bool SimpleSkeletalAnimatedObject_cl::GetHeadRotation(const hkvVec3 &vTargetPos, hkvQuat &targetRot, float fMaxViewAngle)
{
hkvMat4 objectMatrix;
hkvMat4 parentObjectMatrix;
// get the index of the parent bone
int parentBoneIndex = GetMesh()->GetSkeleton()->GetBone(m_iHeadBoneIndex)->m_iParentIndex;
// get the 4x4 matrix of the head bone and its parent in object space
const VisSkeletalAnimResult_cl* pObjectSpaceResult = GetAnimConfig()->GetFinalResult()->GetCurrentObjectSpaceResult();
pObjectSpaceResult->GetBoneTransformationMatrix(m_iHeadBoneIndex, objectMatrix);
pObjectSpaceResult->GetBoneTransformationMatrix(parentBoneIndex, parentObjectMatrix);
// get the current position of the target in object space
hkvVec3 targetObjectspace = TransformToObjectSpace(vTargetPos);
// get the current position of the bone in object space
hkvVec3 boneObjectspace = objectMatrix.getTranslation();
// calculate the look at direction in object space (just for the angle constraints)
hkvVec3 vLookAtDir = targetObjectspace - boneObjectspace;
// check if we are within the angle constraints
vLookAtDir.normalizeIfNotZero();
float fCosMax = hkvMath::cosDeg (180.f - fMaxViewAngle);
bool bInsideFocusArea = vLookAtDir.y < fCosMax;
if (!bInsideFocusArea)
{
// out of focus, so just look straight ahead
targetRot.setIdentity();
return false;
}
// create a LookAt-3x3 matrix from the bone position and the target position (both are in object space)
// This orientation is later used to replace(!) the head bone in local space (VIS_REPLACE_BONE|VIS_LOCAL_SPACE flags)
hkvMat3 lookatMatrix(hkvNoInitialization);
lookatMatrix.setLookInDirectionMatrix(targetObjectspace - boneObjectspace);
// Apply our custom rotation matrix that tells the bone to focus with the eyes and not with the back of the head :-)
lookatMatrix = lookatMatrix.multiply(m_RelativeHeadOrientation);
// Invert the parents object matrix rotation to cancel out all parent bone orientations
hkvMat3 parentObject = parentObjectMatrix.getRotationalPart();
parentObject.invert();
// put everything together
hkvMat3 result = parentObject;
result = result.multiply (lookatMatrix);
// ...and make a quaternion from it
targetRot.setFromMat3 (result);
return true;
}
// Draw the scene to the screen
void draw() {
// Update time
time += 2.0f * timeStep;
/////////////////////// PART 1: SIMULATION /////////////////////////////////
// Grab buffers for OpenCL
acquireGLBuffer(particles.particleBuffer[particles.currentBuffer]);
acquireGLBuffer(particles.particleBuffer[1 - particles.currentBuffer]);
// Prepare to run some kernels
cl_int numParticles = NUM_PARTICLES;
cl_int gridElements = GRID_SIZE * GRID_SIZE * GRID_SIZE;
cl_uint workSize[3] = {numParticles, 0, 0};
cl_uint gridWorkSize[3] = {gridElements, 0, 0};
cl_uint workgroupSize[3] = {256, 0, 0};
// Clear grid
clSetKernelArg(openCLKernels.gridClearKernel, 0, sizeof(cl_mem), &particles.gridSizeBuffer);
clSetKernelArg(openCLKernels.gridClearKernel, 1, sizeof(cl_int), &gridElements);
clRunKernel(openCLKernels.gridClearKernel, gridWorkSize, workgroupSize);
// Compute grid positions
clSetKernelArg(openCLKernels.gridKernel, 0, sizeof(cl_mem), &particles.particleBuffer[particles.currentBuffer]);
clSetKernelArg(openCLKernels.gridKernel, 1, sizeof(cl_mem), &particles.offsetBuffer);
clSetKernelArg(openCLKernels.gridKernel, 2, sizeof(cl_mem), &particles.gridSizeBuffer);
clSetKernelArg(openCLKernels.gridKernel, 3, sizeof(cl_int), &numParticles);
clRunKernel(openCLKernels.gridKernel, workSize, workgroupSize);
// Compute prefix sum for grid
clSetKernelArg(openCLKernels.prefixSumKernel, 2, sizeof(cl_uint), (void*)&gridElements);
int pingpong = 0;
for(cl_int offset = 1; offset <= gridElements; offset *= 2) {
if(offset == 1) {
clSetKernelArg(openCLKernels.prefixSumKernel, 0, sizeof(cl_mem), (void*)&particles.gridSizeBuffer);
}
else {
clSetKernelArg(openCLKernels.prefixSumKernel, 0, sizeof(cl_mem), (void*)&(pingpong == 0 ? particles.gridBuffer[0] : particles.gridBuffer[1]));
}
clSetKernelArg(openCLKernels.prefixSumKernel, 1, sizeof(cl_mem), (void*)&(pingpong == 0 ? particles.gridBuffer[1] : particles.gridBuffer[0]));
clSetKernelArg(openCLKernels.prefixSumKernel, 3, sizeof(cl_int), (void*)&offset);
clRunKernel(openCLKernels.prefixSumKernel, gridWorkSize, workgroupSize);
pingpong = 1 - pingpong;
}
// Reorganize particles
clSetKernelArg(openCLKernels.gridReorderKernel, 0, sizeof(cl_mem), &particles.particleBuffer[particles.currentBuffer]);
clSetKernelArg(openCLKernels.gridReorderKernel, 1, sizeof(cl_mem), &particles.particleBuffer[1 - particles.currentBuffer]);
clSetKernelArg(openCLKernels.gridReorderKernel, 2, sizeof(cl_mem), &particles.velocityBuffer[particles.currentBuffer]);
clSetKernelArg(openCLKernels.gridReorderKernel, 3, sizeof(cl_mem), &particles.velocityBuffer[1 - particles.currentBuffer]);
clSetKernelArg(openCLKernels.gridReorderKernel, 4, sizeof(cl_mem), &particles.offsetBuffer);
clSetKernelArg(openCLKernels.gridReorderKernel, 5, sizeof(cl_mem), (void*)&particles.gridSizeBuffer);
clSetKernelArg(openCLKernels.gridReorderKernel, 6, sizeof(cl_mem), (void*)&(pingpong == 0 ? particles.gridBuffer[0] : particles.gridBuffer[1]));
clSetKernelArg(openCLKernels.gridReorderKernel, 7, sizeof(cl_int), &numParticles);
clRunKernel(openCLKernels.gridReorderKernel, workSize, workgroupSize);
particle* testData = (particle*)malloc(sizeof(cl_float) * numParticles * 4);
// Swap particle buffers
particles.currentBuffer = 1 - particles.currentBuffer;
// Send new cell select buffer
int cellSelect[27];
for(int i = 0; i < 27; i++) {
cellSelect[i] = i;
}
shuffle(cellSelect, 27);
clEnqueueWriteBuffer(clCommandQueue(), particles.cellSelectBuffer, true, 0, 27 * sizeof(cl_int), cellSelect, 0, 0, 0);
clFinish(clCommandQueue());
// Recalculate densities and normalized pressure derivatives
clSetKernelArg(openCLKernels.dataKernel, 0, sizeof(cl_mem), &particles.particleBuffer[particles.currentBuffer]);
clSetKernelArg(openCLKernels.dataKernel, 1, sizeof(cl_mem), &particles.dataBuffer);
clSetKernelArg(openCLKernels.dataKernel, 2, sizeof(cl_mem), (void*)&particles.gridSizeBuffer);
clSetKernelArg(openCLKernels.dataKernel, 3, sizeof(cl_mem), (void*)&(pingpong == 0 ? particles.gridBuffer[1] : particles.gridBuffer[0]));
clSetKernelArg(openCLKernels.dataKernel, 4, sizeof(cl_mem), (void*)&particles.cellSelectBuffer);
clSetKernelArg(openCLKernels.dataKernel, 5, sizeof(cl_int), &numParticles);
clRunKernel(openCLKernels.dataKernel, workSize, workgroupSize);
// Send new cell select buffer
cellSelect[27];
for(int i = 0; i < 27; i++) {
cellSelect[i] = i;
}
shuffle(cellSelect, 27);
clEnqueueWriteBuffer(clCommandQueue(), particles.cellSelectBuffer, true, 0, 27 * sizeof(cl_int), cellSelect, 0, 0, 0);
clFinish(clCommandQueue());
// Integrate position
float dT = timeStep;
clSetKernelArg(openCLKernels.simulationKernel, 0, sizeof(cl_mem), &particles.particleBuffer[particles.currentBuffer]);
clSetKernelArg(openCLKernels.simulationKernel, 1, sizeof(cl_mem), &particles.particleBuffer[1 - particles.currentBuffer]);
clSetKernelArg(openCLKernels.simulationKernel, 2, sizeof(cl_mem), &particles.velocityBuffer[particles.currentBuffer]);
clSetKernelArg(openCLKernels.simulationKernel, 3, sizeof(cl_mem), &particles.velocityBuffer[1 - particles.currentBuffer]);
clSetKernelArg(openCLKernels.simulationKernel, 4, sizeof(cl_mem), &particles.dataBuffer);
clSetKernelArg(openCLKernels.simulationKernel, 5, sizeof(cl_mem), (void*)&particles.gridSizeBuffer);
clSetKernelArg(openCLKernels.simulationKernel, 6, sizeof(cl_mem), (void*)&(pingpong == 0 ? particles.gridBuffer[1] : particles.gridBuffer[0]));
clSetKernelArg(openCLKernels.simulationKernel, 7, sizeof(cl_mem), (void*)&particles.cellSelectBuffer);
clSetKernelArg(openCLKernels.simulationKernel, 8, sizeof(cl_float), &dT);
clSetKernelArg(openCLKernels.simulationKernel, 9, sizeof(cl_float), &time);
clSetKernelArg(openCLKernels.simulationKernel, 10, sizeof(cl_int), &numParticles);
clSetKernelArg(openCLKernels.simulationKernel, 11, sizeof(cl_mem), &particles.terrainBuffer);
PaddedVector windDir = PadVector(
TransformVector(YAxisRotationMatrix(particles.windAngle), MakeVector(1.0f, 0.0f, 0.0f))
);
clSetKernelArg(openCLKernels.simulationKernel, 12, sizeof(cl_float) * 4, &windDir);
clSetKernelArg(openCLKernels.simulationKernel, 13, sizeof(cl_float), &particles.windPower);
clRunKernel(openCLKernels.simulationKernel, workSize, workgroupSize);
// Release buffers back to OpenGL
releaseGLBuffer(particles.particleBuffer[particles.currentBuffer]);
releaseGLBuffer(particles.particleBuffer[1 - particles.currentBuffer]);
// Swap particle buffers
particles.currentBuffer = 1 - particles.currentBuffer;
//////////////////////// PART 2: RENDERIING ////////////////////////////////
// Clear everything first thing.
glClearColor(0.0f, 0.0f, 0.0f, 1.0f);
glBindFramebuffer(GL_FRAMEBUFFER, framebuffers.backgroundFBO);
glClear(GL_COLOR_BUFFER_BIT | GL_DEPTH_BUFFER_BIT);
// Activate shader.
glUseProgram(objectShader.shaderProgram);
// Set projection
Matrix projection = PerspectiveMatrix(
45.0f,
(float)WINDOW_WIDTH/(float)WINDOW_HEIGHT,
0.01f,
100.0f
);
MatrixAsUniform(objectShader.projectionMatrix, projection);
// Vertices
glBindBuffer(GL_ARRAY_BUFFER, terrain.vertexBuffer);
glVertexAttribPointer(
objectShader.vertexPosition,
4,
GL_FLOAT,
GL_FALSE,
sizeof(GLfloat) * 4,
(void*)0
);
glEnableVertexAttribArray(objectShader.vertexPosition);
// Set modelview according to camera
Matrix rot = lookatMatrix(camera.pos, VectorAdd(camera.pos, camera.front), camera.up);
rot = MatrixMul(RotationMatrix(camera.elevation, MakeVector(1, 0, 0)), rot);
Matrix trans = TranslationMatrix(-camera.pos.x, -camera.pos.y, -camera.pos.z);
Matrix modelview = MatrixMul(rot, trans);
MatrixAsUniform(objectShader.modelviewMatrix, modelview);
// Normal view matrix - inverse transpose of modelview.
Matrix normalview = MatrixTranspose(FastMatrixInverse(modelview));
MatrixAsUniform(objectShader.normalviewMatrix, normalview);
// Set heightmap texture
glActiveTexture(GL_TEXTURE0);
glBindTexture(GL_TEXTURE_2D, terrain.heightTexture );
glUniform1i(objectShader.terrainTexture, 0);
// Set color textures
glActiveTexture(GL_TEXTURE1);
glBindTexture(GL_TEXTURE_2D, terrain.lowTexture);
glUniform1i(objectShader.lowTexture, 1);
glActiveTexture(GL_TEXTURE2);
glBindTexture(GL_TEXTURE_2D, terrain.highTexture);
glUniform1i(objectShader.highTexture, 2);
// Turn off culling
glDisable(GL_CULL_FACE);
// Send element buffer to GPU and draw.
glBindBuffer(GL_ELEMENT_ARRAY_BUFFER, terrain.elementBuffer);
glDrawElements(
GL_TRIANGLES,
512 * 512 * 6,
GL_UNSIGNED_INT,
(void*)0
);
// Turn culling back on
glEnable(GL_CULL_FACE);
// Switch to low-res viewport
glViewport(0, 0, WINDOW_WIDTH / RESOLUTION_DIVIDER, WINDOW_HEIGHT / RESOLUTION_DIVIDER);
// Low-res projection matrix
Matrix projectionLowres = PerspectiveMatrix(
45.0f,
(float)(WINDOW_WIDTH / RESOLUTION_DIVIDER) / (float)(WINDOW_HEIGHT / RESOLUTION_DIVIDER),
0.01f,
100.0f
);
// Activate particle depth FBO
glClearColor(0.0f, 0.0f, 0.0f, 0.0f);
glBindFramebuffer(GL_FRAMEBUFFER, framebuffers.particleFBO[0]);
glClear(GL_COLOR_BUFFER_BIT | GL_DEPTH_BUFFER_BIT);
// Activate shader
glUseProgram(particleShader.shaderProgram);
// Set textures
glActiveTexture(GL_TEXTURE0);
glBindTexture(GL_TEXTURE_2D, framebuffers.backgroundTexture);
glUniform1i(particleShader.terrainTexture, 0);
// Send uniforms
MatrixAsUniform(particleShader.modelviewMatrix, modelview);
MatrixAsUniform(particleShader.projectionMatrix, projectionLowres);
glUniform2f(particleShader.screenSize, WINDOW_WIDTH / RESOLUTION_DIVIDER, WINDOW_HEIGHT / RESOLUTION_DIVIDER);
// Bind new buffer and set up arrtibutes
glBindBuffer(GL_ARRAY_BUFFER, particles.vertexBuffer[particles.currentBuffer]);
glVertexAttribPointer(
particleShader.vertexPosition,
4,
GL_FLOAT,
GL_FALSE,
sizeof(GLfloat) * 4,
(void*)0
);
glEnableVertexAttribArray(particleShader.vertexPosition);
// Bind element buffer and draw particles
glBindBuffer(GL_ELEMENT_ARRAY_BUFFER, particles.elementBuffer);
glDrawElements(
GL_POINTS,
NUM_PARTICLES,
GL_UNSIGNED_INT,
(void*)0
);
// Activate particle thickness FBO
glBindFramebuffer(GL_FRAMEBUFFER, framebuffers.particleThicknessFBO[0]);
glClear(GL_COLOR_BUFFER_BIT | GL_DEPTH_BUFFER_BIT);
// Activate shader
glUseProgram(particleThicknessShader.shaderProgram);
// Send uniforms
MatrixAsUniform(particleThicknessShader.modelviewMatrix, modelview);
MatrixAsUniform(particleThicknessShader.projectionMatrix, projectionLowres);
glUniform2f(particleThicknessShader.screenSize, WINDOW_WIDTH / RESOLUTION_DIVIDER, WINDOW_HEIGHT / RESOLUTION_DIVIDER);
// Set textures
glActiveTexture(GL_TEXTURE0);
glBindTexture(GL_TEXTURE_2D, framebuffers.backgroundTexture);
glUniform1i(particleThicknessShader.terrainTexture, 0);
// Bind new buffer and set up arrtibutes
glBindBuffer(GL_ARRAY_BUFFER, particles.vertexBuffer[particles.currentBuffer]);
glVertexAttribPointer(
particleThicknessShader.vertexPosition,
4,
GL_FLOAT,
GL_FALSE,
sizeof(GLfloat) * 4,
(void*)0
);
glEnableVertexAttribArray(particleThicknessShader.vertexPosition);
// Enable additive blending and disable depth test
glEnable(GL_BLEND);
glBlendFunc(GL_ONE, GL_ONE);
glDisable(GL_DEPTH_TEST);
// Bind element buffer and draw particles, this time rendering thickness map
glBindBuffer(GL_ELEMENT_ARRAY_BUFFER, particles.elementBuffer);
glDrawElements(
GL_POINTS,
NUM_PARTICLES,
GL_UNSIGNED_INT,
(void*)0
);
// Turn blending back off and depth test back on
glDisable(GL_BLEND);
glEnable(GL_DEPTH_TEST);
// Activate particle velocity FBO
glBindFramebuffer(GL_FRAMEBUFFER, framebuffers.velocityFBO);
glClear(GL_COLOR_BUFFER_BIT | GL_DEPTH_BUFFER_BIT);
// Activate shader
glUseProgram(particleVelocityShader.shaderProgram);
// Send uniforms
MatrixAsUniform(particleVelocityShader.modelviewMatrix, modelview);
MatrixAsUniform(particleVelocityShader.projectionMatrix, projectionLowres);
glUniform2f(particleVelocityShader.screenSize, WINDOW_WIDTH / RESOLUTION_DIVIDER, WINDOW_HEIGHT / RESOLUTION_DIVIDER);
// Bind new buffer and set up arrtibutes
glBindBuffer(GL_ARRAY_BUFFER, particles.vertexBuffer[particles.currentBuffer]);
glVertexAttribPointer(
particleVelocityShader.vertexPosition,
4,
GL_FLOAT,
GL_FALSE,
sizeof(GLfloat) * 4,
(void*)0
);
glEnableVertexAttribArray(particleVelocityShader.vertexPosition);
// Bind element buffer and draw particles, this time rendering velocity map
glBindBuffer(GL_ELEMENT_ARRAY_BUFFER, particles.elementBuffer);
glDrawElements(
GL_POINTS,
NUM_PARTICLES,
GL_UNSIGNED_INT,
(void*)0
);
// Curvature flow smoothing begins
glUseProgram(curvatureFlowShader.shaderProgram);
// Send uniforms
glUniform1i(curvatureFlowShader.particleTexture, 0);
glUniform2f(curvatureFlowShader.screenSize, WINDOW_WIDTH / RESOLUTION_DIVIDER, WINDOW_HEIGHT / RESOLUTION_DIVIDER);
MatrixAsUniform(curvatureFlowShader.projectionMatrix, projectionLowres);
// Prepare state
glActiveTexture(GL_TEXTURE0);
glBindBuffer(GL_ARRAY_BUFFER, screenQuad.vertexBuffer);
glVertexAttribPointer(
curvatureFlowShader.vertexPosition,
3,
GL_FLOAT,
GL_FALSE,
sizeof(GLfloat) * 3,
(void*)0
);
glEnableVertexAttribArray(curvatureFlowShader.vertexPosition);
glBindBuffer(GL_ELEMENT_ARRAY_BUFFER, screenQuad.elementBuffer);
// Smoothing loop
glDisable(GL_DEPTH_TEST);
pingpong = 0;
for(int i = 0; i < smoothingIterations; i++) {
// Bind no FBO
glBindFramebuffer(GL_FRAMEBUFFER, 0);
// Bind texture
glBindTexture(GL_TEXTURE_2D, framebuffers.particleTexture[pingpong]);
// Activate proper FBO and clear
glBindFramebuffer(GL_FRAMEBUFFER, framebuffers.particleFBO[1 - pingpong]);
// Draw a quad
glDrawElements(
GL_TRIANGLES,
6,
GL_UNSIGNED_INT,
(void*)0
);
// Switch buffers
pingpong = 1 - pingpong;
}
glEnable(GL_DEPTH_TEST);
// Activate particle color FBO
glBindFramebuffer(GL_FRAMEBUFFER, framebuffers.particleColorFBO);
glClear(GL_COLOR_BUFFER_BIT | GL_DEPTH_BUFFER_BIT);
// Liquid shading shader
glUseProgram(liquidShadeShader.shaderProgram);
// Bind and set textures
glActiveTexture(GL_TEXTURE0);
glBindTexture(GL_TEXTURE_2D, framebuffers.particleTexture[0]);
glUniform1i(liquidShadeShader.particleTexture, 0);
glActiveTexture(GL_TEXTURE1);
glBindTexture(GL_TEXTURE_2D, framebuffers.particleThicknessTexture[0]);
glUniform1i(liquidShadeShader.particleThicknessTexture, 1);
glActiveTexture(GL_TEXTURE2);
glBindTexture(GL_TEXTURE_2D, terrain.envTexture);
glUniform1i(liquidShadeShader.environmentTexture, 2);
glActiveTexture(GL_TEXTURE3);
glBindTexture(GL_TEXTURE_2D, framebuffers.particleVelocityTexture);
glUniform1i(liquidShadeShader.velocityTexture, 3);
// Send uniforms
glUniform2f(liquidShadeShader.screenSize, WINDOW_WIDTH, WINDOW_HEIGHT);
MatrixAsUniform(liquidShadeShader.modelviewMatrix, modelview);
MatrixAsUniform(liquidShadeShader.projectionMatrix, projection);
glUniform1i(liquidShadeShader.useThickness, useThickness);
// Draw a quad
glDisable(GL_DEPTH_TEST);
glBindBuffer(GL_ARRAY_BUFFER, screenQuad.vertexBuffer);
glVertexAttribPointer(
liquidShadeShader.vertexPosition,
3,
GL_FLOAT,
GL_FALSE,
sizeof(GLfloat) * 3,
(void*)0
);
glEnableVertexAttribArray(liquidShadeShader.vertexPosition);
glBindBuffer(GL_ELEMENT_ARRAY_BUFFER, screenQuad.elementBuffer);
glDrawElements(
GL_TRIANGLES,
6,
GL_UNSIGNED_INT,
(void*)0
);
glEnable(GL_DEPTH_TEST);
// Switch back to full-res viewport
glViewport(0, 0, WINDOW_WIDTH, WINDOW_HEIGHT);
// Deactivate FBOs
glBindFramebuffer(GL_FRAMEBUFFER, 0);
glClear(GL_COLOR_BUFFER_BIT | GL_DEPTH_BUFFER_BIT);
// Compose shader
glUseProgram(compositionShader.shaderProgram);
// Bind and set textures
glActiveTexture(GL_TEXTURE0);
glBindTexture(GL_TEXTURE_2D, terrain.envTexture);
glUniform1i(compositionShader.backgroundTexture, 0);
glActiveTexture(GL_TEXTURE1);
glBindTexture(GL_TEXTURE_2D, framebuffers.particleColorTexture);
glUniform1i(compositionShader.particleTexture, 1);
glActiveTexture(GL_TEXTURE2);
glBindTexture(GL_TEXTURE_2D, framebuffers.backgroundTexture);
glUniform1i(compositionShader.terrainTexture, 2);
// Send uniforms
MatrixAsUniform(compositionShader.modelviewMatrix, modelview);
// Draw a quad
glDisable(GL_DEPTH_TEST);
glBindBuffer(GL_ARRAY_BUFFER, screenQuad.vertexBuffer);
glVertexAttribPointer(
compositionShader.vertexPosition,
3,
GL_FLOAT,
GL_FALSE,
sizeof(GLfloat) * 3,
(void*)0
);
glEnableVertexAttribArray(compositionShader.vertexPosition);
glBindBuffer(GL_ELEMENT_ARRAY_BUFFER, screenQuad.elementBuffer);
glDrawElements(
GL_TRIANGLES,
6,
GL_UNSIGNED_INT,
(void*)0
);
glEnable(GL_DEPTH_TEST);
// Switch drawing area and displayed area.
glutSwapBuffers();
}