Jump to content

  • Log In with Google      Sign In   
  • Create Account


Member Since 10 Oct 2005
Offline Last Active Today, 03:22 PM

#5236402 manging lifetime of textures/buffers etc

Posted by on 23 June 2015 - 12:48 PM

I would have to somehow know that nothing is using a particual Texture any more and manually call mDevice->DeleteTexture(textureHandle).
So do people implement some form of reference counting inside the Device to count CreateTexure and DeleteTexture on each texture handle? It seems if I'm going to do that I might as well just use a shared_ptr anyway to count it?

Many engines manage memory not in a totally dynamic way. Instead they define memory budgets for particular purposes and they define lifetimes a priory. For example, some assets are loaded at game start and unloaded when the game ends, i.e. they have game lifetime. Level based or region based lifetime are other popular methods. Streaming may also be used in which case buckets of memory are allocated once and then cleared and re-used in dependence on vicinity or whatever. However, the clue is that the lifetime is not measured but prescribed.


I do not know what exactly the Molecule Engine does, but the articles and comments in that blog mention such things, too.

#5236273 Landmarks

Posted by on 23 June 2015 - 01:10 AM

Castle Settings


* awesome tapestry

* torture chamber


Cavern Settings


* abyss

* subsurface flow (short piece of)

* opening in the ceiling where sun / moon light comes in

* lair (e.g. of bats)

* occurrences of natural resources in the walls


Sewer Settings


* entrances / exits

* catchment basin 

* some kind of machinery (pumping, forwarding)

* any bigger room as opposed to the narrowness of the regular sewers

* (e.g.) rat nest


#5235824 Finding near and far plane from projection matrix

Posted by on 20 June 2015 - 03:58 AM

f - 5 = f / 1.02564

-5 = f / 1.02564 - f

    = f ( 1 / 1.02564 - 1 )


f = -5 / ( 1 / 1.02564 - 1 )

#5233555 Moving objects in 3D

Posted by on 08 June 2015 - 10:57 AM

I've looked into it a bit more, and I might be onto something, but I'm not sure, and I'd like some advice please.  I found a function called Transformation, which takes as inputs:
- a translation vector
- a vector for the rotation point, and a quaternion to rotate
- a vector for the point to scale around, a vector for the amount of scaling, and a quaternion for rotating the axes along which to scale (so it doesn't necessarily scale on the default axes)

This sounds like the transform function somewhat known from X3D. There are also the functions named AffineTransform with less parameters. Both can be used. It is probably so that for your use case

* the scalingCenter is (0,0,0)

* the scalingRotation is (1,0,0,0)

* the scaling is (1,1,1)

* the rotationCenter is (0,0,0)

* the rotation is (your-rotation)

* the translation is (your-translation)


However: It would be easier to just set the elements of the transform matrix from the parameters you already have. Assuming that SlimDX uses row vectors, the 1st row stores the sideways vector, the 2nd row stores the upward vector, the 3rd row stores the forward vector, and the 4th row stores the position vector. (Notice please that the rows stores homogeneous coordinate vectors, so that each of the direction vectors has a 0 as its last element, while the position vector has a 1.)


So, if you have pairwise orthogonal vectors right, up, and forward then you can use the setters Matrix.Mxx for that purpose. If they are not pairwise orthogonal then you need to apply the cross product before using the vectors.

#5232353 OpenGL nearPlane, farPlane and virtualScreen

Posted by on 02 June 2015 - 05:18 AM

Well, I'm not totally sure, but when the illustration from the OP is the basis, then the left and right parameters of frustum(….) must not be symmetrical. They probably would be symmetrical if the look-at point is chosen so 

glm::lookAt(Position - eyeOffset, Position + Forward, Up); // left eye
glm::lookAt(Position + eyeOffset, Position + Forward, Up); // right eye

(assuming that eyeOffset is a vector given in the correct co-ordinate space, of course). BUT then the view planes for the left eye and for the right eye would not be coincident but would be crossed a bit. Hence they would not form the shown virtual screen. So ...


Exerting trigonometry, the virtual screen has a width of

     s := 2 * convergence * aspect * tan( 0.5 * fovy )


Let e be the absolute value of the eye offset. Then the irregular parallelogram build of Position, left eye, top intersection of left and right frustum, and the center of the virtual screen allows us to say that for the left eye

   left := -1 * near * ( s/2 - e ) / convergence

   right := +1 * near * ( s/2 + e ) / convergence

and analogously for the right eye

   left := -1 * near * ( s/2 + e ) / convergence

   right := +1 * near * ( s/2 - e ) / convergence

Well, I think at least it could be the solution. What do you mean?

#5230799 What design pattern should I go for if I want to "sort of" convert th...

Posted by on 25 May 2015 - 03:38 AM

The OP doesn't contain exhaustive informations, so to say. In principle I would / do arguing as follows:


The navigation system is a service provider for AI (can I go there?) and the movement system (how to go there?), i.e. it is at lower level than those, and those are at a lower level than game objects (as a whole). When different navigation solutions are available and probably even need different data structures, then integrating all that into game objects would be messy. A possible solution is to use an agent attached to the game object, so that the agent utilizes the services, holds the appropriate data structures, and mediates between them and the game object.

#5230414 Highlight 2d polygon border with shader

Posted by on 22 May 2015 - 08:04 AM

There are (at least) two possible approaches:


a) The original polygon is the dark area inside the green border, and the border is additional geometry build from extending the original polygon. The additional geometry can then be colored by standard interpolation of vertex color.


b) The polygon area reaches to the outside where the green border fades out into the background, and no additional geometry is generated. Then the vertex shader is used to compute the distance of the current pixel to the edges in screen space, and the color of the pixel is chosen as function of the distance. Obviously, you need to send a vertex attribute which is used to distinguish whether or not a vertex counts to an edge that is to be highlighted. The theory can be read in several articles around the web.

#5229297 Another basic array C problem

Posted by on 16 May 2015 - 07:03 AM

I am self studying this 

Okay, then you not should but want to solve the problem by yourself. :) So what is your concrete question regarding the issues I've already noted?


Debugger in dev cpp is not working on my Windows 8.1 64 bit.

Then you shall think to turn to another developer environment. You definitely will need a debugger. The given situation is only the first of many others that will come ...

#5229291 Another basic array C problem

Posted by on 16 May 2015 - 04:20 AM

The forum rules do not allow us to solve other peoples homework, but it allows to answer concrete questions where we can see from that you've tried hard to get a solution yourself.


Well, the program code shown in the OP has at least 10 mistakes, some of which are just flaws, some are errors, and two of them are logical mistakes regarding how a solution may work, inclusive how the correct solution may look like at all. Not wanting to discourage you, but you should restart at the problem analysis: (1) What is the goal, (2) how can it be reached with a 2D array, and (3) how can that be programmed.


As wintertime has excellently stated above, the debugger is a valuable tool. It allows you to peek into the working of computer programs, so lets you examine how statement how, how the resulting program flow is, and how variables change. Use it to your advantage!


I wanted to sort that array first and then find the lowest and the highest array.

This is possible and fine by itself, of course. But ask yourself what you get by doing so. You get the lowest and the highest temperature. The goal accordingly to (1) is to output the day and the city of two specific temperatures. With your thinking considering (2), the indices of the array are used to denote the specific days and cities, resp. You associate the belonging temperature with a specific day and city by writing the temperature value into the array place that is addressed by that indices; that is fine so far. Now, what happens to that association if you sort the array as is shown in the OP?

#5229285 Ray-cylinder intersection (and others)

Posted by on 16 May 2015 - 02:39 AM

In the days back when I programmed a ray tracer (we wrote the year 2000 as I've seen in my documents), I had used a method similar to what Randy Gaul has mentioned above: Intersection computations started with transforming the problem into a normalized local object space. This puts down intersections of translated / rotated / scaled objects to their respective standard form. The transformation is applied just to the origin and direction of the parametric ray. It would appear in a full intersection formula anyway, so does not enhance the computational costs.


In case of a cylinder the standard form, as was used in my ray tracer, meant: The axis of the cylinder is is the local y axis, the basic area is circular, the height is symmetrical to the x/z plane. In this case the cylinder could be hollow if either or both of the ceiling and floor planes are omitted. The belonging code is this (please don't look too close to the coding style):

Bool RegCylinderShape::check(const Ray& aRay,SpanList& Spans) const {
   Bool HitExists = false;
   // localize ray
   // compute parameters
   register Real64 T1 = _localRay.Origin.x/_Rx;
   register Real64 T2 = _localRay.Origin.z/_Rz;
   register Real64 T3 = _localRay.Direction.x/_Rx;
   register Real64 T4 = _localRay.Direction.z/_Rz;
   register Real64 C0 = T1*T1+T2*T2-1.0;
   register Real64 C1 = T1*T3+T2*T4;
   register Real64 C2 = T3*T3+T4*T4;
   C0 = C1*C1-C0*C2;
   // if C0 is greater than zero, the ray hits the infinitely long cylinder
   if(C0>0.0) {
      register Bool B1p,B1n,B2p,B2n;
      // compute the travel coordinates of the near and far hits
      C0 = sqrt(C0);
      T1 = (-C1-C0)/C2;
      T2 = (-C1+C0)/C2;
      // compute y components of both hits in local coordinates
      C1 = T1*_localRay.Direction.y+_localRay.Origin.y;
      C2 = T2*_localRay.Direction.y+_localRay.Origin.y;
      // check whether or not the ray enters/leaves through ceiling/floor planes
      B1p = C1<=_halfHeight;
      B2p = C2<=_halfHeight;
      B1n = C1>=-_halfHeight;
      B2n = C2>=-_halfHeight;
      // if the ray hits the finite cylinder shape ...
      if((B1p||B2p)&&(B1n||B2n)) {
         _HitCode = 0x0;
         // if the ray enters through the ceiling or else the floor plane, the corresponding travel
         // coordinate is adapted accordingly; necessity of adapting is reported in setting the hit
         // code bit #0
         if(!B1p) {
            _HitCode |= 0x1;
            T1 = (_halfHeight-_localRay.Origin.y)/_localRay.Direction.y;
         } else if(!B1n) {
            _HitCode |= 0x1;
            T1 = (-_halfHeight-_localRay.Origin.y)/_localRay.Direction.y;
         // if the ray leaves through the ceiling or else the floor plane, the corresponding travel
         // coordinate is adapted accordingly; necessity of adapting is reported in setting the hit
         // code bit #1
         if(!B2p) {
            _HitCode |= 0x2;
            T2 = (_halfHeight-_localRay.Origin.y)/_localRay.Direction.y;
         } else if(!B2n) {
            _HitCode |= 0x2;
            T2 = (-_halfHeight-_localRay.Origin.y)/_localRay.Direction.y;
         // if the ceiling or floor plane exists, or the ray hits the cylinder outside these planes
         // (either when entering or when leaving), then real hits occur which must be reported to
         // the span list
         if((_HitCode!=0x3)||_Ceiling||_Floor) {
            HitExists = true;
            // if either the ceiling or the floor plane lacks existance, but the ray hits those
            // plane, the corresponding travel coordinate and hit code must be adapted
            if(!((B1p||_Ceiling)&&(B1n||_Floor))) {
               T1 = T2;
               _HitCode &= 0x2;
               _HitCode |= _HitCode>>1;
            } else if(!((B2p||_Ceiling)&&(B2n||_Floor))) {
               T2 = T1;
               _HitCode &= 0x1;
               _HitCode |= _HitCode<<1;
            _HitSeparation = 0.5*(T1+T2);
            // report hits to span list
   return HitExists;

Some explanations: The result of an intersection computation is a SpanList. Here a span is the distance from the ray's origin along its direction until the ray enters and then leaves a volume. If the object is hollow, then the span's enter and leave distances are the same. (This concept was chosen to allow even for multiple intersection as might occur for e.g. a torus.) The SpanList then is an ordered list of spans.


The distances along the ray are stored as multiples of the ray's direction vector. This allows to apply the distances in any space, since the ray's direction vector is not re-normalized after transforming into a local space. So the actual intersection points in e.g. world space are calculated by setting in the distances as the world space's ray parameter.

#5229092 Sequence animation by codes or using .gif image

Posted by on 15 May 2015 - 01:34 AM

Since you asked for opinions: GIF as in-game format is a no-go for me. I mean … it is GIF, isn't it?


From a more technical point of view, you need to have an animation system anyway. It should work considering delta frame times which is not included in GIF animations (AFAIK). Having the own code controlling things gives the best flexibility. So, why not using this same mechanism for all instead of introducing a second mechanism for the gain of, well, nothing in the end.

#5228739 Intersection between circle and line segment

Posted by on 13 May 2015 - 04:27 AM

But in this scenario it doesn't work:

Problem was that condition check: In case that (tM && tA) is true, then also (tM || tA) is true. Hence the then branch is entered and within it the checks exclude (tM && tA) from being a result. The correct code is hence (jsfiddle is also updated if you want to look at it):

                       if(tM || tA){
                            if(tM && !tA){
                                this.intersectPoints = [
                                    new vector(f0,f1)
                            if(tA && !tM){
                                this.intersectPoints = [
                                    new vector(g0,g1)
                            if(tM && tA){
                                this.intersectPoints = [
                                    new vector(f0,f1),
                                    new vector(g0,g1)
                            return true;

#5228730 Intersection between circle and line segment

Posted by on 13 May 2015 - 03:22 AM

Well, it is cumbersome to locate the problem without any hints, just by looking at code. Please construct a simple situation like a segment (0,0) to (10,0) and a circle at (5,0) with radius 1, and write out the resulting values of eDistAtoB (should be 10) and t (should be 5). Further dt should be 1, and hence tMdt == 4 and tAdt == 6. Obviously, this should result in true for both tM and tA. Where does this break, and with which values?

#5228726 Intersection between circle and line segment

Posted by on 13 May 2015 - 03:01 AM

Oh, I've not considered that vector d is normalized. So t (or t-dt or t+dt) need not range from 0 to 1 but from 0 to eDistAtoB … I think at least ...


So try

                       var tM = tMdt >= 0 && tMdt <= eDistAtoB, tA = tAdt >= 0 && tAdt <= eDistAtoB;

instead of

                       var tM = tMdt >= 0 && tMdt <= 1, tA = tAdt >= 0 && tAdt <= 1;
and furthermore
                       if(t >= 0 && t <= eDistAtoB){
instead of
                       if(t >= 0 && t <= 1){

#5228719 Intersection between circle and line segment

Posted by on 13 May 2015 - 01:59 AM

a) In case of 2 intersections: You need to check both intersections not together but on their own. Also there is an error with the condition on tAdt!


b) In case of 1 intersection: You should use t >= 0 && t <= 1 instead of t > 0 && t < 1