Primitive Pad

Pads an array with a given padding value.

Usage

prim_pad(
  x,
  padding_value,
  edge_padding_low,
  edge_padding_high,
  interior_padding
)

Arguments

x

(arrayish)
The array to pad. Can be any data type. x and padding_value must have the same data type. An R value among them assumes the data type of the others when it is in its data type category, and its default data type when none of them has one.

padding_value

(arrayish)
Scalar value to use for padding. Shares x's data type.

edge_padding_low

(integer())
Amount of padding to add at the start of each axis.

edge_padding_high

(integer())
Amount of padding to add at the end of each axis.

interior_padding

(integer())
Amount of padding to add between elements in each axis.

Value

(arrayish)
Has the data type the operands agreed on. Each axis grows by edge_padding_low + edge_padding_high, plus interior_padding between every pair of elements; negative edge padding trims (see hlo_pad()).

Implemented Rules

• stablehlo

• quickr

• reverse

StableHLO

Lowers to hlo_pad(), specified under pad.

See also

nv_pad()

Examples

x <- nv_array(1:3)
# one element before, two after
prim_pad(x, 0L, edge_padding_low = 1L, edge_padding_high = 2L, interior_padding = 0L)
#> AnvlArray
#>  0
#>  1
#>  2
#>  3
#>  0
#>  0
#> [ CPUi32{6} ] 

# interior padding goes between the elements
prim_pad(x, 0L, 0L, 0L, 1L)
#> AnvlArray
#>  1
#>  0
#>  2
#>  0
#>  3
#> [ CPUi32{5} ] 

# negative edge padding trims instead
prim_pad(x, 0L, -1L, 0L, 0L)
#> AnvlArray
#>  2
#>  3
#> [ CPUi32{2} ] 

# one padding amount per axis
prim_pad(nv_matrix(1:4, nrow = 2), 0L, c(1L, 0L), c(0L, 1L), c(0L, 0L))
#> AnvlArray
#>  0 0 0
#>  1 3 0
#>  2 4 0
#> [ CPUi32{3,3} ] 

# the R padding value is built at x's data type
prim_pad(nv_array(c(1.5, 2.5), dtype = "f64"), 0, 1L, 1L, 0L)
#> AnvlArray
#>  0.0000
#>  1.5000
#>  2.5000
#>  0.0000
#> [ CPUf64{4} ]