9 replies to this topic

[londer]

which one of these paradigms is better to follow in gamedev? or maybe both in combination?

[frob]

They are buzzwords.

Both are labels for different methods of encapsulation, which has always been a good practice regardless of language.

[Radikalizm]

If you're familiar with object oriented-designs there should really be no reason for switching to data-oriented desings until you actually find a valid reason to do so, there's no reason to trouble yourself with adapting to a new paradigm or to try and combine the two if you don't have any problems with the paradigm you're currently using.

Data-oriented designs are all about shifting your focus from code to data, and they are often used to restructure code and data layout so they are more 'cache-friendly'. If you don't have any bottlenecks with data updates there's probably no reason for you to trouble yourself with trying to switch to a data-oriented design.

So basically, stick with what you know best. If what you know best one day doesn't cut it anymore for the problem you're trying to solve (and I think it'll be a long while before that day comes), consider switching to another approach.

[Hodgman]

Why not use both? A lot of the time, they're not mutually exclusive choices.

OOD is basically a big collection of wisdom in writing robust software.
DOD is basically the statement "please stop and think about what exactly it is that you're telling the CPU, Cache and RAM to do, and then keep it simple".

The way I see it, DOD is basically there as a "check and balance" to make sure you don't end up getting too carried away with OOD abstractions, to the point where you forget that at the end of the day, all you're doing is transforming blobs of data into other blobs of data.

[Bregma]

Red. Paint your bike shed red. That way, your game will get done sooner and there will be fewer bugs.

There are a number of books available that describe the best colour to paint your bike shed. Purchasing them will make the authors wealthy and give you grounds to prove someone on the internet is wrong.

[snowmanZOMG]

which one of these paradigms is better to follow in gamedev? or maybe both in combination?

Use what's most appropriate for the problem you're trying to solve. It may be difficult to know which solution is "more appropriate" if you don't have the requisite experience. You can often use pure DoD and avoid OOP but you might have to work much harder to get the result you want. The same can be said of the other direction, but the cost of "work much harder" might appear to you in a different way. One may just be awkward to write the code for or you may have to write excessive amounts of code. The other may obscure the purpose of code unnecessarily or require extreme amounts of detailed knowledge to be able to use properly.

Data oriented designs can lend itself to very high performance code, but you need to do it correctly. OOP can make things more convenient, but you may end up with a monstrous hierarchy with strange dependencies. And you may have poor performing code.

Understand your tools well and use them wisely.

[Xai]

I actual practice something I always thought of as "data oriented design" ... but it is not the same thing as what these other posters (and probably the OP) are talking about. When modeling and then designing my program, I feel in my heart that the "data" is the only true thing that absolutely has to exist in a computer. everything else are fictions invented for the programming. And as I model, I focus on the inherent association between different pieces of data among the "objects" in my model. I have a very "data model" oriented approach as I'm starting out. Usually even drawing up simplified ERD style diagram for the objects in my system. Then I attach the most fundamental API to these, based on the data they wrap and how they should "protect" it or what useful features they can expose. As I flesh out more and more functionality among the classes / objects in my hierarchy when I get confused in the mess of abstraction or complexity I may have created, I go back to basics and ask "which class owns the data?", "what do I need to do to that data?", "what methods should I add to do this?" - and "who should own that method?". And the "who should own" is usually based on what data is needed to drive the logic of the method.

For instance a generic "Board" class wouldn't know how to setup a chess game, because a board doesn't know the rules or pieces of chess. But a "ChessBoard", "ChessGame", "ChessGameEngine", etc ... would all be acceptable places to put this logic. Even though the data being written is probably "owned" by a "Board" object, the logic of doing so is not ... "Chess" is a board game, hence a client of the concept of a board.

Now my last sentence sounds all OO right? ... and it is ... but I got to that design from a data first process, that ENDS in OO, but doesn't start there.

[superman3275]

What's interesting to me is pointers applying to this:

If a lot of different parts of your program need certain data, You should try to seperate the data into the certain parts of it those classes need. Then, if there actually is shared data, you'll need a pointer. Now here's the problem: Which class actually holds the data? And to solve this, all I have to do is think:

Which class will need more (Ex: You have a Soundmanager Class and a SoundPlayer class. The Soundmanager class keeps track of sounds size, length, type, etc. while SoundPlayer plays it. Which one will need more of the actual data? Soundmanager, while all Soundplayer wants is the audio file, it doesn't care about the rest of the stuff you tie to itj.)
Which class will use it the most.

Now, you may say: But why can't Soundplayer just have the audio file and Soundmanager have the data. Well, how do you think SoundManager calculates the data? But wait: If all I want is for Soundmanager to have the data points, then I could just have SoundManager keep the actual audio file while Soundmanager uses a pointer to the file to get the information and store it. Problem Solved!

Many beginners don't see the importance of pointers, however they become needed when you have shared data (And dynamically allocated memory, however unless you're in a cutting edge game you generally don't need to utilize this because of the increase in size and decrease in cost of RAM.)

[Goran Milovanovic]

... I'll just upvote hodgman.

Now my last sentence sounds all OO right? ... and it is ... but I got to that design from a data first process, that ENDS in OO, but doesn't start there.

There is a distinction to be made between "Data Oriented" and "Data Driven"; I think your post describes the latter, which is basically a facet of good OOD.

Edited by Goran Milovanovic, 21 November 2012 - 12:53 PM.

[Hodgman]

Now my last sentence sounds all OO right? ... and it is ... but I got to that design from a data first process, that ENDS in OO, but doesn't start there.

There is a distinction to be made between "Data Oriented" and "Data Driven"; I think your post describes the latter, which is basically a facet of good OOD.

A better description of DOD than the one I gave earlier might be "think about the data first, then the code later".
Under that description, Xai sounds like he's focussing on (or orienting his thoughts around) the data as a primary concern.

e.g. with an OOD renderer, when setting material values, you might think "ok, I'll need a pointer to an IMaterialDataSource interface... what methods will it have?".
With a DOD renderer, you might instead consider that if the signature to your function was void SetMaterialData(const byte* allTheDataYouNeed, int size), then what would that buffer of bytes contain, and how would be be laid out?

Often, if you design around the data first, and then slap an OO layer on top of the data, then it becomes very data-driven as a side effect.

e.g. I've got an OOP interface for reflecting on shader programs, such as looking up a techqnique by name:

struct ShaderPack : NonCopyable, NoCreate
{
	int FindTechniqueByIndex( const char* name ) const;
	int GetTechniqueBlah( int index ) const;
};

But long before I wrote that interface and what methods it could have, I first sat down and designed a bunch of classes with no methods and just data. There's also other classes that can consume this data besides the above interface.

template<class T> struct ArrayOffset  { s32 offset; /*operators to act like a T[]    */ };
                  struct StringOffset { s32 offset; /*operators to act like a char*  */ };

struct Technique
{
	u32 blah ... ;
};
struct ShaderPackBlob
{
	u32                         numTechniques;
	ArrayOffset<Technique>      techniques;
	ArrayOffset<StringOffset>   techniqueNames;
};

Edited by Hodgman, 21 November 2012 - 08:42 PM.