View page source

warp.randf#

warp.randf(state: uint32) → float#

  • Kernel

  • Python

Generate a uniform random float in the range [0.0, 1.0).

In a kernel, advances state in place, so successive calls return different values; called from the Python scope, it does not modify state, so repeated calls with the same state return the same value (see rand_init()). Values are multiples of 2^-24, matching the 24 bits of precision in a 32-bit float.

Random built-ins take an RNG state produced by rand_init(); initialize it once per thread, then draw values:

@wp.kernel
def random_floats(seed: int, out: wp.array[wp.vec2]):
    i = wp.tid()
    rng = wp.rand_init(seed, i)
    # Each call advances rng, so the two draws differ.
    out[i] = wp.vec2(wp.randf(rng), wp.randf(rng))

randi(), randu(), and randn() are used the same way: initialize rng once with rand_init(), then pass it to each call.

warp.randf(
state: uint32,
low: float32,
high: float32,
) → float

  • Kernel

  • Python

Generate a uniform random float in the range [low, high).

In a kernel, advances state in place, so successive calls return different values; called from the Python scope, it does not modify state, so repeated calls with the same state return the same value (see rand_init()). Equivalent to low + (high - low) * wp.randf(state).