Create a backend-aware dense matrix — adgeMatrix • amatrix

Converts a base R matrix or Matrix::dgeMatrix to an adgeMatrix with the specified backend, policy, and precision. This is the primary user-facing constructor for dense amatrix objects.

Usage

adgeMatrix(
  x,
  mode = NULL,
  backend = NULL,
  preferred_backend = NULL,
  policy = NULL,
  precision = NULL
)

Arguments

x: A base R matrix, dgeMatrix, or any denseMatrix coercible to dgeMatrix.
mode: Single string shortcut accepted by .amatrix_resolve_mode(); used to set backend, policy, and precision together. Pass NULL to use the individual arguments instead. In particular, mode = "fast" requests reduced precision and prefers an available fast-capable accelerator automatically, with CPU fallback when none is available.
backend: Alias for preferred_backend; ignored when preferred_backend is non-NULL.
preferred_backend: Single string naming the preferred compute backend, e.g. "cpu", "mlx", or "metal".
policy: Single string; one of "auto", "cpu", "mlx", "metal", "arrayfire", "opencl".
precision: Single string; "strict" for full double-precision accuracy or "fast" to allow reduced precision on GPU backends.

Value

An adgeMatrix with the data from x and the requested backend metadata.

Examples

m <- matrix(1:6, nrow = 2)
A <- adgeMatrix(m)
A
#> An amatrix dense matrix [cpu|policy=auto|precision=strict]
#> 2 x 3 Matrix of class "adgeMatrix"
#>      [,1] [,2] [,3]
#> [1,]    1    3    5
#> [2,]    2    4    6