feat: add move_pos and update cuda demo

.github/workflows/workflow.yml

@@ -122,40 +122,41 @@ jobs:
           files: build/coverage/coverage.info
           verbose: true
 
+
+  check_nvidia_gpu:
+    runs-on: ubuntu-latest  # use own runner with NVIDIA GPU
+    outputs:
+      has_gpu: ${{ steps.check_gpu.outputs.has_gpu }}
+    steps:
+      - name: Check for NVIDIA GPU
+        id: check_gpu
+        run: |
+          if lspci | grep -i nvidia; then
+            echo "NVIDIA GPU found."
+            echo "::set-output name=has_gpu::true"
+          else
+            echo "No NVIDIA GPU found."
+            echo "::set-output name=has_gpu::false"
+          fi
+
   build_ubuntu_with_cuda:
-    runs-on: ubuntu-latest
+    runs-on: ubuntu-latest  # use own runner with NVIDIA GPU
+    needs: check_nvidia_gpu
+    if: needs.check_nvidia_gpu.outputs.has_gpu == 'true'
     steps:
       - uses: actions/checkout@v3
 
-      - uses: Jimver/[email protected]
+      - name: Set up CUDA
+        uses: Jimver/[email protected] # https://github.com/marketplace/actions/cuda-toolkit
         with:
-          # Cuda version
-          cuda: 11.7.0 # optional, default is 12.5.0
-          # Only installs specified subpackages, must be in the form of a JSON array. For example, if you only want to install nvcc and visual studio integration: ["nvcc", "visual_studio_integration"] double quotes required! Note that if you want to use this on Linux, 'network' method MUST be used.
-          sub-packages: '[]' # optional, default is []
-          # Only installs specified subpackages that do not have the cuda prefix, must be in the form of a JSON array. For example, if you only want to install libcublas and libcufft: ["libcublas", "libcufft"] double quotes required! Note that this only works with 'network' method on only on Linux.
-          non-cuda-sub-packages: '[]' # optional, default is []
-          # Installation method, can be either 'local' or 'network'. 'local' downloads the entire installer with all packages and runs that (you can still only install certain packages with sub-packages on Windows). 'network' downloads a smaller executable which only downloads necessary packages which you can define in subPackages
-          method: network # optional, default is local
-          # (Linux and 'local' method only) override arguments for the linux .run installer. For example if you don't want samples use ["--toolkit"] double quotes required!
-          linux-local-args: '["--toolkit", "--samples"]' # optional, default is ["--toolkit", "--samples"]
-          # Use GitHub cache to cache downloaded installer on GitHub servers
-          use-github-cache: true # optional, default is true
-          # Use local cache to cache downloaded installer on the local runner
-          use-local-cache: true # optional, default is true
-          # Suffix of log file name in artifact
-          log-file-suffix: log.txt # optional, default is log.txt    
-
-      - name: Check for CUDA
-        id: check_cuda
-        run: |
-          if command -v nvcc &> /dev/null; then
-            echo "CUDA is installed."
-            echo "has_cuda=true" >> $GITHUB_ENV
-          else
-            echo "CUDA is not installed."
-            echo "has_cuda=false" >> $GITHUB_ENV
-          fi
+          cuda: 11.7.0
+          sub-packages: '[]'
+          non-cuda-sub-packages: '[]'
+          method: network
+          linux-local-args: '["--toolkit", "--samples"]'
+          use-github-cache: true
+          use-local-cache: true
+          log-file-suffix: log.txt
 
       - name: Install dependencies
         run: |

test/system_test/MassSpring3D/ClothSimulation.cpp

@@ -22,8 +22,16 @@ const ClothSimulation::ClothMesh ClothSimulation::GetClothMesh() const {
     return _cloth_mesh;
 };
 
-void ClothSimulation::move(const dtk::dtkDouble2& v) {
+void ClothSimulation::move(const dtk::dtkDouble3& v) {
+    // Update camera position
+    g_camera_position += glm::vec3(v.x, v.y, v.z);
 
+    // Recalculate the ModelView matrix
+    g_ModelViewMatrix = glm::lookAt(
+        g_camera_position,
+        g_camera_target,
+        g_camera_up
+    ) * glm::translate(glm::mat4(1), glm::vec3(0.0f, 0.0f, _param.w / 4));
 };
 
 void ClothSimulation::CleanUp() {
@@ -104,11 +112,15 @@ void ClothSimulation::InitCloth() {
 }
 
 void ClothSimulation::InitScene() {
+    g_camera_position = glm::vec3(0.618, -0.786, 0.3f) * g_camera_distance;
+    g_camera_target = glm::vec3(0.0f, 0.0f, -1.0f);
+    g_camera_up = glm::vec3(0.0f, 0.0f, 1.0f);
+
     g_ModelViewMatrix = glm::lookAt(
-        glm::vec3(0.618, -0.786, 0.3f) * g_camera_distance,
-        glm::vec3(0.0f, 0.0f, -1.0f),
-        glm::vec3(0.0f, 0.0f, 1.0f)
-    ) * glm::translate(glm::mat4(1), glm::vec3(0.0f, 0.0f, _param.w / 4));
+        g_camera_position,
+        g_camera_position + g_camera_target,
+        g_camera_up
+    );
     g_ProjectionMatrix = glm::perspective(PI / 4.0f, g_windowWidth * 1.0f / g_windowHeight, 0.01f, 1000.0f);
 };
 
@@ -143,9 +155,6 @@ void ClothSimulation::UpdateRenderTarget() {
     const std::vector<dtk::dtkDouble3>& normalData = _cloth_mesh->GetVertexNormals();
     float* normalBuffer = new float[vertexBufferSize];
     for (dtk::dtkID i = 0;i < mPts->GetNumberOfPoints();i++) {
-        // vertexBuffer[3 * i + 0] = (float)mPts->GetPoint(i)[0];
-        // vertexBuffer[3 * i + 1] = (float)mPts->GetPoint(i)[1];
-        // vertexBuffer[3 * i + 2] = (float)mPts->GetPoint(i)[2];
         normalBuffer[3 * i + 0] = (float)normalData[i].x;
         normalBuffer[3 * i + 1] = (float)normalData[i].y;
         normalBuffer[3 * i + 2] = (float)normalData[i].z;

test/system_test/MassSpring3D/dtkPhysMassSpringSolver.cpp

@@ -227,10 +227,13 @@ void dtk::dtkPhysMassSpringSolver::step() {
 
 
 void dtk::dtkPhysMassSpringSolver::satisfy(ClothDropType type) {
-    if (type == Sphere) {
-        const float radius = 0.64f;
-        const Eigen::Vector3f center(0, 0, -1);
-
+    auto satisfySphere = [&](const float& radius, const Eigen::Vector3f& center) {
+#ifdef DTK_CUDA
+        int num_points = _system->GetNumberOfMassPoints();
+        size_t size = num_points * 3 * sizeof(float);
+        run_satisfy_sphere_with_cuda(d_current_state, num_points, radius, make_float3(center[0], center[1], center[2]));
+        cudaMemcpy(_current_state.data(), d_current_state, size, cudaMemcpyDeviceToHost);
+#else
         for (int i = 0; i < _system->GetNumberOfMassPoints(); i++) {
             Vector3f p(
                 _current_state[3 * i + 0] - center[0],
@@ -248,10 +251,12 @@ void dtk::dtkPhysMassSpringSolver::satisfy(ClothDropType type) {
                 _current_state[3 * i + j] = p[j] + center[j];
             }
         }
-    }
-    else {
-        int fix_idx = 0;
-        for (int i = 0; i < 3; i++)
-            _current_state[fix_idx + i] = _initial_state[fix_idx + i];
+#endif
+        };
+
+
+    if (type == Sphere) {
+        satisfySphere(0.64f, Eigen::Vector3f(-0.5, 0, -1));
+        satisfySphere(0.64f, Eigen::Vector3f(0, 0.5, -2));
     }
 }

test/system_test/MassSpring3D/include/ClothSimulation.h

@@ -70,7 +70,8 @@ class ClothSimulation : public Scene {
 
     const ClothMesh GetClothMesh() const;
 
-    void move(const dtk::dtkDouble2& v);
+    void move(const dtk::dtkDouble3& v);
+    void rotate_view(const float& deltaX, const float& deltaY);
 
     void CleanUp();
 
@@ -95,6 +96,9 @@ class ClothSimulation : public Scene {
     // Camera
     dtk::dtkMatrix44 g_ModelViewMatrix;
     dtk::dtkMatrix44 g_ProjectionMatrix;
+    glm::vec3 g_camera_position;
+    glm::vec3 g_camera_target;
+    glm::vec3 g_camera_up;
     const float g_camera_distance = 4.2f;
     const float PI = glm::pi<float>();
 

test/system_test/MassSpring3D/include/Scene.h

@@ -14,7 +14,7 @@ class Scene {
     virtual void Update(float dt) = 0;
     virtual void Render() = 0;
     virtual void CleanUp() = 0;
-    virtual void move(const dtk::dtkDouble2& v) = 0;
+    virtual void move(const dtk::dtkDouble3& v) = 0;
 
     bool IsPause() const {
         return State == SCENE_PAUSE;

test/system_test/MassSpring3D/include/step.cuh

@@ -7,4 +7,10 @@ void run_local_step_with_cuda(
     const float* d_rest_lengths,                                       
     int num_springs
 );
+void run_satisfy_sphere_with_cuda(
+    float* d_current_state, 
+    int num_points, 
+    float radius, 
+    float3 center
+);
 #endif // __LOCALSTEP_CUH__

test/system_test/MassSpring3D/main.cpp

@@ -99,6 +99,7 @@ void display() {
     draw_text(5, 40, "Push [1-1] to switch scene");
     draw_text(w - 150, h - 20, "refer: apollonia");
 
+    draw_text(5, h - 60, "Mass Point: %d", I1_EDGE_NUM * I1_EDGE_NUM);
     draw_text(5, h - 40, "%.2f FPS", fps);
     if (current_scene->IsPause())
         draw_text(5, h - 20, "dt: %.2f ms PAUSED", dt * 1000);
@@ -118,26 +119,37 @@ void reshape(int width, int height) {
     gluPerspective(45.0, width / (float)height, 0.1, 100.0);
 }
 
-void mouse(int button, int state, int x, int y) {}
-
-void move_pos(const dtk::dtkDouble2& v) { current_scene->move(v); }
+void move_pos(const dtk::dtkDouble3& v) {
+    if (current_scene == nullptr) {
+        return;
+    }
+    current_scene->move(v);
+}
 
 void keyboard(unsigned char key, int x, int y) {
+    const float moveSpeed = 0.1f;
     switch (key) {
     case '1':
         current_scene = new ClothSimulation(WINDOW_WIDTH, WINDOW_HEIGHT, { 0, -9.8 }, I1_EDGE_NUM);
+        move_pos(dtk::dtkDouble3(0, 0, 0));
         break;
     case 'w':
-        move_pos(dtk::dtkDouble2(0, 1));
+        move_pos(dtk::dtkDouble3(0, moveSpeed, 0)); // Move forward
         break;
     case 'a':
-        move_pos(dtk::dtkDouble2(-1, 0));
+        move_pos(dtk::dtkDouble3(-moveSpeed, 0, 0)); // Move left
         break;
     case 's':
-        move_pos(dtk::dtkDouble2(0, -1));
+        move_pos(dtk::dtkDouble3(0, -moveSpeed, 0)); // Move backward
         break;
     case 'd':
-        move_pos(dtk::dtkDouble2(1, 0));
+        move_pos(dtk::dtkDouble3(moveSpeed, 0, 0)); // Move right
+        break;
+    case 'q':
+        move_pos(dtk::dtkDouble3(0, 0, moveSpeed)); // Move up
+        break;
+    case 'e':
+        move_pos(dtk::dtkDouble3(0, 0, -moveSpeed)); // Move down
         break;
     case ' ':
         current_scene->SetPause(!current_scene->IsPause());
@@ -150,6 +162,8 @@ void keyboard(unsigned char key, int x, int y) {
     }
 }
 
+void mouse(int button, int state, int x, int y) {}
+
 void motion(int x, int y) {}
 
 void special(int key, int x, int y) {}

test/system_test/MassSpring3D/step.cu

@@ -41,5 +41,39 @@ void run_local_step_with_cuda(
     );
 
     // Wait for all threads to complete
+    cudaDeviceSynchronize();
+}
+
+extern "C" __global__ void satisfySphereKernel(float* current_state, int num_points, float radius, float3 center) {
+    int i = blockIdx.x * blockDim.x + threadIdx.x;
+    if (i < num_points) {
+        float3 p = make_float3(
+            current_state[3 * i + 0] - center.x,
+            current_state[3 * i + 1] - center.y,
+            current_state[3 * i + 2] - center.z
+        );
+
+        float norm = sqrtf(p.x * p.x + p.y * p.y + p.z * p.z);
+        if (norm < radius) {
+            p.x /= norm;
+            p.y /= norm;
+            p.z /= norm;
+            p.x *= radius;
+            p.y *= radius;
+            p.z *= radius;
+
+            current_state[3 * i + 0] = p.x + center.x;
+            current_state[3 * i + 1] = p.y + center.y;
+            current_state[3 * i + 2] = p.z + center.z;
+        }
+    }
+}
+
+void run_satisfy_sphere_with_cuda(float* d_current_state, int num_points, float radius, float3 center) {
+    int blockSize = 256;
+    int numBlocks = (num_points + blockSize - 1) / blockSize;
+
+    satisfySphereKernel<<<numBlocks, blockSize>>>(d_current_state, num_points, radius, center);
+
     cudaDeviceSynchronize();
 }