/// Elementwise operator /
AutoDiffBlock operator/(const AutoDiffBlock& rhs) const
{
if (jac_.empty() && rhs.jac_.empty()) {
return constant(val_ / rhs.val_);
}
if (jac_.empty()) {
return val_ / rhs;
}
if (rhs.jac_.empty()) {
return *this / rhs.val_;
}
int num_blocks = numBlocks();
std::vector<M> jac(num_blocks);
assert(numBlocks() == rhs.numBlocks());
typedef Eigen::DiagonalMatrix<Scalar, Eigen::Dynamic> D;
D D1 = val_.matrix().asDiagonal();
D D2 = rhs.val_.matrix().asDiagonal();
D D3 = (1.0/(rhs.val_*rhs.val_)).matrix().asDiagonal();
for (int block = 0; block < num_blocks; ++block) {
assert(jac_[block].rows() == rhs.jac_[block].rows());
assert(jac_[block].cols() == rhs.jac_[block].cols());
jac[block] = D3 * (D2*jac_[block] - D1*rhs.jac_[block]);
}
return function(val_ / rhs.val_, jac);
}
string AES128CTR(string in, string key, string nonce){
string cipher=newString(NULL, in.len);
string keystream;
string counter=newString(NULL,8);
int i,j,n;
n=numBlocks(in,16);
if (nonce.len!=8){
nonce=newString(NULL,8);
}
j=0;
for(i=0; i<n; i++){
keystream=AES128EncodeBlock(stringCat(nonce, counter), key);
do{
if(j<in.len){
cipher.c[j]=keystream.c[j%16] ^ in.c[j];
j++;
}
else break;
}while(j%16);
littleEndianIncrement(&counter);
}
return cipher;
}
/// Sizes (number of columns) of Jacobian blocks.
std::vector<int> blockPattern() const
{
const int nb = numBlocks();
std::vector<int> bp(nb);
for (int block = 0; block < nb; ++block) {
bp[block] = jac_[block].cols();
}
return bp;
}
void sobel1(int *h_result, unsigned int *h_pic, int xsize, int ysize, int thresh)
{
int *d_result;
unsigned int *d_pic;
int resultSize = xsize * ysize * 3 * sizeof(int);
int picSize = xsize * ysize * sizeof(int);
cudaMalloc( (void**)&d_result, resultSize);
if( !d_result) {
exit(-1);
}
cudaMalloc( (void**)&d_pic, picSize);
if( !d_pic) {
exit(-1);
}
cudaMemcpy(d_result, h_result, resultSize, cudaMemcpyHostToDevice);
cudaMemcpy(d_pic, h_pic, picSize, cudaMemcpyHostToDevice);
dim3 threadsPerBlock(BLOCKSIZE, BLOCKSIZE);
dim3 numBlocks(ceil((float)ysize/(float)threadsPerBlock.x), ceil((float)xsize/(float)threadsPerBlock.y));
cudaEvent_t start, stop;
cudaEventCreate(&start);
cudaEventCreate(&stop);
{ __set_CUDAConfig(numBlocks, threadsPerBlock );
d_sobel1 (d_result, d_pic, xsize, ysize, thresh);}
cudaEventSynchronize(stop);
float elapsedTime;
cudaEventElapsedTime(&elapsedTime, start, stop);
cudaEventDestroy(start);
cudaEventDestroy(stop);
cudaMemcpy(h_result, d_result, resultSize, cudaMemcpyDeviceToHost);
cudaMemcpy(h_pic, d_pic, picSize, cudaMemcpyDeviceToHost);
cudaFree(d_result);
cudaFree(d_pic);
}
/// Elementwise operator +=
AutoDiffBlock& operator+=(const AutoDiffBlock& rhs)
{
if (jac_.empty()) {
jac_ = rhs.jac_;
} else if (!rhs.jac_.empty()) {
assert (numBlocks() == rhs.numBlocks());
assert (value().size() == rhs.value().size());
const int num_blocks = numBlocks();
for (int block = 0; block < num_blocks; ++block) {
assert(jac_[block].rows() == rhs.jac_[block].rows());
assert(jac_[block].cols() == rhs.jac_[block].cols());
jac_[block] += rhs.jac_[block];
}
}
val_ += rhs.val_;
return *this;
}
/// Elementwise operator -
AutoDiffBlock operator-(const AutoDiffBlock& rhs) const
{
if (jac_.empty() && rhs.jac_.empty()) {
return constant(val_ - rhs.val_);
}
if (jac_.empty()) {
return val_ - rhs;
}
if (rhs.jac_.empty()) {
return *this - rhs.val_;
}
std::vector<M> jac = jac_;
assert(numBlocks() == rhs.numBlocks());
int num_blocks = numBlocks();
for (int block = 0; block < num_blocks; ++block) {
assert(jac[block].rows() == rhs.jac_[block].rows());
assert(jac[block].cols() == rhs.jac_[block].cols());
jac[block] -= rhs.jac_[block];
}
return function(val_ - rhs.val_, jac);
}
string AES128DecodeCBC(string in, string key, string IV){
if(in.len%16){
return NULLSTRING;
}
string out = NULLSTRING;
string *blocks = blockString(in, 16);
int num = numBlocks(in, 16);
for(int i=num-1; i>=0; i--){
blocks[i] = AES128DecodeBlock(blocks[i],key);
if(i==0){
out = stringCat(stringXOR(blocks[i], IV), out);
} else{
out = stringCat(stringXOR(blocks[i-1], blocks[i]),out);
}
}
return out;
}
string AES128EncodeCBC(string in, string key, string IV){
string out = NULLSTRING;
if(!validatePKCS7Padding(in)){
in = PKCS7PadString(in, 16);
}
string *blocks = blockString(in, 16);
int num = numBlocks(in, 16);
for(int i=0; i<num; i++){
if(i==0){
blocks[0] = stringXOR(blocks[0], IV);
}else{
blocks[i] = stringXOR(blocks[i], newString(&out.c[(i-1)*16],16));
}
out = stringCat(out, AES128EncodeBlock(blocks[i], key));
}
return out;
}
void DynamicMarchingTetrahedra::update(GLContext * gl) {
GLContext::Program * prog = gl->getProgram(m_sceneShader.c_str());
prog->use();
Vec3i numBlocks(64);
Vec3i threadBlockSize(4);
Vec3i gridSize = (numBlocks + threadBlockSize - 1) / threadBlockSize;
gl->setUniform(prog->getUniformLoc("cubeInfo"), m_cubeInfo);
gl->setUniform(prog->getUniformLoc("isPrefixSumPass"), true);
gl->setUniform(prog->getUniformLoc("numCubes"), numBlocks);
gl->setUniform(prog->getUniformLoc("sync1"), m_disp1);
gl->setUniform(prog->getUniformLoc("sync2"), m_disp2);
gl->setUniform(prog->getUniformLoc("sync3"), m_disp3);
gl->setUniform(prog->getUniformLoc("sync4"), m_disp4);
gl->setUniform(prog->getUniformLoc("maxTetrahedras"), 6);
glBindBufferBase(GL_SHADER_STORAGE_BUFFER, 2, m_buffers[DMT_Buffer_Types::INDEX_BUFFER]);
glDispatchCompute(gridSize.x, gridSize.y, gridSize.z);
glMemoryBarrier(GL_SHADER_STORAGE_BARRIER_BIT);
GPUPrefixScan::scan(gl, m_buffers[DMT_Buffer_Types::INDEX_BUFFER], m_buffers[DMT_Buffer_Types::BLOCK_BUFFER], 100*100*100);
prog->use();
gl->setUniform(prog->getUniformLoc("isPrefixSumPass"), false);
glBindBufferBase(GL_SHADER_STORAGE_BUFFER, 1, m_buffers[DMT_Buffer_Types::MESH_BUFFER]);
glDispatchCompute(gridSize.x, gridSize.y, gridSize.z);
glMemoryBarrier(GL_SHADER_STORAGE_BARRIER_BIT);
m_numTriangles = GPUPrefixScan::getSum(gl, m_buffers[DMT_Buffer_Types::BLOCK_BUFFER]);
}
void
SideSetsAroundSubdomain::modify()
{
// Reference the the libMesh::MeshBase
MeshBase & mesh = _mesh_ptr->getMesh();
// Extract the 'first' block ID
SubdomainID block_id = *blockIDs().begin();
// Extract the SubdomainID
if (numBlocks() > 1)
mooseWarning("SideSetsAroundSubdomain only acts on a single subdomain, but multiple were provided: only the " << block_id << "' subdomain is being used.");
// Create the boundary IDs from the list of names provided (the true flag creates ids from unknown names)
std::vector<BoundaryID> boundary_ids = _mesh_ptr->getBoundaryIDs(_boundary_names, true);
// construct the FE object so we can compute normals of faces
setup();
Point face_normal;
bool add_to_bdy = true;
// Get a reference to our BoundaryInfo object for later use
BoundaryInfo & boundary_info = mesh.get_boundary_info();
// Loop over the elements
MeshBase::const_element_iterator el = mesh.active_elements_begin();
const MeshBase::const_element_iterator end_el = mesh.active_elements_end();
for (; el != end_el ; ++el)
{
const Elem* elem = *el;
SubdomainID curr_subdomain = elem->subdomain_id();
// We only need to loop over elements in the source subdomain
if (curr_subdomain != block_id)
continue;
for (unsigned int side = 0; side < elem->n_sides(); ++side)
{
const Elem * neighbor = elem->neighbor(side);
if (neighbor == NULL || // element on boundary OR
neighbor->subdomain_id() != block_id) // neighboring element is on a different subdomain
{
if (_using_normal)
{
_fe_face->reinit(elem, side);
face_normal = _fe_face->get_normals()[0];
add_to_bdy = (_normal*face_normal >= 1.0 - _normal_tol);
}
// Add the boundaries, if appropriate
if (add_to_bdy)
for (const auto & boundary_id : boundary_ids)
boundary_info.add_side(elem, side, boundary_id);
}
}
}
finalize();
// Assign the supplied names to the newly created side sets
for (unsigned int i = 0; i < boundary_ids.size(); ++i)
boundary_info.sideset_name(boundary_ids[i]) = _boundary_names[i];
}
// Number of sets.
int numSets() const { return _comm->getSize() / numBlocks(); }
