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

Mojo function

coalesce

def coalesce[LayoutType: TensorLayout, //](layout: LayoutType) -> Layout[*?, *?]

Simplifies a layout by merging contiguous dimensions.

Iterates over the flattened (shape, stride) pairs and:

1. Skips shape-1 dimensions.
2. Merges a dimension into the previous one when prev_shape * prev_stride == current_stride (contiguous).
3. Otherwise starts a new dimension.

The result is the simplest layout that maps coordinates to the same linear offsets as the original.

Example:

from layout.tile_layout import Layout, coalesce, row_major, Idx

# A row-major 2x4 layout has contiguous strides -> coalesces to 1D
var layout = row_major[2, 4]()
var coalesced = coalesce(layout)  # shape (8,), stride (1,)

Parameters:

• ​LayoutType (TensorLayout): The type of the input layout.

Args:

• ​layout (LayoutType): The layout to coalesce.

Returns:

Layout[*?, *?]: A new layout with contiguous dimensions merged.