Hi,

in the SMAA low preset the parameter SMAA_MAX_SEARCH_STEPS_DIAG is set to 0. Manually doing that in SweetFX 2.0 results in an error and setting it to 1 works with a warning. There are no such issue in previous SweetFX versions.

I'll look into it.

Previous SweetFX versions used SMAA 2.6 but with SweetFX 2.0 I've integrated SMAA 2.8 instead .. it's likely because of some changes in that.

As a workaround : Don't set SMAA_MAX_SEARCH_STEPS_DIAG to 0 in any SweetFX version, it negatively affects the quality of SMAA when you do that.

Even a low setting like 2 looks much better than 0.

Thx 4 looking.

A higher SMAA setting causes flickering I not observed @low preset where this parameter is set to 0 and also combined with downsampling it's nice to have a setting with potentially less performance impact which affects only the most obvious edges (and nice to have for testing in general).

Anyway, after further testing flickering seems to depend almost only on the threshold. So positive diag settings are useful and higher thresholds like 0.15, too.

The FXAA threshold seems to lead to similar results btw.

PS: Is SMAA/FXAA actually separated from the other effects in SweetFX 2.0 and applied afterwards?

Interesting - I must remember to do some testing to see how flicking can be reduced.

Yes SMAA , FXAA and all other effects that require it are now correctly separated into their own pass as needed so they can be applied without conflicting with each other.

The order is :

Cartoon
LumaSharpen
SMAA
FXAA
Bloom
HDR
Explosion
Advanced CRT
PixelArt CRT
+ the rest in a single shared pass, because they don't need to have their own separate pass and so running them in the same pass is faster because every time you do an extra pass there is a slight performance overhead.

I try to list the effects in the settings in the same order they are executed but the order doesn't currently completely match the code, because I've been moving some of the effects. I will try to remember to make sure they match when the final is released.