For the complete documentation index, see llms.txt. Markdown versions of all pages are available by appending .md to any URL (e.g. /max/get-started.md).

Mojo function

max

def max[dtype: DType, input_fn: def[width: Int, rank: Int](IndexList[rank]) capturing thin -> SIMD[dtype, width], output_fn: def[width: SIMDLength, rank: Int](IndexList[rank], SIMD[dtype, width]) capturing thin -> None, /, target: StringSlice[ImmStaticOrigin] = StringSlice("cpu"), *, reduce_dim: Int](input_shape: Coord, context: Optional[DeviceContext] = None)

Computes the max across the input and output shape.

This performs the max computation on the domain specified by input_shape, loading the inputs using the input_fn. The results are stored using the output_fn.

Parameters:

dtype (DType): The dtype of the input and output.
input_fn (def[width: Int, rank: Int](IndexList[rank]) capturing thin -> SIMD[dtype, width]): The function to load the input.
output_fn (def[width: SIMDLength, rank: Int](IndexList[rank], SIMD[dtype, width]) capturing thin -> None): The function to store the output.
target (StringSlice[ImmStaticOrigin]): The target to run on.
reduce_dim (Int): The axis to perform the max on.

Args:

input_shape (Coord): The input shape.
context (Optional[DeviceContext]): The pointer to DeviceContext.

Raises:

If the operation fails.

def max[dtype: DType](src: Span[Scalar[dtype]]) -> Scalar[dtype]

Computes the max element in a buffer.

Parameters:

dtype (DType): The dtype of the input.

Args:

src (Span[Scalar[dtype]]): The buffer.

Returns:

Scalar[dtype]: The maximum of the buffer elements.

Raises:

If the operation fails.