Sign in to follow this  
the dodger uk

sorting array of float3s from camera

Recommended Posts

Sort them based on the squared distance from the camera (i.e. magnitude squared of  positions[n] - cameraPosition). Good sort candidates would be radix sort or just a std::sort.

 

But aren't you going to take into account the direction the camera is looking? Then you need to dot product the camera direction with (positions[n] - cameraPosition) and reject any that are behind the camera (dot product negative).

 

If you can get the points in clip space before that you may be able to just sort on the Z (or whatever into the screen is) value in clip space but you can't do that if you transform into clip space in a vertex shader.

 

EDIT: Wrong way round for the vector subtractions, line from A to B is B - A ;)

Edited by Paradigm Shifter

Share this post


Link to post
Share on other sites

You need to write a functor (or a lambda function in C++ 11) to provide the sort comparison function for the elements.

 

I've had too much cheap Lidl beer to explain it fully. I'm thinking a functor which you initialise with the camera position (and direction if required, I think it will be) so it can store those as state in the function object.

 

Something like this

struct SortVec3Helper
{
    SortVec3Helper(const vec3& camPos) 
       : m_camPos(camPos)   
    {}

    bool operator() (const vec3& lhs, const vec3& rhs)
    {
         return /* squared distance from cam to point lhs is less than squared distance to rhs */
    }

    vec3 m_camPos;
};

And pass one of those objects as the comparison function object to std::sort (3rd argument, I think).

 

Read up on functors if you aren't sure how to use them. I'm sure you can do it with lambdas as well but I'm not a C++ 11 guru.

Edited by Paradigm Shifter

Share this post


Link to post
Share on other sites

I'm sure you can do it with lambdas as well but I'm not a C++ 11 guru.

It would be something like
float3 pos[200];
Initialize(pos, 200);

std::sort(pos + 0, pos + 200, [](const float3 &o, const float3 &i) { return dot(o, o) < dot(i, i); });

 

Note: for most efficient vector-oriented float3 implementations, you must use pass-by-reference. Pass-by-copy would likely violate alignment rules. This is the case for Bullet vector class.

Share this post


Link to post
Share on other sites

Well as long as the compiler respects any alignment requirements for pass by value.

 

Probably a better way is to pack your positions into a float4 and use the spare w component for your sort criteria, then sort the array based on the w value.

 

Did you understand what I was saying about the functor to do the sort? Notice my functor had state (i.e. the camera position) but if you use the w component for precalculated distance from camera you don't need the state and can sort with a normal function for the comparison func if you like.

 

EDIT: Precalculating the distance from the camera as a prepass is going to be more efficient as well since you wont end up recalculating the distance for every pairwise comparison the sort does, so that has to be the best plan.

Edited by Paradigm Shifter

Share this post


Link to post
Share on other sites

I feel there is some information missing in that question. Without knowing why he wants those vectors to be sorted people wanting to help cant know if there is some way around sorting them. For example if its just for drawing back to front to have it look right, then not sorting them and activating the depth buffer would be the better solution.

Share this post


Link to post
Share on other sites

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

Sign in to follow this