It's more on the texture distortion or perturbation effect for EMBM, but that is another topic to discuss.
I'd be interested to see the new optimization methods that you have mentioned above, but for overall feedback, I'd say it's normal performance wise considering there's still alpha textures involved.
btw, any new z-sorting code in relic hunter? or same old tried and tested work? 😉 haven't checked that part yet..
All the grass, trees and bushes are masked mode. Only a few things are alpha, like the fake shadows (see latest screenies), but they stick on the ground, so there's is barely any z-fight. Also, those grass islands under a grass mesh, they are masked too and slightly elevated uniquely. To prevent z-fights, they are elevated depending on their x/y position (the fraction after the comma, divided by 20 or so).
So there's no z-sorting by me. I did Z-sort alpha in the past, by repositioning the meshes in the order of distance to the camera, which works, but as resolutions increase and I learn more about good masking, it seems kind of unnecessary.
I had the intel DSS export plugin for photoshop years ago, that allowed to make masked textures that fould fade in as smooth as alpha ones. Now I use paint dot net, still good dss export options, but not as good as the ones o the other one.
Anyway, in case you didn't know, this is the most important thing when working with masked textures, dds, png or tga: duplicate the actual layer, then blur the one below, gaussian 4 or so, then set it's alpha to about 3%. This will help DX to fade the outline to the correct color (otherwise to black)
