Quad isn't the best?

While reading some articles over the internet I am deeply starting to thing, that the quad-based LoD algorithm wasn't the best idea. I think so mainly because using such technique on a rather flat surfaces I still will have a lot of unneeded detail.

To demonstrate what I mean, I have found some pictures on the internet.

The picture to the left illustrates something like a hill near the flat ground. The hill itself is more "extruded" from the surface, and it needs more vertexes to represent all it's slopes and caves. It's more logical to add extra detail to the hills, and do not extra detail where the area is completely flat. This is the algorithm I need to implement if I want more complex terrain look more natural. This algorithm is called Optimally Adapted Meshes (OAM).

Unlike previous picture, this one exactly represents the technique I use at the moment. It can be seen with naked eyes that there are way more vertexes used on the areas, where you actually do not need them.

The OAM algorithm, or it's implementation to visualize terrains in realtime (Real-time Optimally Adapted Meshes, or ROAM) uses a height map to determine for each vertex it's error value - the heights around it so see if the area around it is flat or it needs more detail. Based on error value it determines where the extra detail should be added and where it's ok to leave the same amount of vertexes at the moment.

I also need to think in parallel the scaling problem, and implement it inside.

Links

Roam info & papers: 1, 2, 3, 4

 

Glitches

I have realized, that something needs to be changed in my code. I did some tests trying to see how program responses to big numbers. I have created a planet with the size exact to Earth's, and tried to fly here and there to see the performance, and then I ran to some unpleasant glitches.

This is taken from the stationary orbit at 1100 km from the surface of the virtual Earth. Strange thing that this artifacts not always appears (maybe I should call them UFO? :) ). It seems that this happens when some other processes run in background and use CPU resources a lot. I haven't tested it on other computers, maybe it's hardware specific.

I also noticed, that memory allocation and garbage collection isn't quite good, because the project constantly drains memory while runs. The problem is in LoD code, definitely.

Scales again

I've been thinking a lot about scales recently, once again, and those thoughts driving me nuts slowly.

Space is a very very very large thing. Our planet is about 6'300 km in radius, that is 12'600 km in diameter, and that is 12'600'000 meters, 8 digits. Assuming one world unit in the close distance is 1 meter of the planet, I have only 100 more sized object to be able to handle in close distance. So if I will be crazy enough to fly close to the sun (so that is should be rendered in real scale), I would have an float overflow because the radius of the sun is abou 110 times bigger, that the Earth's radius. We can ask why do we need such big numbers to store if we are 10 meters from the surface? Well, even if the part of a surface if 10 meters from you, you still have to take in mind that the center point of an asteroidal body is way far away from you and it is needed to make a proper curvature of a planet. To solve this somehow I have only 3 options in my mind:

1. Store vertex coordinates in doubles and apply exponential space transformations before converting them to floats.
2. Store vertex coordinates in doubles and cull vertexes that are very far away.
3. Make up something else.

And in my opinion, the only reasonable option is the last one..

There are couple of planetary engine demos around the web, I wonder how they solve this problem.

Wierd things

Last night at about 1 a.m. I was trying to put some tests to reveal the difference in calculating floats and doubles. The idea was simple - fix a time when the test stars, put in a loop for 1 million times some equations like multiplying and finding a square root of the number, and then fix the end time so do subtract it from the start time.

In first test floats were used. The number like 123456.78 was multiplied, divided, square rooted and etc. for fixed amount of iterations. And then, using doubles (big numbers from previous post) the same type of test was performed.

I was shocked that rather small float numbers had a higher results than doubles!

This is truly the result that let me think more closely to use doubles instead of floats without any doubts regarding performance.