Created on 07.10.2005 - 12:01

Guest(r), please, see yours "Private Msgs".
Bye

Created on 10.10.2005 - 17:26

You can try to convert it as ARBfp/vp... I have heard that ARBfp/vp is much faster on Ati cards...
Here a short guide:

rapidshare.de/files/6111181/ARBfpvp.txt.html

Bye

[Dieser Beitrag wurde am 10.10.2005 - 17:26 von ShadX aktualisiert]

Created on 13.10.2005 - 04:21

This looks really nice!

Before - 1280x1024



After - 1280x1024

Created on 16.10.2005 - 01:59

Interesting. I'm gonna toy with those settings see what I can come up with.

You guys make some great shaders. I have virtually every one of them in my shader folder.

Much thanks for all your work.

Created on 17.10.2005 - 06:35

Here's a Front Mission comparison using these values.

float mx = 0.3; // max smoothing factor
float mn = 0.5; // max sharpening factor

Grid View:

No Shader

Shader On

3D Models/Combat View:

No Shader

Shader On

Created on 18.10.2005 - 01:18

I've been messing with the smoothing and sharpness values to see if I could reduce smoothing a bit but retain a non-pixelated look.

Here's what I've got with significantly less smoothing but a high sharpness value. These shots are a value of .05 Smoothing and Sharpness value at .50.


(Lower Smoothing-Grid)
(No Shader-Grid)

(Lower Smoothing-3D Combat)
(No Shader-3D Combat)


And that's Hoshigami not FFT in the shader thread. They look almost the same though.

I'm really happy with these settings I think. It retains a lot, almost all, of the detail and removes almost all of the pixelation.

Created on 18.10.2005 - 10:47

Very good job, Guest(r).
Only I don't understand why you use mx, mn and k...
I think is better to write:
float mx = 0.3; // max smoothing factor
//float mn = 0.3; // max sharpening factor
//float k = -(mn+mx)/3.0;
float k = -0.2;// -(0.3+0.3)/3.0;
You save one ADD, one RCP and one MUL...
Bye

Created on 18.10.2005 - 12:37

Ok, Guest, but if you can, is better if you don't use divisions. Every division stall the shader. If you use:
float k = -(mn+mx)*0.333;
you have nearly the same result without stall.
Bye