* Spy Party 
<http://www.eurogamer.net/articles/gdc-experimental-games-shown-playable> - 
Game where you're trying to blend in with NPCs while carrying out missions, 
and another player is trying to guess which NPC is you. As above, link is 
to an article 
* Moon Stories - A lot of different experimental gameplay ideas and 
tiny games, most of them involving playing with time as a game mechanic and 
romantic relationships also as a game mechanic. Here are links to three of 
them that actually let you play: I wish I were the moon 
<http://www.kongregate.com/games/danielben/i-wish-i-were-the-moon> , The 
<http://www.ludomancy.com/games/Trials.html> Trials , and Storyteller 
<http://www.ludomancy.com/games/StoryTeller.html>  
* Flower <http://en.wikipedia.org/wiki/Flower_(game)> - The 
developers walked us through the evolution of this game's gameplay. The 
endpoint is you fly around as a cloud of petals, collecting new petals that 
give you powers, and spreading the petals and using them to pass obstacles. 
Pretty, ambient, and not challenging. Like a kid's game. 
* Achron <http://achrongame.com/> - The RTS where you can go back in 
time or send things back in time. The most fun-looking and groundbreaking 
of the lot IMHO. 
* Rom Check Fail <http://rom-check-fail.en.softonic.com/> - You're a 
classic NES game hero up against a classic NES enemy. But which hero and 
which both randomize every few seconds. Link is to a place where you can 
download it. 
* Where is my Heart 
<http://bushghost.blogspot.com/2008/12/student-igf-release.html> - You 
control 3 little monsters as they cooperate to get through levels. Similar 
to the lost vikings 
<http://www.blizzard.com/us/blizzclassic/vikingsdemo.html> . The game is 
presented on one screen in comic-book panels, and it's not clear without 
exploration which panels connect to which others in what ways. Link is to a 
place where you can download it. 
* Closure <http://www.kongregate.com/games/GlaielGamer/closure> - In 
this, every level is dark, and there are few light sources. However, your 
character does not physically interact with anything in the dark - so if you 
move a light away from a wall, you can walk through the wall. Or fall off 
the bottom of the screen, through the dark floor. Link is to its Kongregate 
page, where you can play it 
* Spelunky <http://forums.tigsource.com/index.php?topic=4017.0> - 
Actually, this talk was about roguelikes and extreme consequences in 
general. The key point is that if the level is procedurally generated, 
death is permanent, and you can't save your game, then every time you play 
the game, that is the only time anyone will ever play the level you're on. 
That makes it intense and important-feeling. This one was given by 
Timothy's friend Derek Yu. Link is to tigsource, where you can download it. 
Timothy does ops for tigsource. 
 
Here are my notes on the ones that I got to see. 
 
First half - Unfinished Swan, Shadow Physics, Miegakure 4D, Spy Party, Moon Stories 
Second half 
Flower - Jenova Chen 
-flow +3D -AI 
-safe, free, love - peace and harmony - petal blowing - growing - orbs 
-many prototypes - mind-teleport - ground, seed, golf 
-needed depth and challenge - collect petals, gain spells, time limit 
Achron - in development for 10 years 
-RTS where all units can time travel 
-timewave (past changes gradually affect future) 
-chrono energy - further back you travel, the higher the energy drain 
-return to present to recharge 
-achrongame.com 
Closure - Tyler Glail 
-borne out of a frustration with 'dark levels' - areas/rooms with only small light and limited vision 
-idea - had to resort to puzzle game - can only interact with what you can see 
Where is my Heart 
-Comic Panel effect 
-multi avatar 
-one that can see dark world 
Rom Check Fail 
-Gameplay Variation 
-standard pacing chart - gameplay variance induces spikes in intensity 
-vary hero too, not just weapon or enemy 
Roguelikes 
-Rogue - 1980 - berkeley - permanent death + randomly generated levels 
-chess rogue, gangsta rogue, real lives 2007, dwarf fortress, etc 
-Spelunky - more approachable, platformer rogue 
 
 
 
Negotiation 101 
 
This was a great talk and had lots of good points that everyone seeking to have a good career should know.  
 
 
First rule of negotiation is Preparation 
-Negotiations can be won or lost before they even start 
-Know thyself - objectives, parameters 
-Plan B so you can walk out if offer is not good 
-What will they want? Company & Person - know person's role 
-Can you give what they could want? 
-New terms added when agreed with lower peons 
-Legal department = bad - only see risks 
-talk to marketing! 
-_Practice_ 
-intro, question, points, exits 
-comfortable, confident 
-Research culture issues 
-cards for Japanese 
-order in which to shake hands (leader first in China) 
-Practice with big purchases (TV, car, etc.) 
Negotiation 
-in-person, remote? usually best is their place- they can't walk out on you in their own office 
-make sure key decision makers are involved 
-one person per party - less people = easier negotiations - see "Getting to Yes" Book 
Strategies 
-Soft - getting deal no matter what 
-Hard - fight all points to get the best deal 
-Hard is obviously better, but if everyone is hard all the time, nothing would ever get done 
-Emphathize - discover actual interests - discuss interests, _not positions_ 
-Listen - blame? stay calm, just listen - "Make your fist in your pocket." 
-Cooldown - keep up momentum 
-if no deal, try to preserve the relationship 
 
 
 
Social Games 
 
Decent talk with some interesting games. Only caught part of this one as the conference was winding down.  
 
 
Parking Wars 
-parasite cheat (won't ticket you if you don't ticket me) 
-fake accounts (park all there) 
BackChannel 
-comments on TV show while it is going - clicked ones grow 
-similar tag matching 
Kelly's Bags 
-electrolux, social dilemma, events, being hunted, bags, choose the most popular one, you win 
 
 
 
Iterative Level Design 
 
Great talk from the Bioware guys on iterative level design. Also could only catch the end of this one. 
 
 
Bioware - the Richards - nice guys 
Corey Andruko - coray@bioware.com 
Dusty Everman - dusty@bioware.com 
 
Documentation 
Mass Effect 1 - Kismet -> wiki export 
Mass Effect 2 - SharePoint -> Perforce