• Advertisement
  • entries
  • comments
  • views

D3D10 Stuff

Sign in to follow this  


Pulp Fiction
I don't need you to tell me how f*****g good my coffee is, okay? I'm the one who buys it. I know how good it is. When Bonnie goes shopping she buys S**T. Me, I buy the gourmet expensive stuff because when I drink it I want to taste it.
So it's a little after 20 past midnight on a Caturday evening (or is it now Bunday morning?) and I'm enjoying some cheese on toast. Not just generic cheap cheese of course - Sainsburys "Taste The Difference" Isle of Bute Cheddar no less. Being a working lad I'm buying the expensive products because I appreciate the taste - this cheese tastes so much better than the cheap own-brand rubbish. The coffee I buy tastes so much better than brand-name instant-filth as well.

So, why discuss such things?

You sit and watch the reference rasterizer for more than a few frames and you'll quite quickly get the urge to grab a snack or a drink. I wonder if the RefRast contributes to obesity.

Anyway, I think I've cracked relief mapping. The results I'm getting seem reasonable if a little different to the research paper. The code I've got is working with inverse heights yet generates correct results - if I flip them around to match the research paper I get garbage. Whether this means I've got a mistake elsewhere (possible)) or the paper is missing something I don't know...

Just for novelty value I've got the following animation for you:

4.21 seconds per frame for regular RM
4.02 seconds per frame for displaying sample count

The first half of the animation shows the intended effect but from a developers point of view the second half is more interesting. I've written my code to early-out where possible, so the brighter a pixel the more samples required to compute the final colour (white pixels require upto 38 samples in this situation); an ideal situation would be as dark as possible - indicating as minimal texture bandwidth as possible...

The following were rendered at 1024x768 but resized for the web:

59.75 seconds per frame

74.98 seconds per frame

Points for the first person to comment with what the text actually says [wink]
Sign in to follow this  


Recommended Comments

Shouldn't the parts that stick up the highest, like the top of those bricks, be the first ones to early out? If you are casting a ray toward the relief mapped surface from the eye point, whatever is closer to it should be detected the most quickly - or am I misinterpreting how you implemented the algorithm?

Share this comment

Link to comment

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now

  • Advertisement