/external/eigen/Eigen/src/SparseCore/ |
H A D | MappedSparseMatrix.h | 42 inline MappedSparseMatrix(Index rows, Index cols, Index nnz, StorageIndex* outerIndexPtr, StorageIndex* innerIndexPtr, Scalar* valuePtr, StorageIndex* innerNonZeroPtr = 0) argument 43 : Base(rows, cols, nnz, outerIndexPtr, innerIndexPtr, valuePtr, innerNonZeroPtr)
|
H A D | ConservativeSparseSparseProduct.h | 41 // Therefore, we have nnz(lhs*rhs) = nnz(lhs) + nnz(rhs) 51 Index nnz = 0; local 64 indices[nnz] = i; 65 ++nnz; 74 for(Index k=0; k<nnz; ++k) 87 // FIXME reserve nnz non zeros 88 // FIXME implement faster sorting algorithms for very small nnz 93 if((nnz<20 [all...] |
H A D | SparseMap.h | 124 inline SparseMapBase(Index rows, Index cols, Index nnz, IndexPointer outerIndexPtr, IndexPointer innerIndexPtr, argument 126 : m_outerSize(IsRowMajor?rows:cols), m_innerSize(IsRowMajor?cols:rows), m_zero_nnz(0,internal::convert_index<StorageIndex>(nnz)), m_outerIndex(outerIndexPtr), 131 inline SparseMapBase(Index size, Index nnz, IndexPointer innerIndexPtr, ScalarPointer valuePtr) argument 132 : m_outerSize(1), m_innerSize(size), m_zero_nnz(0,internal::convert_index<StorageIndex>(nnz)), m_outerIndex(m_zero_nnz.data()), 195 inline SparseMapBase(Index rows, Index cols, Index nnz, StorageIndex* outerIndexPtr, StorageIndex* innerIndexPtr, argument 197 : Base(rows, cols, nnz, outerIndexPtr, innerIndexPtr, valuePtr, innerNonZerosPtr) 201 inline SparseMapBase(Index size, Index nnz, StorageIndex* innerIndexPtr, Scalar* valuePtr) argument 202 : Base(size, nnz, innerIndexPtr, valuePtr) 237 /** Constructs a read-write Map to a sparse matrix of size \a rows x \a cols, containing \a nnz non-zero coefficients, 245 inline Map(Index rows, Index cols, Index nnz, StorageInde argument 270 Map(Index rows, Index cols, Index nnz, const StorageIndex* outerIndexPtr, const StorageIndex* innerIndexPtr, const Scalar* valuePtr, const StorageIndex* innerNonZerosPtr = 0) argument [all...] |
H A D | SparseBlock.h | 46 Index nnz = 0; local 50 ++nnz; 51 return nnz; 132 Index nnz = tmp.nonZeros(); local 145 if(nnz>free_size) 148 typename SparseMatrixType::Storage newdata(m_matrix.data().allocatedSize() - block_size + nnz); 153 internal::smart_copy(tmp.valuePtr() + tmp_start, tmp.valuePtr() + tmp_start + nnz, newdata.valuePtr() + start); 154 internal::smart_copy(tmp.innerIndexPtr() + tmp_start, tmp.innerIndexPtr() + tmp_start + nnz, newdata.indexPtr() + start); 156 internal::smart_copy(matrix.valuePtr()+end, matrix.valuePtr()+end + tail_size, newdata.valuePtr()+start+nnz); 157 internal::smart_copy(matrix.innerIndexPtr()+end, matrix.innerIndexPtr()+end + tail_size, newdata.indexPtr()+start+nnz); 489 Index nnz = m_block.nonZeros(); local [all...] |
H A D | SparseSelfAdjointView.h | 467 Index nnz = count.sum(); local 470 dest.resizeNonZeros(nnz);
|
/external/tensorflow/tensorflow/core/kernels/ |
H A D | sparse_softmax_op.cc | 70 const int nnz = static_cast<int>(indices_t->dim_size(0)); variable 76 OP_REQUIRES_OK(context, context->allocate_output(0, TensorShape({nnz}),
|
H A D | sparse_tensor_dense_matmul_op_test.cc | 42 static Graph* SparseTensorDenseMatmul(int nnz, int m, int k, int n, argument 45 Tensor a_values(DT_FLOAT, TensorShape({nnz})); 46 Tensor a_indices(DT_INT64, TensorShape({nnz, 2})); 57 for (int32 i = 0; i < nnz; ++i) {
|
H A D | reshape_util.cc | 54 const int64 nnz = input_indices_in.shape().dim_size(0); local 124 TensorShape({nnz, output_rank}), 128 for (int i = 0; i < nnz; ++i) {
|
H A D | sparse_dense_binary_op_shared.cc | 31 // The only output is a vector of flat values with shape [nnz], since this op 109 const int nnz = static_cast<int>(indices_t->dim_size(0)); variable 111 ctx->allocate_output(0, TensorShape({nnz}), &output_values)); 113 ctx, ctx->allocate_temp(DataTypeToEnum<T>::value, TensorShape({nnz}), 132 for (int i = 0; i < nnz; ++i) { \
|
H A D | sparse_tensor_dense_matmul_op_gpu.cu.cc | 31 __global__ void SparseTensorDenseMatMulKernel(int nnz, int m, int b_rows, argument 39 CUDA_1D_KERNEL_LOOP(index, nnz * p) { 73 int nnz = a_values.size(); local 80 // TODO(ebrevdo): Should this be alpha * nnz instead of 81 // out.size()? Perhaps p * nnz ? 82 CudaLaunchConfig config = GetCudaLaunchConfig(p * nnz, d); 86 nnz, m, b_rows, b_cols, p, a_indices.data(), a_values.data(),
|
H A D | sparse_reduce_op.cc | 270 int64 nnz = 0; variable 273 nnz++; 279 0, TensorShape({nnz, reduction.reduced_shape.dims()}), 288 ctx->allocate_output(1, TensorShape({nnz}), &out_values_t));
|
H A D | sparse_tensor_dense_matmul_op.cc | 68 const int64 nnz = a_indices->shape().dim_size(0); variable 69 OP_REQUIRES(ctx, nnz == a_values->NumElements(), 111 OP_REQUIRES(ctx, FastBoundsCheck(nnz * outer_right, int32max), 113 "Cannot use GPU when output.shape[1] * nnz(a) > 2^31")); 251 const std::size_t nnz = a_values.size(); local 267 for (std::size_t i = 0; i < nnz; ++i) { 287 for (std::size_t i = 0; i < nnz; ++i) { \ 303 // columns in the nnz loop.
|
H A D | sparse_conditional_accumulator.h | 164 const int64 nnz = grad_idx->dim_size(0); variable 169 accum_idx_vec_->reserve(nnz); 170 for (int i = 0; i < nnz; i++) { 186 count_element_ = new std::vector<int>(nnz, 1); 323 const int64 nnz = count_element_->size(); local 333 &accum_flat(0,0), &accum_flat(nnz,0), &accum_flat(0,0), 340 for (int64 i = 0; i < nnz; i++) { 387 const int64 nnz = grad_idx_tensor->dim_size(0); local 391 OP_REQUIRES_BOOLEAN(ctx, grad_val_tensor->dim_size(0) == nnz, 392 errors::InvalidArgument("Expected ", nnz, 424 const int64 nnz = accum_idx_vec_->size(); local [all...] |
/external/eigen/Eigen/src/OrderingMethods/ |
H A D | Ordering.h | 134 StorageIndex nnz = StorageIndex(mat.nonZeros()); local 136 StorageIndex Alen = internal::colamd_recommended(nnz, m, n); 144 for(StorageIndex i=0; i < nnz; i++) A(i) = mat.innerIndexPtr()[i];
|
H A D | Eigen_Colamd.h | 189 the COLAMD_RECOMMENDED (nnz, n_row, n_col) macro. It returns -1 if any 190 argument is negative. 2*nnz space is required for the row and column 194 and nnz/5 more space is recommended for run time efficiency. 251 * \param nnz nonzeros in A 257 inline IndexType colamd_recommended ( IndexType nnz, IndexType n_row, IndexType n_col) argument 259 if ((nnz) < 0 || (n_row) < 0 || (n_col) < 0) 262 return (2 * (nnz) + colamd_c (n_col) + colamd_r (n_row) + (n_col) + ((nnz) / 5)); 327 IndexType nnz ; /* nonzeros in A */ local 385 nnz [all...] |
/external/eigen/bench/ |
H A D | sparse_setter.cpp | 107 std::cout << "nnz = " << coords.size() << "\n"; 302 const int nnz, 312 for (int n = 0; n < nnz; n++){ 316 //cumsum the nnz per row to get Bp[] 322 Bp[n_row] = nnz; 325 for(int n = 0; n < nnz; n++){ 384 I nnz = 0; local 397 Aj[nnz] = j; 398 Ax[nnz] = x; 399 nnz 300 coo_tocsr(const int n_row, const int n_col, const int nnz, const Coordinates Aij, const Values Ax, int Bp[], int Bj[], T Bx[]) argument [all...] |
/external/eigen/test/ |
H A D | sparse_basic.cpp | 73 Index nnz = internal::random<int>(1,int(rows)/2); local 77 m2.reserve(VectorXi::Constant(m2.outerSize(), int(nnz))); 79 m2.reserve(m2.outerSize() * nnz); 84 for (Index k=0; k<nnz; ++k)
|
/external/eigen/Eigen/src/IterativeLinearSolvers/ |
H A D | IncompleteCholesky.h | 217 Index nnz = m_L.nonZeros(); local 218 Map<VectorSx> vals(m_L.valuePtr(), nnz); //values 219 Map<VectorIx> rowIdx(m_L.innerIndexPtr(), nnz); //Row indices 329 vals = Map<const VectorSx>(L_save.valuePtr(), nnz); 330 rowIdx = Map<const VectorIx>(L_save.innerIndexPtr(), nnz);
|
/external/eigen/Eigen/src/SparseLU/ |
H A D | SparseLU.h | 434 // Apply the permutation and compute the nnz per column. 537 Index nnz = m_mat.nonZeros(); local 541 Index info = Base::memInit(m, n, nnz, lwork, m_perfv.fillfactor, m_perfv.panel_size, m_glu);
|
/external/libavc/encoder/ |
H A D | ih264e_cavlc.c | 100 * total nnz. 145 /* Compute Runs of zeros for all nnz coefficients except the last 3 */ 298 DEBUG("\n[%d numcoeff, %d numtrailing ones, %d nnz]",u4_total_coeff, 0, u4_nc); 359 DEBUG("\n[%d numcoeff, %d numtrailing ones, %d nnz]",u4_total_coeff, u4_trailing_ones, u4_nc); 593 * pointer to the buffer containing nnz of all the subblks to the top 596 * pointer to the buffer containing nnz of all the subblks to the left 754 /* estimate nnz for the current mb */ 1233 UWORD32 *nnz; local 1244 /* set nnz to zero */ 1246 nnz 1544 UWORD32 *nnz; local [all...] |
/external/eigen/Eigen/src/SuperLUSupport/ |
H A D | SuperLUSupport.h | 139 union {int nnz;int lda;}; member in union:Eigen::SluMatrix::__anon6431::__anon6432 212 res.storage.nnz = internal::convert_index<int>(mat.nonZeros()); 271 res.storage.nnz = mat.nonZeros(); 721 m_l.resizeNonZeros(Lstore->nnz); 723 m_u.resizeNonZeros(Ustore->nnz);
|