void
LiGL2D::setVbo(int spaceVect)
{
GLuint oldVbo = 0;
GLuint newVbo = 0;
if(vbo != 0){
oldVbo = vbo;
vbo = 0;
}
if(iw != 0 && ih !=0){
GLint bsize;
// create buffer object
unsigned int size = ((int)iw/(spaceVect+1))*((int)ih/(spaceVect+1)) * 6 * sizeof(float2);
glGenBuffers( 1, &newVbo);
glBindBuffer( GL_ARRAY_BUFFER, newVbo);
// initialize buffer object
glBufferData( GL_ARRAY_BUFFER, size, 0, GL_DYNAMIC_DRAW);
glGetBufferParameterivARB(GL_ARRAY_BUFFER_ARB, GL_BUFFER_SIZE_ARB, &bsize);
glBindBuffer( GL_ARRAY_BUFFER, 0);
// register buffer object with CUDA
CUDA_SAFE_CALL(cudaGLRegisterBufferObject(newVbo));
sVbo = ((int)iw/(spaceVect+1))*((int)ih/(spaceVect+1))*6;
vbo = newVbo;
emit sendVbo(vbo);
}
if(oldVbo != 0){
CUDA_SAFE_CALL(cudaGLUnregisterBufferObject(oldVbo));
glDeleteBuffers(1, &oldVbo);
}
}
// Map for data access
U8* LLVertexBuffer::mapBuffer(S32 access)
{
LLMemType mt(LLMemType::MTYPE_VERTEX_DATA);
if (mFinal)
{
llwarns << "LLVertexBuffer::mapBuffer() called on a finalized buffer." << llendl;
}
if (!useVBOs() && !mMappedData && !mMappedIndexData)
{
llwarns << "LLVertexBuffer::mapBuffer() called on unallocated buffer." << llendl;
}
if (!mLocked && useVBOs())
{
setBuffer(0);
mLocked = TRUE;
stop_glerror();
mMappedData = (U8*) glMapBufferARB(GL_ARRAY_BUFFER_ARB, GL_WRITE_ONLY_ARB);
stop_glerror();
mMappedIndexData = (U8*) glMapBufferARB(GL_ELEMENT_ARRAY_BUFFER_ARB, GL_WRITE_ONLY_ARB);
stop_glerror();
if (!mMappedData)
{
//--------------------
//print out more debug info before crash
llinfos << "vertex buffer size: (num verts : num indices) = " << getNumVerts() << " : " << getNumIndices() << llendl ;
GLint size ;
glGetBufferParameterivARB(GL_ARRAY_BUFFER_ARB, GL_BUFFER_SIZE_ARB, &size) ;
llinfos << "GL_ARRAY_BUFFER_ARB size is " << size << llendl ;
//--------------------
GLint buff;
glGetIntegerv(GL_ARRAY_BUFFER_BINDING_ARB, &buff);
if (buff != mGLBuffer)
{
llwarns << "Invalid GL vertex buffer bound: " << buff << llendl;
}
llwarns << "glMapBuffer returned NULL (no vertex data)" << llendl;
}
if (!mMappedIndexData)
{
GLint buff;
glGetIntegerv(GL_ELEMENT_ARRAY_BUFFER_BINDING_ARB, &buff);
if (buff != mGLIndices)
{
llwarns << "Invalid GL index buffer bound: " << buff << llendl;
}
llwarns << "glMapBuffer returned NULL (no index data)" << llendl;
}
sMappedCount++;
}
return mMappedData;
}
void VBO::unload(bool save)
{
// Clean up buffer_copy, if it exists.
delete[] buffer_copy;
buffer_copy = 0;
// Save data before unloading.
if (save)
{
VertexBuffer::Bind bind(*this);
GLint size;
glGetBufferParameterivARB(getTarget(), GL_BUFFER_SIZE, &size);
const char *src = static_cast<char *>(map());
if (src)
{
buffer_copy = new char[size];
memcpy(buffer_copy, src, size);
unmap();
}
}
glDeleteBuffers(1, &vbo);
vbo = 0;
}
void VBO::unload(bool save)
{
// Save data before unloading.
if (save)
{
VertexBuffer::Bind bind(*this);
GLint size;
glGetBufferParameterivARB(getTarget(), GL_BUFFER_SIZE, &size);
map(); // saves buffer content to memory_map.
unmap();
}
glDeleteBuffersARB(1, &vbo);
vbo = 0;
}
void
LiGL2D::init(int image_width, int image_height)
{
GLint bsize;
iw = image_width;
ih = image_height;
imW = iw;
imH = ih;
makeCurrent();
// allocation du pbo
glGenBuffers(1, &pbo);
glBindBuffer(GL_ARRAY_BUFFER, pbo);
glBufferData(GL_ARRAY_BUFFER, image_height*image_width* 4*sizeof(GLubyte),NULL, GL_DYNAMIC_DRAW);
glGetBufferParameterivARB(GL_ARRAY_BUFFER_ARB, GL_BUFFER_SIZE_ARB, &bsize);
glBindBuffer(GL_ARRAY_BUFFER, 0);
CUDA_SAFE_CALL(cudaGLRegisterBufferObject(pbo));
setVbo(10);
// allocation de la texture d'affichage
createTexture(&tex, image_width, image_height);
glBindBufferARB(GL_PIXEL_UNPACK_BUFFER_ARB, 0);
initialise = true;
std::cout << "buffer pixel num " << pbo << " taille : " << bsize << "\n";
emit sendPbo(pbo);
}
static void init_vbo(ParticleEmitter* emitter)
{
if(!GL_ARB_vertex_buffer_object)
{
rb_raise(rb_eRuntimeError, "Ashton::ParticleEmitter requires GL_ARB_vertex_buffer_object, which is not supported by your OpenGL");
}
int num_vertices = emitter->max_particles * VERTICES_IN_PARTICLE;
emitter->color_array = ALLOC_N(Color_i, num_vertices);
emitter->color_array_offset = 0;
emitter->texture_coords_array = ALLOC_N(Vertex2d, num_vertices);
emitter->texture_coords_array_offset = sizeof(Color_i) * num_vertices;
emitter->vertex_array = ALLOC_N(Vertex2d, num_vertices);
emitter->vertex_array_offset = (sizeof(Color_i) + sizeof(Vertex2d)) * num_vertices;
// Create the VBO, but don't upload any data yet.
int data_size = (sizeof(Color_i) + sizeof(Vertex2d) + sizeof(Vertex2d)) * num_vertices;
glGenBuffersARB(1, &emitter->vbo_id);
glBindBufferARB(GL_ARRAY_BUFFER_ARB, emitter->vbo_id);
glBufferDataARB(GL_ARRAY_BUFFER_ARB, data_size, NULL, GL_STREAM_DRAW_ARB);
// Check the buffer was actually created.
int buffer_size = 0;
glGetBufferParameterivARB(GL_ARRAY_BUFFER_ARB, GL_BUFFER_SIZE_ARB, &buffer_size);
if(buffer_size != data_size)
{
rb_raise(rb_eRuntimeError, "Failed to create a VBO [%d bytes] to hold emitter data.", data_size);
}
glBindBufferARB(GL_ARRAY_BUFFER_ARB, 0);
return;
}
int main(int argc, char **argv)
{
int devID;
cudaDeviceProp deviceProps;
printf("%s Starting...\n\n", sSDKname);
printf("[%s] - [OpenGL/CUDA simulation] starting...\n", sSDKname);
// 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.
if (false == initGL(&argc, argv))
{
exit(EXIT_SUCCESS);
}
// use command-line specified CUDA device, otherwise use device with highest Gflops/s
devID = findCudaGLDevice(argc, (const char **)argv);
// get number of SMs on this GPU
checkCudaErrors(cudaGetDeviceProperties(&deviceProps, devID));
printf("CUDA device [%s] has %d Multi-Processors\n",
deviceProps.name, deviceProps.multiProcessorCount);
// automated build testing harness
if (checkCmdLineFlag(argc, (const char **)argv, "file"))
{
getCmdLineArgumentString(argc, (const char **)argv, "file", &ref_file);
}
// Allocate and initialize host data
GLint bsize;
sdkCreateTimer(&timer);
sdkResetTimer(&timer);
hvfield = (cData *)malloc(sizeof(cData) * DS);
memset(hvfield, 0, sizeof(cData) * DS);
// Allocate and initialize device data
cudaMallocPitch((void **)&dvfield, &tPitch, sizeof(cData)*DIM, DIM);
cudaMemcpy(dvfield, hvfield, sizeof(cData) * DS,
cudaMemcpyHostToDevice);
// Temporary complex velocity field data
cudaMalloc((void **)&vxfield, sizeof(cData) * PDS);
cudaMalloc((void **)&vyfield, sizeof(cData) * PDS);
setupTexture(DIM, DIM);
bindTexture();
// Create particle array
particles = (cData *)malloc(sizeof(cData) * DS);
memset(particles, 0, sizeof(cData) * DS);
initParticles(particles, DIM, DIM);
// Create CUFFT transform plan configuration
cufftPlan2d(&planr2c, DIM, DIM, CUFFT_R2C);
cufftPlan2d(&planc2r, DIM, DIM, CUFFT_C2R);
// TODO: update kernels to use the new unpadded memory layout for perf
// rather than the old FFTW-compatible layout
cufftSetCompatibilityMode(planr2c, CUFFT_COMPATIBILITY_FFTW_PADDING);
cufftSetCompatibilityMode(planc2r, CUFFT_COMPATIBILITY_FFTW_PADDING);
glGenBuffersARB(1, &vbo);
glBindBufferARB(GL_ARRAY_BUFFER_ARB, vbo);
glBufferDataARB(GL_ARRAY_BUFFER_ARB, sizeof(cData) * DS,
particles, GL_DYNAMIC_DRAW_ARB);
glGetBufferParameterivARB(GL_ARRAY_BUFFER_ARB, GL_BUFFER_SIZE_ARB, &bsize);
if (bsize != (sizeof(cData) * DS))
goto EXTERR;
glBindBufferARB(GL_ARRAY_BUFFER_ARB, 0);
checkCudaErrors(cudaGraphicsGLRegisterBuffer(&cuda_vbo_resource, vbo, cudaGraphicsMapFlagsNone));
getLastCudaError("cudaGraphicsGLRegisterBuffer failed");
if (ref_file)
{
autoTest(argv);
cleanup();
cudaDeviceReset();
printf("[fluidsGL] - Test Results: %d Failures\n", g_TotalErrors);
exit(g_TotalErrors == 0 ? EXIT_SUCCESS : EXIT_FAILURE);
}
else
{
atexit(cleanup);
glutMainLoop();
}
cudaDeviceReset();
if (!ref_file)
{
exit(EXIT_SUCCESS);
}
return 0;
EXTERR:
printf("Failed to initialize GL extensions.\n");
cudaDeviceReset();
exit(EXIT_FAILURE);
}
int main(int argc, char **argv)
{
int devID;
cudaDeviceProp deviceProps;
printf("%s Starting...\n\n", sSDKname);
printf("[%s] - [OpenGL/CUDA simulation] starting...\n", sSDKname);
// 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.
if (false == initGL(&argc, argv))
{
exit(EXIT_SUCCESS);
}
// use command-line specified CUDA device, otherwise use device with highest Gflops/s
#ifndef OPTIMUS
devID = findCudaGLDevice(argc, (const char **)argv);
#else
devID = gpuGetMaxGflopsDeviceId();
#endif
// get number of SMs on this GPU
checkCudaErrors(cudaGetDeviceProperties(&deviceProps, devID));
printf("CUDA device [%s] has %d Multi-Processors\n",
deviceProps.name, deviceProps.multiProcessorCount);
// automated build testing harness
if (checkCmdLineFlag(argc, (const char **)argv, "file"))
{
getCmdLineArgumentString(argc, (const char **)argv, "file", &ref_file);
}
// Allocate and initialize host data
GLint bsize;
sdkCreateTimer(&timer);
sdkResetTimer(&timer);
hvfield = (cData *)malloc(sizeof(cData) * DS);
memset(hvfield, 0, sizeof(cData) * DS);
// Allocate and initialize device data
cudaMallocPitch((void **)&dvfield, &tPitch, sizeof(cData)*DIM, DIM);
cudaMemcpy(dvfield, hvfield, sizeof(cData) * DS,
cudaMemcpyHostToDevice);
// Temporary complex velocity field data
cudaMalloc((void **)&vxfield, sizeof(cData) * PDS);
cudaMalloc((void **)&vyfield, sizeof(cData) * PDS);
setupTexture(DIM, DIM);
bindTexture();
// Create particle array in host memory
particles = (cData *)malloc(sizeof(cData) * DS);
memset(particles, 0, sizeof(cData) * DS);
#ifdef BROADCAST
int step = 1;
// Broadcasted visualization stepping.
if (argc > 3)
step = atoi(argv[3]);
// Create additional space to store particle packets
// for broadcasting.
wstep = step; hstep = step;
int npackets = sizeof(float) * (DIM / wstep) * (DIM / hstep) / UdpBroadcastServer::PacketSize;
if (sizeof(float) * (DIM / wstep) * (DIM / hstep) % UdpBroadcastServer::PacketSize)
npackets++;
packets = (char*)malloc(npackets *
(UdpBroadcastServer::PacketSize + sizeof(unsigned int)));
#endif
initParticles(particles, DIM, DIM);
#if defined(OPTIMUS) || defined(BROADCAST)
// Create particle array in device memory
cudaMalloc((void **)&particles_gpu, sizeof(cData) * DS);
cudaMemcpy(particles_gpu, particles, sizeof(cData) * DS, cudaMemcpyHostToDevice);
#endif
// Create CUFFT transform plan configuration
cufftPlan2d(&planr2c, DIM, DIM, CUFFT_R2C);
cufftPlan2d(&planc2r, DIM, DIM, CUFFT_C2R);
// TODO: update kernels to use the new unpadded memory layout for perf
// rather than the old FFTW-compatible layout
cufftSetCompatibilityMode(planr2c, CUFFT_COMPATIBILITY_FFTW_PADDING);
cufftSetCompatibilityMode(planc2r, CUFFT_COMPATIBILITY_FFTW_PADDING);
glGenBuffersARB(1, &vbo);
glBindBufferARB(GL_ARRAY_BUFFER_ARB, vbo);
glBufferDataARB(GL_ARRAY_BUFFER_ARB, sizeof(cData) * DS,
particles, GL_DYNAMIC_DRAW_ARB);
glGetBufferParameterivARB(GL_ARRAY_BUFFER_ARB, GL_BUFFER_SIZE_ARB, &bsize);
if (bsize != (sizeof(cData) * DS))
goto EXTERR;
glBindBufferARB(GL_ARRAY_BUFFER_ARB, 0);
#ifndef OPTIMUS
checkCudaErrors(cudaGraphicsGLRegisterBuffer(&cuda_vbo_resource, vbo, cudaGraphicsMapFlagsNone));
getLastCudaError("cudaGraphicsGLRegisterBuffer failed");
#endif
if (ref_file)
{
autoTest(argv);
cleanup();
// 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();
printf("[fluidsGL] - Test Results: %d Failures\n", g_TotalErrors);
exit(g_TotalErrors == 0 ? EXIT_SUCCESS : EXIT_FAILURE);
}
else
{
#ifdef BROADCAST
const char *sv_addr = "127.0.0:9097";
const char *bc_addr = "127.255.255.2:9097";
// Server address
if (argc > 2)
sv_addr = argv[2];
// Broadcast address
if (argc > 1)
bc_addr = argv[1];
server.reset(new UdpBroadcastServer(sv_addr, bc_addr));
// Listen to clients' feedbacks in a separate thread.
{
pthread_t tid;
pthread_create(&tid, NULL, &feedback_listener, &step);
}
// Broadcast the particles state in a separate thread.
{
pthread_t tid;
pthread_create(&tid, NULL, &broadcaster, &step);
}
#endif
#if defined (__APPLE__) || defined(MACOSX)
atexit(cleanup);
#else
glutCloseFunc(cleanup);
#endif
glutMainLoop();
}
// 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();
if (!ref_file)
{
exit(EXIT_SUCCESS);
}
return 0;
EXTERR:
printf("Failed to initialize GL extensions.\n");
// 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_FAILURE);
}
// Map for data access
volatile U8* LLVertexBuffer::mapVertexBuffer(S32 type, S32 access)
{
LLMemType mt(LLMemType::MTYPE_VERTEX_DATA);
if (mFinal)
{
llerrs << "LLVertexBuffer::mapVeretxBuffer() called on a finalized buffer." << llendl;
}
if (!useVBOs() && !mMappedData && !mMappedIndexData)
{
llerrs << "LLVertexBuffer::mapVertexBuffer() called on unallocated buffer." << llendl;
}
if (!mVertexLocked && useVBOs())
{
{
setBuffer(0, type);
mVertexLocked = TRUE;
stop_glerror();
if(sDisableVBOMapping)
{
allocateClientVertexBuffer() ;
}
else
{
mMappedData = (U8*) glMapBufferARB(GL_ARRAY_BUFFER_ARB, GL_WRITE_ONLY_ARB);
}
stop_glerror();
}
if (!mMappedData)
{
if(!sDisableVBOMapping)
{
//--------------------
//print out more debug info before crash
llinfos << "vertex buffer size: (num verts : num indices) = " << getNumVerts() << " : " << getNumIndices() << llendl ;
GLint size ;
glGetBufferParameterivARB(GL_ARRAY_BUFFER_ARB, GL_BUFFER_SIZE_ARB, &size) ;
llinfos << "GL_ARRAY_BUFFER_ARB size is " << size << llendl ;
//--------------------
GLint buff;
glGetIntegerv(GL_ARRAY_BUFFER_BINDING_ARB, &buff);
if ((GLuint)buff != mGLBuffer)
{
llerrs << "Invalid GL vertex buffer bound: " << buff << llendl;
}
llerrs << "glMapBuffer returned NULL (no vertex data)" << llendl;
}
else
{
llerrs << "memory allocation for vertex data failed." << llendl ;
}
}
sMappedCount++;
}
return mMappedData;
}
JNIEXPORT void JNICALL Java_org_lwjgl_opengl_ARBBufferObject_nglGetBufferParameterivARB(JNIEnv *env, jclass clazz, jint target, jint pname, jlong params, jlong function_pointer) {
GLint *params_address = (GLint *)(intptr_t)params;
glGetBufferParameterivARBPROC glGetBufferParameterivARB = (glGetBufferParameterivARBPROC)((intptr_t)function_pointer);
glGetBufferParameterivARB(target, pname, params_address);
}
// Map for data access
volatile U8* LLVertexBuffer::mapVertexBuffer(S32 type, S32 index)
{
LLMemType mt(LLMemType::MTYPE_VERTEX_DATA);
if (mFinal)
{
llerrs << "LLVertexBuffer::mapVeretxBuffer() called on a finalized buffer." << llendl;
}
if (!useVBOs() && !mMappedData && !mMappedIndexData)
{
llerrs << "LLVertexBuffer::mapVertexBuffer() called on unallocated buffer." << llendl;
}
if (!mVertexLocked && useVBOs())
{
{
setBuffer(0, type);
mVertexLocked = TRUE;
stop_glerror();
if(sDisableVBOMapping)
{
allocateClientVertexBuffer() ;
}
else
{
U8* src = NULL;
{
src = (U8*) glMapBufferARB(GL_ARRAY_BUFFER_ARB, GL_WRITE_ONLY_ARB);
}
mMappedData = LL_NEXT_ALIGNED_ADDRESS<U8>(src);
mAlignedOffset = mMappedData - src;
}
stop_glerror();
}
if (!mMappedData)
{
log_glerror();
//check the availability of memory
U32 avail_phy_mem, avail_vir_mem;
LLMemoryInfo::getAvailableMemoryKB(avail_phy_mem, avail_vir_mem) ;
llinfos << "Available physical mwmory(KB): " << avail_phy_mem << llendl ;
llinfos << "Available virtual memory(KB): " << avail_vir_mem << llendl;
if(!sDisableVBOMapping)
{
//--------------------
//print out more debug info before crash
llinfos << "vertex buffer size: (num verts : num indices) = " << getNumVerts() << " : " << getNumIndices() << llendl ;
GLint size ;
glGetBufferParameterivARB(GL_ARRAY_BUFFER_ARB, GL_BUFFER_SIZE_ARB, &size) ;
llinfos << "GL_ARRAY_BUFFER_ARB size is " << size << llendl ;
//--------------------
GLint buff;
glGetIntegerv(GL_ARRAY_BUFFER_BINDING_ARB, &buff);
if ((GLuint)buff != mGLBuffer)
{
llerrs << "Invalid GL vertex buffer bound: " << buff << llendl;
}
llerrs << "glMapBuffer returned NULL (no vertex data)" << llendl;
}
else
{
llerrs << "memory allocation for vertex data failed." << llendl ;
}
}
sMappedCount++;
}
return mMappedData+mOffsets[type]+ (mIsStrided ? mStride : sTypeSize[type])*index;
}
JNIEXPORT void JNICALL Java_org_lwjgl_opengl_ARBBufferObject_nglGetBufferParameterivARB(JNIEnv *env, jclass clazz, jint target, jint pname, jobject params, jint params_position, jlong function_pointer) {
GLint *params_address = ((GLint *)(*env)->GetDirectBufferAddress(env, params)) + params_position;
glGetBufferParameterivARBPROC glGetBufferParameterivARB = (glGetBufferParameterivARBPROC)((intptr_t)function_pointer);
glGetBufferParameterivARB(target, pname, params_address);
}
