My project uses DirectX 10 and some of its boilerplate to render a scene, however, it crashes with an error message "Could not initialize the model object." As far as I understand, making it up to this point means that, at the very least, the model has been successfully created, so the error must be in one of the files below, which is fortunate as the most difficult tasks are handled by the FallBodyClass.cpp that hosts OpenCL API interactions. If needed, I can try attaching parts of it in a later edit.
During debug, my IDE shows that all components of m_Model (m_vertexBuffer, m_indexBuffer etc) are shown as with _vfptr . I do not know what to make of it, but it does seem to confirm that modelclass.cpp is the point of failure.
graphicsclass.cpp
GraphicsClass::GraphicsClass()
{
m_Direct3D = 0;
m_Model = 0;
m_ColorShader = 0;
m_bodies = BODIES;
}
GraphicsClass::GraphicsClass(const GraphicsClass& other)
{}
GraphicsClass::~GraphicsClass()
{}
bool GraphicsClass::Initialize(int screenWidth, int screenHeight, HWND hwnd)
{
bool result;
// Create the Direct3D object.
m_Direct3D = new D3DClass;
if (!m_Direct3D)
{
return false;
}
// Initialize the Direct3D object.
result = m_Direct3D->Initialize(screenWidth, screenHeight, VSYNC_ENABLED, hwnd, FULL_SCREEN, SCREEN_DEPTH, SCREEN_NEAR);
if (!result)
{
MessageBox(hwnd, L"Could not initialize Direct3D", L"Error", MB_OK);
return false;
}
// Create the model object.
m_Model = new ModelClass(m_bodies);
if (!m_Model)
{
return false;
}
// Initialize the model object.
result = m_Model->Initialize(m_Direct3D->GetDevice());
if (!result)
{
MessageBox(hwnd, L"Could not initialize the model object.", L"Error", MB_OK);
return false;
}
modelclass.cpp
ModelClass::ModelClass(int bodies)
{
m_vertexBuffer = 0;
m_indexBuffer = 0;
m_positions = 0;
m_velocities = 0;
m_bodySystem = 0;
m_bodies = bodies;
}
ModelClass::ModelClass(const ModelClass& other)
{}
ModelClass::~ModelClass()
{}
bool ModelClass::Initialize(ID3D10Device* device)
{
bool result;
TwoLines twoLinesConstants = CalculateLinesConstants(M_PI_4);
m_positions = new float[COORD_DIM * m_bodies];
m_velocities = new float[VEL_DIM * m_bodies];
m_bodySystem = new class FallBodyClass(m_bodies, &m_positions, &m_velocities, twoLinesConstants, result);
if (!result) {
return false;
}
// Initialize the vertex and index buffer that hold the geometry for the triangle.
result = InitializeBuffers(device, twoLinesConstants);
if(!result)
{
return false;
}
return true;
}
FallBodyclass.cpp
FallBodyClass::FallBodyClass(int bodies, float ** positionsCPU, float ** velocitiesCPU, TwoLines twoLines, bool & success)
:bodies(bodies)
{
cl_int ret;
// getting the first available platform
cl_platform_id clPlatformID[2];
cl_platform_id GPUplatform;
cl_uint num_platforms;
//char str[1024];
ret = clGetPlatformIDs(2, clPlatformID, &num_platforms);
GPUplatform = clPlatformID[0]; //choose GPU platform
//error |= clGetPlatformInfo(GPUplatform, CL_PLATFORM_NAME, 0, NULL, NULL);
//clGetPlatformInfo(GPUplatform, CL_PLATFORM_VENDOR, sizeof(str), str, NULL);
// getting the first GPU device
ret |= clGetDeviceIDs(GPUplatform, CL_DEVICE_TYPE_GPU, 1, &device, NULL);
if (ret != CL_SUCCESS)
{
success = false;
return;
}
//clGetDeviceInfo(device, CL_DEVICE_NAME, sizeof(str), str, NULL);
// creating the context
context = clCreateContext(0, 1, &device, NULL, NULL, &ret);
if (ret != CL_SUCCESS)
{
success = false;
return;
}
cl_queue_properties props[] = {
CL_QUEUE_PROFILING_ENABLE
};
// creating the command queue
queue = clCreateCommandQueueWithProperties(context, device, props, &ret);
if (ret != CL_SUCCESS)
{
success = false;
return;
}
// setting the local variables
// (at the same time one of them supposed to be 0 and another to be 1)
read = 0;
write = 1;
// reading the kernel
FILE * f = NULL;
char fileName[18] = "kernel.cl";
f = fopen(fileName, "rb");
if(f == NULL)
{
success = false;
return;
}
// getting the length of the source code for the kernel
fseek(f, 0, SEEK_END);
size_t codeLength = ftell(f);
rewind(f);
char * code = (char *)malloc(codeLength + 1);
if (fread(code, codeLength, 1, f) != 1)
{
fclose(f);
free(code);
success = false;
return;
}
// closing the file and 0-terminating the source code
fclose(f);
code[codeLength] = '\0';
// creating the program
program = clCreateProgramWithSource(context, 1, (const char **)&code, &codeLength, &ret);
if (ret != CL_SUCCESS)
{
success = false;
return;
}
// clearing the memory
free(code);
// building the program
ret |= clBuildProgram(program, 0, NULL, NULL, NULL, NULL);
// creating the kernel
kernel = clCreateKernel(program, "impactManager", &ret);
// setting the local size of the group the largest possible in order to load all computational units
int numGroups;
ret |= clGetDeviceInfo(device, CL_DEVICE_MAX_COMPUTE_UNITS, sizeof(numGroups), &numGroups, NULL);
localSize = bodies / numGroups;
// allocating pinned buffers for velocities and positions, and stuck
positionsCPUBuffer = clCreateBuffer(context, CL_MEM_READ_WRITE | CL_MEM_ALLOC_HOST_PTR, COORD_DIM * bodies * sizeof(float) , NULL, NULL);
velocitiesCPUBuffer = clCreateBuffer(context, CL_MEM_READ_WRITE | CL_MEM_ALLOC_HOST_PTR, VEL_DIM * bodies * sizeof(float) , NULL, NULL);
linesCPUBuffer = clCreateBuffer(context, CL_MEM_READ_WRITE | CL_MEM_ALLOC_HOST_PTR, 8 * sizeof(float), NULL, NULL);
// get pointers to arrays to operate with the buffers (array map buffers here (to program) as float-arrays)
*positionsCPU = (float *)clEnqueueMapBuffer(queue, positionsCPUBuffer, CL_TRUE, CL_MAP_WRITE, 0, COORD_DIM * bodies * sizeof(float), 0, NULL, NULL, NULL);
*velocitiesCPU = (float *)clEnqueueMapBuffer(queue, velocitiesCPUBuffer, CL_TRUE, CL_MAP_WRITE, 0, VEL_DIM * bodies * sizeof(float), 0, NULL, NULL, NULL);
float * linesCPU = (float *)clEnqueueMapBuffer(queue, linesCPUBuffer, CL_TRUE, CL_MAP_WRITE, 0, 8 * sizeof(float), 0, NULL, NULL, NULL);
// initialization of the bodies' positions and velocities, and stuck
initBodies(*positionsCPU, *velocitiesCPU);
initLines(twoLines, linesCPU);
// unmapping the pointers to arrays (invalidates array pointers)
clEnqueueUnmapMemObject(queue, positionsCPUBuffer, *positionsCPU, 0, NULL, NULL);
clEnqueueUnmapMemObject(queue, velocitiesCPUBuffer, *velocitiesCPU, 0, NULL, NULL);
clEnqueueUnmapMemObject(queue, linesCPUBuffer, linesCPU, 0, NULL, NULL);
// allocate two arrays on GPU for positions and velocities
for (int i = 0; i < 2; ++i) {
positionsGPU[i] = clCreateBuffer(context, CL_MEM_READ_WRITE, COORD_DIM * bodies * sizeof(float), NULL, NULL);
ret |= clEnqueueWriteBuffer(queue, positionsGPU[i], CL_TRUE, 0, COORD_DIM * bodies * sizeof(float), *positionsCPU, 0, NULL, NULL);
velocitiesGPU[i] = clCreateBuffer(context, CL_MEM_READ_WRITE, VEL_DIM * bodies * sizeof(float), NULL, NULL);
ret |= clEnqueueWriteBuffer(queue, velocitiesGPU[i], CL_TRUE, 0, VEL_DIM * bodies * sizeof(float), *velocitiesCPU, 0, NULL, NULL);
}
linesGPU = clCreateBuffer(context, CL_MEM_READ_WRITE, 8 * sizeof(float), NULL, NULL);
ret |= clEnqueueWriteBuffer(queue, linesGPU, CL_TRUE, 0, 8 * sizeof(float), linesCPU, 0, NULL, NULL);
if (ret != CL_SUCCESS)
{
success = false;
return;
}
}
void FallBodyClass::initLines(IN TwoLines l, OUT float *linesCPU)
{
linesCPU[0] = l.a1;
linesCPU[1] = l.b1;
linesCPU[2] = l.R1.x;
linesCPU[3] = l.R1.y;
linesCPU[4] = l.a2;
linesCPU[5] = l.b2;
linesCPU[6] = l.R2.x;
linesCPU[7] = l.R2.y;
}
// initialization of the bodies' positions and velocities
void FallBodyClass::initBodies(float * positionsCPU, float * velocitiesCPU)
{
float scale = 0.20f;
// initialization of the memory
memset(positionsCPU, 0, COORD_DIM * bodies * sizeof(float));
memset(velocitiesCPU, 0, VEL_DIM * bodies * sizeof(float));
// for the randomization
srand((unsigned int)time(NULL));
for (int i = 0; i < bodies; i++)
{
positionsCPU[COORD_DIM * i] = 1.8*((rand() / (float)RAND_MAX) - 0.5); //x axis
positionsCPU[COORD_DIM * i + 1] = 0.9; //y axis
positionsCPU[COORD_DIM * i + 2] = 0.0f; //z axis
positionsCPU[COORD_DIM * i + 3] = 0.0f; // stuck variable
// velocities are zeros
velocitiesCPU[VEL_DIM* i] = 0.0;
velocitiesCPU[VEL_DIM* i + 1] = -2 * (rand() / (float)RAND_MAX);
velocitiesCPU[VEL_DIM* i + 2] = 0.0;
}
}
// updating the bodies' positions and velocities. Stuck is updated inside too
void FallBodyClass::update(float dt, float * positionsCPU, float * velocitiesCPU, bool & success)
{
cl_int error = CL_SUCCESS;
size_t global_work_size;
size_t local_work_size;
success = true;
if (localSize > bodies)
localSize = bodies;
local_work_size = localSize;
global_work_size = bodies;
// passing the arguments
// we write the new positions and velocities and read the previous ones
error |= clSetKernelArg(kernel, 0, sizeof(cl_mem), (void *)&positionsGPU[write]);
error |= clSetKernelArg(kernel, 1, sizeof(cl_mem), (void *)&velocitiesGPU[write]);
error |= clSetKernelArg(kernel, 2, sizeof(cl_mem), (void *)&positionsGPU[read]);
error |= clSetKernelArg(kernel, 3, sizeof(cl_mem), (void *)&velocitiesGPU[read]);
error |= clSetKernelArg(kernel, 4, sizeof(cl_float), (void *)&dt);
error |= clSetKernelArg(kernel, 5, sizeof(cl_mem), (void *)&linesGPU);
// just swap read and write in order not to copy the arrays
int temp;
temp = write;
write = read;
read = temp;
// executing the kernel
error |= clEnqueueNDRangeKernel(queue, kernel, 1, NULL, &global_work_size, &local_work_size, 0, NULL, NULL);
// synchronization
clFinish(queue);
// asynchronously reading the updated values
error |= clEnqueueReadBuffer(queue, positionsGPU[read], CL_FALSE, 0, COORD_DIM * bodies * sizeof(float), positionsCPU, 0, NULL, NULL);
if (error != CL_SUCCESS)
{
success = false;
}
error |= clEnqueueReadBuffer(queue, velocitiesGPU[read], CL_FALSE, 0, VEL_DIM * bodies * sizeof(float), velocitiesCPU, 0, NULL, NULL);
if (error != CL_SUCCESS)
{
success = false;
}
///////////
bool toReboot = positionsCPU[3]; //fourth index of the [0] first element
//bool toReboot = false;
////////////
if (toReboot) {
positionsCPU = (float *)clEnqueueMapBuffer(queue, positionsCPUBuffer, CL_TRUE, CL_MAP_WRITE, 0, COORD_DIM * bodies * sizeof(float), 0, NULL, NULL, NULL);
velocitiesCPU = (float *)clEnqueueMapBuffer(queue, velocitiesCPUBuffer, CL_TRUE, CL_MAP_WRITE, 0, VEL_DIM * bodies * sizeof(float), 0, NULL, NULL, NULL);
initBodies(positionsCPU, velocitiesCPU);
// unmapping the pointers
clEnqueueUnmapMemObject(queue, positionsCPUBuffer, positionsCPU, 0, NULL, NULL);
clEnqueueUnmapMemObject(queue, velocitiesCPUBuffer, velocitiesCPU, 0, NULL, NULL);
//update values on GPU side
error |= clEnqueueWriteBuffer(queue, positionsGPU[read], CL_TRUE, 0, COORD_DIM * bodies * sizeof(float), positionsCPU, 0, NULL, NULL);
error |= clEnqueueWriteBuffer(queue, velocitiesGPU[read], CL_TRUE, 0, VEL_DIM * bodies * sizeof(float), velocitiesCPU, 0, NULL, NULL);
}
return;
}
FallBodyClass::~FallBodyClass(void)
{
// synchronization (if something has to be done)
clFinish(queue);
// releasing all objects
clReleaseMemObject(linesGPU);
clReleaseMemObject(linesCPUBuffer);
clReleaseMemObject(velocitiesGPU[0]);
clReleaseMemObject(velocitiesGPU[1]);
clReleaseMemObject(positionsGPU[0]);
clReleaseMemObject(positionsGPU[1]);
clReleaseMemObject(positionsCPUBuffer);
clReleaseMemObject(velocitiesCPUBuffer);
clReleaseKernel(kernel);
clReleaseProgram(program);
clReleaseCommandQueue(queue);
clReleaseContext(context);
}