Sign in to follow this  
JDUK

Where to get more info on Buffer locking in MDX9

Recommended Posts

JDUK    100
Hi im using Managed DirectX 9 and C#. & I'm reading through Tom Millers excellent book but im not quite grasping the intricacies of locking buffers (the why's and when to's, etc) as described in Chapter 8. Does any one know where i can get a more indepth tutorial on the topic? Thanks.

Share this post


Link to post
Share on other sites
IFooBar    906
the documnetation that comes with the DX SDK has excellent information on vertex/index buffers and how to use them most efficiently.

In a nutshell. By locking a vb/ib (vertex buffer/index buffer), you're telling d3d that "Im using these vertices/indices". If d3d knows this then it wont do any processing on the vertices you are using. When you unlock, you are telling d3d that you're done using those vertices, so now d3d can do whatever it wants to it.

You have to lcok vertex buffers to put data inside them. When they are created, all the vertices have garbage values. SO you need to lock the buffer to get a pointer to the buffer, then through this pointer you can put data into the buffer.

As for how to lock, well you have two types of data that you put in a buffer. Static data, and dynamic data. Static would be the vetices of a room (usually dosen't change) and dynamic would be a particle system (usually changes every frame).

For static data you just usually lock teh buffer once and fill the whole thing at load time. So there isn't really very many things to take into consideration - abuse load time as much as you can :)

Dynamic data on the other hand has to be filled in almost everyframe. Every frame you'd lock the buffer and tell d3d that "Im using this portion of the buffer, dont touch it, do what you want with the rest of the buffer". You do that by specifying the D3DLOCK_NOOVERWRITE flag when locking a buffer. This lets you update a portion of the buffer (the changed portion) and also lets d3d access teh rest of the buffer. This is necessarry because usually other parts of your buffer are still being rendered, so if you locked the whole buffer, youd cause a stall in the graphics pipeline (you'd have to wait till the hardware was done rendering and only then would you be allowed access). But with the NOOVERWRITE flag, you allow the hw to work in parallel with you.

Again the sdk docs describe three different ways of making use of dynamic vb/ib. Make sure you read it all.

Share this post


Link to post
Share on other sites
JDUK    100
Big thx! great description :) I totaly get what your saying .... it was just the only part of the book that sounded a bit esoteric or heavy going.

Just to clarify: a room (a simple one that does not change) would be static while a player model is dynamic as it moves and animates ?

Is there much of a performance gain to be had by carefully managing these resources or are they only a consideration when wanting to take games as near to the cutting edge as managed code gets?

thanks again :)

Share this post


Link to post
Share on other sites
onnel    124
Great info IFooBar!

A couple of follow on questions:

1. How do you manage the info to keep track of where the info for a given mesh is (either vertex or index respectively) in a buffer? I.E., my static walls are from 0 to 15000 15001 to 16355 is my first model, etc...

2. How many different Index and Vertex Buffers do you use? Is it worth having one for static data which never changes and another for dynamic? What about multiple buffers for different dynamic info?

3. How would you recommend handling data that changes often (in this case my terrain triangles, which are determined using a dynmanic LOD)?

I'm currently clearing my buffers with every render of every object (i.e., always rewriting to the buffers starting at offset 0). This obviously is not the most efficient way to go!

Thanks for the very lucid explanation!

Onnel

Share this post


Link to post
Share on other sites
IFooBar    906
JDUK: That is correct, the room would be static and the character would be dynamic. And yes there is a big performance gain to be had by carefully managing buffers. I don't have the exact links handy but nVidia's developer site and ATI's developer site have a lot of documents on optimizations and how to properly manage data in a buffer. They have all these charts that show you how much of a performance increase was achieved with various tests. the nvidia and ati developer sites have a lot of similarly useful documents/presentations on different aspects of graphics.


onnel:

1) Well there are many ways to keep track of where your data is within a buffer. It can be as simple as keeping a variable that tells where in the buffer it starts and another variable that keeps track of how many vertices is part of the object.

2) Definetly if you have data that is never going to be changing then storing all of that in a single static buffer is better then storing them in different buffers. Dynamic buffers are usually a "See which case gives best performance" issue. See link below.

3) If you haven't seen the VertexBuffer Optimization yet, I suggest you take a look at it. *Very* good info in there, and some of the main docs I mentioned above are directly linked from there as well.

Share this post


Link to post
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now

Sign in to follow this