Deferred Rendering With 11.6

Question

Hi, everyone.Since the first beta of adobe flash 11.6 released, I have been working on the new features, it is truly amazing that, Stage3D 2 is almost the SM3.0 standard.Here is the screenShot of using deferred rendering with Staeg3D 2. There are 1000 animated point lights with full Phong shdaing(Specular, Diffuse , Attanuation, Color) enabled in the scene. It runs pretty well in my GT 130M laptop:)Here are the technique I used in this demo:1) MRT(Multiple Render Targets) with 4 channels used: Position(RGBA Half_Float), Normal, Specular, Color.2) Rectangle texture for storing those G-Buffers with 1000 x 600 resolutions.3) Out put depth for the final composite pass, in order to perform hardware depth testing with the forward rendering objects in the back buffer.Here are the comments and may be bugs I found during the developing.1) If the rectangle texture are used for render targets, all of the render targets must be rectangle texture, and all of them have to be the same resolution, otherwise, the rendering would messed up, some of them did not rendered after insert a none- rectangle texture.i.e. G_Buffer pass-->all rectangle texture. Then insert  a none- rectangle texture to lighting pass. the G_buffer would never be rendered since that. I dont know if this is the hardware limitations, I am using GT 130M, the driver is the latest one.2) I suggest  Adobe could implement the depth-stencil texture in the 11.7 release. Since Currently doing the soft shadow mapping is less presicion and more expensive.With depth - stencil we could do the hardware soft shadow mapping, the format should looks like this: D24S8.3) So far, the performence and quality are very good, good job guys:)4) If there is any chance I can get the latest Stage3D 2 beta version, please let me know. My email is :Jasonhuang1988@gmail.com. I am currently working with Flare3D team, Thanks a lot:)And here is the SSAO screen shot with 16 samples and seperated 4x4 box filtering, running at the same condition:)

jingchao_Huang · Answer

Well, this is an early preview of Stage3D2, I am afraid that currently, it can not be found the beta version in Internet , but I guess it will be released in near future.

I finally finished the deferred rendering, with SSAO+ FXAA + Tonemapping. But still I did not use shadow maaping in this case, as the resulution issue till can not be fixed.

Here is the screen shot:

There are 500 point lights, and fully SSAO pass( two blur with depth and normal) , HDR pass and FXAA pass. it is running at GTX580 in 1280 * 720 resulotion, and we get solid 60 FPS:)

Regarding the resulotion issue, I found the problem:

In XNA we do something like create a new depth stencil buffer for MRT render targets, something like:

MRTs[0].depthStencil = new depthSencil

MRTs[1] .....

But in stage3D2 current version we still do not have a such thing, therefore, when we use 1024*768 for MRT render targets, that means the depth stencil buffer is also created by that resolution, So later we set a larger size of render target, the depth stencil will be skewed..

Hope, you guys could add the accessing of depth and stencil buffer for textures.

Regards:)

Sign up

To post, reply, or follow discussions, please sign in with your Adobe ID.

Sign in to Adobe Community

To post, reply, or follow discussions, please sign in with your Adobe ID.