Two new demos on my site

Humus, you are on fire! ATI should just point developer.ati.com to your site, since at the current rate, your site will have more samples soon!
 
Very cool!

Your demos are getting a lot more creative now, as opposed to just plain DP3 or EMBM. I haven't tried it out, but wouldn't your monochrome method create a lot of shimmering if the light was moving? Sorry if I'm nitpicking...

EDIT:
Just tried it, and the shimmering is a lot. It would be neat if you could implement it so that as an area got brighter, instead of drawing a completely new pattern, just make it so that more white dots appeared. Now that I think about it, this is a lot like Tom's method, isn't it. I guess all he needs is a better random pattern over the screen.
 
Why aren't you clearing the back buffer? Your scene doesn't cover the whole screen so I get to see the contents of the previous frames :-?
 
Support nVidia cards too. :)
Ahh hell I hope my 9700 arrives soon, don't wanna go back to my 8500 just to see these demos. ;)
 
DemoCoder said:
Humus, you are on fire! ATI should just point developer.ati.com to your site, since at the current rate, your site will have more samples soon!
Click to expand...

Heh :LOL:
Just got a lot of time these days.
 
Mintmaster said:
Very cool!

Your demos are getting a lot more creative now, as opposed to just plain DP3 or EMBM. I haven't tried it out, but wouldn't your monochrome method create a lot of shimmering if the light was moving? Sorry if I'm nitpicking...

EDIT:
Just tried it, and the shimmering is a lot. It would be neat if you could implement it so that as an area got brighter, instead of drawing a completely new pattern, just make it so that more white dots appeared. Now that I think about it, this is a lot like Tom's method, isn't it. I guess all he needs is a better random pattern over the screen.
Click to expand...

Sure it shimmering a lot. Doing something like what you're proposing would of course look better, but it will be extremely hard to implement, I'm not sure if it's doable at all with the amount of math available in a fragment shader.
While the basic idea is the same as Tom uses there's a key difference between my approach and his. With his method the final black or white color is choosen based on color and position. In my method it's only based on color with the mapping function done such that a bright pixel is more likely to turn white while a dark is more likely to turn black.
 
fresh said:
Why aren't you clearing the back buffer? Your scene doesn't cover the whole screen so I get to see the contents of the previous frames :-?
Click to expand...

It does cover the whole screen. It's one fat quad over the whole screen.
 
Ante P said:
Support nVidia cards too. :)
Click to expand...

The dither demo would be possible, but I'm not so sure about the laplace demo. It uses five simultaneous textures and requires a range of [-4, 4]. Also, the math is not so straightforward to break up into multipass.
 
Humus:
Try a screen-sized texture with white noice as the dithering offset.

Oh, forgot to say:
Nice demos. The screenshots at least, since I can't run them.
 
They run pretty fast both on my RadeOn 8500.

215-220 for dither (standard res)
180 laplace (standard res)

Humus just stop releasing such cute demos, that makes me angry, my engine can't render that atm. :p

(well it can't render anything atm, since I'm coding an abstract layer for API indepency)
 
800! That's quite impressive. I get 150 fps in laplace and 200 in dither.

Basic said:
Humus:
Try a screen-sized texture with white noice as the dithering offset.
Click to expand...

That's a good idea. Would solve the problem of shimmering too I guess. Will experiment a little with it.
 
Basic, you're genious :) Tried it, and it works extremely well! Very little shimmering, but still the full effect, and no dependent texture read neccesary.
I'll upload a new version when my roommate puts the server back online again ... not sure what my roommate have done now, he's playing around with another server in the switchroom right now ...
This new version should also work for GF3/4.
 
:oops:
Always feel good to hear such things, even if it is somewhat exaggerated. But feel free to say that to the ATI folks when everything's sorted out and you're over there. :D

I think it's possible to make some interesting effects with that offset texture. If the "whiteness" of the noice is changed over the texture (ie completely random in the center, and gradually get grouped into high- and low-value groups to the sides), you'd get an effect where some parts are nicely dithered, and others get a rough raster.
Could be used as an "artistic" way to put emphasis on part of the screen.

I'm not sure how to make such variable dither texture. But maybe it could be done by putting white noice in one channel, and a very clearly grouped dither pattern in another, and then use a second (stretched low res) texture to give an interpolation factor. Painting in the "interpolation factor texture" would move the "emphasis" over the image.

- Just some brainstorming.
 
Demos

What can I say but, wow! Amazing stuff Humus.

The 7.77s now render the shadows perfectly.

One, probably numb-nuts question, but could the laplace technique be used as the basis for a Parhelia style FAA?
 
You must log in or register to reply here.

Similar threads

G
I found a folder with a bunch of old scene demos on my harddrive...
Replies
13
Views
4K
Kurt
K
A
Variance Shadow Maps Demo (D3D10)
2 3 4
Replies
66
Views
111K
homerdog
homerdog
H
Something to bring your Radeon 9700 to it's knees.
2
Replies
30
Views
5K
Humus
H
H
Shadow mapping for point lights demo
2
Replies
29
Views
4K
Humus
H
H
My site is up again and with a new volumetric fogging demo
2 3
Replies
48
Views
5K
Humus
H
Back
Top