Curious why no hybrid MSAA/SSAA?

[Xmas]

Doing that breaks gamma corrected AA. 'Nuf said

In fact, it doesn't. "Gamma corrected AA" is gamma corrected downsampling, and this method doesn't affect downsampling at all.

 

[Mintmaster]

Doing that breaks gamma corrected AA. 'Nuf said

Could you explain how? Doesn't the card still do the final downsampling the same?

Also, could you modify the repeating geometry in your FSAA viewer so that we can see the effects of centroid sampling?

EDIT: Nevermind, I see you provided the source code. I'll give it a shot. And I see Xmas just brought up the same point about GCAA.

 

[SteveHill]

here's a pic of some alpha test errors in action...

Sorry, but to me it looks like a sorting issue. (Rendering stuff in the wrong order.)
Of course I might be wrong but if there's also transparency on the "ice", which I believe is the case, it really sounds like a sorting issue.

Yup, I agree. I spotted a few sorting errors in the demo and I was told that they (artists) didn't bother/forgot/chose not to* flag said surfaces/materials as "requiring sorting".

* A lot of the time you don't notice the errors; I only spotted significant problems in conjunction with particles -- rocket trails are a glaring example.

 

[KimB]

* A lot of the time you don't notice the errors; I only spotted significant problems in conjunction with particles -- rocket trails are a glaring example.

You'd think that rocket trails would be one of the easiest to sort correctly.

 

[Colourless]

As to my comment about it breaking GCAA, it's a matter of Direct3D specifications. The hardware MUST average the subsamples when you use multisampling masking. It is not allowed to perform other post processing such as GCAA. I noticed this behaviour ages ago occuring within ATI's drivers and queried it and that is what I was told (by OpenGL guy?)

 

[OpenGL guy]

As to my comment about it breaking GCAA, it's a matter of Direct3D specifications. The hardware MUST average the subsamples when you use multisampling masking. It is not allowed to perform other post processing such as GCAA. I noticed this behaviour ages ago occuring within ATI's drivers and queried it and that is what I was told (by OpenGL guy?)

That was what the old DCT kit required, not sure now. Anyway, if you use this algorithm in the driver, there's nothing to stop you from doing GCAA.

 

[Mintmaster]

OGL guy, could you please share the answer to my centroid question? When using multisample masking, is the hardware capable of texture sampling at the centroid of only the enabled samples? Apologies if I'm sounding pushy.

 

[OpenGL guy]

OGL guy, could you please share the answer to my centroid question? When using multisample masking, is the hardware capable of texture sampling at the centroid of only the enabled samples? Apologies if I'm sounding pushy.

It should work, but I've never tried it.

 

[Mintmaster]

OGL guy, could you please share the answer to my centroid question? When using multisample masking, is the hardware capable of texture sampling at the centroid of only the enabled samples? Apologies if I'm sounding pushy.

It should work, but I've never tried it.

Cool. If so, then Xmas' suggestion (draw alpha-tested objects multiple times with multisample masking) should work without the shifting. Or am I missing something?

 

[OpenGL guy]

OGL guy, could you please share the answer to my centroid question? When using multisample masking, is the hardware capable of texture sampling at the centroid of only the enabled samples? Apologies if I'm sounding pushy.

It should work, but I've never tried it.

Cool. If so, then Xmas' suggestion (draw alpha-tested objects multiple times with multisample masking) should work without the shifting. Or am I missing something?

I think I tried this idea before, can't recall why it didn't work. I'll have to look into it.