Package index • cudaverse

Device and provenance

cuda_available(): Detect a usable CUDA backend
cuda_diagnostics(): Diagnose the optional CUDA runtime
cuda_memory_info(): Inspect CUDA memory
cuda_select_device(): Select a computation device without hiding fallback
cuda_stage(): Record one compute stage
cuda_provenance(): Inspect actual compute provenance

Dense tensors

cuda_tensor(): Create a GPU-aware tensor
Ops(<cudatensor>) `%*%`(<cudatensor>): Arithmetic operators for GPU-aware tensors
`[`(<cudatensor>) `[<-`(<cudatensor>): Subset and replace tensor values
dimnames(<cudatensor>): Inspect tensor dimension labels
tensor_broadcast_to(): Broadcast a tensor to a compatible shape
tensor_device(): Inspect tensor device and backend
tensor_matmul(): Matrix multiplication for tensors
tensor_reshape(): Reshape a tensor without changing its values
tensor_shape(): Inspect tensor shape
tensor_sum() tensor_mean(): Tensor reductions
to_cpu(): Transfer tensor data to base R
to_device(): Transfer a tensor to a device

Sparse matrices

cuda_sparse(): Create a GPU-aware sparse matrix
as_coo() as_csr(): Convert sparse storage format
dimnames(<cudasparse>): Inspect sparse matrix dimension labels
t(<cudasparse>): Transpose a GPU-aware sparse matrix
to_dgCMatrix(): Convert to an R sparse matrix
sparse_info(): Inspect sparse matrix metadata
sparse_matmul_dense(): Sparse matrix by dense matrix multiplication
sparse_matvec(): Sparse matrix-vector multiplication
sparse_normalize(): Normalize sparse rows or columns without densifying
sparse_row_sums() sparse_col_sums(): Sparse row and column reductions

Numerical algorithms

cuda_svd(): GPU-aware singular value decomposition
cuda_pca(): GPU-aware principal component analysis
cuda_distance(): Pairwise distances with an optional CUDA backend
cuda_knn(): k-nearest neighbours
cuda_kmeans(): GPU-aware k-means clustering
predict(<cuda_pca>): Project observations with a fitted CUDA-aware PCA model
predict(<cuda_kmeans>): Assign observations with a fitted CUDA-aware k-means model

Graph workflows

cuda_knn_graph(): Build a sparse graph from nearest neighbours
as_adjacency_matrix(): Extract a graph adjacency matrix
cuda_louvain(): Cluster a graph with Louvain
cuda_leiden(): Cluster a graph with Leiden

Embeddings

cuda_umap(): UMAP embedding
cuda_tsne(): t-SNE embedding
cuda_diffusion_map(): Diffusion-map-style embedding
embedding_coordinates(): Extract embedding coordinates