Transpose a GPU-aware sparse matrix

Swaps sparse rows and columns while preserving stored values, logical format, dimension labels, and the actual device. The native CUDA backend transposes its CSR backing storage on the device. Compatibility backends rebuild same-device storage from the stable public COO metadata.

Usage

# S3 method for class 'cudasparse'
t(x)

Arguments

x

A cudasparse matrix.

Value

A transposed cudasparse matrix on the same device as x.

Examples

x <- cuda_sparse(matrix(c(1, 0, 2, 0, 3, 0), 2), device = "cpu")
t(x)
#> <cudasparse[3x2] nnz=3 format=csr device=cpu backend=Matrix>
#> 3 x 2 sparse Matrix of class "dgCMatrix"
#>         
#> [1,] 1 .
#> [2,] 2 .
#> [3,] 3 .