I used Texture2D's heavily with the usage RenderTarget, however in Beta 2 its like a whole new API.
I understand that Surface no longer exists and RenderTarget/2D should be used. The problem is, the whole mentality is backwards compared to what im used to.
From what I see I should create a RenderTarget2D and set that as a render target, then do rendertarget.GetTexture() if I want to set it. Thats fine.
However, how do I copy between render targets
There is no StretchRect or equivelent function I can see, which I used alot before. Surely I dont have to go about getting a texture, setting it, creating a 'copy' effect, setting a new render target, resolving the new render target just to copy from one to another
Is there no alternative to StretchRect If not its a sad loss that makes things cumbersome, or even tiresome now.
Rendering To Textures in Beta2
WinFormsUser13232
Sorry, but im still lost with this.
I have a brightpass and bloom post process effect.
I need to get the current backbuffer into a RenderTarget2D...
Webbert
juanraul
"Resolve" basically just means "copy current backbuffer contents to a texture".
The reason you can only do this from a backbuffer, and not copy any arbitrary texture into some other texture, is that on Xbox there is only one current backbuffer, which lives in special memory on the GPU. All rendering happens into this special memory, and the only thing you can do at the end is to copy the results back into a regular texture. We looked at trying to implement Windows Direct3D9 style semantics on Xbox, but realised that was going to be way too slow and unpredictable to be useful, wheras with the current API, we can make things both consistent and fast on both platforms.
To copy a Texture2D into a cubemap, you can create a RenderTargetCube, set one of its faces as the active rendertarget, draw your Texture2D as a fullscreen quad, then ResolveRenderTarget to get the result back into the cubemap. On Xbox that will actually do a copy from the special framebuffer memory, while on Windows it will generally just be able to share the underlying rendertarget memory so no extra copying is required.
Pimpom
We were using SretchRect to copy from the BackBuffer to a texture for some post process effects.
Also used it to get into the correct format for heightmap processing.
Also for extracting cubmap surfaces into a Texture2D, and visa versa.
And yes, also for taking SnapShots of the scene (screenshots).
As you can see we use it quite alot. If it's not possible to get texture2D surfaces and copy between them, can you at least post a little of equivelent of how to do it via the new RenderTarget2D and Resolve function
SpaceCadet
Farrokh Moori
walking fish
Like in the previous post - the rendertarget2d is receiving the rendering calls. Call SetRenderTarget before calling Clear, Draw, etc. When you're done drawing into the RenderTarget2D, call ResolveRenderTarget and then use the texture from the render target object (obtained by calling GetTexture() on the object) as an input to your next pass (which you probably want to render to another render target, or to the implicit render target, i.e. the back buffer, by calling SetRenderTarget(0, null)).
e_LA
- Set destination rendertarget
- Draw fullscreen quad using source rendertarget texture
- ResolveRenderTarget
- Unset rendertarget
If the thing you want to resolve to happens to be the same size as the backbuffer,
you don't need to use rendertargets at all:
- Create a texture with the ResolveTarget usage
- Render fullscreen quad to backbuffer
- ResolveBackBuffer, passing it your texture
That's sometimes a bit more direct, but less flexible. There won't be any perf difference between the two approaches.
John123
If you really want to have the rendered scene in a RenderTarget2D, you need to render it into that rendertarget, then ResolveRenderTarget when you are done.
If you want to render into the main backbuffer, use ResolveBackBuffer, which resolves into a regular texture, not a RenderTarget2D.
mroctober
You're right, there is no StretchRect any more. You have to Resolve a rendertarget after you are done drawing into it to make the data available as a texture. That's how the Xbox hardware works, as as a neat benefit of emulating this on Windows, the new API makes multisampled rendertargets work much more reliably and consistently than they used to.
Why are you needing to copy between rendertargets, btw
ICP-Fan
Alexander Petukhov
I just posted a code snippet on how to use render targets in Beta 2 in this topic: http://forums.microsoft.com/MSDN/ShowPost.aspx PostID=882859&SiteID=1
To copy between render targets, you could use a SpriteBatch to render the data from one render target (obtained with GetTexture) onto another render target. I don't know if this will give you any good performance though, but it's worth a shot.
Marimuthu_r
Wait, so if I want to copy from one RenderTarget2D/Texture2D to another I do this:
Set RenderTarget B on the device
Set RenderTarget A Texture
Draw a Full Screen Quad
ResolveRenderTarget B;
Set Device render target to NULL;
use Render Target B Texture whever I want
OR
Set RenderTarget B on the device
ResolveBackbuffer into texture
Set Device render target to NULL;
use texture wherever I like
Which one is correct or best
Tatworth
Ok, one more question. Does ResolveBackBuffer() copy the real 'hidden' backbuffer into a texture, or only from the last RenderTarget2D you set
If its not the hidden backbuffer im stuck again.