I looked on the net for some detailed article on WPF Architecture. I found the MSDN Documentation but its too detailed and goes into the namespaces even before explaining the basics. I found a video on Channel9 of Chris Anderson doing a whiteboard session explaining the architecture. Greg Schechter also has an hour long video on Channel9 where he talks about WPF architecture , Windows Vista DWM and many other things.
I thought all that I learned from watching these videos could be summarized in an article to explain the architecture and save others the trouble of watching the videos (only if you dont have the time for it, otherwise you would learn a lot from them).
In WPF you have a Visual Tree of all the objects in your XAML file. This is a tree with all the visible UIElements as well as the other visual core components that make up each object. Now the UI thread running your application manages the Visual Tree. WPF has a Single Threaded Model and your UI thread is the only thread you as a developer have to manage. Any UI element must be accessed via this thread. Apart from the UI thread, there is a Render thread running in the background. The Render thread is responsible for creating and managing a Composition Tree based on the Visual Tree. Remember that in WPF a button can have its content as a Grid which might have an Image as the content of its each cell. In other words, elements can be composed of other elements and this composition can be as nested as you want it to be. The Composition Tree layouts the entire compositon of all the elements in the XAML file. The Render Thread is responsible for navigating through this entire tree and rendering the elements.
An important point to note here is that as changes happen on the visual tree e.g.: a radio button is selected, only those deltas are communicated to the render thread and accordingly the composition tree is updated. Also the re-rendering happens only for those deltas instead re-rendering or repainting the whole screen. The render thread is responsible for talking to Direct3D and rendering the content. Lets see how that happens.
The above diagram explains the WPF architecture. At the lower most level you have the Kernel , your graphics drivers etc… on top of it you have 2 modules. USER32 is reponsible for deciding what window goes where on your screen. So each window gets its allocated space by this guy. Direct3D is responsible for rendering the content of each of these windows. Direct3D talks to the graphics drivers and sets the pixels on the monitor. milcore is not a publically exposed API but futured versions of WPF might see it opened up to developers maybe via other interfaces than the ones explained below.
On top of Direct3D, the WPF team built milcore.dll. MIL is an acronym for Media Integration Library. This is an unmanaged module and is basically the composition engine thats providing features like 2D, 3D, animation etc… Its the one that directly communicates to Direct3D and can be thought of as a middle man between the rest of WPF and Direct3D. There is also a WindCodecs.dll which houses the unmanaged implementation of various imaging codecs like png, jpeg etc…
On top of this you have the .NET 3.0 Common Language Runtime. Above this layer is the PresentationCore which is a mix of managed and unmanaged code. This is a low level API exposed by WPF providing features for 2D, 3D, Geometry etc… Above this is the managed PresentationFramework. I provides access to high level features like applications, controls, styles etc.. Also you have access to layouts, data, content and actions which help you build up your entire application.
So thats the WPF architecture.
