Search Results

Search found 33291 results on 1332 pages for 'development environment'.

Page 495/1332 | < Previous Page | 491 492 493 494 495 496 497 498 499 500 501 502  | Next Page >

  • Multiple textures on a mesh created in blender and imported in xna

    - by alecnash
    I created a cube in blender which has multiple images applied to its faces. I am trying to import the model into xna and get the same results as shown when rendering the model in blender. I go through every mesh (for the cube its only one) and through every part but only the first image used in blender is displayed in every face. The code I am using to fetch the texture looks like that: foreach (ModelMesh m in model.Meshes) { foreach (Effect e in m.Effects) { foreach (var part in m.MeshParts) { e.CurrentTechnique = e.Techniques["Lambert"]; e.Parameters["view"].SetValue(camera.viewMatrix); e.Parameters["projection"].SetValue(camera.projectionMatrix); e.Parameters["colorMap"].SetValue(modelTextures[part.GetHashCode()]); } } m.Draw(); } Am I missing something?

    Read the article

  • Multi-Threaded Pipelined Game Engine Data Synchronization Questions

    - by Douglas
    Let's say I'm setting up a worker pool based game engine with pipelining. Let's say I have 4 stages in my pipeline as such: Stage 1: Physics Stage 2: AI/Input Stage 3: Game Logic Stage 4: Rendering Now let's say that the physics detects a collision between a bullet and a character in stage 1. Two frames later the game logic may choose to remove that bullet from the simulation, however none of the other copies of the data for the other pipeline stages will get this information. How is this sort of thing and other things like it get handled? Do you generally make changes like this to every pipeline stage's data at the end of a frame?

    Read the article

  • Proper way to use a RenderTarget2D to draw multiple textures?

    - by TheBroodian
    In the process of trying to resolve a split screen issue, I've been trying to use a RenderTarget2D to draw a portion of my scene to a Texture2D, and then again to another Texture2D, but the end result of both Texture2D's is coming out the same. Can anybody tell me what I'm doing wrong? Texture2D camera1Render; Texture2D camera2Render; GraphicsDevice.SetRenderTarget(RenderTarget); GraphicsDevice.Clear(Color.Transparent); map.Draw(mapDisplayDevice, Camera1, new Location(0, 0), false); camera1Render = RenderTarget; GraphicsDevice.Clear(Color.Transparent); map.Draw(mapDisplayDevice, Camera2, new Location(0, 0), false); camera2Render = RenderTarget; SetRenderTarget(null);

    Read the article

  • Data structures for a 2D multi-layered and multi-region map?

    - by DevilWithin
    I am working on a 2D world editor and a world format subsequently. If I were to handle the game "world" being created just as a layered set of structures, either in top or side views, it would be considerably simple to do most things. But, since this editor is meant for 3rd parties, I have no clue how big worlds one will want to make and I need to keep in mind that eventually it will become simply too much to check, handling and comparing stuff that are happening completely away from the player position. I know the solution for this is to subdivide my world into sub regions and stream them on the fly, loading and unloading resources and other data. This way I know a virtually infinite game area is achievable. But, while I know theoretically what to do, I really have a few questions I'd hoped to get answered for some hints about the topic. The logic way to handle the regions is some kind of grid, would you pick evenly distributed blocks with equal sizes or would you let the user subdivide areas by taste with irregular sized rectangles? In case of even grids, would you use some kind of block/chunk neighbouring system to check when the player transposes the limit or just put all those in a simple array? Being a region a different data structure than its owner "game world", when streaming a region, would you deliver the objects to the parent structures and track them for unloading later, or retain the objects in each region for a more "hard-limit" approach? Introducing the subdivision approach to the project, and already having a multi layered scene graph structure on place, how would i make it support the new concept? Would you have the parent node have the layers as children, and replicate in each layer node, a node per region? Or the opposite, parent node owns all the regions possible, and each region has multiple layers as children? Or would you just put the region logic outside the graph completely(compatible with the first suggestion in Q.3) When I say virtually infinite worlds, I mean it of course under the contraints of the variable sizes and so on. Using float positions, a HUGE world can already be made. Do you think its sane to think beyond that? Because I think its ok to stick to this limit since it will never be reached so easily.. As for when to stream a region, I'm implementing it as a collection of watcher cameras, which the streaming system works with to know what to load/unload. The problem here is, i will be needing some kind of warps/teleports built in for my game, and there is a chance i will be teleporting a player to a unloaded region far away. How would you approach something like this? Is it sane to load any region to memory which can be teleported to by a warp within a radius from the player? Sorry for the huge question, any answers are helpful!

    Read the article

  • Precision loss when transforming from cartesian to isometric

    - by Justin Skiles
    My goal is to display a tile map in isometric projection. This tile map has 25 tiles across and 25 tiles down. Each tile is 32x32. See below for how I'm accomplishing this. World Space World Space to Screen Space Rotation (45 degrees) Using a 2D rotation matrix, I use the following: double rotation = Math.PI / 4; double rotatedX = ((tileWorldX * Math.Cos(rotation)) - ((tileWorldY * Math.Sin(rotation))); double rotatedY = ((tileWorldX * Math.Sin(rotation)) + (tileWorldY * Math.Cos(rotation))); World Space to Screen Space Scale (Y-axis reduced by 50%) Here I simply scale down the Y value by a factor of 0.5. Problem And it works, kind of. There are some tiny 1px-2px gaps between some of the tiles when rendering. I think there's some precision loss somewhere, or I'm not understanding how to get these tiles to fit together perfectly. I'm not truncating or converting my values to non-decimal types until I absolutely have to (when I pass to the render method, which only takes integers). I'm not sure how to guarantee pixel perfect rendering precision when I'm rotating and scaling on a level of higher precision. Any advice? Do I need to supply for information?

    Read the article

  • Frame timing for GLFW versus GLUT

    - by linello
    I need a library which ensures me that the timing between frames are more constant as possible during an experiment of visual psychophics. This is usually done synchronizing the refresh rate of the screen with the main loop. For example if my monitor runs at 60Hz I would like to specify that frequency to my framework. For example if my gameloop is the following void gameloop() { // do some computation printDeltaT(); Flip buffers } I would like to have printed a constant time interval. Is it possible with GLFW?

    Read the article

  • How to find the window size in XNA

    - by Nick Van Hoogenstyn
    I just wanted to know if there was a way to find out the size of the window in XNA. I don't want to set it to a specific size; I would like to know what dimensions it currently displays as automatically. Is there a way to find this information out? I realize I probably should have found this information out (or set it myself manually) before working on the game, but I'm a novice and am now hoping to work within the dimensions I have already become invested in. Thanks!

    Read the article

  • How do I create a third Person View using DXUTCamera in DX10?

    - by David
    I am creating a 3d flying game and using DXUTCamera for my view. I can get the camera to take on the characters position, But I would like to view my character in the 3rd person. Here is my code for first person view: //Put the camera on the object. D3DXVECTOR3 viewerPos; D3DXVECTOR3 lookAtThis; D3DXVECTOR3 up ( 5.0f, 1.0f, 0.0f ); D3DXVECTOR3 newUp; D3DXMATRIX matView; //Set the viewer's position to the position of the thing. viewerPos.x = character->x; viewerPos.y = character->y; viewerPos.z = character->z; // Create a new vector for the direction for the viewer to look character->setUpWorldMatrix(); D3DXVECTOR3 newDir, lookAtPoint; D3DXVec3TransformCoord(&newDir, &character->initVecDir, &character->matAllRotations); // set lookatpoint D3DXVec3Normalize(&lookAtPoint, &newDir); lookAtPoint.x += viewerPos.x; lookAtPoint.y += viewerPos.y; lookAtPoint.z += viewerPos.z; g_Camera.SetViewParams(&viewerPos, &lookAtPoint); So does anyone have an ideas how I can move the camera to the third person view? preferably timed so there is a smooth action in the camera movement. (I'm hoping I can just edit this code instead of bringing in another camera class)

    Read the article

  • Making more complicated systems(entity-component-system model question)

    - by winch
    I'm using a model where entities are collections of components and components are just data. All the logic goes into systems which operate on components. Making basic systems(for Rendering and handling collision) was easy. But how do I do more compilcated systems? For example, in a CollisionSystem I can check if entity A collides with entity B. I have this code in CollisionSystem for checking if B damages A: if(collides(a, b)) { HealthComponent* hc = a->get<HealthComponent(); hc.reduceHealth(b->get<DamageComponent>()->getDamage()); But I feel that this code shouldn't belong to Collision system. Where should code like this be and which additional systems should I create to make this code generic?

    Read the article

  • Child transforms problem when loading 3DS models using assimp

    - by MhdSyrwan
    I'm trying to load a textured 3d model into my scene using assimp model loader. The problem is that child meshes are not situated correctly (they don't have the correct transformations). In brief: all the mTansform matrices are identity matrices, why would that be? I'm using this code to render the model: void recursive_render (const struct aiScene *sc, const struct aiNode* nd, float scale) { unsigned int i; unsigned int n=0, t; aiMatrix4x4 m = nd->mTransformation; m.Scaling(aiVector3D(scale, scale, scale), m); // update transform m.Transpose(); glPushMatrix(); glMultMatrixf((float*)&m); // draw all meshes assigned to this node for (; n < nd->mNumMeshes; ++n) { const struct aiMesh* mesh = scene->mMeshes[nd->mMeshes[n]]; apply_material(sc->mMaterials[mesh->mMaterialIndex]); if (mesh->HasBones()){ printf("model has bones"); abort(); } if(mesh->mNormals == NULL) { glDisable(GL_LIGHTING); } else { glEnable(GL_LIGHTING); } if(mesh->mColors[0] != NULL) { glEnable(GL_COLOR_MATERIAL); } else { glDisable(GL_COLOR_MATERIAL); } for (t = 0; t < mesh->mNumFaces; ++t) { const struct aiFace* face = &mesh->mFaces[t]; GLenum face_mode; switch(face->mNumIndices) { case 1: face_mode = GL_POINTS; break; case 2: face_mode = GL_LINES; break; case 3: face_mode = GL_TRIANGLES; break; default: face_mode = GL_POLYGON; break; } glBegin(face_mode); for(i = 0; i < face->mNumIndices; i++)// go through all vertices in face { int vertexIndex = face->mIndices[i];// get group index for current index if(mesh->mColors[0] != NULL) Color4f(&mesh->mColors[0][vertexIndex]); if(mesh->mNormals != NULL) if(mesh->HasTextureCoords(0))//HasTextureCoords(texture_coordinates_set) { glTexCoord2f(mesh->mTextureCoords[0][vertexIndex].x, 1 - mesh->mTextureCoords[0][vertexIndex].y); //mTextureCoords[channel][vertex] } glNormal3fv(&mesh->mNormals[vertexIndex].x); glVertex3fv(&mesh->mVertices[vertexIndex].x); } glEnd(); } } // draw all children for (n = 0; n < nd->mNumChildren; ++n) { recursive_render(sc, nd->mChildren[n], scale); } glPopMatrix(); } What's the problem in my code ? I've added some code to abort the program if there's any bone in the meshes, but the program doesn't abort, this means : no bones, is that normal? if (mesh->HasBones()){ printf("model has bones"); abort(); } Note: I am using openGL & SFML & assimp

    Read the article

  • Simple project - make a 3D box tumble and fall to the ground [closed]

    - by Dominic Bou-Samra
    Possible Duplicate: Resources to learn programming rigid body simulation Hi guys, I want to try learning rigid-body dynamic simulation. I have done a fluid and cloth simulation before, but never anything rigid. My maths knowledge is limited in that I don't know the notation that well. Are there any good cliff-notes, tutorials, guides on how I would accomplish a simple task like this? I don't want a super complex pdf that's only a little relevant. Thanks.

    Read the article

  • vector rotations for branches of a 3d tree

    - by freefallr
    I'm attempting to create a 3d tree procedurally. I'm hoping that someone can check my vector rotation maths, as I'm a bit confused. I'm using an l-system (a recursive algorithm for generating branches). The trunk of the tree is the root node. It's orientation is aligned to the y axis. In the next iteration of the tree (e.g. the first branches), I might create a branch that is oriented say by +10 degrees in the X axis and a similar amount in the Z axis, relative to the trunk. I know that I should keep a rotation matrix at each branch, so that it can be applied to child branches, along with any modifications to the child branch. My questions then: for the trunk, the rotation matrix - is that just the identity matrix * initial orientation vector ? for the first branch (and subsequent branches) - I'll "inherit" the rotation matrix of the parent branch, and apply x and z rotations to that also. e.g. using glm::normalize; using glm::rotateX; using glm::vec4; using glm::mat4; using glm::rotate; vec4 vYAxis = vec4(0.0f, 1.0f, 0.0f, 0.0f); vec4 vInitial = normalize( rotateX( vYAxis, 10.0f ) ); mat4 mRotation = mat4(1.0); // trunk rotation matrix = identity * initial orientation vector mRotation *= vInitial; // first branch = parent rotation matrix * this branches rotations mRotation *= rotate( 10.0f, 1.0f, 0.0f, 0.0f ); // x rotation mRotation *= rotate( 10.0f, 0.0f, 0.0f, 1.0f ); // z rotation Are my maths and approach correct, or am I completely wrong? Finally, I'm using the glm library with OpenGL / C++ for this. Is the order of x rotation and z rotation important?

    Read the article

  • Question about BoundingSpheres and Ray intersections

    - by NDraskovic
    I'm working on a XNA project (not really a game) and I'm having some trouble with picking algorithm. I have a few types of 3D models that I draw to the screen, and one of them is a switch. So I'm trying to make a picking algorithm that would enable the user to click on the switch and that would trigger some other function. The problem is that the BoundingSphere.Intersect() method always returns null as result. This is the code I'm using: In the declaration section: ` //Basic matrices private Matrix world = Matrix.CreateTranslation(new Vector3(0, 0, 0)); private Matrix view = Matrix.CreateLookAt(new Vector3(10, 10, 10), new Vector3(0, 0, 0), Vector3.UnitY); private Matrix projection = Matrix.CreatePerspectiveFieldOfView(MathHelper.ToRadians(45), 800f / 600f, 0.01f, 100f); //Collision detection variables Viewport mainViewport; List<BoundingSphere> spheres = new List<BoundingSphere>(); Ray ControlRay; Vector3 nearPoint, farPoint, nearPlane, farPlane, direction; ` And then in the Update method: ` nearPlane = new Vector3((float)Mouse.GetState().X, (float)Mouse.GetState().Y, 0.0f); farPlane = new Vector3((float)Mouse.GetState().X, (float)Mouse.GetState().Y, 10.0f); nearPoint = GraphicsDevice.Viewport.Unproject(nearPlane, projection, view, world); farPoint = GraphicsDevice.Viewport.Unproject(farPlane, projection, view, world); direction = farPoint - nearPoint; direction.Normalize(); ControlRay = new Ray(nearPoint, direction); if (spheres.Count != 0) { for (int i = 0; i < spheres.Count; i++) { if (spheres[i].Intersects(ControlRay) != null) { Window.Title = spheres[i].Center.ToString(); } else { Window.Title = "Empty"; } } ` The "spheres" list gets filled when the 3D object data gets loaded (I read it from a .txt file). For every object marked as switch (I use simple numbers to determine which object is to be drawn), a BoundingSphere is created (center is on the coordinates of the 3D object, and the diameter is always the same), and added to the list. The objects are drawn normally (and spheres.Count is not 0), I can see them on the screen, but the Window title always says "Empty" (of course this is just for testing purposes, I will add the real function when I get positive results) meaning that there is no intersection between the ControlRay and any of the bounding spheres. I think that my basic matrices (world, view and projection) are making some problems, but I cant figure out what. Please help.

    Read the article

  • What is a correct step by step logic of exporting scene with baked occlusion for loading it at runtime?

    - by myWallJSON
    I wonder what is a correct step by step logic of exporting scene with baked occlusion (Culling data) for loading that scene at runtime (on fly from the internet for example))? So currently my plan looks like this: I create prefabs Place them onto my scene (into Hierarchy) (say create 20 buffolows and some hourses and some buildings) Create empty prefab and drag all my scene objects from hierarchy onto it Export prefab So generally I put all my scene objects into one large prefab and export it but it seems that all objects that were marked as static get this property turned off when loading them at runtime and so no Frustrum Culling, and no Occlusion culling happens. So I wonder what is a correct way of exporting Sceen + Objecrts + Occlusion (and onther culing) data for future load of such scene at runtime? I wonder about current 3.5.2 Pro and future 4 Pro versions of U3D.

    Read the article

  • How do I tackle top down RPG movement?

    - by WarmWaffles
    I have a game that I am writing in Java. It is a top down RPG and I am trying to handle movement in the world. The world is largely procedural and I am having a difficult time tackling how to handle character movement around the world and render the changes to the screen. I have the world loaded in blocks which contains all the tiles. How do I tackle the character movement? I am stumped and can't figure out where I should go with it. EDIT: Well I was abstract with the issue at hand. Right now I can only think of rigidly sticking everything into a 2D array and saving the block ID and the player offset in the Block or I could just "float" everything and move about between tiles so to speak.

    Read the article

  • Collision detection with entities/AI

    - by James Williams
    I'm making my first game in Java, a top down 2D RPG. I've handled basic collision detection, rendering and have added an NPC, but I'm stuck on how to handle interaction between the player and the NPC. Currently I'm drawing out my level and then drawing characters, NPCs and animated tiles on top of this. The problem is keeping track of the NPCs so that my Character class can interact with methods in the NPC classes on collision. I'm not sure my method of drawing the level and drawing everything else on top is a good one - can anyone shed any light on this topic?

    Read the article

  • How to modify VBO data

    - by Romeo
    I am learning LWJGL so i can start working on my game. In order to learn LWJGL I got the idea to implement the map builder so I can get comfortable with graphics programming. Now, for the map creation tool I need to draw new elements or draw the old one's with different coordinates. Let me explain this: My game will be a 2D scroller. The map will be consisting of multiple rectangles ( 2 strip triangles). When I click my left-mouse button i want to start the rectangle and when I release it I want to stop the rectangle bottom-right at that position. As I want to use VBOs I want to know how to modify data inside the VBO based on user input. Should i have a copy of a vertex array and then add the whole array to the VBO at each user input? How is usually implemented the VBO update?

    Read the article

  • Surface normal to screen angle

    - by Tannz0rz
    I've been struggling to get this working. I simply wish to take a surface normal and convert it to a screen angle. As an example, assuming we're working with the highlighted surface on the sphere below, where the arrow is the normal, the 2D angle would obviously be PI/4 radians. Here's one of the many things I've tried to no avail: float4 A = v.vertex; float4 B = v.vertex + float4(v.normal, 0.0); A = mul(VP, A); B = mul(VP, B); A.xy = (0.5 * (A.xy / A.w)) + 0.5; B.xy = (0.5 * (B.xy / B.w)) + 0.5; o.theta = atan2(B.y - A.y, B.x - A.x); I'm finally at my wit's end. Thanks for any and all help.

    Read the article

  • Multiplayer Network Game - Interpolation and Frame Rate

    - by J.C.
    Consider the following scenario: Let's say, for sake of example and simplicity, that you have an authoritative game server that sends state to its clients every 45ms. The clients are interpolating state with an interpolation delay of 100 ms. Finally, the clients are rendering a new frame every 15ms. When state is updated on the client, the client time is set from the incoming state update. Each time a frame renders, we take the render time (client time - interpolation delay) and identify a previous and target state to interpolate from. To calculate the interpolation amount/factor, we take the difference of the render time and previous state time and divide by the difference of the target state and previous state times: var factor = ((renderTime - previousStateTime) / (targetStateTime - previousStateTime)) Problem: In the example above, we are effectively displaying the same interpolated state for 3 frames before we collected the next server update and a new client (render) time is set. The rendering is mostly smooth, but there is a dash of jaggedness to it. Question: Given the example above, I'd like to think that the interpolation amount/factor should increase with each frame render to smooth out the movement. Should this be considered and, if so, what is the best way to achieve this given the information from above?

    Read the article

  • Rotate 3D Model from a custom position

    - by Nipuna Silva
    I have a 3D Model like above in which i want to rotate it from a given location(pointed in red) but I can only rotate it from the middle. How can I rotate it from a custom point. Edit: I successfully able to rotate the model from the below position by getting the radius of the model and applying it to the world matrix Vector3 point = new Vector3(-radius, 0, 0); world = Matrix.CreateTranslation(-radius, 0, 0); But now I cannot change the position of the object and it always centered in middle of the screen. I think that's because i applied the above code. How can I place it anywhere I want?

    Read the article

  • Low complexity shader to indicate the sides of a polyline

    - by Pris
    I have a bunch of polylines that I draw using GL_LINES. They can have thousands of points. They actually represent the separation of land and water on a map. I don't have complete polygons, just the ordered set of points. I'm looking for a neat but efficient way to visually convey Side A and Side B as being different. For example I could offset the polyline in one direction a few times and fade it out (but every offset is doubling the number of points), or offset it once to make a "ribbon" and give one side a 'glow' like effect to mimic the outer glow or shadow of a polygon). This is for a mobile application and I'm using OpenGL ES 2. I'd like to keep the effect as simple as possible from a complexity stand point. I'm looking for some additional ideas; maybe there's a clever shader technique out there or a visual effect I haven't considered.

    Read the article

  • Best way to implement an AI for Dominion? [on hold]

    - by j will
    I'm creating a desktop client and server backend for the game, Dominion, by Donald X. Vaccarino. I've been reading up on AI techniques and algorithms and I just wanted to what is the best way to implement an AI for such a game? Would it better to look at neural networks, genetic algorithms, decision trees, fuzzy logic, or any other methodology? For those who do not know how Dominion works, check out this part of the wikipedia article: http://en.wikipedia.org/wiki/Dominion_(card_game)#Gameplay

    Read the article

  • Problems with moving 2D circle/box collision detection

    - by dario3004
    This is my first game ever and I'm a newbie in computer physics. I've got this code for the collision detection and it works fine for BOTTOM and TOP collision.It miss the collision detection with the paddle's edge and angles so I've (roughly) tried to implement it. Main method that is called for bouncing, it checks if it bounce with wall, or with top (+ right/left side) or with bottom (+ right/left side): protected void handleBounces(float px, float py) { handleWallBounce(px, py); if(mBall.y < getHeight()/4){ if (handleRedFastBounce(mRed, px, py)) return; if (handleRightSideBounce(mRed,px,py)) return; if (handleLeftSideBounce(mRed,px,py)) return; } if(mBall.y > getHeight()/4 * 3){ if (handleBlueFastBounce(mBlue, px, py)) return; if (handleRightSideBounce(mBlue,px,py)) return; if (handleLeftSideBounce(mBlue,px,py)) return; } } This is the code for the BOTTOM bounce: protected boolean handleRedFastBounce(Paddle paddle, float px, float py) { if (mBall.goingUp() == false) return false; // next position tx = mBall.x; ty = mBall.y - mBall.getRadius(); // actual position ptx = px; pty = py - mBall.getRadius(); dyp = ty - paddle.getBottom(); xc = tx + (tx - ptx) * dyp / (ty - pty); if ((ty < paddle.getBottom() && pty > paddle.getBottom() && xc > paddle.getLeft() && xc < paddle.getRight())) { mBall.x = xc; mBall.y = paddle.getBottom() + mBall.getRadius(); mBall.bouncePaddle(paddle); playSound(mPaddleSFX); increaseDifficulty(); return true; } else return false; } As long as I understood it should be something like this: So I tried to make the "left side" and "right side" bounce method: protected boolean handleLeftSideBounce(Paddle paddle, float px, float py){ // next position tx = mBall.x + mBall.getRadius(); ty = mBall.y; // actual position ptx = px + mBall.getRadius(); pty = py; dyp = tx - paddle.getLeft(); yc = ty + (pty - ty) * dyp / (ptx - tx); if (ptx < paddle.getLeft() && tx > paddle.getLeft()){ System.out.println("left side bounce1"); System.out.println("yc: " + yc + "top: " + paddle.getTop() + " bottom: " + paddle.getBottom()); if (yc > paddle.getTop() && yc < paddle.getBottom()){ System.out.println("left side bounce2"); mBall.y = yc; mBall.x = paddle.getLeft() - mBall.getRadius(); mBall.bouncePaddle(paddle); playSound(mPaddleSFX); increaseDifficulty(); return true; } } return false; } I think I'm quite near to the solution but I'm having big troubles with the new "yc" formula. I tried so many versions of it but since I don't know the theory behind it I can't adjust for the Y axis. Since the Y axis is inverted I even tried this: yc = ty - (pty - ty) * dyp / (ptx - tx); I tried Googling it but I can't seem to find a solution for it. Also this method fails when ball touches the angle and I don't think is a nice way because it just test "one" point of the ball and probably there will be many cases in which the ball won't bounce.

    Read the article

  • Efficiently separating Read/Compute/Write steps for concurrent processing of entities in Entity/Component systems

    - by TravisG
    Setup I have an entity-component architecture where Entities can have a set of attributes (which are pure data with no behavior) and there exist systems that run the entity logic which act on that data. Essentially, in somewhat pseudo-code: Entity { id; map<id_type, Attribute> attributes; } System { update(); vector<Entity> entities; } A system that just moves along all entities at a constant rate might be MovementSystem extends System { update() { for each entity in entities position = entity.attributes["position"]; position += vec3(1,1,1); } } Essentially, I'm trying to parallelise update() as efficiently as possible. This can be done by running entire systems in parallel, or by giving each update() of one system a couple of components so different threads can execute the update of the same system, but for a different subset of entities registered with that system. Problem In reality, these systems sometimes require that entities interact(/read/write data from/to) each other, sometimes within the same system (e.g. an AI system that reads state from other entities surrounding the current processed entity), but sometimes between different systems that depend on each other (i.e. a movement system that requires data from a system that processes user input). Now, when trying to parallelize the update phases of entity/component systems, the phases in which data (components/attributes) from Entities are read and used to compute something, and the phase where the modified data is written back to entities need to be separated in order to avoid data races. Otherwise the only way (not taking into account just "critical section"ing everything) to avoid them is to serialize parts of the update process that depend on other parts. This seems ugly. To me it would seem more elegant to be able to (ideally) have all processing running in parallel, where a system may read data from all entities as it wishes, but doesn't write modifications to that data back until some later point. The fact that this is even possible is based on the assumption that modification write-backs are usually very small in complexity, and don't require much performance, whereas computations are very expensive (relatively). So the overhead added by a delayed-write phase might be evened out by more efficient updating of entities (by having threads work more % of the time instead of waiting). A concrete example of this might be a system that updates physics. The system needs to both read and write a lot of data to and from entities. Optimally, there would be a system in place where all available threads update a subset of all entities registered with the physics system. In the case of the physics system this isn't trivially possible because of race conditions. So without a workaround, we would have to find other systems to run in parallel (which don't modify the same data as the physics system), other wise the remaining threads are waiting and wasting time. However, that has disadvantages Practically, the L3 cache is pretty much always better utilized when updating a large system with multiple threads, as opposed to multiple systems at once, which all act on different sets of data. Finding and assembling other systems to run in parallel can be extremely time consuming to design well enough to optimize performance. Sometimes, it might even not be possible at all because a system just depends on data that is touched by all other systems. Solution? In my thinking, a possible solution would be a system where reading/updating and writing of data is separated, so that in one expensive phase, systems only read data and compute what they need to compute, and then in a separate, performance-wise cheap, write phase, attributes of entities that needed to be modified are finally written back to the entities. The Question How might such a system be implemented to achieve optimal performance, as well as making programmer life easier? What are the implementation details of such a system and what might have to be changed in the existing EC-architecture to accommodate this solution?

    Read the article

  • Dynamic libraries are not allowed on iOS but what about this?

    - by tapirath
    I'm currently using LuaJIT and its FFI interface to call C functions from LUA scripts. What FFI does is to look at dynamic libraries' exported symbols and let the developer use it directly form LUA. Kind of like Python ctypes. Obviously using dynamic libraries is not permitted in iOS for security reasons. So in order to come up with a solution I found the following snippet. /* (c) 2012 +++ Filip Stoklas, aka FipS, http://www.4FipS.com +++ THIS CODE IS FREE - LICENSED UNDER THE MIT LICENSE ARTICLE URL: http://forums.4fips.com/viewtopic.php?f=3&t=589 */ extern "C" { #include <lua.h> #include <lualib.h> #include <lauxlib.h> } // extern "C" #include <cassert> // Please note that despite the fact that we build this code as a regular // executable (exe), we still use __declspec(dllexport) to export // symbols. Without doing that FFI wouldn't be able to locate them! extern "C" __declspec(dllexport) void __cdecl hello_from_lua(const char *msg) { printf("A message from LUA: %s\n", msg); } const char *lua_code = "local ffi = require('ffi') \n" "ffi.cdef[[ \n" "const char * hello_from_lua(const char *); \n" // matches the C prototype "]] \n" "ffi.C.hello_from_lua('Hello from LUA!') \n" // do actual C call ; int main() { lua_State *lua = luaL_newstate(); assert(lua); luaL_openlibs(lua); const int status = luaL_dostring(lua, lua_code); if(status) printf("Couldn't execute LUA code: %s\n", lua_tostring(lua, -1)); lua_close(lua); return 0; } // output: // A message from LUA: Hello from LUA! Basically, instead of using a dynamic library, the symbols are exported directly inside the executable file. The question is: is this permitted by Apple? Thanks.

    Read the article

< Previous Page | 491 492 493 494 495 496 497 498 499 500 501 502  | Next Page >