int main(int argc, char** argv)
{
// Launch CUDA/GL
init(argc, argv);
cudaGLSetGLDevice( compat_getMaxGflopsDeviceId() );
cudaGLRegisterBufferObject( planetVBO );
cudaGLRegisterBufferObject( velocityVBO );
#if VISUALIZE == 1
initCuda(N_FOR_VIS);
#else
initCuda(2*128);
#endif
projection = glm::perspective(fovy, float(width)/float(height), zNear, zFar);
view = glm::lookAt(cameraPosition, glm::vec3(0.0, 0.0, 0), glm::vec3(0,1,0));
projection = projection * view;
initShaders(program);
glEnable(GL_DEPTH_TEST);
glutDisplayFunc(display);
glutKeyboardFunc(keyboard);
glutMotionFunc(mouseMotion);
glutMainLoop();
return 0;
}
int main(){
//Change this line to use your name!
yourName = "Karl Li";
// Needed in OSX to force use of OpenGL3.2
glfwOpenWindowHint(GLFW_OPENGL_VERSION_MAJOR, 3);
glfwOpenWindowHint(GLFW_OPENGL_VERSION_MINOR, 2);
glfwOpenWindowHint(GLFW_OPENGL_FORWARD_COMPAT, GL_TRUE);
glfwOpenWindowHint(GLFW_OPENGL_PROFILE, GLFW_OPENGL_CORE_PROFILE);
init();
initCuda();
CUT_CHECK_ERROR_GL();
initVAO();
initTextures();
GLuint passthroughProgram;
passthroughProgram = initShader("shaders/passthroughVS.glsl", "shaders/passthroughFS.glsl");
glUseProgram(passthroughProgram);
glActiveTexture(GL_TEXTURE0);
// send into GLFW main loop
while(1){
display();
if (glfwGetKey(GLFW_KEY_ESC) == GLFW_PRESS || !glfwGetWindowParam( GLFW_OPENED )){
exit(0);
}
}
glfwTerminate();
return 0;
}
// Main program
int main(int argc, char** argv)
{
// Create the CUTIL timer
cutilCheckError( cutCreateTimer( &timer));
if (CUTFalse == initGL(argc, argv)) {
return CUTFalse;
}
initCuda(argc, argv);
CUT_CHECK_ERROR_GL();
// register callbacks
glutDisplayFunc(fpsDisplay);
glutKeyboardFunc(keyboard);
glutMouseFunc(mouse);
glutMotionFunc(motion);
// start rendering mainloop
glutMainLoop();
// clean up
cudaThreadExit();
cutilExit(argc, argv);
}
int main(int argc, char** argv)
{
if (Device::isCuda())
{
gpu::GLUTImageViewers::init(argc, argv);
// Device::printAll();
Device::printAllSimple();
// Server Cuda1: in [0,5]
// Server Cuda2: in [0,2]
int deviceId = 0;
initCuda(deviceId);
int isOk = start();
//cudaDeviceReset causes the driver to clean up all state.
// While not mandatory in normal operation, it is good practice.
HANDLE_ERROR(cudaDeviceReset());
return isOk;
}
else
{
return EXIT_FAILURE;
}
}
// Main program
int main(int argc, char** argv)
{
if( argc > 1 )
{
int width = atoi(argv[1]);
int height = atoi(argv[2]);
if( width > 100 && height > 100 )
{
window_width = width;
window_height = height;
}
}
if (false == initGL(argc, argv)) {
return false;
}
initCuda();
// register callbacks
glutDisplayFunc(fpsDisplay);
glutKeyboardFunc(keyboard);
glutSpecialFunc(spkeyboard);
glutMouseFunc(mouse);
glutPassiveMotionFunc(motion);
// start rendering mainloop
glutMainLoop();
}
int main(int argc, char *argv[])
{
// Read command line files
// Init cuda
CUcontext cuContext;
initCuda(cuContext);
// Image buffer allocation
unsigned int depth =4;
unsigned int width=960;
unsigned int height=1080;
size_t bufferSize = sizeof(unsigned char)*width*height*depth;
unsigned char *imgRight = (unsigned char*)malloc(bufferSize);
unsigned char *imgLeft= (unsigned char*)malloc(bufferSize);
// Read the list of test files
std::vector<std::string> testsFiles;
readTestsFiles(testsFiles, "./tests.txt");
// Launch tests
for(int i=0; i<testsFiles.size(); i++)
{
std::stringstream testImage1;
testImage1 << "./" << testsFiles[i] << "_1.dat";
std::stringstream testImage2;
testImage2 << "./" << testsFiles[i] << "_2.dat";
readTestImage( testImage1.str(), imgRight, bufferSize);
readTestImage( testImage2.str(), imgLeft, bufferSize);
VertexBufferObject rightPoints;
VertexBufferObject leftPoints;
DescriptorData rightDescriptors;
DescriptorData leftDescriptors;
computeDescriptorsLane(imgRight, depth, width, height, rightPoints, rightDescriptors);
computeDescriptorsLane(imgLeft, depth, width, height, leftPoints, leftDescriptors);
UInt2 imgSize(1920,1080);
vector<CvPoint2D32f> leftMatchedPts;
vector<CvPoint2D32f> rightMatchedPts;
leftMatchedPts.reserve(10000);
rightMatchedPts.reserve(10000);
computeMatching( leftDescriptors, rightDescriptors, leftMatchedPts, rightMatchedPts, imgSize);
}
// finalize cuda
CUresult cerr = cuCtxDestroy(cuContext);
checkError(cerr);
return 0;
}
// This test specifies a single test (where you specify radius and/or iterations)
int runSingleTest(char *ref_file, char *exec_path)
{
int nTotalErrors = 0;
char dump_file[256];
printf("[runSingleTest]: [%s]\n", sSDKsample);
initCuda();
unsigned int *dResult;
unsigned int *hResult = (unsigned int *)malloc(width * height * sizeof(unsigned int));
size_t pitch;
checkCudaErrors(cudaMallocPitch((void **)&dResult, &pitch, width*sizeof(unsigned int), height));
// run the sample radius
{
printf("%s (radius=%d) (passes=%d) ", sSDKsample, filter_radius, iterations);
bilateralFilterRGBA(dResult, width, height, euclidean_delta, filter_radius, iterations, kernel_timer);
// check if kernel execution generated an error
getLastCudaError("Error: bilateralFilterRGBA Kernel execution FAILED");
checkCudaErrors(cudaDeviceSynchronize());
// readback the results to system memory
cudaMemcpy2D(hResult, sizeof(unsigned int)*width, dResult, pitch,
sizeof(unsigned int)*width, height, cudaMemcpyDeviceToHost);
sprintf(dump_file, "nature_%02d.ppm", filter_radius);
sdkSavePPM4ub((const char *)dump_file, (unsigned char *)hResult, width, height);
if (!sdkComparePPM(dump_file, sdkFindFilePath(ref_file, exec_path), MAX_EPSILON_ERROR, 0.15f, false))
{
printf("Image is Different ");
nTotalErrors++;
}
else
{
printf("Image is Matching ");
}
printf(" <%s>\n", ref_file);
}
printf("\n");
free(hResult);
checkCudaErrors(cudaFree(dResult));
return nTotalErrors;
}
int main(int argc, char** argv)
{
// Launch CUDA/GL
init(argc, argv);
cudaGLSetGLDevice( compat_getMaxGflopsDeviceId() );
initPBO(&pbo);
cudaGLRegisterBufferObject( planetVBO );
#if VISUALIZE == 1
initCuda(N_FOR_VIS, glm::vec4 (cameraPosition, 1));
#else
initCuda(20*120);
#endif
// setDevicePrefetch (prefetchEnabled);
perspMat = glm::perspective(fovy, float(width)/float(height), zNear, zFar);
view = glm::lookAt(cameraPosition, glm::vec3(0), glm::vec3(0,0,1));
projection = perspMat * view;
GLuint passthroughProgram;
initShaders(program);
glUseProgram(program[HEIGHT_FIELD]);
glActiveTexture(GL_TEXTURE0);
glEnable(GL_DEPTH_TEST);
glutDisplayFunc(display);
glutKeyboardFunc(keyboard);
glutMainLoop();
return 0;
}
bool init(obj *mesh) {
glfwSetErrorCallback(errorCallback);
if (!glfwInit()) {
return false;
}
width = 350;
height = 350;
window = glfwCreateWindow(width, height, "CIS 565 Pathtracer", NULL, NULL);
if (!window) {
glfwTerminate();
return false;
}
glfwMakeContextCurrent(window);
glfwSetKeyCallback(window, keyCallback);
// Set up GL context
glewExperimental = GL_TRUE;
if (glewInit() != GLEW_OK) {
return false;
}
// Initialize other stuff
initVAO();
initTextures();
initCuda();
initPBO();
float cbo[] = {
0.0, 1.0, 0.0,
0.0, 0.0, 1.0,
1.0, 0.0, 0.0
};
rasterizeSetBuffers(mesh->getBufIdxsize(), mesh->getBufIdx(),
mesh->getBufPossize() / 3,
mesh->getBufPos(), mesh->getBufNor(), mesh->getBufCol());
GLuint passthroughProgram;
passthroughProgram = initShader();
glUseProgram(passthroughProgram);
glActiveTexture(GL_TEXTURE0);
return true;
}
////////////////////////////////////////////////////////////////////////////////
//! Run a simple benchmark test for CUDA
////////////////////////////////////////////////////////////////////////////////
void
runBenchmark( int argc, char **argv )
{
int devID = 0;
shrLog("[runBenchmark]: [%s]\n", sSDKsample);
devID = cutilChooseCudaDevice(argc, argv);
loadImageData(argc, argv);
initCuda();
g_CheckRender = new CheckBackBuffer(width, height, 4, false);
g_CheckRender->setExecPath(argv[0]);
unsigned int *d_result;
cutilSafeCall( cudaMalloc( (void **)&d_result, width*height*sizeof(unsigned int)) );
// warm-up
boxFilterRGBA(d_img, d_temp, d_temp, width, height, filter_radius, iterations, nthreads);
cutilSafeCall( cutilDeviceSynchronize() );
// Start round-trip timer and process iCycles loops on the GPU
iterations = 1; // standard 1-pass filtering
const int iCycles = 150;
double dProcessingTime = 0.0;
shrLog("\nRunning BoxFilterGPU for %d cycles...\n\n", iCycles);
shrDeltaT(2);
for (int i = 0; i < iCycles; i++)
{
dProcessingTime += boxFilterRGBA(d_img, d_temp, d_img, width, height, filter_radius, iterations, nthreads);
}
// check if kernel execution generated an error and sync host
cutilCheckMsg("Error: boxFilterRGBA Kernel execution FAILED");
cutilSafeCall(cutilDeviceSynchronize());
// Get average computation time
dProcessingTime /= (double)iCycles;
// log testname, throughput, timing and config info to sample and master logs
shrLogEx(LOGBOTH | MASTER, 0, "boxFilter-texture, Throughput = %.4f M RGBA Pixels/s, Time = %.5f s, Size = %u RGBA Pixels, NumDevsUsed = %u, Workgroup = %u\n",
(1.0e-6 * width * height)/dProcessingTime, dProcessingTime,
(width * height), 1, nthreads);
shrLog("\n");
}
void loadVolumeData(char *exec_path)
{
// load volume data
const char *path = sdkFindFilePath(volumeFilename, exec_path);
if (path == NULL)
{
fprintf(stderr, "Error unable to find 3D Volume file: '%s'\n", volumeFilename);
exit(EXIT_FAILURE);
}
size_t size = volumeSize.width*volumeSize.height*volumeSize.depth;
uchar *h_volume = loadRawFile(path, size);
initCuda(h_volume, volumeSize);
sdkCreateTimer(&timer);
free(h_volume);
}
////////////////////////////////////////////////////////////////////////////////
//! Run a simple benchmark test for CUDA
////////////////////////////////////////////////////////////////////////////////
int runBenchmark(int argc, char **argv)
{
printf("[runBenchmark]: [%s]\n", sSDKsample);
loadImageData(argc, argv);
initCuda();
unsigned int *dResult;
size_t pitch;
checkCudaErrors(cudaMallocPitch((void **)&dResult, &pitch, width*sizeof(unsigned int), height));
sdkStartTimer(&kernel_timer);
// warm-up
bilateralFilterRGBA(dResult, width, height, euclidean_delta, filter_radius, iterations, kernel_timer);
checkCudaErrors(cudaDeviceSynchronize());
// Start round-trip timer and process iCycles loops on the GPU
iterations = 1; // standard 1-pass filtering
const int iCycles = 150;
double dProcessingTime = 0.0;
printf("\nRunning BilateralFilterGPU for %d cycles...\n\n", iCycles);
for (int i = 0; i < iCycles; i++)
{
dProcessingTime += bilateralFilterRGBA(dResult, width, height, euclidean_delta, filter_radius, iterations, kernel_timer);
}
// check if kernel execution generated an error and sync host
getLastCudaError("Error: bilateralFilterRGBA Kernel execution FAILED");
checkCudaErrors(cudaDeviceSynchronize());
sdkStopTimer(&kernel_timer);
// Get average computation time
dProcessingTime /= (double)iCycles;
// log testname, throughput, timing and config info to sample and master logs
printf("bilateralFilter-texture, Throughput = %.4f M RGBA Pixels/s, Time = %.5f s, Size = %u RGBA Pixels, NumDevsUsed = %u\n",
(1.0e-6 * width * height)/dProcessingTime, dProcessingTime, (width * height), 1);
printf("\n");
return 0;
}
void runAutoTest(int argc, char **argv)
{
int devID = 0;
shrLog("[runAutoTest]: [%s] (automated testing w/ readback)\n", sSDKsample);
devID = cutilChooseCudaDevice(argc, argv);
loadImageData(argc, argv);
initCuda();
g_CheckRender = new CheckBackBuffer(width, height, 4, false);
g_CheckRender->setExecPath(argv[0]);
unsigned int *d_result;
cutilSafeCall( cudaMalloc( (void **)&d_result, width*height*sizeof(unsigned int)) );
for(int i = 0; i < 4; i++)
{
shrLog("[AutoTest]: %s (radius=%d)", sSDKsample, filter_radius );
bilateralFilterRGBA(d_result, width, height, euclidean_delta, filter_radius, iterations, nthreads);
// check if kernel execution generated an error
cutilCheckMsg("Error: bilateralFilterRGBA Kernel execution FAILED");
cutilSafeCall( cutilDeviceSynchronize() );
cudaMemcpy(g_CheckRender->imageData(), d_result, width*height*sizeof(unsigned int), cudaMemcpyDeviceToHost);
g_CheckRender->savePPM(sOriginal[i], false, NULL);
if (!g_CheckRender->PPMvsPPM(sOriginal[i], sReference[i], MAX_EPSILON_ERROR, 0.15f)) {
g_TotalErrors++;
}
gaussian_delta += 1.0f;
euclidean_delta *= 1.25f;
updateGaussian(gaussian_delta, filter_radius);
}
cutilSafeCall( cudaFree( d_result ) );
delete g_CheckRender;
}
////////////////////////////////////////////////////////////////////////////////
//! Run a simple benchmark test for CUDA
////////////////////////////////////////////////////////////////////////////////
int runBenchmark()
{
printf("[runBenchmark]: [%s]\n", sSDKsample);
initCuda(true);
unsigned int *d_result;
checkCudaErrors(cudaMalloc((void **)&d_result, width*height*sizeof(unsigned int)));
// warm-up
boxFilterRGBA(d_img, d_temp, d_temp, width, height, filter_radius, iterations, nthreads, kernel_timer);
checkCudaErrors(cudaDeviceSynchronize());
sdkStartTimer(&kernel_timer);
// Start round-trip timer and process iCycles loops on the GPU
iterations = 1; // standard 1-pass filtering
const int iCycles = 150;
double dProcessingTime = 0.0;
printf("\nRunning BoxFilterGPU for %d cycles...\n\n", iCycles);
for (int i = 0; i < iCycles; i++)
{
dProcessingTime += boxFilterRGBA(d_img, d_temp, d_img, width, height, filter_radius, iterations, nthreads, kernel_timer);
}
// check if kernel execution generated an error and sync host
getLastCudaError("Error: boxFilterRGBA Kernel execution FAILED");
checkCudaErrors(cudaDeviceSynchronize());
sdkStopTimer(&kernel_timer);
// Get average computation time
dProcessingTime /= (double)iCycles;
// log testname, throughput, timing and config info to sample and master logs
printf("boxFilter-texture, Throughput = %.4f M RGBA Pixels/s, Time = %.5f s, Size = %u RGBA Pixels, NumDevsUsed = %u, Workgroup = %u\n",
(1.0e-6 * width * height)/dProcessingTime, dProcessingTime,
(width * height), 1, nthreads);
printf("\n");
return 0;
}
bool init(int argc, char* argv[]) {
glfwSetErrorCallback(errorCallback);
if (!glfwInit()) {
return false;
}
width = 800;
height = 800;
window = glfwCreateWindow(width, height, "CIS 565 Rasterizer", NULL, NULL);
if (!window){
glfwTerminate();
return false;
}
glfwMakeContextCurrent(window);
glfwSetKeyCallback(window, keyCallback);
glfwSetMouseButtonCallback(window,MouseClickCallback);
glfwSetCursorEnterCallback(window,CursorEnterCallback);
glfwSetCursorPosCallback(window,CursorCallback);
// Set up GL context
glewExperimental = GL_TRUE;
if(glewInit()!=GLEW_OK){
return false;
}
// Initialize other stuff
initVAO();
initTextures();
initCuda();
initPBO();
GLuint passthroughProgram;
passthroughProgram = initShader();
glUseProgram(passthroughProgram);
glActiveTexture(GL_TEXTURE0);
return true;
}
int main(int argc, char* argv[]){
//Change this line to use your name!
yourName = "Aparajith Sairam";
init(argc, argv);
initVAO();
initTextures();
initCuda();
GLuint passthroughProgram;
passthroughProgram = initShader("shaders/passthroughVS.glsl", "shaders/passthroughFS.glsl");
glUseProgram(passthroughProgram);
glActiveTexture(GL_TEXTURE0);
glutDisplayFunc(display);
glutKeyboardFunc(keyboard);
glutMainLoop();
return 0;
}
////////////////////////////////////////////////////////////////////////////////
// Program main
////////////////////////////////////////////////////////////////////////////////
int
main( int argc, char** argv)
{
//start logs
shrSetLogFileName ("volumeRender.txt");
shrLog("%s Starting...\n\n", argv[0]);
if (cutCheckCmdLineFlag(argc, (const char **)argv, "qatest") ||
cutCheckCmdLineFlag(argc, (const char **)argv, "noprompt"))
{
g_bQAReadback = true;
fpsLimit = frameCheckNumber;
}
if (cutCheckCmdLineFlag(argc, (const char **)argv, "glverify"))
{
g_bQAGLVerify = true;
fpsLimit = frameCheckNumber;
}
if (g_bQAReadback) {
// use command-line specified CUDA device, otherwise use device with highest Gflops/s
if( cutCheckCmdLineFlag(argc, (const char**)argv, "device") ) {
cutilDeviceInit(argc, argv);
} else {
cudaSetDevice( cutGetMaxGflopsDeviceId() );
}
} else {
// First initialize OpenGL context, so we can properly set the GL for CUDA.
// This is necessary in order to achieve optimal performance with OpenGL/CUDA interop.
initGL( &argc, argv );
// use command-line specified CUDA device, otherwise use device with highest Gflops/s
if( cutCheckCmdLineFlag(argc, (const char**)argv, "device") ) {
cutilGLDeviceInit(argc, argv);
} else {
cudaGLSetGLDevice( cutGetMaxGflopsDeviceId() );
}
/*
int device;
struct cudaDeviceProp prop;
cudaGetDevice( &device );
cudaGetDeviceProperties( &prop, device );
if( !strncmp( "Tesla", prop.name, 5 ) ) {
shrLog("This sample needs a card capable of OpenGL and display.\n");
shrLog("Please choose a different device with the -device=x argument.\n");
cutilExit(argc, argv);
}
*/
}
// parse arguments
char *filename;
if (cutGetCmdLineArgumentstr( argc, (const char**) argv, "file", &filename)) {
volumeFilename = filename;
}
int n;
if (cutGetCmdLineArgumenti( argc, (const char**) argv, "size", &n)) {
volumeSize.width = volumeSize.height = volumeSize.depth = n;
}
if (cutGetCmdLineArgumenti( argc, (const char**) argv, "xsize", &n)) {
volumeSize.width = n;
}
if (cutGetCmdLineArgumenti( argc, (const char**) argv, "ysize", &n)) {
volumeSize.height = n;
}
if (cutGetCmdLineArgumenti( argc, (const char**) argv, "zsize", &n)) {
volumeSize.depth = n;
}
// load volume data
char* path = shrFindFilePath(volumeFilename, argv[0]);
if (path == 0) {
shrLog("Error finding file '%s'\n", volumeFilename);
exit(EXIT_FAILURE);
}
size_t size = volumeSize.width*volumeSize.height*volumeSize.depth*sizeof(VolumeType);
void *h_volume = loadRawFile(path, size);
initCuda(h_volume, volumeSize);
free(h_volume);
cutilCheckError( cutCreateTimer( &timer));
shrLog("Press '=' and '-' to change density\n"
" ']' and '[' to change brightness\n"
" ';' and ''' to modify transfer function offset\n"
" '.' and ',' to modify transfer function scale\n\n");
// calculate new grid size
gridSize = dim3(iDivUp(width, blockSize.x), iDivUp(height, blockSize.y));
if (g_bQAReadback) {
g_CheckRender = new CheckBackBuffer(width, height, 4, false);
g_CheckRender->setPixelFormat(GL_RGBA);
g_CheckRender->setExecPath(argv[0]);
g_CheckRender->EnableQAReadback(true);
uint *d_output;
cutilSafeCall(cudaMalloc((void**)&d_output, width*height*sizeof(uint)));
cutilSafeCall(cudaMemset(d_output, 0, width*height*sizeof(uint)));
float modelView[16] =
{
1.0f, 0.0f, 0.0f, 0.0f,
0.0f, 1.0f, 0.0f, 0.0f,
0.0f, 0.0f, 1.0f, 0.0f,
0.0f, 0.0f, 4.0f, 1.0f
};
invViewMatrix[0] = modelView[0]; invViewMatrix[1] = modelView[4]; invViewMatrix[2] = modelView[8]; invViewMatrix[3] = modelView[12];
invViewMatrix[4] = modelView[1]; invViewMatrix[5] = modelView[5]; invViewMatrix[6] = modelView[9]; invViewMatrix[7] = modelView[13];
invViewMatrix[8] = modelView[2]; invViewMatrix[9] = modelView[6]; invViewMatrix[10] = modelView[10]; invViewMatrix[11] = modelView[14];
// call CUDA kernel, writing results to PBO
copyInvViewMatrix(invViewMatrix, sizeof(float4)*3);
// Start timer 0 and process n loops on the GPU
int nIter = 10;
for (int i = -1; i < nIter; i++)
{
if( i == 0 ) {
cudaThreadSynchronize();
cutStartTimer(timer);
}
render_kernel(gridSize, blockSize, d_output, width, height, density, brightness, transferOffset, transferScale);
}
cudaThreadSynchronize();
cutStopTimer(timer);
// Get elapsed time and throughput, then log to sample and master logs
double dAvgTime = cutGetTimerValue(timer)/(nIter * 1000.0);
shrLogEx(LOGBOTH | MASTER, 0, "volumeRender, Throughput = %.4f MTexels/s, Time = %.5f s, Size = %u Texels, NumDevsUsed = %u, Workgroup = %u\n",
(1.0e-6 * width * height)/dAvgTime, dAvgTime, (width * height), 1, blockSize.x * blockSize.y);
cutilCheckMsg("Error: render_kernel() execution FAILED");
cutilSafeCall( cudaThreadSynchronize() );
cutilSafeCall( cudaMemcpy(g_CheckRender->imageData(), d_output, width*height*4, cudaMemcpyDeviceToHost) );
g_CheckRender->savePPM(sOriginal[g_Index], true, NULL);
if (!g_CheckRender->PPMvsPPM(sOriginal[g_Index], sReference[g_Index], MAX_EPSILON_ERROR, THRESHOLD)) {
shrLog("\nFAILED\n\n");
} else {
shrLog("\nPASSED\n\n");
}
cudaFree(d_output);
freeCudaBuffers();
if (g_CheckRender) {
delete g_CheckRender; g_CheckRender = NULL;
}
} else {
// This is the normal rendering path for VolumeRender
glutDisplayFunc(display);
glutKeyboardFunc(keyboard);
glutMouseFunc(mouse);
glutMotionFunc(motion);
glutReshapeFunc(reshape);
glutIdleFunc(idle);
initPixelBuffer();
if (g_bQAGLVerify) {
g_CheckRender = new CheckBackBuffer(width, height, 4);
g_CheckRender->setPixelFormat(GL_RGBA);
g_CheckRender->setExecPath(argv[0]);
g_CheckRender->EnableQAReadback(true);
}
atexit(cleanup);
glutMainLoop();
}
cudaThreadExit();
shrEXIT(argc, (const char**)argv);
}
////////////////////////////////////////////////////////////////////////////////
// Program main
////////////////////////////////////////////////////////////////////////////////
int
main(int argc, char **argv)
{
int devID = 0;
char *ref_file = NULL;
pArgc = &argc;
pArgv = argv;
// start logs
printf("%s Starting...\n\n", argv[0]);
if (checkCmdLineFlag(argc, (const char **)argv, "help"))
{
printHelp();
exit(EXIT_SUCCESS);
}
// use command-line specified CUDA device, otherwise use device with highest Gflops/s
if (argc > 1)
{
if (checkCmdLineFlag(argc, (const char **)argv, "threads"))
{
nthreads = getCmdLineArgumentInt(argc, (const char **) argv, "threads");
}
if (checkCmdLineFlag(argc, (const char **)argv, "radius"))
{
filter_radius = getCmdLineArgumentInt(argc, (const char **) argv, "radius");
}
if (checkCmdLineFlag(argc, (const char **)argv, "passes"))
{
iterations = getCmdLineArgumentInt(argc, (const char **) argv, "passes");
}
if (checkCmdLineFlag(argc, (const char **)argv, "file"))
{
getCmdLineArgumentString(argc, (const char **)argv, "file", (char **)&ref_file);
}
}
// load image to process
loadImageData(argc, argv);
if (checkCmdLineFlag(argc, (const char **)argv, "benchmark"))
{
// This is a separate mode of the sample, where we are benchmark the kernels for performance
devID = findCudaDevice(argc, (const char **)argv);
// Running CUDA kernels (boxfilter) in Benchmarking mode
g_TotalErrors += runBenchmark();
exit(g_TotalErrors == 0 ? EXIT_SUCCESS : EXIT_FAILURE);
}
else if (checkCmdLineFlag(argc, (const char **)argv, "radius") ||
checkCmdLineFlag(argc, (const char **)argv, "passes"))
{
// This overrides the default mode. Users can specify the radius used by the filter kernel
devID = findCudaDevice(argc, (const char **)argv);
g_TotalErrors += runSingleTest(ref_file, argv[0]);
exit(g_TotalErrors == 0 ? EXIT_SUCCESS : EXIT_FAILURE);
}
else
{
// Default mode running with OpenGL visualization and in automatic mode
// the output automatically changes animation
printf("\n");
// First initialize OpenGL context, so we can properly set the GL for CUDA.
// This is necessary in order to achieve optimal performance with OpenGL/CUDA interop.
initGL(&argc, argv);
int dev = findCapableDevice(argc, argv);
if (dev != -1)
{
cudaGLSetGLDevice(dev);
}
else
{
cudaDeviceReset();
exit(EXIT_SUCCESS);
}
// Now we can create a CUDA context and bind it to the OpenGL context
initCuda();
initGLResources();
// sets the callback function so it will call cleanup upon exit
atexit(cleanup);
printf("Running Standard Demonstration with GLUT loop...\n\n");
printf("Press '+' and '-' to change filter width\n"
"Press ']' and '[' to change number of iterations\n"
"Press 'a' or 'A' to change animation ON/OFF\n\n");
// Main OpenGL loop that will run visualization for every vsync
glutMainLoop();
cudaDeviceReset();
exit(g_TotalErrors == 0 ? EXIT_SUCCESS : EXIT_FAILURE);
}
}
////////////////////////////////////////////////////////////////////////////////
// Program main
////////////////////////////////////////////////////////////////////////////////
int main(int argc, char **argv)
{
// start logs
int devID;
char *ref_file = NULL;
printf("%s Starting...\n\n", argv[0]);
#if defined(__linux__)
setenv ("DISPLAY", ":0", 0);
#endif
// use command-line specified CUDA device, otherwise use device with highest Gflops/s
if (argc > 1)
{
if (checkCmdLineFlag(argc, (const char **)argv, "radius"))
{
filter_radius = getCmdLineArgumentInt(argc, (const char **) argv, "radius");
}
if (checkCmdLineFlag(argc, (const char **)argv, "passes"))
{
iterations = getCmdLineArgumentInt(argc, (const char **)argv, "passes");
}
if (checkCmdLineFlag(argc, (const char **)argv, "file"))
{
getCmdLineArgumentString(argc, (const char **)argv, "file", (char **)&ref_file);
}
}
// load image to process
loadImageData(argc, argv);
if (checkCmdLineFlag(argc, (const char **)argv, "benchmark"))
{
// This is a separate mode of the sample, where we are benchmark the kernels for performance
devID = findCudaDevice(argc, (const char **)argv);
// Running CUDA kernels (bilateralfilter) in Benchmarking mode
g_TotalErrors += runBenchmark(argc, argv);
// cudaDeviceReset causes the driver to clean up all state. While
// not mandatory in normal operation, it is good practice. It is also
// needed to ensure correct operation when the application is being
// profiled. Calling cudaDeviceReset causes all profile data to be
// flushed before the application exits
cudaDeviceReset();
exit(g_TotalErrors == 0 ? EXIT_SUCCESS : EXIT_FAILURE);
}
else if (checkCmdLineFlag(argc, (const char **)argv, "radius") ||
checkCmdLineFlag(argc, (const char **)argv, "passes"))
{
// This overrides the default mode. Users can specify the radius used by the filter kernel
devID = findCudaDevice(argc, (const char **)argv);
g_TotalErrors += runSingleTest(ref_file, argv[0]);
// cudaDeviceReset causes the driver to clean up all state. While
// not mandatory in normal operation, it is good practice. It is also
// needed to ensure correct operation when the application is being
// profiled. Calling cudaDeviceReset causes all profile data to be
// flushed before the application exits
cudaDeviceReset();
exit(g_TotalErrors == 0 ? EXIT_SUCCESS : EXIT_FAILURE);
}
else
{
// Default mode running with OpenGL visualization and in automatic mode
// the output automatically changes animation
printf("\n");
// First initialize OpenGL context, so we can properly set the GL for CUDA.
// This is necessary in order to achieve optimal performance with OpenGL/CUDA interop.
initGL(argc, (char **)argv);
int dev = findCapableDevice(argc, argv);
if (dev != -1)
{
dev = gpuGLDeviceInit(argc, (const char **)argv);
if (dev == -1)
{
exit(EXIT_FAILURE);
}
}
else
{
// cudaDeviceReset causes the driver to clean up all state. While
// not mandatory in normal operation, it is good practice. It is also
// needed to ensure correct operation when the application is being
// profiled. Calling cudaDeviceReset causes all profile data to be
// flushed before the application exits
cudaDeviceReset();
exit(EXIT_SUCCESS);
}
// Now we can create a CUDA context and bind it to the OpenGL context
initCuda();
initGLResources();
// sets the callback function so it will call cleanup upon exit
#if defined (__APPLE__) || defined(MACOSX)
atexit(cleanup);
#else
glutCloseFunc(cleanup);
#endif
printf("Running Standard Demonstration with GLUT loop...\n\n");
printf("Press '+' and '-' to change filter width\n"
"Press ']' and '[' to change number of iterations\n"
"Press 'e' and 'E' to change Euclidean delta\n"
"Press 'g' and 'G' to changle Gaussian delta\n"
"Press 'a' or 'A' to change Animation mode ON/OFF\n\n");
// Main OpenGL loop that will run visualization for every vsync
glutMainLoop();
}
}
int main(int argc, char**argv)
{
capture.open(0);
if (capture.isOpened() == false)
{
std::cerr << "no capture device found" << std::endl;
return 1;
}
capture.set(CV_CAP_PROP_FRAME_WIDTH, vgaSize.width);
capture.set(CV_CAP_PROP_FRAME_HEIGHT, vgaSize.height);
if (capture.get(cv::CAP_PROP_FRAME_WIDTH) != (double)vgaSize.width || capture.get(cv::CAP_PROP_FRAME_HEIGHT) != (double)vgaSize.height)
{
std::cerr << "current device doesn't support " << vgaSize.width << "x" << vgaSize.height << " size" << std::endl;
return 2;
}
cv::Mat image;
capture >> image;
cv::namedWindow(windowName);
cv::imshow(windowName, image);
initCuda();
initArray(image);
char key = -1;
enum device statusDevice = useCpuSimd;
enum precision statusPrecision = precisionFloat;
int index = 1;
cv::Mat stub = cv::imread(imagePath[index][0], cv::IMREAD_UNCHANGED);
cv::Mat gain = cv::Mat(stub.rows, stub.cols/2, CV_16SC1, stub.data);
double elapsedTime;
while (isFinish(key) == false)
{
capture >> image;
switch (key)
{
case 'h':
case 'H':
// switch to half precision
statusPrecision = precisionHalf;
std::cout << std::endl << header << "half " << std::endl;
stub = cv::imread(imagePath[index][0], cv::IMREAD_UNCHANGED);
gain = cv::Mat(stub.rows, stub.cols/2, CV_16SC1, stub.data);
break;
case 'f':
case 'F':
// switch to single precision
statusPrecision = precisionFloat;
std::cout << std::endl << header << "single" << std::endl;
stub = cv::imread(imagePath[index][1], cv::IMREAD_UNCHANGED);
gain = cv::Mat(stub.rows, stub.cols, CV_32FC1, stub.data);
break;
case 'b':
case 'B':
// switch to gray gain
statusPrecision = precisionByte;
std::cout << std::endl << header << "char" << std::endl;
gain = cv::imread(imagePath[index][2], cv::IMREAD_GRAYSCALE);
break;
case '0':
case '1':
index = key - '0';
switch (statusPrecision)
{
case precisionHalf:
// precision half
stub = cv::imread(imagePath[index][0], cv::IMREAD_UNCHANGED);
gain = cv::Mat(stub.rows, stub.cols/2, CV_16SC1, stub.data);
break;
case precisionFloat:
// precision single
stub = cv::imread(imagePath[index][1], cv::IMREAD_UNCHANGED);
gain = cv::Mat(stub.rows, stub.cols, CV_32FC1, stub.data);
break;
case precisionByte:
// precision single
gain = cv::imread(imagePath[index][2], cv::IMREAD_GRAYSCALE);
break;
default:
break;
}
break;
case 'c':
case 'C':
std::cout << std::endl << "Using CPU SIMD " << std::endl;
statusDevice = useCpuSimd;
break;
case 'g':
case 'G':
std::cout << std::endl << "Using GPU " << std::endl;
statusDevice = useGpu;
break;
default:
break;
}
if (statusDevice == useCpuSimd)
{
elapsedTime = multiplyImage(image, gain);
}
else
{
#ifdef HAVE_CUDA
// CUDA
elapsedTime = multiplyImageCuda(image, gain);
#endif // HAVE_CUDA
}
computeStatistics(elapsedTime, key);
if (key == 's' || key == 'S')
{
cv::imwrite(dumpFilename, image);
}
cv::imshow(windowName, image);
key = cv::waitKey(1);
}
std::cout << std::endl;
cv::destroyAllWindows();
releaseArray();
return 0;
}
int main(int argc, char** argv){
bool loadedScene = false;
//for(int i=1; i<argc; i++){
string header; string data;
//istringstream liness(argv[i]);
//getline(liness, header, '='); getline(liness, data, '=');
//if(strcmp(header.c_str(), "mesh")==0){
//renderScene = new scene(data);
//laoding file for obj
data = "../../objs/plane.obj";
mesh = new obj();
objLoader* loader = new objLoader(data, mesh);
mesh->buildVBOs();
meshVector.push_back(mesh);
data = "../../objs/bunny.obj";
mesh = new obj();
loader = new objLoader(data, mesh);
mesh->buildVBOs();
meshVector.push_back(mesh);
delete loader;
loadedScene = true;
//}
//}
stencilBuffer = new int[width*height];
#if STENCIL == 1
isStencil = true;
#else
isStencil = false;
#endif
firstObj = 0;
secondObj = FIRST;
if(!loadedScene){
cout << "Usage: mesh=[obj file]" << endl;
return 0;
}
frame = 0;
seconds = time (NULL);
fpstracker = 0;
// Launch CUDA/GL
#ifdef __APPLE__
// Needed in OSX to force use of OpenGL3.2
glfwOpenWindowHint(GLFW_OPENGL_VERSION_MAJOR, 3);
glfwOpenWindowHint(GLFW_OPENGL_VERSION_MINOR, 2);
glfwOpenWindowHint(GLFW_OPENGL_FORWARD_COMPAT, GL_TRUE);
glfwOpenWindowHint(GLFW_OPENGL_PROFILE, GLFW_OPENGL_CORE_PROFILE);
init();
#else
init(argc, argv);
#endif
initCuda();
initVAO();
initTextures();
GLuint passthroughProgram;
passthroughProgram = initShader("shaders/passthroughVS.glsl", "shaders/passthroughFS.glsl");
glUseProgram(passthroughProgram);
glActiveTexture(GL_TEXTURE0);
#ifdef __APPLE__
// send into GLFW main loop
while(1){
display();
if (glfwGetKey(GLFW_KEY_ESC) == GLFW_PRESS || !glfwGetWindowParam( GLFW_OPENED )){
kernelCleanup();
cudaDeviceReset();
exit(0);
}
}
glfwTerminate();
#else
glutDisplayFunc(display);
glutKeyboardFunc(keyboard);
glutMainLoop();
#endif
kernelCleanup();
return 0;
}
int main(int argc, char** argv){
bool loadedScene = false;
for(int i=1; i<argc; i++){
string header; string data;
istringstream liness(argv[i]);
getline(liness, header, '='); getline(liness, data, '=');
if(strcmp(header.c_str(), "mesh")==0){
//renderScene = new scene(data);
mesh = new obj();
objLoader* loader = new objLoader(data, mesh);
mesh->buildVBOs();
delete loader;
loadedScene = true;
}
}
if(!loadedScene){
cout << "Usage: mesh=[obj file]" << endl;
return 0;
}
// Initialization of camera parameters
cam.position = glm::vec3(0.0f, 1.0f, 1.0f);
cam.up = glm::vec3(0.0f, 1.0f, 0.0f);
cam.view = glm::normalize(-cam.position);
cam.right = glm::normalize(glm::cross(cam.view, cam.up));
cam.fovy = 45.0f;
// Initialize transformation
model = new glm::mat4(utilityCore::buildTransformationMatrix(glm::vec3(0.0f, -0.2f, 0.0f), glm::vec3(0.0f), glm::vec3(0.7f)));
view = new glm::mat4(glm::lookAt(cam.position, glm::vec3(0.0f), cam.up));
projection = new glm::mat4(glm::perspective(cam.fovy, (float)width / height, zNear, zFar));
transformModel2Projection = new cudaMat4(utilityCore::glmMat4ToCudaMat4(*projection * *view * *model));
// Initialize viewport in the model space
viewPort = glm::normalize(utilityCore::multiplyMat(utilityCore::glmMat4ToCudaMat4(*projection * *view), glm::vec4(cam.view, 1.0f)));
frame = 0;
seconds = time (NULL);
fpstracker = 0;
// Launch CUDA/GL
#ifdef __APPLE__
// Needed in OSX to force use of OpenGL3.2
glfwOpenWindowHint(GLFW_OPENGL_VERSION_MAJOR, 3);
glfwOpenWindowHint(GLFW_OPENGL_VERSION_MINOR, 2);
glfwOpenWindowHint(GLFW_OPENGL_FORWARD_COMPAT, GL_TRUE);
glfwOpenWindowHint(GLFW_OPENGL_PROFILE, GLFW_OPENGL_CORE_PROFILE);
init();
#else
init(argc, argv);
#endif
initCuda();
initVAO();
initTextures();
GLuint passthroughProgram;
passthroughProgram = initShader("shaders/passthroughVS.glsl", "shaders/passthroughFS.glsl");
glUseProgram(passthroughProgram);
glActiveTexture(GL_TEXTURE0);
#ifdef __APPLE__
// send into GLFW main loop
while(1){
display();
if (glfwGetKey(GLFW_KEY_ESC) == GLFW_PRESS || !glfwGetWindowParam( GLFW_OPENED )){
kernelCleanup();
cudaDeviceReset();
exit(0);
}
}
glfwTerminate();
#else
glutDisplayFunc(display);
glutKeyboardFunc(keyboard);
glutSpecialFunc(specialFunction);
glutMouseFunc(mouseClick);
glutMotionFunc(mouseMotion);
glutMainLoop();
#endif
kernelCleanup();
delete model;
delete view;
delete projection;
delete transformModel2Projection;
return 0;
}
int main(int argc, char** argv){
#ifdef __APPLE__
// Needed in OSX to force use of OpenGL3.2
glfwOpenWindowHint(GLFW_OPENGL_VERSION_MAJOR, 3);
glfwOpenWindowHint(GLFW_OPENGL_VERSION_MINOR, 2);
glfwOpenWindowHint(GLFW_OPENGL_FORWARD_COMPAT, GL_TRUE);
glfwOpenWindowHint(GLFW_OPENGL_PROFILE, GLFW_OPENGL_CORE_PROFILE);
#endif
// Set up pathtracer stuff
bool loadedScene = false;
finishedRender = false;
targetFrame = 0;
singleFrameMode = false;
// Load scene file
for(int i=1; i<argc; i++){
string header; string data;
istringstream liness(argv[i]);
getline(liness, header, '='); getline(liness, data, '=');
if(strcmp(header.c_str(), "scene")==0){
renderScene = new scene(data);
loadedScene = true;
}else if(strcmp(header.c_str(), "frame")==0){
targetFrame = atoi(data.c_str());
singleFrameMode = true;
}
}
if(!loadedScene){
cout << "Error: scene file needed!" << endl;
return 0;
}
// Set up camera stuff from loaded pathtracer settings
iterations = 0;
renderCam = &renderScene->renderCam;
parameterSet = &renderScene->parameterSet;
width = renderCam->resolution[0];
height = renderCam->resolution[1];
textures=new m_BMP[renderScene->bmps.size()];
int i,j,k;
for(i=0;i<renderScene->bmps.size();i++)
{
//int w=renderScene->bmps[i]->Width;
//int h=renderScene->bmps[i]->Height;
BMP now;
now.ReadFromFile(renderScene->bmps[i].c_str());
int h=now.TellHeight();int w=now.TellWidth();
textures[i].resolution=glm::vec2(w,h);
textures[i].colors=new glm::vec3[w*h];
for(j=0;j<w;j++)for(k=0;k<h;k++)
{
RGBApixel current=now.GetPixel(j,k);
textures[i].colors[j*h+k]=glm::vec3(current.Red,current.Green,current.Blue)*(1.0f/255.0f);
}
}
if(targetFrame>=renderCam->frames){
cout << "Warning: Specified target frame is out of range, defaulting to frame 0." << endl;
targetFrame = 0;
}
// Launch CUDA/GL
#ifdef __APPLE__
init();
#else
init(argc, argv);
#endif
initCuda();
initVAO();
initTextures();
GLuint passthroughProgram;
passthroughProgram = initShader("shaders/passthroughVS.glsl", "shaders/passthroughFS.glsl");
glUseProgram(passthroughProgram);
glActiveTexture(GL_TEXTURE0);
starttime=clock();
#ifdef __APPLE__
// send into GLFW main loop
while(1){
display();
if (glfwGetKey(GLFW_KEY_ESC) == GLFW_PRESS || !glfwGetWindowParam( GLFW_OPENED )){
exit(0);
}
}
glfwTerminate();
#else
glutDisplayFunc(display);
glutKeyboardFunc(keyboard);
glutMainLoop();
#endif
return 0;
}
////////////////////////////////////////////////////////////////////////////////
//! Run a simple test for CUDA
////////////////////////////////////////////////////////////////////////////////
void
runTest(int argc, char** argv)
{
CUcontext cuContext;
// initialize CUDA
CUfunction pk = NULL;
const char cubin_name [] = "pass_kernel.cubin";
const char kernel_name [] = "pass_kernel";
CU_SAFE_CALL(initCuda(cuContext, argv[0], &pk, argc, argv, cubin_name, kernel_name));
printf("initCuda-returned CUfunction:\n");
// cuParamSetx, x=i f v
// http://visionexperts.blogspot.com/2010/07/cuda-parameter-alignment.html - check alignment
#define ALIGN_UP(offset, alignment) \
(offset) = ((offset) + (alignment) - 1) & ~((alignment) - 1)
size_t offset = 0;
// input integers
// CU paramset i.
for(int i = 0 ; i < NUM_ARG ; i++)
{
int align = __alignof(int);
ALIGN_UP(offset, align);
cuParamSeti(pk, offset, i);
printf ("offset %d = %d\n", i, offset);
offset += sizeof(int);
}
// return array for updated inputs
int size_int = sizeof(int);
int size_array = size_int * NUM_ARG;
CUdeviceptr d_return_values;
cuMemAlloc (&d_return_values, size_array);
void* ptr = (void*)(size_t)d_return_values;
int align = __alignof(ptr);
ALIGN_UP(offset, align);
cuParamSetv(pk, offset, &ptr, sizeof(ptr));
printf("return values offset:%d\n", offset);
offset += sizeof(ptr);
CUdeviceptr d_return_N;
cuMemAlloc(&d_return_N, size_int);
void* ptrN = (void*)(size_t)d_return_N;
int alignN = __alignof(ptrN);
ALIGN_UP(offset, alignN);
cuParamSetv(pk, offset, &ptrN, sizeof(ptr));
printf("return int offset:%d\n", offset);
offset += sizeof(ptrN);
// Calling kernel
int BLOCK_SIZE_X = NUM_ARG;
int BLOCK_SIZE_Y = 1;
int BLOCK_SIZE_Z = 1;
int GRID_SIZE = 1;
cutilDrvSafeCallNoSync(cuFuncSetBlockShape(pk, BLOCK_SIZE_X, BLOCK_SIZE_Y, BLOCK_SIZE_Z));
printf("paramsetsize:%d\n", offset);
CU_SAFE_CALL(cuParamSetSize(pk, offset));
CU_SAFE_CALL(cuLaunchGrid(pk, GRID_SIZE, GRID_SIZE));
int* h_return_values = (int*)malloc(NUM_ARG * sizeof(int));
CU_SAFE_CALL(cuMemcpyDtoH((void*)h_return_values, d_return_values, size_array));
CU_SAFE_CALL(cuMemFree(d_return_values));
for(int i=0;i<NUM_ARG;i++)
printf("%dth value = %d\n", i, h_return_values[i]);
free(h_return_values);
int* h_return_N = (int*)malloc(sizeof(int));
CU_SAFE_CALL(cuMemcpyDtoH((void*)h_return_N, d_return_N, size_int));
CU_SAFE_CALL(cuMemFree(d_return_N));
printf("%d sizeof array\n", *h_return_N);
if(cuContext !=NULL) cuCtxDetach(cuContext);
}
////////////////////////////////////////////////////////////////////////////////
// Program main
////////////////////////////////////////////////////////////////////////////////
int main( int argc, char** argv)
{
shrQAStart(argc, argv);
// start logs
shrSetLogFileName ("bilateralFilter.txt");
shrLog("%s Starting...\n\n", argv[0]);
// use command-line specified CUDA device, otherwise use device with highest Gflops/s
cutGetCmdLineArgumenti( argc, (const char**) argv, "threads", &nthreads );
cutGetCmdLineArgumenti( argc, (const char**) argv, "radius", &filter_radius);
// load image to process
loadImageData(argc, argv);
if (cutCheckCmdLineFlag(argc, (const char **)argv, "qatest") ||
cutCheckCmdLineFlag(argc, (const char **)argv, "noprompt"))
{
// Running CUDA kernel (bilateralFilter) without visualization (QA Testing/Verification)
runAutoTest(argc, argv);
shrQAFinishExit(argc, (const char **)argv, (g_TotalErrors == 0 ? QA_PASSED : QA_FAILED));
}
else if (cutCheckCmdLineFlag(argc, (const char **)argv, "benchmark"))
{
// Running CUDA kernel (bilateralFilter) in Benchmarking Mode
runBenchmark(argc, argv);
shrQAFinishExit(argc, (const char **)argv, (g_TotalErrors == 0 ? QA_PASSED : QA_FAILED));
}
else
{
// Running CUDA kernel (bilateralFilter) in CUDA + OpenGL Visualization Mode
if ( cutCheckCmdLineFlag(argc, (const char **)argv, "device")) {
printf("[%s]\n", argv[0]);
printf(" Does not explicitly support -device=n in OpenGL mode\n");
printf(" To use -device=n, the sample must be running w/o OpenGL\n\n");
printf(" > %s -device=n -qatest\n", argv[0]);
printf("exiting...\n");
exit(0);
}
// First initialize OpenGL context, so we can properly set the GL for CUDA.
// This is necessary in order to achieve optimal performance with OpenGL/CUDA interop.
initGL( argc, argv );
if ( cutCheckCmdLineFlag(argc, (const char **)argv, "device")) {
cutilGLDeviceInit(argc, argv);
} else {
cudaGLSetGLDevice (cutGetMaxGflopsDeviceId() );
}
initCuda();
initOpenGL();
}
atexit(cleanup);
printf("Running Standard Demonstration with GLUT loop...\n\n");
printf("Press '+' and '-' to change number of iterations\n"
"Press LEFT and RIGHT change euclidean delta\n"
"Press UP and DOWN to change gaussian delta\n"
"Press '1' to show original image\n"
"Press '2' to show result\n\n");
glutMainLoop();
cutilDeviceReset();
shrQAFinishExit(argc, (const char **)argv, (g_TotalErrors == 0 ? QA_PASSED : QA_FAILED));
}
int main(int argc, char** argv){
#ifdef __APPLE__
// Needed in OSX to force use of OpenGL3.2
glfwOpenWindowHint(GLFW_OPENGL_VERSION_MAJOR, 3);
glfwOpenWindowHint(GLFW_OPENGL_VERSION_MINOR, 2);
glfwOpenWindowHint(GLFW_OPENGL_FORWARD_COMPAT, GL_TRUE);
glfwOpenWindowHint(GLFW_OPENGL_PROFILE, GLFW_OPENGL_CORE_PROFILE);
#endif
// Set up pathtracer stuff
bool loadedScene = false;
finishedRender = false;
targetFrame = 0;
singleFrameMode = false;
// Load scene file
for(int i=1; i<argc; i++){
string header; string data;
istringstream liness(argv[i]);
getline(liness, header, '='); getline(liness, data, '=');
if(strcmp(header.c_str(), "scene")==0){
renderScene = new scene(data);
loadedScene = true;
}else if(strcmp(header.c_str(), "frame")==0){
targetFrame = atoi(data.c_str());
singleFrameMode = true;
}
else if(strcmp(header.c_str(), "mblur")==0){
mblur = atoi(data.c_str());
}
else if(strcmp(header.c_str(), "dof")==0){
dof = atoi(data.c_str());
}
else if(strcmp(header.c_str(), "textureMode")==0){
textureMode = atoi(data.c_str());
}
}
if(!loadedScene){
cout << "Error: scene file needed!" << endl;
return 0;
}
// Set up camera stuff from loaded pathtracer settings
iterations = 0;
renderCam = &renderScene->renderCam;
width = renderCam->resolution[0];
height = renderCam->resolution[1];
if(targetFrame>=renderCam->frames){
cout << "Warning: Specified target frame is out of range, defaulting to frame 0." << endl;
targetFrame = 0;
}
// Launch CUDA/GL
#ifdef __APPLE__
init();
#else
init(argc, argv);
#endif
initCuda();
initVAO();
initTextures();
GLuint passthroughProgram;
passthroughProgram = initShader("shaders/passthroughVS.glsl", "shaders/passthroughFS.glsl");
glUseProgram(passthroughProgram);
glActiveTexture(GL_TEXTURE0);
#ifdef __APPLE__
// send into GLFW main loop
while(1){
display();
if (glfwGetKey(GLFW_KEY_ESC) == GLFW_PRESS || !glfwGetWindowParam( GLFW_OPENED )){
exit(0);
}
}
glfwTerminate();
#else
glutDisplayFunc(display);
glutKeyboardFunc(keyboard);
glutMouseFunc(mouse);
glutMainLoop();
#endif
return 0;
}
int main(int argc, char** argv){
bool loadedScene = false;
for(int i=1; i<argc; i++){
string header; string data;
istringstream liness(argv[i]);
getline(liness, header, '='); getline(liness, data, '=');
if(strcmp(header.c_str(), "mesh")==0){
//renderScene = new scene(data);
mesh = new obj();
objLoader* loader = new objLoader(data, mesh);
mesh->buildVBOs();
delete loader;
loadedScene = true;
}
}
if(!loadedScene){
cout << "Usage: mesh=[obj file]" << endl;
return 0;
}
frame = 0;
seconds = time (NULL);
fpstracker = 0;
// Launch CUDA/GL
#ifdef __APPLE__
// Needed in OSX to force use of OpenGL3.2
glfwOpenWindowHint(GLFW_OPENGL_VERSION_MAJOR, 3);
glfwOpenWindowHint(GLFW_OPENGL_VERSION_MINOR, 2);
glfwOpenWindowHint(GLFW_OPENGL_FORWARD_COMPAT, GL_TRUE);
glfwOpenWindowHint(GLFW_OPENGL_PROFILE, GLFW_OPENGL_CORE_PROFILE);
init();
#else
init(argc, argv);
#endif
// Initialize camera position
cam = glm::translate( cam, glm::vec3( 0.0, 0.0, 2.0f ) );
initCuda();
initVAO();
initTextures();
GLuint passthroughProgram;
passthroughProgram = initShader("shaders/passthroughVS.glsl", "shaders/passthroughFS.glsl");
glUseProgram(passthroughProgram);
glActiveTexture(GL_TEXTURE0);
#ifdef __APPLE__
// send into GLFW main loop
while(1){
display();
if (glfwGetKey(GLFW_KEY_ESC) == GLFW_PRESS || !glfwGetWindowParam( GLFW_OPENED )){
kernelCleanup();
cudaDeviceReset();
exit(0);
}
}
glfwTerminate();
#else
glutDisplayFunc(display);
glutKeyboardFunc(keyboard);
glutMouseFunc(mouse);
glutMotionFunc(mouse_motion);
glutMainLoop();
#endif
kernelCleanup();
return 0;
}
////////////////////////////////////////////////////////////////////////////////
// Program main
////////////////////////////////////////////////////////////////////////////////
int
main( int argc, char** argv)
{
pArgc = &argc;
pArgv = argv;
shrQAStart(argc, argv);
// start logs
shrSetLogFileName ("boxFilter.txt");
shrLog("%s Starting...\n\n", argv[0]);
// use command-line specified CUDA device, otherwise use device with highest Gflops/s
if (argc > 1) {
cutGetCmdLineArgumenti( argc, (const char**) argv, "threads", &nthreads );
cutGetCmdLineArgumenti( argc, (const char**) argv, "radius", &filter_radius);
if (cutCheckCmdLineFlag(argc, (const char **)argv, "glverify"))
{
g_bOpenGLQA = true;
fpsLimit = frameCheckNumber;
}
if (cutCheckCmdLineFlag(argc, (const char **)argv, "fbo")) {
g_bFBODisplay = true;
}
}
// load image to process
loadImageData(argc, argv);
if (cutCheckCmdLineFlag(argc, (const char **)argv, "qatest") ||
cutCheckCmdLineFlag(argc, (const char **)argv, "noprompt"))
{
// Running CUDA kernel (boxFilter) without visualization (QA Testing/Verification)
runAutoTest(argc, argv);
shrQAFinishExit(argc, (const char **)argv, (g_TotalErrors == 0 ? QA_PASSED : QA_FAILED));
}
else if (cutCheckCmdLineFlag(argc, (const char **)argv, "benchmark"))
{
// Running CUDA kernels (boxfilter) in Benchmarking mode
runBenchmark(argc, argv);
shrQAFinishExit(argc, (const char **)argv, (g_TotalErrors == 0 ? QA_PASSED : QA_FAILED));
}
else
{
// Running CUDA kernels (boxFilter) with OpenGL visualization
if (g_bFBODisplay) shrLog("[FBO Display] ");
if (g_bOpenGLQA) shrLog("[OpenGL Readback Comparisons] ");
shrLog("\n");
if ( cutCheckCmdLineFlag(argc, (const char **)argv, "device")) {
printf(" This SDK does not explicitly support -device=n when running with OpenGL.\n");
printf(" When specifying -device=n (n=0,1,2,....) the sample must not use OpenGL.\n");
printf(" See details below to run without OpenGL:\n\n");
printf(" > %s -device=n -qatest\n\n", argv[0]);
printf("exiting...\n");
shrQAFinishExit(argc, (const char **)argv, QA_WAIVED);
}
// First initialize OpenGL context, so we can properly set the GL for CUDA.
// This is necessary in order to achieve optimal performance with OpenGL/CUDA interop.
initGL( &argc, argv );
int dev = findCapableDevice(argc, argv);
if( dev != -1 ) {
cudaGLSetGLDevice( dev );
} else {
cutilDeviceReset();
shrQAFinishExit(argc, (const char **)argv, QA_WAIVED);
}
// Now we can create a CUDA context and bind it to the OpenGL context
initCuda();
initGLResources();
if (g_bOpenGLQA) {
if (g_bFBODisplay) {
g_CheckRender = new CheckFBO(width, height, 4, g_FrameBufferObject);
} else {
g_CheckRender = new CheckBackBuffer(width, height, 4);
}
g_CheckRender->setPixelFormat(GL_RGBA);
g_CheckRender->setExecPath(argv[0]);
g_CheckRender->EnableQAReadback(true);
}
}
// sets the callback function so it will call cleanup upon exit
atexit(cleanup);
shrLog("Running Standard Demonstration with GLUT loop...\n\n");
shrLog("Press '+' and '-' to change filter width\n"
"Press ']' and '[' to change number of iterations\n\n");
// Main OpenGL loop that will run visualization for every vsync
glutMainLoop();
cutilDeviceReset();
shrQAFinishExit(argc, (const char **)argv, (g_TotalErrors == 0 ? QA_PASSED : QA_FAILED));
}
int main(int argc, char** argv)
{
// Launch CUDA/GL
init(argc, argv);
cudaGLSetGLDevice( compat_getMaxGflopsDeviceId() );
cudaGLRegisterBufferObject( planetVBO );
#if VISUALIZE == 1
initCuda(N_FOR_VIS);
#else
initCuda(2*128);
#endif
projection = glm::perspective(fovy, float(width)/float(height), zNear, zFar);
view = camera.getViewMatrix();
projection = projection * view;
GLuint passthroughProgram;
initShaders(program);
glUseProgram(program[HEIGHT_FIELD]);
glActiveTexture(GL_TEXTURE0);
glEnable(GL_DEPTH_TEST);
glutDisplayFunc(display);
glutKeyboardFunc(keyboard);
glutMainLoop();
return 0;
}
int main(int argc, char** argv)
{
// Launch CUDA/GL
init(argc, argv);
cudaGLSetGLDevice(0);
initPBO(&pbo);
cudaGLRegisterBufferObject( planetVBO );
initCuda(N_FOR_VIS);
projection = glm::perspective(fovy, float(width)/float(height), zNear, zFar);
view = glm::lookAt(cameraPosition, glm::vec3(0), glm::vec3(0,0,1));
projection = projection * view;
GLuint passthroughProgram;
initShaders(program);
glUseProgram(program[HEIGHT_FIELD]);
glActiveTexture(GL_TEXTURE0);
glEnable(GL_DEPTH_TEST);
glutDisplayFunc(display);
glutKeyboardFunc(keyboard);
glutMainLoop();
return 0;
}
