Skip to contents

Solve triangular systems with mlx arrays

Source: R/linalg_adv.R
mlx_solve_triangular.Rd

Solve triangular systems with mlx arrays

Usage

mlx_solve_triangular(a, b, upper = FALSE, device = NULL)

backsolve(r, x, k = NULL, upper.tri = TRUE, transpose = FALSE, ...)

# Default S3 method
backsolve(r, x, k = NULL, upper.tri = TRUE, transpose = FALSE, ...)

# S3 method for class 'mlx'
backsolve(
  r,
  x,
  k = NULL,
  upper.tri = TRUE,
  transpose = FALSE,
  ...,
  device = NULL
)

Arguments

a

An mlx triangular matrix.

b

Right-hand side matrix or vector.

upper

Logical; if TRUE, a is upper triangular, otherwise lower.

device

Execution target for APIs that expose a one-off device or stream override. Supply "gpu", "cpu", or an mlx_stream created via mlx_new_stream(). Ordinary array operations use the current mlx_device() instead.

r

Triangular system matrix passed to backsolve().

x

Right-hand side supplied to backsolve().

k

Number of columns of r to use.

upper.tri

Logical; indicates if r is upper triangular.

transpose

Logical; if TRUE, solve t(r) %*% x = b.

...

Additional arguments forwarded to the corresponding base R implementation for signature compatibility.

Value

An mlx array solution.

Details

As of MLX 0.31.1, this operation only runs on CPU. Run it inside with_device() or local_device(), or pass device = "cpu".

See also

mlx.linalg.solve_triangular

Examples

a <- mlx_matrix(c(2, 1, 0, 3), 2, 2)
b <- mlx_matrix(c(1, 5), 2, 1)
mlx_solve_triangular(a, b, upper = FALSE, device = "cpu")
#> mlx array [2 x 1]
#>   dtype: float32
#>   values:
#>      [,1]
#> [1,]  0.5
#> [2,]  1.5