Voxel Traversal Algorithm (Ray Casting)

Published August 02, 2018
Advertisement

For more information and updates about the game, which is a voxel colony simulation / survival, please subscribe to r/CheesyGames.

World Generation

image.thumb.png.17fa74b357314a335ce832b6bba44677.png

This is a simple world made of chunks of 32³ voxels. The world itself is not static : as you can see on the top of the image, chunks only exist if there is at least one solid voxel in them. In other words, the world is dynamic and can contain as many chunks as the player's computer can handle.

In this particular screenshot, the world is generated with the simple vectorial gradient equation that I invented in high school but which I suppose already existed. Here's the basic equation :

\[\overrightarrow{ voxel \: value } = \frac{ \overrightarrow{ position } \cdot \overrightarrow{ gradient } }{ \overrightarrow{ gradient } \cdot \overrightarrow{ gradient} }\]

Preview of LaTeX equation :

That's the equation I came up with and remembered. The gradient * gradient can be simplified for the magnitude (length) of the gradient power squared.

\[\overrightarrow{ voxel \: value } = \frac{ \overrightarrow{ position } \cdot \overrightarrow{ gradient } }{ \left \| \overrightarrow{ gradient } \right \| ^{2} }\]

Preview of LaTeX equation :

In conclusion, this gives an N dimensional gradient which gives a single decimal number.

Voxel Traversal Algorithm

As for the voxel traversal algorithm, I decided to go with the most popular one, which was made by John Amanatides and Andrew Woo. As much as I like research papers, I also despise them because they lack simplicity, examples and full source code. That's why I had to google implementations of it and later on remembered that I had actually already implemented this algorithm a few years ago.

Summary

The simplest way to understand the algorithm is to imagine a line in an 3D world made of blocks. Which blocks does the line touch? Then, in which order are they touched based on the line's start and end positions? The goal is to traverse iteratively the blocks that are touched by the line .

More simply, the logic of the algorithm can be summed making a distinction between the ray's direction's components. Those three define the importance of their axes in terms of how many blocks need to be traversed in what direction. Think of this with integers : two people are running to reach a goal; the fastest runs a 5 km/h, while the slowest runs at 1 km/h. For each time step, i.e. an hour, how many kilometers have each runner traveled? The ratio is 5 : 1, so, to merge the analogy, a ray would traverse each step 5 blocks on the X axis and 1 block on the Y axis. Of course, it's more complicated than that, as there are more parameters to it, especially because of exceptions such as what to do when each component is equal with one another?

image.png.795262d06175e7eef086822b9a19cb46.png

Implementation

The first thing to know about my implementation is that each voxel index is centered within the voxel itself. In other words, the voxel at the position (0, 0, 0) starts at (-0.5, -0.5, -0.5) inclusively and ends at (0.5, 0.5, 0.5) exclusively. This is for a cube extent of 1, naturally. The original research paper doesn't work that way as it starts at the lowest corner, i.e. the voxel spans from (0, 0, 0) to (1, 1, 1). Without any further delay, here is the class for the VoxelRay :

package com.cheesygames.colonysimulation.math.bounding.ray;

import com.cheesygames.colonysimulation.math.MathExt;
import com.cheesygames.colonysimulation.math.bounding.VoxelWorldUtils;
import com.cheesygames.colonysimulation.math.vector.Vector3i;
import com.cheesygames.colonysimulation.world.World;
import com.jme3.math.Vector3f;
import com.jme3.scene.plugins.blender.math.Vector3d;

/**
 * Ray for ray casting inside a voxel world. Each voxel is considered as a cube within this ray. A ray consists of a starting position, a direction and a length. The voxel distance
 * is computed once the method {@link #rayCastLocal(double, VoxelRayOnTraversing, Vector3i)} or {@link #rayCast(double, VoxelRayOnTraversing)} is called.
 */
public class VoxelRay {

    private Vector3d m_start;
    private Vector3d m_offsettedStart;
    private Vector3d m_direction;
    private double m_length;
    private int m_voxelDistance;
    private boolean m_wasStopped;

    /**
     * Constructs an invalid {@link VoxelRay} as its direction and length are null. The setters must be called after constructing a {@link VoxelRay} with this constructors.
     */
    public VoxelRay() {
        this.m_start = new Vector3d();
        this.m_offsettedStart = new Vector3d();
        this.m_direction = new Vector3d();
        this.m_length = 0;
    }

    /**
     * Constructs a {@link VoxelRay} from two points : start and end.
     *
     * @param start The absolute starting position of the ray.
     * @param end   The absolute ending position of the ray.
     */
    public VoxelRay(Vector3d start, Vector3d end) {
        this.m_start = new Vector3d(start);
        this.m_offsettedStart = new Vector3d();
        this.m_direction = end.subtract(start);
        this.m_length = m_direction.length();
        this.m_direction.normalizeLocal();
    }

    /**
     * Constructs a {@link VoxelRay} from two points : start and end.
     *
     * @param start The absolute starting position of the ray.
     * @param end   The absolute ending position of the ray.
     */
    public VoxelRay(Vector3f start, Vector3f end) {
        this.m_start = new Vector3d(start);
        this.m_offsettedStart = new Vector3d();
        this.m_direction = new Vector3d(end).subtractLocal(m_start);
        this.m_length = m_direction.length();
        this.m_direction.normalizeLocal();
    }

    /**
     * Constructs a {@link VoxelRay} from a start, a direction and a length.
     *
     * @param start     The absolute starting position of the ray.
     * @param direction The direction of the ray. Must be normalized.
     * @param length    The length of the ray.
     */
    public VoxelRay(Vector3d start, Vector3d direction, double length) {
        this.m_start = new Vector3d(start);
        this.m_offsettedStart = new Vector3d();
        this.m_direction = new Vector3d(direction);
        this.m_length = length;
    }

    /**
     * Constructs a {@link VoxelRay} from a start, a direction and a length.
     *
     * @param start     The absolute starting position of the ray.
     * @param direction The direction of the ray. Must be normalized.
     * @param length    The length of the ray.
     */
    public VoxelRay(Vector3f start, Vector3f direction, float length) {
        this.m_start = new Vector3d(start);
        this.m_offsettedStart = new Vector3d();
        this.m_direction = new Vector3d(direction);
        this.m_length = length;
    }

    /**
     * Casts the ray from its starting position towards its direction whilst keeping in mind its length. A lambda parameter is supplied and called each time a voxel is traversed.
     * This allows the lambda to stop anytime the algorithm to continue its loop.
     *
     * @param onTraversingVoxel The operation to execute when traversing a voxel. This method called the same number of times as the value of {@link #getVoxelDistance()}. The
     *                          supplied {@link Vector3i} parameter is not a new instance but a local instance, so it is a reference. The return value {@link Boolean} defines if
     *                          the algorithm should stop.
     *
     * @see <a href="http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.42.3443&rep=rep1&type=pdf">A Fast Voxel Traversal Algorithm</a>
     */
    public void rayCast(VoxelRayOnTraversing onTraversingVoxel) {
        rayCastLocal(World.VOXEL_HALF_EXTENT, onTraversingVoxel, new Vector3i());
    }

    /**
     * Casts the ray from its starting position towards its direction whilst keeping in mind its length. A lambda parameter is supplied and called each time a voxel is traversed.
     * This allows the lambda to stop anytime the algorithm to continue its loop.
     *
     * @param voxelHalfExtent   The half extent (radius) of a voxel.
     * @param onTraversingVoxel The operation to execute when traversing a voxel. This method called the same number of times as the value of {@link #getVoxelDistance()}. The
     *                          supplied {@link Vector3i} parameter is not a new instance but a local instance, so it is a reference. The return value {@link Boolean} defines if
     *                          the algorithm should stop.
     *
     * @see <a href="http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.42.3443&rep=rep1&type=pdf">A Fast Voxel Traversal Algorithm</a>
     */
    public void rayCast(double voxelHalfExtent, VoxelRayOnTraversing onTraversingVoxel) {
        rayCastLocal(voxelHalfExtent, onTraversingVoxel, new Vector3i());
    }

    /**
     * Casts the ray from its starting position towards its direction whilst keeping in mind its length. A lambda parameter is supplied and called each time a voxel is traversed.
     * This allows the lambda to stop anytime the algorithm to continue its loop.
     * <p>
     * This method is local because the parameter voxelIndex is locally changed to avoid creating a new instance of {@link Vector3i}.
     *
     * @param onTraversingVoxel The operation to execute when traversing a voxel. This method called the same number of times as the value of {@link #getVoxelDistance()}. The
     *                          supplied {@link Vector3i} parameter is not a new instance but a local instance, so it is a reference. The return value {@link Boolean} defines if
     *                          the algorithm should stop.
     * @param voxelIndex        The voxel index to locally modify in order to traverse voxels. This parameter exists simply to avoid creating a new {@link Vector3i} instance.
     *
     * @see <a href="http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.42.3443&rep=rep1&type=pdf">A Fast Voxel Traversal Algorithm</a>
     */
    public void rayCastLocal(VoxelRayOnTraversing onTraversingVoxel, Vector3i voxelIndex) {
        rayCastLocal(World.VOXEL_HALF_EXTENT, onTraversingVoxel, voxelIndex);
    }

    /**
     * Casts the ray from its starting position towards its direction whilst keeping in mind its length. A lambda parameter is supplied and called each time a voxel is traversed.
     * This allows the lambda to stop anytime the algorithm to continue its loop.
     * <p>
     * This method is local because the parameter voxelIndex is locally changed to avoid creating a new instance of {@link Vector3i}.
     *
     * @param voxelHalfExtent   The half extent (radius) of a voxel.
     * @param onTraversingVoxel The operation to execute when traversing a voxel. This method called the same number of times as the value of {@link #getVoxelDistance()}. The
     *                          supplied {@link Vector3i} parameter is not a new instance but a local instance, so it is a reference. The return value {@link Boolean} defines if
     *                          the algorithm should stop.
     * @param voxelIndex        The voxel index to locally modify in order to traverse voxels. This parameter exists simply to avoid creating a new {@link Vector3i} instance.
     *
     * @see <a href="http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.42.3443&rep=rep1&type=pdf">A Fast Voxel Traversal Algorithm</a>
     */
    public void rayCastLocal(double voxelHalfExtent, VoxelRayOnTraversing onTraversingVoxel, Vector3i voxelIndex) {
        assert !Double.isNaN(voxelHalfExtent);

        assert !Double.isNaN(m_start.x);
        assert !Double.isNaN(m_start.y);
        assert !Double.isNaN(m_start.z);

        assert !Double.isNaN(m_direction.x);
        assert !Double.isNaN(m_direction.y);
        assert !Double.isNaN(m_direction.z);

        assert !Double.isNaN(m_length);

        m_wasStopped = false;
        final double voxelExtent = voxelHalfExtent * 2;

        // This id of the first/current voxel hit by the ray.
        m_offsettedStart.set(m_start).addLocal(voxelHalfExtent, voxelHalfExtent, voxelHalfExtent);
        VoxelWorldUtils.getVoxelIndexNoOffsetLocal(voxelExtent, m_offsettedStart, voxelIndex);

        computeVoxelDistance(voxelExtent, voxelIndex);
        assert !Double.isNaN(m_voxelDistance);

        // In which direction the voxel ids are incremented.
        int stepX = (int) MathExt.getSignZeroPositive(m_direction.x);
        int stepY = (int) MathExt.getSignZeroPositive(m_direction.y);
        int stepZ = (int) MathExt.getSignZeroPositive(m_direction.z);

        // Distance along the ray to the next voxel border from the current position (tMaxX, tMaxY, tMaxZ).
        double nextVoxelBoundaryX = (voxelIndex.x + (MathExt.getNegativeSign(stepX) + 1)) * voxelExtent;
        double nextVoxelBoundaryY = (voxelIndex.y + (MathExt.getNegativeSign(stepY) + 1)) * voxelExtent;
        double nextVoxelBoundaryZ = (voxelIndex.z + (MathExt.getNegativeSign(stepZ) + 1)) * voxelExtent;

        // tMaxX, tMaxY, tMaxZ -- distance until next intersection with voxel-border
        // the value of t at which the ray crosses the first vertical voxel boundary
        double tMaxX = (m_direction.x != 0) ? (nextVoxelBoundaryX - m_offsettedStart.x) / m_direction.x : Double.MAX_VALUE;
        double tMaxY = (m_direction.y != 0) ? (nextVoxelBoundaryY - m_offsettedStart.y) / m_direction.y : Double.MAX_VALUE;
        double tMaxZ = (m_direction.z != 0) ? (nextVoxelBoundaryZ - m_offsettedStart.z) / m_direction.z : Double.MAX_VALUE;

        // tDeltaX, tDeltaY, tDeltaZ --
        // how far along the ray we must move for the horizontal component to equal the width of a voxel
        // the direction in which we traverse the grid
        // can only be FLT_MAX if we never go in that direction
        double tDeltaX = (m_direction.x != 0) ? stepX * voxelExtent / m_direction.x : Double.MAX_VALUE;
        double tDeltaY = (m_direction.y != 0) ? stepY * voxelExtent / m_direction.y : Double.MAX_VALUE;
        double tDeltaZ = (m_direction.z != 0) ? stepZ * voxelExtent / m_direction.z : Double.MAX_VALUE;

        onTraversingVoxel.startRayCast();

        if (onTraversingVoxel.apply(voxelIndex)) {
            m_wasStopped = true;
            return;
        }

        int traversedVoxelCount = 0;
        while (++traversedVoxelCount < m_voxelDistance) {
            if (tMaxX < tMaxY && tMaxX < tMaxZ) {
                voxelIndex.x += stepX;
                tMaxX += tDeltaX;
            }
            else if (tMaxY < tMaxZ) {
                voxelIndex.y += stepY;
                tMaxY += tDeltaY;
            }
            else {
                voxelIndex.z += stepZ;
                tMaxZ += tDeltaZ;
            }

            if (onTraversingVoxel.apply(voxelIndex)) {
                m_wasStopped = true;
                break;
            }
        }
    }

    /**
     * Computes the voxel distance, a.k.a. the number of voxel to traverse, for the ray cast.
     *
     * @param voxelExtent The extent of a voxel, which is the equivalent for a cube of a sphere's radius.
     * @param startIndex  The starting position's index.
     */
    private void computeVoxelDistance(double voxelExtent, Vector3i startIndex) {
        m_voxelDistance = 1 + MathExt.abs(VoxelWorldUtils.getVoxelIndexNoOffset(voxelExtent, m_offsettedStart.x + m_direction.x * m_length) - startIndex.x) + MathExt.abs(
            VoxelWorldUtils.getVoxelIndexNoOffset(voxelExtent, m_offsettedStart.y + m_direction.y * m_length) - startIndex.y) + MathExt.abs(
            VoxelWorldUtils.getVoxelIndexNoOffset(voxelExtent, m_offsettedStart.z + m_direction.z * m_length) - startIndex.z);
    }

    public Vector3d getStart() {
        return m_start;
    }

    public Vector3d getDirection() {
        return m_direction;
    }

    public double getLength() {
        return m_length;
    }

    public int getVoxelDistance() {
        return m_voxelDistance;
    }

    public void setStart(Vector3d start) {
        m_start.set(start);
    }

    public void setStart(Vector3f start) {
        m_start.set(start);
    }

    /**
     * Sets the direction.
     *
     * @param direction The direction to set to the ray. Must be normalized.
     */
    public void setDirection(Vector3d direction) {
        m_direction.set(direction);
    }

    /**
     * Sets the direction.
     *
     * @param direction The direction to set to the ray. Must be normalized.
     */
    public void setDirection(Vector3f direction) {
        m_direction.set(direction);
    }

    /**
     * Sets the length of the ray.
     *
     * @param length The new length of the ray. Must be positive.
     */
    public void setLength(double length) {
        m_length = length;
    }

    /**
     * Sets the end position of the ray, which is not a real variable but a way to set the direction and the length at the same time. The start position does matter for this
     * method.
     *
     * @param end Where the ray ends.
     */
    public void setEnd(Vector3d end) {
        m_direction.set(end).subtractLocal(m_start);
        m_length = m_direction.length();
        m_direction.normalizeLocal();
    }

    /**
     * Gets if the voxel ray cast was stopped by the "onTraversingVoxel" method call.
     *
     * @return True if the voxel ray cast was stopped by the "onTraversingVoxel" method call, false otherwise.
     */
    public boolean wasStopped() {
        return m_wasStopped;
    }
}

Here are the external static methods :

/**
 * Gets the voxel index of the specified position. This method suppose that the parameter position is already offsetted with + voxel half extent. This method local because the supplied voxel index will be locally modified and returned.
 *
 * @param voxelExtent The  extent of a voxel, which is the equivalent to a cube of a sphere's diameter.
 * @param position    The position to get the voxel index from. Must already be offsetted with + voxel half extent * @param voxelIndex  Where to store the voxel index.
 * @return The voxel index parameter that is set to the supplied position's voxel index.
 */
public static Vector3i getVoxelIndexNoOffsetLocal(double voxelExtent, Vector3d position, Vector3i voxelIndex) {
    return voxelIndex.set(getVoxelIndexNoOffset(voxelExtent, position.x), getVoxelIndexNoOffset(voxelExtent, position.y), getVoxelIndexNoOffset(voxelExtent, position.z));
}

/**
 * Gets the sign of the supplied number. The method being "zero position" means that the sign of zero is 1.
 *
 * @param number The number to get the sign from.
 * @return The number's sign.
 */
public static long getSignZeroPositive(double number) {
    assert !Double.isNaN(number);
    return getNegativeSign(number) | 1;
}

/**
 * Gets the negative sign of the supplied number. So, in other words, if the number is negative, -1 is returned but if the number is positive or zero, then zero is returned.
 * It does not check if the parameter is NaN.
 *
 * @param number The number to get its negative sign.
 * @return -1 if the number is negative, 0 otherwise.
 */
public static long getNegativeSign(double number) {
    assert !Double.isNaN(number);
    return Double.doubleToRawLongBits(number) >> BIT_COUNT_EXCLUDING_SIGN_64;
}

The important parts to adjust the algorithm to fit my voxel boundaries are the following :

m_offsettedStart.set(m_start).addLocal(voxelHalfExtent, voxelHalfExtent, voxelHalfExtent);

It is mandatory to add the half extent to the starting position.

double nextVoxelBoundaryX = (voxelIndex.x + (MathExt.getNegativeSign(stepX) + 1)) * voxelExtent; double nextVoxelBoundaryY = (voxelIndex.y + (MathExt.getNegativeSign(stepY) + 1)) * voxelExtent; double nextVoxelBoundaryZ = (voxelIndex.z + (MathExt.getNegativeSign(stepZ) + 1)) * voxelExtent;

What the MathExt method call does could be programmed as : (stepX >= 0 ? 1 : 0).

I don't know how to express how it is delightful when everything starts to fit and work properly :')

Here are some screenshots :

Capture-2.thumb.png.d01ec6b16fe9960734bdf464692cb352.png
Capture-4.thumb.png.d894ebdd7869ab90ec51542e74c0a421.png
Capture-3.thumb.png.ef840ead6db6c7b7cb4e31eadea4e50c.png

2 likes 6 comments

Comments

swiftcoder

Don't the gradients cancel in your formula? You should end up with just position/gradient?

July 31, 2018 11:57 PM
bdubreuil
14 minutes ago, swiftcoder said:

Don't the gradients cancel in your formula? You should end up with just position/gradient?

They are not scalars but vectors; that's what the arrow means on top of a variable. So it's not scalar multiplications but dot products.

 

My friend just told me that I can use TeX instead of images for the equations : I'll update that at the same time that I'll upload the video.

August 01, 2018 12:16 AM
JoeJ
Quote


        // In which direction the voxel ids are incremented.
        double stepX = MathExt.getSignZeroPositive(m_direction.x);

Shouldn't the steps be integers? Actually there is double to int conversion in the inner loop caused by this.

August 01, 2018 05:06 AM
bdubreuil
10 hours ago, JoeJ said:

Shouldn't the steps be integers? Actually there is double to int conversion in the inner loop caused by this.

You're right, thank you. I'll fix it.

August 01, 2018 03:10 PM
Awoken

Very impressive.  Nicely done.  If you don't mind saying, what language did you write this code in?  I can't follow or understand it.  I know English isn't your native language, but would you be willing to offer up a simplified explanation of the logic you used for this blog post? 

August 01, 2018 09:48 PM
bdubreuil
6 hours ago, Awoken said:

If you don't mind saying, what language did you write this code in? 

It's in Java but everything remains the same as other C languages, except the calls to my personal methods, which is kind of egoistic of me. I just haven't tried yet if they are more performant than Java's JDK math methods.

 

6 hours ago, Awoken said:

ould you be willing to offer up a simplified explanation of the logic you used for this blog post

For the gradient or the algorithm? Well, first of all, for the algorithm, you need to at least read the research paper to get a grasp of what was the intent and the result wanted by the authors, even if you don't understand everything.

Second of all, the simplest way to understand the algorithm is to imagine a line in an 3D world made of blocks. Which blocks does the line touch? Then, in which order are they touched based on the line's start and end positions? The goal is to traverse iteratively the blocks that are touched by the line .

Third of all, the logic of the algorithm can be summed making a distinction between the ray's direction's components. Those three define the importance of their axes in terms of how many blocks need to be traversed in what direction. Think of this with integers : two people are running to reach a goal; the fastest runs a 5 km/h, while the slowest runs at 1 km/h. For each time step, i.e. an hour, how many kilometers have each runner traveled? The ratio is 5 : 1, so, to merge the analogy, a ray would traverse each step 5 blocks on the X axis and 1 block on the Y axis. Of course, it's more complicated than that, as there are more parameters to it, especially because of exceptions such as what to do when each component is equal with one another?

I'll add this explanation to the article ^^

August 02, 2018 04:19 AM
Awoken

I'm still confused.  Which part of the code you posted actually determines which voxels are hit?

August 02, 2018 09:18 PM
bdubreuil
4 hours ago, Awoken said:

I'm still confused.  Which part of the code you posted actually determines which voxels are hit?


        if (onTraversingVoxel.apply(voxelIndex)) {
            m_wasStopped = true;
            return;
        }

        int traversedVoxelCount = 0;
        while (++traversedVoxelCount < m_voxelDistance) {
            if (tMaxX < tMaxY && tMaxX < tMaxZ) {
                voxelIndex.x += stepX;
                tMaxX += tDeltaX;
            }
            else if (tMaxY < tMaxZ) {
                voxelIndex.y += stepY;
                tMaxY += tDeltaY;
            }
            else {
                voxelIndex.z += stepZ;
                tMaxZ += tDeltaZ;
            }

            if (onTraversingVoxel.apply(voxelIndex)) {
                m_wasStopped = true;
                break;
            }
        }

Look more specifically at onTraversingVoxel.apply(voxelIndex) : this line of code applies a Java functional interface, which is a lambda. The supplied parameter is the absolute (world) index of the voxel touched.

August 03, 2018 01:58 AM
Awoken

Oh fantastic explanation.  I now understand the logic and it's remarkably simple, which is a good thing.  Great blog post, very educational.

August 03, 2018 06:55 PM
SuperVGA

It's interesting that neither this thread nor the linked paper mentions it: I'm convinced this is just a 3D bresenham implementation. (credit where due)
Back in early 2010, I played around with implementing bresenham in a fragment shader working on a volume texture, and it actually did pretty good.
It's sort of a nice way to avoid using polygons, if you're into that sort of thing, but it was a little too much branching for my GPU at the time...

October 10, 2018 09:46 AM
bdubreuil
3 hours ago, SuperVGA said:

It's interesting that neither this thread nor the linked paper mentions it: I'm convinced this is just a 3D bresenham implementation. (credit where due)
Back in early 2010, I played around with implementing bresenham in a fragment shader working on a volume texture, and it actually did pretty good.
It's sort of a nice way to avoid using polygons, if you're into that sort of thing, but it was a little too much branching for my GPU at the time...

I didn't know that algorithm. However, the result of the Bresenham's method is the first of two possible choices, which can traverse diagonally. For my algorithm, I need the second choice because I don't want the ray cast to select voxels diagonally, as it might select hidden voxels.

October 10, 2018 01:24 PM
SuperVGA
49 minutes ago, thecheeselover said:

I didn't know that algorithm. However, the result of the Bresenham's method is the first of two possible choices, which can traverse diagonally. For my algorithm, I need the second choice because I don't want the ray cast to select voxels diagonally, as it might select hidden voxels.

Ah, right. Yeah, I don't think it's something that requires a whole lot of thought either, so it's fair enough. Mostly I was just puzzled as to why the paper didn't reference it.

October 10, 2018 02:16 PM
You must log in to join the conversation.
Don't have a GameDev.net account? Sign up!
Profile
Author
Advertisement
Advertisement