| Matrix multiplication for adgCMatrix | %*%,adgCMatrix,adgCMatrix-method %*%,adgCMatrix,adgeMatrix-method %*%,adgCMatrix,ANY-method %*%,adgCMatrix,dgCMatrix-method %*%,adgCMatrix,dgeMatrix-method %*%,adgCMatrix,Matrix-method %*%,adgCMatrix,matrix-method |
| Scaled matrix multiply with optional bias: alpha*(A%*%B) + beta*C | addmm |
| Create a backend-aware sparse matrix | adgCMatrix |
| Sparse column-compressed matrix with backend-dispatch metadata | adgCMatrix-class |
| Create a backend-aware dense matrix | adgeMatrix |
| Dense general matrix with backend-dispatch metadata | adgeMatrix-class |
| Sparse logical matrix with backend-dispatch metadata | adlgCMatrix-class |
| Dense logical matrix with backend-dispatch metadata | adlgeMatrix-class |
| Row and column argmax/argmin | am_argreduce am_colargmax am_colargmin am_rowargmax am_rowargmin |
| In-place elementwise operation on a resident handle | am_ewise_inplace |
| QR decomposition of an amatrix object | am_qr |
| Scatter mean by group labels | am_scatter_mean |
| Backend-dispatched sweep | am_sweep |
| In-place broadcast sweep on a resident handle | am_sweep_inplace |
| Query the capabilities of a registered backend | amatrix_backend_capabilities |
| Query the features of a registered backend | amatrix_backend_features |
| Run a canary health probe against a registered backend | amatrix_backend_health_probe |
| Tabulate dispatch plans across multiple operations | amatrix_backend_matrix |
| List names of all registered backends | amatrix_backend_names |
| Compute the dispatch plan for a single operation | amatrix_backend_plan |
| Query the precision modes supported by a registered backend | amatrix_backend_precision_modes |
| Summarise the status of registered backends | amatrix_backend_status |
| Report amatrix benchmark status across ops and backends | amatrix_benchmark_report |
| Bind an amatrix object to resident backend storage | amatrix_bind_resident |
| Get or set the model cache maximum size | amatrix_cache_max_size amatrix_set_cache_max_size |
| Calibrate GPU dispatch thresholds for this machine | amatrix_calibrate |
| Retrieve the current calibration state | amatrix_calibration_info |
| Compile a reusable matrix-product plan | amatrix_compile_product |
| Get the session-level default dispatch policy | amatrix_default_policy |
| Get the session-level default precision mode | amatrix_default_precision |
| Low-level backend dispatch for a single operation | amatrix_dispatch_op |
| Collect full dispatch information for an aMatrix object | amatrix_execution_info |
| Explain dispatch decisions for an aMatrix operation | amatrix_explain |
| Return the amatrix backend fallback log | amatrix_fallback_log |
| Clear the amatrix backend fallback log | amatrix_fallback_log_reset |
| Free stale GPU residency entries and optionally flush the model cache | amatrix_gc |
| GPU backend status: why am I (not) on the GPU? | amatrix_gpu_status |
| Force materialization of an aMatrix to a host Matrix object | amatrix_materialize_host |
| Report GPU residency and model cache usage | amatrix_memory_stats |
| Prepare operands for a repeated matrix product | amatrix_prepare_operands |
| Register a backend with the amatrix dispatch system | amatrix_register_backend |
| Release GPU-resident data held by an amatrix object | amatrix_release_resident |
| Query GPU residency state of an aMatrix object | amatrix_residency_info |
| Choose a residency-capable accelerator backend for a hot path | amatrix_resident_backend_for |
| Set the session-level default dispatch policy | amatrix_set_default_policy |
| Set the session-level default precision mode | amatrix_set_default_precision |
| Enable GPU acceleration for this session | amatrix_use_gpu |
| Warm up GPU backends to eliminate cold-start latency | amatrix_warm |
| Virtual base class for backend-aware matrices | aMatrix-class |
| Cholesky factorization result | amChol-class |
| LU factorization result for general square matrices | amLU-class |
| Truncated SVD factorization result | amSVD-class |
| Fit linear models with array-shaped response | array_lm |
| Coerce an object to adgCMatrix | as_adgCMatrix |
| Coerce an object to adgeMatrix | as_adgeMatrix |
| Convert a resident handle back to an adgeMatrix | as_adgeMatrix.resident_handle |
| Coerce amatrix objects to base R types | as.array,adgCMatrix-method as.array,adgeMatrix-method as.matrix,adgCMatrix-method as.matrix,adgeMatrix-method as.matrix,amChol-method as.matrix,aTransposeView-method as.matrix,KronMatrix-method as.matrix.adgCMatrix as.matrix.adgeMatrix as.matrix.aTransposeView as.numeric,adgeMatrix-method as.vector,adgeMatrix-method |
| Lazy transpose view of an adgeMatrix | aTransposeView-class |
| Batch Cholesky factorization | batch_chol |
| Batch crossproduct | batch_crossprod |
| Batch triangular solve | batch_solve |
| Block Lanczos SVD via block Krylov iteration | block_lanczos block_svd |
| Extract the diagonal of a Cholesky factor | chol_diag |
| Compute the Cholesky factorization of an adgeMatrix | chol_factor |
| Log-determinant from a Cholesky factor | chol_logdet |
| Solve a linear system using a Cholesky factor | chol_solve |
| Solve many right-hand-side batches with one Cholesky factor | chol_solve_batches |
| Cholesky factorization for adgCMatrix | chol,adgCMatrix-method |
| Cholesky factorization for adgeMatrix | chol,adgeMatrix-method |
| Compute a correlation matrix | correlation |
| Covariance-to-correlation methods for amatrix objects | cov2cor,adgCMatrix-method cov2cor,adgeMatrix-method |
| Backend-dispatched covariance matrix | covariance |
| Cross-product plus diagonal perturbation | crossprod_add_diag |
| Weighted cross-product X'WX | crossprod_weighted |
| Cross-product methods for adgCMatrix | crossprod,adgCMatrix,adgCMatrix-method crossprod,adgCMatrix,adgeMatrix-method crossprod,adgCMatrix,ANY-method crossprod,adgCMatrix,dgCMatrix-method crossprod,adgCMatrix,dgeMatrix-method crossprod,adgCMatrix,Matrix-method crossprod,adgCMatrix,matrix-method crossprod,adgCMatrix,missing-method tcrossprod,adgCMatrix,adgCMatrix-method tcrossprod,adgCMatrix,adgeMatrix-method tcrossprod,adgCMatrix,ANY-method tcrossprod,adgCMatrix,dgCMatrix-method tcrossprod,adgCMatrix,dgeMatrix-method tcrossprod,adgCMatrix,Matrix-method tcrossprod,adgCMatrix,matrix-method tcrossprod,adgCMatrix,missing-method |
| Cross-product methods for adgeMatrix | crossprod,adgeMatrix,ANY-method crossprod,adgeMatrix,missing-method tcrossprod,adgeMatrix,ANY-method tcrossprod,adgeMatrix,missing-method |
| GPU-accelerated pairwise distance matrix | dist_matrix |
| Inner product of two vectors or matrices | dot |
| Eigendecomposition for adgCMatrix | eigen,adgCMatrix-method |
| Eigendecomposition for adgeMatrix | eigen,adgeMatrix-method |
| Symmetric eigendecomposition | eigh |
| Element-wise operations | ewise |
| Generalised matrix multiply (BLAS DGEMM interface) | gemm |
| GPU-accelerated truncated SVD via irlba | irlba |
| GPU-native truncated SVD via Lanczos bidiagonalization | irlba_native |
| GPU-accelerated pairwise kernel matrix | kernel_matrix |
| Eager Kronecker product | kron |
| Construct a lazy Kronecker product | kron_matrix |
| Kronecker product of backend-aware matrices | kronecker,adgCMatrix,adgCMatrix-method kronecker,adgCMatrix,adgeMatrix-method kronecker,adgCMatrix,matrix-method kronecker,adgeMatrix,adgCMatrix-method kronecker,adgeMatrix,adgeMatrix-method kronecker,adgeMatrix,matrix-method kronecker,matrix,adgCMatrix-method kronecker,matrix,adgeMatrix-method kronecker-methods |
| Lazy Kronecker product of two matrices | KronMatrix-class |
| Fit a single linear model | lm_fit |
| Leave-one-out cross-validation for linear models | lm_loo_cv |
| Store a general square matrix for LU-based solving | lu_factor |
| Solve a linear system using an LU factor | lu_solve |
| Fit multiple linear models against a shared design matrix | many_lm |
| Matrix functions via symmetric eigendecomposition | mat_fun mat_log mat_pow mat_sqrt |
| Matrix multiplication | matmul |
| Matrix multiplication for adgeMatrix | %*%,adgeMatrix,adgCMatrix-method %*%,adgeMatrix,adgeMatrix-method %*%,adgeMatrix,ANY-method %*%,adgeMatrix,aTransposeView-method %*%,adgeMatrix,dgCMatrix-method %*%,adgeMatrix,dgeMatrix-method %*%,adgeMatrix,Matrix-method %*%,adgeMatrix,matrix-method %*%,aTransposeView,adgeMatrix-method %*%,aTransposeView,ANY-method %*%,aTransposeView,aTransposeView-method %*%,aTransposeView,matrix-method %*%,dgCMatrix,adgCMatrix-method %*%,dgeMatrix,adgCMatrix-method %*%,KronMatrix,matrix-method %*%,KronMatrix,numeric-method %*%,matrix,adgCMatrix-method %*%,matrix,adgeMatrix-method %*%,matrix,aTransposeView-method %*%,matrix,KronMatrix-method %*%,numeric,adgCMatrix-method %*%,numeric,adgeMatrix-method %*%,numeric,KronMatrix-method matmul-methods |
| Nearest-centroid assignment via fused squared-distance computation | pairwise_sqdist_argmin |
| Project and reconstruct data using a truncated SVD | pca_coef |
| QR downdate after removing one row | qr_downdate |
| Inspect an amQR factorization object | qr_info |
| Evaluate a quadratic form using a Cholesky factor | quad_form |
| Create a mutable GPU-resident handle | resident_handle |
| Column sums of a GPU-resident handle | rh_colSums |
| Row sums of a GPU-resident handle | rh_rowSums |
| Fit a single ridge regression model | ridge_fit |
| Compute a ridge regression solution path | ridge_path |
| Row and column means | colmeans rowmeans |
| Row and column diagonal scaling | colscale rowscale |
| Row and column sums | colsums rowsums |
| Row and column summary methods for adgCMatrix | colMeans,adgCMatrix-method colSums,adgCMatrix-method rowMeans,adgCMatrix-method rowSums,adgCMatrix-method |
| Row and column summary methods for adgeMatrix | colMeans,adgeMatrix-method colSums,adgeMatrix-method rowMeans,adgeMatrix-method rowSums,adgeMatrix-method |
| GPU-native randomized SVD (Halko et al. 2011) | rsvd |
| Segment mean by group labels | segment_mean |
| Segment sum by group labels | segment_sum |
| Doubly-stochastic scaling via Sinkhorn-Knopp iterations | sinkhorn |
| Solve a triangular linear system | solve_triangular |
| Solve a linear system for adgCMatrix | solve,adgCMatrix,ANY-method solve,adgCMatrix,missing-method |
| Solve a linear system for adgeMatrix | solve,adgeMatrix,ANY-method solve,adgeMatrix,missing-method |
| Compute a truncated SVD of an aMatrix | svd_factor |
| Project new data onto SVD left singular vectors | svd_project |
| Reconstruct data from SVD coordinates | svd_reconstruct |
| Singular value decomposition for adgeMatrix | svd svd,adgeMatrix-method svd-methods |
| Singular value decomposition for adgCMatrix | svd,adgCMatrix-method |
| Symmetrise a matrix | sym |
| Weighted outer cross-product XWX' | tcrossprod_weighted |
| Matrix trace | trace |
| Stochastic trace estimator (Hutchinson) | trace_estim |
| Evaluate code with temporary amatrix defaults | with_amatrix |
| Fit a weighted least squares model | wls_fit |
| Log-determinant via the Woodbury matrix determinant lemma | woodbury_logdet |
| Solve a linear system using the Woodbury matrix identity | woodbury_solve |
| Weighted cross-product X'Wy | xty_weighted |
