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

advanced_indexing_setitem_inplace

def advanced_indexing_setitem_inplace[index_rank: Int, updates_rank: Int, input_type: DType, index_type: DType, //, start_axis: Int, num_index_tensors: Int, target: StringSlice[StaticConstantOrigin], trace_description: StringSlice[StaticConstantOrigin], updates_tensor_fn: def[width: Int](IndexList[updates_rank]) capturing -> SIMD[input_type, width], indices_fn: def[indices_index: Int](IndexList[index_rank]) capturing -> Scalar[index_type]](input_tensor: TileTensor[input_type, Storage=input_tensor.Storage, address_space=input_tensor.address_space, linear_idx_type=input_tensor.linear_idx_type, element_size=input_tensor.element_size], index_tensor_shape: IndexList[index_rank], updates_tensor_strides: IndexList[updates_rank], ctx: DeviceContext)

Implement basic numpy-style advanced indexing with assignment.

This is designed to be fused with other view-producing operations to implement full numpy-indexing semantics.

This assumes the dimensions in input_tensor not indexed by index tensors are ":", ie selecting all indices along the slice. For example in numpy:

# rank(indices1) == 2
# rank(indices2) == 2
# rank(updates) == 2
input_tensor[:, :, :, indices1, indices2, :, :] = updates

We calculate the following for all valid valued indexing variables:

input_tensor[
    a, b, c,
    indices1[i, j],
    indices2[i, j],
    d, e
] = updates[i, j]

In this example start_axis = 3 and num_index_tensors = 2.

In terms of implementation details, our strategy is to iterate over all indices over a common iteration range. The idea is we can map indices in this range to the write location in input_tensor as well as the data location in updates. An update can illustrate how this is possible best:

Imagine the input_tensor shape is [A, B, C, D] and we have indexing tensors I1 and I2 with shape [M, N, K]. Assume I1 and I2 are applied to dimensions 1 and 2.

I claim an appropriate common iteration range is then (A, M, N, K, D). Note we expect updates to be the shape [A, M, N, K, D]. We will show this by providing the mappings into updates and input_tensor:

Consider an arbitrary set of indices in this range (a, m, n, k, d): - The index into updates is (a, m, n, k, d). - The index into input_tensor is (a, I1[m, n, k], I2[m, n, k], d).

TODO(GEX-1951): Support boolean tensor mask support TODO(GEX-1952): Support non-contiguous indexing tensor case TODO(GEX-1953): Support fusion (especially view-fusion) TODO(GEX-1954): Unify getitem and setitem using generic views. (Requires non-strided view functions).

Parameters:

index_rank (Int): The rank of the indexing tensors.
updates_rank (Int): The rank of the updates tensor.
input_type (DType): The dtype of the input tensor.
index_type (DType): The dtype of the indexing tensors.
start_axis (Int): The first dimension in input where the indexing tensors are applied. It is assumed the indexing tensors are applied in consecutive dimensions.
num_index_tensors (Int): The number of indexing tensors.
target (StringSlice[StaticConstantOrigin]): The target architecture to operation on.
trace_description (StringSlice[StaticConstantOrigin]): For profiling, the trace name the operation will appear under.
updates_tensor_fn (def[width: Int](IndexList[updates_rank]) capturing -> SIMD[input_type, width]): Fusion lambda for the update tensor.
indices_fn (def[indices_index: Int](IndexList[index_rank]) capturing -> Scalar[index_type]): Fusion lambda for the indices tensors.

Args:

input_tensor (TileTensor[input_type, Storage=input_tensor.Storage, address_space=input_tensor.address_space, linear_idx_type=input_tensor.linear_idx_type, element_size=input_tensor.element_size]): The input tensor being indexed into and modified in-place.
index_tensor_shape (IndexList[index_rank]): The shape of each index tensor.
updates_tensor_strides (IndexList[updates_rank]): The strides of the update tensor.
ctx (DeviceContext): The device context as prepared by the graph compiler.