Your Wish List : Things that didn't make D3D10

[Reverend]

Well, out with it then. A number of things were shot down "outright", a number of things were "under consideration" but ultimately was excluded (helps if we can name the reasons), a number of things were "really close" to being included...

It would be much more interesting to read/discuss "official reasons for exclusions", I suppose.

 

[digitalwanderer]

Mandatory AA compatibility/support built into it.

Don't know if that's possible or not, but I'd really have liked to see a bit more effort with it.

(Just IMHO and all that, don't kill me!)

 

[memberSince97]

I think deferred rendering and AA is possible in DX10

 

[SuperCow]

In my opinion the most glarant missing feature in DX10 is the unability to resolve a multisampled depth buffer (either through ResolveSubResource() or manually in a shader accessing samples). This will be causing special cases in quite a few engines that were looking forward to treat MSAA as an orthogonal feature in their engine.

 

[armchair_architect]

In my opinion the most glarant missing feature in DX10 is the unability to resolve a multisampled depth buffer (either through ResolveSubResource() or manually in a shader accessing samples). This will be causing special cases in quite a few engines that were looking forward to treat MSAA as an orthogonal feature in their engine.

Out of curiosity, how would you merge the depth samples within a pixel? Averaging (like resolves of color buffers) isn't going to give you anything useful, for the same reason that normal bilinear filtering of shadowmaps is wrong.

 

[Demirug]

Mandatory AA compatibility/support built into it.

Don't know if that's possible or not, but I'd really have liked to see a bit more effort with it.

(Just IMHO and all that, don't kill me!)

This is currently planed for D3D 10.1

 

[digitalwanderer]

This is currently planed for D3D 10.1

Are you serious? Or are you playing a modern "dx 9.1" joke on the old Diggers?

 

[SuperCow]

Out of curiosity, how would you merge the depth samples within a pixel? Averaging (like resolves of color buffers) isn't going to give you anything useful, for the same reason that normal bilinear filtering of shadowmaps is wrong.

You're perfectly right that there is no "right" way of resolving a depth buffer. However a "gross" resolve operation that e.g. picks a single sample would be enough to cover a lot of cases like soft particles, depth-based fog and even depth of field. The preferred option would be to let the pixel shader access individual samples in the depth buffer like any other multisampled surfaces however this isn't supported in DX10.

 

[Demirug]

Are you serious? Or are you playing a modern "dx 9.1" joke on the old Diggers?

No the “codenameâ€￾ for the next DirectX version is currently 10.1 and the shader model is 4.1.

You're perfectly right that there is no "right" way of resolving a depth buffer. However a "gross" resolve operation that e.g. picks a single sample would be enough to cover a lot of cases like soft particles, depth-based fog and even depth of field. The preferred option would be to let the pixel shader access individual samples in the depth buffer like any other multisampled surfaces however this isn't supported in DX10.

Well Direct3D 10 can support any operation for every format. The problem here is that resolves or subsample loads for depth stencil formats are not required operations to claim D3D10 compatibility.

 

[Humus]

Well Direct3D 10 can support any operation for every format.

That's not true. There are some stuff that's mandatory, some that's optional, and some that outright disallowed so that the runtime prevents such functionality. There's a table in the spec which lists what's mandatory, optional and unsupported/disallowed.

 

[Demirug]

That's not true. There are some stuff that's mandatory, some that's optional, and some that outright disallowed so that the runtime prevents such functionality. There's a table in the spec which lists what's mandatory, optional and unsupported/disallowed.

Thanks for this update. I was aware that there is such a list but I didn’t know that there is a disallowed section.

 

[bloodbob]

A teapot primitive.

 

[nicolasb]

Ray-tracing!

 

[Davros]

continuous tessellation that rocks....

 

[Davros]

Looks like ati have been listening to me

lets hope devs use this feature...

 

[Kaotik]

Looks like ati have been listening to me

lets hope devs use this feature...

The problem is the same as with TruForm for example, only one IHV supports it, before it's adopted by Microsoft for DirectX, there will most likely be handful of games at best supporting it.

 

[Demirug]

Looks like ati have been listening to me

lets hope devs use this feature...

It’s not part of the DX10 API. Therefore the needs another API hack to make it accessible.

 

[AlNom]

Why would they include the tessellator then

 

[Demirug]

There was a tessellator in some early D3D10 drafts and ATI had the technology from the Xbox anyway. The still could provide support for it with OpenGL.

 

[Davros]

I thought it was allready in dx9 ?

Heres some text i nicked from x-bit review of the parhelia
http://www.xbitlabs.com/articles/video/display/matrox-parhelia.html

"Adaptive polygon tessellation is one more key feature of Matrox Parhelia. As hardware adaptive polygon tessellation is going to be one of the requirements for DirectX9-compatible graphics cards"

Was it removed or is the programable tessalator in r600 something different ?