Search Results

Search found 19338 results on 774 pages for 'game loop'.

Page 333/774 | < Previous Page | 329 330 331 332 333 334 335 336 337 338 339 340  | Next Page >

  • Does DirectX implement Triple Buffering?

    - by Asik
    As AnandTech put it best in this 2009 article: In render ahead, frames cannot be dropped. This means that when the queue is full, what is displayed can have a lot more lag. Microsoft doesn't implement triple buffering in DirectX, they implement render ahead (from 0 to 8 frames with 3 being the default). The major difference in the technique we've described here is the ability to drop frames when they are outdated. Render ahead forces older frames to be displayed. Queues can help smoothness and stuttering as a few really quick frames followed by a slow frame end up being evened out and spread over more frames. But the price you pay is in lag (the more frames in the queue, the longer it takes to empty the queue and the older the frames are that are displayed). As I understand it, DirectX "Swap Chain" is merely a render ahead queue, i.e. buffers cannot be dropped; the longer the chain, the greater the input latency. At the same time, I find it hard to believe that the most widely used graphics API would not implement such fundamental functionality correctly. Is there a way to get proper triple buffered vertical synchronisation in DirectX?

    Read the article

  • How to use the zoom gesture in libgdx?

    - by user3452725
    I found the example code for the GestureListener class, but I don't understand the zoom method: private float initialScale = 1; public boolean zoom (float originalDistance, float currentDistance) { float ratio = originalDistance / currentDistance; //I get this camera.zoom = initialScale * ratio; //This doesn't make sense to me because it seems like every time you pinch to zoom, it resets to the original zoom which is 1. So basically it wouldn't 'save' the zoom right? System.out.println(camera.zoom); //Prints the camera zoom return false; } Am I not interpreting this right?

    Read the article

  • runtime error: invalid memory address or nil pointer dereference

    - by Klink
    I want to learn OpenGL 3.0 with golang. But when i try to compile some code, i get many errors. package main import ( "os" //"errors" "fmt" //gl "github.com/chsc/gogl/gl33" //"github.com/jteeuwen/glfw" "github.com/go-gl/gl" "github.com/go-gl/glfw" "runtime" "time" ) var ( width int = 640 height int = 480 ) var ( points = []float32{0.0, 0.8, -0.8, -0.8, 0.8, -0.8} ) func initScene() { gl.Init() gl.ClearColor(0.0, 0.5, 1.0, 1.0) gl.Enable(gl.CULL_FACE) gl.Viewport(0, 0, 800, 600) } func glfwInitWindowContext() { if err := glfw.Init(); err != nil { fmt.Fprintf(os.Stderr, "glfw_Init: %s\n", err) glfw.Terminate() } glfw.OpenWindowHint(glfw.FsaaSamples, 1) glfw.OpenWindowHint(glfw.WindowNoResize, 1) if err := glfw.OpenWindow(width, height, 0, 0, 0, 0, 32, 0, glfw.Windowed); err != nil { fmt.Fprintf(os.Stderr, "glfw_Window: %s\n", err) glfw.CloseWindow() } glfw.SetSwapInterval(1) glfw.SetWindowTitle("Title") } func drawScene() { for glfw.WindowParam(glfw.Opened) == 1 { gl.Clear(gl.COLOR_BUFFER_BIT) vertexShaderSrc := `#version 120 attribute vec2 coord2d; void main(void) { gl_Position = vec4(coord2d, 0.0, 1.0); }` vertexShader := gl.CreateShader(gl.VERTEX_SHADER) vertexShader.Source(vertexShaderSrc) vertexShader.Compile() fragmentShaderSrc := `#version 120 void main(void) { gl_FragColor[0] = 0.0; gl_FragColor[1] = 0.0; gl_FragColor[2] = 1.0; }` fragmentShader := gl.CreateShader(gl.FRAGMENT_SHADER) fragmentShader.Source(fragmentShaderSrc) fragmentShader.Compile() program := gl.CreateProgram() program.AttachShader(vertexShader) program.AttachShader(fragmentShader) program.Link() attribute_coord2d := program.GetAttribLocation("coord2d") program.Use() //attribute_coord2d.AttribPointer(size, typ, normalized, stride, pointer) attribute_coord2d.EnableArray() attribute_coord2d.AttribPointer(0, 3, false, 0, &(points[0])) //gl.DrawArrays(gl.TRIANGLES, 0, len(points)) gl.DrawArrays(gl.TRIANGLES, 0, 3) glfw.SwapBuffers() inputHandler() time.Sleep(100 * time.Millisecond) } } func inputHandler() { glfw.Enable(glfw.StickyKeys) if glfw.Key(glfw.KeyEsc) == glfw.KeyPress { //gl.DeleteBuffers(2, &uiVBO[0]) glfw.Terminate() } if glfw.Key(glfw.KeyF2) == glfw.KeyPress { glfw.SetWindowTitle("Title2") fmt.Println("Changed to 'Title2'") fmt.Println(len(points)) } if glfw.Key(glfw.KeyF1) == glfw.KeyPress { glfw.SetWindowTitle("Title1") fmt.Println("Changed to 'Title1'") } } func main() { runtime.LockOSThread() glfwInitWindowContext() initScene() drawScene() } And after that: panic: runtime error: invalid memory address or nil pointer dereference [signal 0xb code=0x1 addr=0x0 pc=0x41bc6f74] goroutine 1 [syscall]: github.com/go-gl/gl._Cfunc_glDrawArrays(0x4, 0x7f8500000003) /tmp/go-build463568685/github.com/go-gl/gl/_obj/_cgo_defun.c:610 +0x2f github.com/go-gl/gl.DrawArrays(0x4, 0x3, 0x0, 0x45bd70) /tmp/go-build463568685/github.com/go-gl/gl/_obj/gl.cgo1.go:1922 +0x33 main.drawScene() /home/klink/Dev/Go/gogl/gopher/exper.go:85 +0x1e6 main.main() /home/klink/Dev/Go/gogl/gopher/exper.go:116 +0x27 goroutine 2 [syscall]: created by runtime.main /build/buildd/golang-1/src/pkg/runtime/proc.c:221 exit status 2

    Read the article

  • How does flash store (represent) movieclips and sprites?

    - by humbleBee
    When we draw any object in flash and convert it into a movieclip or a sprite, how is it stored or represented in flash. I know in vector art it is stored or represented as line segments using formulae. Is there any way to get the vertices of the shape that was drawn? For example, lets say a simple rectangle is drawn and is converted to a movieclip. Is there anyway to obtain the vertices and the line segments from the sprite? So that its shape is obtained. Enough information should be obtained so that the shape can be replicated. That's the key - replication. In simple terms, where does flash store information about a shape that has been drawn so that we can obtain it and attempt to replicate the shape ourselves?

    Read the article

  • Illumination and Shading for computer graphics class

    - by Sam I Am
    I am preparing for my test tomorrow and this is one of the practice questions. I solved it partially but I am confused with the rest. Here is the problem: Consider a gray world with no ambient and specular lighting ( only diffuse lighting). The screen coordinates of a triangle P1,P2,P3, are P1=(100,100), P2= (300,150), P3 = (200, 200). The gray values at P!,P2,P3 are 1/2, 3/4, and 1/4 respectively. The light is at infinity and its direction and gray color are (1,1,1) and 1.0 respectively. The coefficients of diffused reflection is 1/2. The normals of P1,P2,P3 are N1= (0,0,1), N2 = (1,0,0), and N3 = (0,1,0) respectively. Consider the coordinates of three points P1,P2,P3 to be 0. Do not normalize the normals. I have computed that the illumination at the 3 vertices P1,P2,P3 is (1/4,3/8,1/8). Also I computed that interpolation coefficients of a point P inside the triangle whose coordinates are (220, 160) are given by (1/5,2/5,2/5). Now I have 4 more questions regarding this problem. 1) The illumination at P using Gouraud Shading is: i) 1/2 The answer is 1/2, but I have no idea how to compute it.. 2) The interpolated normal at P is given by i) (2/5, 2/5,1/5) ii) (1/2, 1/4, 1/4) iii) (3/5, 1/5, 1/5) 3) The interpolated color at P is given by: i) 1/2 Again, I know the correct answer but no idea how to solve it 4) The illumination at P using Phong Shading is i) 1/4 ii) 9/40 iii) 1/2

    Read the article

  • 3Ds Max is exporting model with more normals than vertices

    - by Delta
    I made a simple teapot with the "Create Standard Primitives" option and exported it as a collada file, ended up with this: < float_array id="Teapot001-POSITION-array" count="1590" < float_array id="Teapot001-Normal0-array" count="9216" For what I know there should be only one normal per vertex, am I wrong? What am I supposed to do with that much normals? Just put them on the normal buffer all at once normally?

    Read the article

  • Cool examples of procedural pixel shader effects?

    - by Robert Fraser
    What are some good examples of procedural/screen-space pixel shader effects? No code necessary; just looking for inspiration. In particular, I'm looking for effects that are not dependent on geometry or the rest of the scene (would look okay rendered alone on a quad) and are not image processing (don't require a "base image", though they can incorporate textures). Multi-pass or single-pass is fine. Screenshots or videos would be ideal, but ideas work too. Here are a few examples of what I'm looking for (all from the RenderMonkey samples): PS - I'm aware of this question; I'm not asking for a source of actual shader implementations but instead for some inspirational ideas -- and the ones at the NVIDIA Shader Library mostly require a scene or are image processing effects. EDIT: this is an open-ended question and I wish there was a good way to split the bounty. I'll award the rep to the best answer on the last day.

    Read the article

  • Maya .IFF plugins for Gimp

    - by Kara Marfia
    Maya's preferred format for saving off a UV Snapshot is its own .IFF format, so I was hoping to find a plugin allowing Gimp 2 (Windows) to read it. I've found plenty of plugins for different linux distros, but none are win-friendly (that I can discern - admittedly I'm no whiz with Gimp). Does anyone know of one? Alternately, .tiff seems to work just fine, so if there's no good reason to bother fiddling with IFFs, I'd appreciate the input there, too. (sorry if this isn't on-topic)

    Read the article

  • Ruby: implementing alpha-beta pruning for tic-tac-toe

    - by DerNalia
    So, alpha-beta pruning seems to be the most efficient algorithm out there aside from hard coding (for tic tac toe). However, I'm having problems converting the algorithm from the C++ example given in the link: http://www.webkinesia.com/games/gametree.php #based off http://www.webkinesia.com/games/gametree.php # (converted from C++ code from the alpha - beta pruning section) # returns 0 if draw LOSS = -1 DRAW = 0 WIN = 1 @next_move = 0 def calculate_ai_next_move score = self.get_best_move(COMPUTER, WIN, LOSS) return @next_move end def get_best_move(player, alpha, beta) best_score = nil score = nil if not self.has_available_moves? return false elsif self.has_this_player_won?(player) return WIN elsif self.has_this_player_won?(1 - player) return LOSS else best_score = alpha NUM_SQUARES.times do |square| if best_score >= beta break end if self.state[square].nil? self.make_move_with_index(square, player) # set to negative of opponent's best move; we only need the returned score; # the returned move is irrelevant. score = -get_best_move(1-player, -beta, -alpha) if (score > bestScore) @next_move = square best_score = score end undo_move(square) end end end return best_score end the problem is that this is returning nil. some support methods that are used above: WAYS_TO_WIN = [[0, 1, 2], [3, 4, 5], [6, 7, 8], [0, 3, 6], [1, 4, 7], [2, 5, 8],[0, 4, 8], [2, 4, 6]] def has_this_player_won?(player) result = false WAYS_TO_WIN.each {|solution| result = self.state[solution[0]] if contains_win?(solution) } return (result == player) end def contains_win?(ttt_win_state) ttt_win_state.each do |pos| return false if self.state[pos] != self.state[ttt_win_state[0]] or self.state[pos].nil? end return true end def make_move(x, y, player) self.set_square(x,y, player) end

    Read the article

  • Calculating instantaneous speed and acceleration for a simple Car software model

    - by Dylan
    I am trying to model a speedometer and tachometer for a simple software model of a car dashboard. I want this to be relatively simple, so for my purposes I won't likely simulate variables such as drag (or, assume that drag is a constant). But I would like to know the general formulas for: 1) Calculating the RPM, depending on a position of a graphical slider representing the accelerator. 2) Using this information to find the instantaneous speed (or, magnitude of instantaneous velocity?). I am not sure, in the case of 2), what other independent variables I need to consider. Do I need to consider the frequency of rotation of the wheels (assuming a fixed radius), in addition to the RPM? If anyone can give me a rough explanation plus relevant formulas, or alternatively direct me to other trusted resources online (I have had a hard time sifting through info and determining the accuracy), it would be much appreciated.

    Read the article

  • Correct Rotation and Translation with a 4x4 matrix

    - by sFuller
    I am using a 4x4 matrix to transform verts in a shader. I multiply an identity matrix by a rotation matrix by a translation matrix. I am trying to first rotate the verts and then translate them, however in my result, it appears that the verts are being transformed and then rotated. My matrix looks something like this: m00 m01 m02 tx m10 m11 m12 ty m20 m21 m22 tz --- --- --- 1 I am not using OpenGL's fixed function pipeline, I am multiplying matrices on the client side, and uploading the matrix to a GLSL shader. If it helps, I am using my own matrix multiplication code, but I have recreated this problem using matrices on my graphing calculator, so I don't believe my matrix code has errors.. I'll include a visual aid if needed.

    Read the article

  • Group arrival steering

    - by ltjax
    I've got group movement implemented pretty much like this: http://www.red3d.com/cwr/steer/CrowdPath.html Basically, that's combining path following and separation. It works nicely as long as units are in transit, but arrival does not work very well at all. Right now, units just cease to use the path following component once the "exit" the path, i.e. when their closest point on the path is on or past the end. This leads to those units bumping into each other and also overshooting the point the player clicked. Ideally, I'd have the units arrive scattered around the finish point (and reasonable close to each other), not all clumped up past the finish line. I'd imagine that some kind of arrival steering might work here, but based on other units and a "fuzzy" classification of the end of the path. Is there any proven way to do this?

    Read the article

  • Shader optimization - cg/hlsl pseudo and via multiplication

    - by teodron
    Since HLSL/Cg do not allow texture fetching inside conditional blocks, I am first checking a variable and performing some computations, afterwards setting a float flag to 0.0 or 1.0, depending on the computations. I'd like to trigger a texture fetch only if the flag is 1.0 or not null, for that matter of fact. I kind of hoped this would do the trick: float4 TU0_atlas_colour = pseudoBool * tex2Dlod(TU0_texture, float4(tileCoord, 0, mipLevel)); That is, if pseudoBool is 0, will the texture fetch function still be called and produce overhead? I was hoping to prevent it from getting executed via this trick that usually works in plain C/C++.

    Read the article

  • (Quaternion based) Trouble moving foward based on model rotation

    - by ChocoMan
    Using quaternions, I'm having trouble moving my model in its facing direction. Currently the model moves can move in all cardinal directions with no problems. The problem comes when I rotate the move as it still travelling in the direction of world space. Meaning, if I'm moving forward, backward or any other direction while rotating the model, the model acts like its a figure skater spinning while traveling in the same direction. How do I update the direction of travel proper with the facing direction of the model? Rotates model on Y-axis: Yaw = pController.ThumbSticks.Right.X * MathHelper.ToRadians(speedAngleMAX); AddRotation = Quaternion.CreateFromYawPitchRoll(yaw, 0, 0); ModelLoad.MRotation *= AddRotation; MOrientation = Matrix.CreateFromQuaternion(ModelLoad.MRotation); Moves model forward: // Move Forward if (pController.IsButtonDown(Buttons.LeftThumbstickUp)) { SpeedX = (float)(Math.Sin(ModelLoad.ModelRotation)) * FWDSpeedMax * pController.ThumbSticks.Left.Y * (float)gameTime.ElapsedGameTime.TotalSeconds; SpeedZ = (float)(Math.Cos(ModelLoad.ModelRotation)) * FWDSpeedMax * pController.ThumbSticks.Left.Y * (float)gameTime.ElapsedGameTime.TotalSeconds; // Update model position ModelLoad._modelPos += Vector3.Forward * SpeedZ; ModelLoad._modelPos += Vector3.Left * SpeedX; }

    Read the article

  • accessing c++ class members with luaplus

    - by cppanda
    i've implemented LuaPlus in my engine eventmanager successfully and really like the flexibility i gained. but i'm still not exactly where i want to by, because i can't link my c++ classes to a lua class. for example i have a Actor class in c++, and i want to be able to create the same class in lua and gain access to members with luaplus, but i can't figure how i can achieve that. Is this actually luaplus built in functionality, or do i have to write my own interface that exchanges data tables between c++ and lua? my current approach would be to fire an event in luascript that creates an new actor class in c++ code, and transfer its id and the data i need to back to lua. when i modify the data i send the modifications back to c++ code again, but i actually thought there's something in luaplus that exposes this functionality already.

    Read the article

  • Low coupling and tight cohesion

    - by hidayat
    Of course it depends on the situation. But when a lower lever object or system communicate with an higher level system, should callbacks or events be preferred to keeping a pointer to higher level object? For example, we have a world class that has a member variable vector<monster> monsters. When the monster class is going to communicate with the world class, should I prefer using a callback function then or should I have a pointer to the world class inside the monster class?

    Read the article

  • Realistic planetary terrain generation with weights

    - by Programmdude
    I need terrain generation for a planet. The planet will be divided up into several hundred hexes, and I need it to be realistic and based on weights. I have dabbled in terrain generation before, but nothing like this. So I figure it would be a good idea to ask the community for answers, recommended articles or the like. By realistic, I mean not just random hexes, but continent shaped things with a few islands. More desert around the equator and more ice around the poles. I also have two weights I need to base it around: ice percentage and water percentage. That means that around XX% of the planet will need to be water. Does anyone have any advice or places to start? Generating arbitrary terrain is easy, but something a bit more "organic" like this seems rather difficult. It also needs to be seamless. Should be obvious since it's a planet, but no harm in pointing it out.

    Read the article

  • android: How to apply pinch zoom and pan to 2D GLSurfaceView

    - by mak_just4anything
    I want to apply pinch zoom and panning effect on GLSurfaceView. It is Image editor, so It would not be 3D object. I tried to implement using these following links: https://groups.google.com/forum/#!topic/android-developers/EVNRDNInVRU Want to apply pinch and zoom to GLSurfaceView(3d Object) http://www.learnopengles.com/android-lesson-one-getting-started/ These all are links for 3D object rendering. I can not use ImageView as I need to work out with OpenGL so, had to implement it on GLSurfaceView. Suggest me or any reference links are there for such implementation. **I need it for 2D only.

    Read the article

  • Friction not working for Vehicle in BulletPhysics

    - by Manmohan Bishnoi
    I am creating a vehicle using bullet-physics engine (v 2.82). I created a ground ( btBoxShape ), a box and a vehicle (following the demo). But friction between ground and vehicle wheels seems not working. As soon as the vehicle is placed in 3d world, it starts moving forward. START : Steering works for the vehicle, but engineForce and brakingForce does not work (i.e. I cannot speed-up or stop the vehicle) : I create physics world like this : void initPhysics() { broadphase = new btDbvtBroadphase(); collisionConfiguration = new btDefaultCollisionConfiguration(); dispatcher = new btCollisionDispatcher(collisionConfiguration); solver = new btSequentialImpulseConstraintSolver(); dynamicsWorld = new btDiscreteDynamicsWorld(dispatcher, broadphase, solver, collisionConfiguration); dynamicsWorld->setGravity(btVector3(0, -9.81, 0)); // Debug Drawer bulletDebugugger.setDebugMode(btIDebugDraw::DBG_DrawWireframe); dynamicsWorld->setDebugDrawer(&bulletDebugugger); //groundShape = new btStaticPlaneShape(btVector3(0, 1, 0), 1); groundShape = new btBoxShape(btVector3(50, 3, 50)); fallShape = new btBoxShape(btVector3(1, 1, 1)); // Orientation and Position of Ground groundMotionState = new btDefaultMotionState(btTransform(btQuaternion(0, 0, 0, 1), btVector3(0, -3, 0))); btRigidBody::btRigidBodyConstructionInfo groundRigidBodyCI(0, groundMotionState, groundShape, btVector3(0, 0, 0)); groundRigidBody = new btRigidBody(groundRigidBodyCI); dynamicsWorld->addRigidBody(groundRigidBody); /////////////////////////////////////////////////////////////////////// // Vehicle Setup /////////////////////////////////////////////////////////////////////// vehicleChassisShape = new btBoxShape(btVector3(1.f, 0.5f, 2.f)); vehicleBody = new btCompoundShape(); localTrans.setIdentity(); localTrans.setOrigin(btVector3(0, 1, 0)); vehicleBody->addChildShape(localTrans, vehicleChassisShape); localTrans.setOrigin(btVector3(3, 0.f, 0)); vehicleMotionState = new btDefaultMotionState(localTrans); //vehicleMotionState = new btDefaultMotionState(btTransform(btQuaternion(0, 0, 0, 1), btVector3(3, 0, 0))); btVector3 vehicleInertia(0, 0, 0); vehicleBody->calculateLocalInertia(vehicleMass, vehicleInertia); btRigidBody::btRigidBodyConstructionInfo vehicleRigidBodyCI(vehicleMass, vehicleMotionState, vehicleBody, vehicleInertia); vehicleRigidBody = new btRigidBody(vehicleRigidBodyCI); dynamicsWorld->addRigidBody(vehicleRigidBody); wheelShape = new btCylinderShapeX(btVector3(wheelWidth, wheelRadius, wheelRadius)); { vehicleRayCaster = new btDefaultVehicleRaycaster(dynamicsWorld); vehicle = new btRaycastVehicle(vehicleTuning, vehicleRigidBody, vehicleRayCaster); // never deactivate vehicle vehicleRigidBody->setActivationState(DISABLE_DEACTIVATION); dynamicsWorld->addVehicle(vehicle); float connectionHeight = 1.2f; bool isFrontWheel = true; vehicle->setCoordinateSystem(rightIndex, upIndex, forwardIndex); // 0, 1, 2 // add wheels // front left btVector3 connectionPointCS0(CUBE_HALF_EXTENT-(0.3*wheelWidth), connectionHeight, 2*CUBE_HALF_EXTENT-wheelRadius); vehicle->addWheel(connectionPointCS0, wheelDirectionCS0, wheelAxleCS, suspensionRestLength, wheelRadius, vehicleTuning, isFrontWheel); // front right connectionPointCS0 = btVector3(-CUBE_HALF_EXTENT+(0.3*wheelWidth), connectionHeight, 2*CUBE_HALF_EXTENT-wheelRadius); vehicle->addWheel(connectionPointCS0, wheelDirectionCS0, wheelAxleCS, suspensionRestLength, wheelRadius, vehicleTuning, isFrontWheel); isFrontWheel = false; // rear right connectionPointCS0 = btVector3(-CUBE_HALF_EXTENT+(0.3*wheelWidth), connectionHeight, -2*CUBE_HALF_EXTENT+wheelRadius); vehicle->addWheel(connectionPointCS0, wheelDirectionCS0, wheelAxleCS, suspensionRestLength, wheelRadius, vehicleTuning, isFrontWheel); // rear left connectionPointCS0 = btVector3(CUBE_HALF_EXTENT-(0.3*wheelWidth), connectionHeight, -2*CUBE_HALF_EXTENT+wheelRadius); vehicle->addWheel(connectionPointCS0, wheelDirectionCS0, wheelAxleCS, suspensionRestLength, wheelRadius, vehicleTuning, isFrontWheel); for (int i = 0; i < vehicle->getNumWheels(); i++) { btWheelInfo& wheel = vehicle->getWheelInfo(i); wheel.m_suspensionStiffness = suspensionStiffness; wheel.m_wheelsDampingRelaxation = suspensionDamping; wheel.m_wheelsDampingCompression = suspensionCompression; wheel.m_frictionSlip = wheelFriction; wheel.m_rollInfluence = rollInfluence; } } /////////////////////////////////////////////////////////////////////// // Orientation and Position of Falling body fallMotionState = new btDefaultMotionState(btTransform(btQuaternion(0, 0, 0, 1), btVector3(-1, 5, 0))); btScalar mass = 1; btVector3 fallInertia(0, 0, 0); fallShape->calculateLocalInertia(mass, fallInertia); btRigidBody::btRigidBodyConstructionInfo fallRigidBodyCI(mass, fallMotionState, fallShape, fallInertia); fallRigidBody = new btRigidBody(fallRigidBodyCI); dynamicsWorld->addRigidBody(fallRigidBody); } I step physics world like this : // does not work vehicle->applyEngineForce(maxEngineForce, WHEEL_REARLEFT); vehicle->applyEngineForce(maxEngineForce, WHEEL_REARRIGHT); // these also do not work vehicle->setBrake(gBreakingForce, WHEEL_REARLEFT); vehicle->setBrake(gBreakingForce, WHEEL_REARRIGHT); // this works vehicle->setSteeringValue(gVehicleSteering, WHEEL_FRONTLEFT); vehicle->setSteeringValue(gVehicleSteering, WHEEL_FRONTRIGHT); dynamicsWorld->stepSimulation(1 / 60.0f, 10); However If I apply brakingForce to all 4 wheels (i.e. including WHEEL_FRONTLEFT and WHEEL_FRONTRIGHT), then my vehicle stops, but keeps sliding/moving forward very very slowly. How do I fix this ?

    Read the article

  • XNA 4: GetData from Texture2D and Set it into Texture3D with specific order

    - by cubrman
    I am trying to convert my color grading 2d lookup texture into 3d LUT. When I simply use: ColorAtlas.GetData(data); ColorAtlas3D.SetData(data); I get this: I tried building my 2d atlass horizontally but it did not helped - the data was messed up in a different way. So my question is how can I influence the order of the data I get from the 2d atlas and how can I properly pass it into my 3d atlas? Update: I know that I can GetData from a specific Rectangular area and put it into several arrays, but the result is still the same. This is what I tried: Color[] data2D = new Color[0]; for (int i = 0; i < 32; i++) { Color[] data = new Color[32 * 32]; GraphicsDevice.SetRenderTarget(null); ColorAtlas.GetData(0, new Rectangle(0, i*32, 32, 32), data, 0, data.Length); int oldLength = data2D.Length; Array.Resize<Color>(ref data2D, oldLength + data.Length); Array.Copy(data, 0, data2D, oldLength, data.Length); } ColorAtlas3D.SetData(data2D);

    Read the article

  • 3d vertex translated onto 2d viewport

    - by Dan Leidal
    I have a spherical world defined by simple trigonometric functions to create triangles that are relatively similar in size and shape throughout. What I want to be able to do is use mouse input to target a range of vertices in the area around the mouse click in order to manipulate these vertices in real time. I read a post on this forum regarding translating 3d world coordinates into the 2d viewport.. it recommended that you should multiply the world vector coordinates by the viewport and then the projection, but they didn't include any code examples, and suffice to say i couldn't get any good results. Further information.. I am using a lookat method for the viewport. Does this cause a problem, and if so is there a solution? If this isn't the problem, does anyone have a simple code example illustrating translating one vertex in a 3d world into a 2d viewspace? I am using XNA.

    Read the article

  • Double sides face with two normals

    - by Marnix
    I think this isn't possible, but I just want to check this: Is it possible to create a face in opengl that has two normals? So: I want the inside and outside of some cilinder to be drawn, but I want the lights to do as expected and not calculate it for the normal given. I was trying to do this with backface culling off, so I would have both faces, but the light was wrongly calculated of course. Is this possible, or do I have to draw an inside and an outside? So draw twice?

    Read the article

  • Why am I not getting an sRGB default framebuffer?

    - by Aaron Rotenberg
    I'm trying to make my OpenGL Haskell program gamma correct by making appropriate use of sRGB framebuffers and textures, but I'm running into issues making the default framebuffer sRGB. Consider the following Haskell program, compiled for 32-bit Windows using GHC and linked against 32-bit freeglut: import Foreign.Marshal.Alloc(alloca) import Foreign.Ptr(Ptr) import Foreign.Storable(Storable, peek) import Graphics.Rendering.OpenGL.Raw import qualified Graphics.UI.GLUT as GLUT import Graphics.UI.GLUT(($=)) main :: IO () main = do (_progName, _args) <- GLUT.getArgsAndInitialize GLUT.initialDisplayMode $= [GLUT.SRGBMode] _window <- GLUT.createWindow "sRGB Test" -- To prove that I actually have freeglut working correctly. -- This will fail at runtime under classic GLUT. GLUT.closeCallback $= Just (return ()) glEnable gl_FRAMEBUFFER_SRGB colorEncoding <- allocaOut $ glGetFramebufferAttachmentParameteriv gl_FRAMEBUFFER gl_FRONT_LEFT gl_FRAMEBUFFER_ATTACHMENT_COLOR_ENCODING print colorEncoding allocaOut :: Storable a => (Ptr a -> IO b) -> IO a allocaOut f = alloca $ \ptr -> do f ptr peek ptr On my desktop (Windows 8 64-bit with a GeForce GTX 760 graphics card) this program outputs 9729, a.k.a. gl_LINEAR, indicating that the default framebuffer is using linear color space, even though I explicitly requested an sRGB window. This is reflected in the rendering results of the actual program I'm trying to write - everything looks washed out because my linear color values aren't being converted to sRGB before being written to the framebuffer. On the other hand, on my laptop (Windows 7 64-bit with an Intel graphics chip), the program prints 0 (huh?) and I get an sRGB default framebuffer by default whether I request one or not! And on both machines, if I manually create a non-default framebuffer bound to an sRGB texture, the program correctly prints 35904, a.k.a. gl_SRGB. Why am I getting different results on different hardware? Am I doing something wrong? How can I get an sRGB framebuffer consistently on all hardware and target OSes?

    Read the article

  • First Minecraft mod not working: make a new sword

    - by yamikoWebs
    I am making my first mod and cannot see what is wrong with it. I am using MCP and Modloader. For my first mod I was going to make swords. I started with making a new EnumToolMaterials WOOD(0, 59, 2.0F, 0, 15), STONE(1, 131, 4.0F, 1, 5), IRON(2, 250, 6.0F, 2, 14), LAPIS(3, 750, 7.0F, 2, 14), OBSIDIAN(3, 1000, 7.5F, 3, 12), EMERALD(3, 1561, 8.0F, 3, 10),//diamond GREEN(3, 2000, 9.0F, 4, 10),//emerald GOLD(0, 200, 12.0F, 0, 22); then here is the mod class public class _Mod_Yamiko extends BaseMod{ /* mod itemts */ public static final Item swordLapis = (new ItemSword(600, EnumToolMaterial.LAPIS)).setItemName("swordLapis"); public static final Item swordObsidian = (new ItemSword(601, EnumToolMaterial.OBSIDIAN)).setItemName("swordObsidian"); public static final Item swordGreen = (new ItemSword(602, EnumToolMaterial.GREEN)).setItemName("swordGreen"); public void load(){ //set images swordLapis.iconIndex = ModLoader.addOverride("/gui/items.png","/gui/swordLapis.png"); ModLoader.addName(swordLapis, "Lapis Sword"); //craft ModLoader.addRecipe(new ItemStack(_Mod_Yamiko.swordLapis, 1), new Object[]{ " X ", " X ", " Y ", 'X', Block.dirt, 'Y', Item.stick }); } public String getVersion(){ return "0.1"; } } Then I made a 16×16 .png image. I am not sure where to save it so I recompiled and reobfuscated, took the mod files and put it in my local Minecraft install, added the image where it be should be. No problems when playing but I cannot make the new sword.

    Read the article

  • Who owns the intellectual property to Fragile Allegiance?

    - by analytik
    Fragile Allegiance was developed by Gremlin Interactive, which was later acquired by Infogrames (Atari). I couldn't find any details of the acquisition though. The only interesting thing I have found online is that the owner of the registered trademark Fragile Allegiance is Interplay, who published Fragile Allegiance. However, the only copyright note I've found was in one installation .ini file, claiming it for Gremlin. What are the common business practices when it comes to old, unused IPs? What do publishers/developers actually need to legally claim an intellectual property? Does anyone have an experience with contacting big publishers with copyright/IP inquiries? Related legal question.

    Read the article

< Previous Page | 329 330 331 332 333 334 335 336 337 338 339 340  | Next Page >