Primitive RNG Bit Generator

Generates pseudo-random numbers using the specified algorithm and returns the updated RNG state together with the generated values.

Usage

prim_rng_bit_generator(state, rng_algorithm = "THREE_FRY", dtype, shape)

Arguments

state: (arrayish)
RNG state: a 1-D array of the ui64 data type. Its length depends on rng_algorithm – exactly 2 for "THREE_FRY", 2 or 3 for "PHILOX", and whatever the implementation wants for "DEFAULT".
rng_algorithm: (character(1))
One of "THREE_FRY" (default), "PHILOX" or "DEFAULT", the last leaving the choice to the implementation.
dtype: (character(1) | DataType)
Data type of the generated random values. Can be any numeric data type.
shape: (integer())
Shape of the result.

Value

(named list of two arrayish)
Elements state, the updated RNG state with the input state's data type and shape, and values, the random values with the given dtype and shape.

Implemented Rules

stablehlo

StableHLO

Lowers to hlo_rng_bit_generator(), specified under rng_bit_generator.

Examples

# THREE_FRY, the default, takes a two-element state
state <- nv_array(c(0L, 0L), dtype = "ui64")
prim_rng_bit_generator(state, dtype = "f32", shape = c(3, 2))
#> $state
#> AnvlArray
#>  0
#>  3
#> [ CPUui64{2} ] 
#> 
#> $values
#> AnvlArray
#>  1.7973e+09 2.5791e+09
#>  1.3515e+09 3.2358e+09
#>  1.6886e+09 4.2293e+09
#> [ CPUf32{3,2} ] 
#> 

# the updated state feeds the next draw, so the two differ
out <- prim_rng_bit_generator(state, dtype = "f32", shape = 3L)
prim_rng_bit_generator(out$state, dtype = "f32", shape = 3L)
#> $state
#> AnvlArray
#>  0
#>  4
#> [ CPUui64{2} ] 
#> 
#> $values
#> AnvlArray
#>  1.6886e+09
#>  4.2293e+09
#>  3.0983e+09
#> [ CPUf32{3} ] 
#> 

# PHILOX also accepts a three-element state
prim_rng_bit_generator(
  nv_array(c(0L, 0L, 0L), dtype = "ui64"), "PHILOX",
  dtype = "i32", shape = 4L
)
#> $state
#> AnvlArray
#>  0
#>  1
#>  0
#> [ CPUui64{3} ] 
#> 
#> $values
#> AnvlArray
#>  1.7139e+09
#>  3.7818e+09
#>  3.1599e+09
#>  2.6005e+09
#> [ CPUi32{4} ] 
#>