cuda.tile.maximum

cuda.tile.maximum(x, y, /, *, flush_to_zero=False, propagate_nan=False)

Elementwise maximum on two tiles.

Can also use builtin operation max(x, y).

Parameters:

• x (Tile) – LHS tile.

• y (Tile) – RHS tile. flush_to_zero (const bool): If True, flushes subnormal inputs and results to sign-preserving zero, default is False. propagate_nan (const bool): Float types only. If False (the default), NaN inputs are ignored. If True, NaN propagates: a NaN makes the result NaN for min/max, and argmin/argmax return the index of the first NaN.

The shape of x and y will be broadcasted and dtype promoted to common dtype.

Return type:

Tile

Examples

SnippetComplete Example

tx = ct.arange(4, dtype=ct.int32)
ty = ct.full((4,), 2, dtype=ct.int32)
print(max(tx, ty))

Copy

import cuda.tile as ct
import torch

@ct.kernel
def kernel():
    tx = ct.arange(4, dtype=ct.int32)
    ty = ct.full((4,), 2, dtype=ct.int32)
    print(max(tx, ty))


torch.cuda.init()
ct.launch(torch.cuda.current_stream(), (1,), kernel, ())
torch.cuda.synchronize()

Output

[2, 2, 2, 3]