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 package

layout

Provides layout and layout tensor types, which abstract memory layout for multidimensional data.

• The Layout type represents a mapping between a set of logical coordinates and a linear index. It can be used, for example, to map logical tensor coordinates to a memory address, or to map GPU threads to tiles of data.

• The LayoutTensor type is a high-performance tensor with explicit memory layout via a Layout.

Modules

• ​coord: Subpackage exposing std.utils.coord as layout.coord for kernel imports.
• ​copy: Provides copy policy traits and utilities for layout memory operations.
• ​element: Provides element-based access to memory using layout-driven vectorization.
• ​int_tuple: Hierarchical integer tuple data structures for high-performance tensor operations.
• ​layout: Provides a high-performance tensor layout system for memory mapping and indexing.
• ​layout_tensor: Provides the LayoutTensor type for representing multidimensional data.
• ​math: Implements math methods that work on layout tensors.
• ​runtime_layout: Provides the RuntimeLayout type and functions for working with it. You can use RuntimeLayout to define a layout where the dimensions are not known at compile time.
• ​runtime_tuple: Provides the RuntimeTuple data structure and related utility functions for handling tuple-like data with both compile-time and runtime elements. RuntimeTuple is designed for high-performance tensor operations, supporting efficient manipulation of multi-dimensional data structures like shapes, indices, and coordinates.
• ​swizzle: Defines swizzle layouts for optimizing memory access patterns.
• ​tensor_core: Tensor Core Module for High-Performance Matrix Operations.
• ​tensor_core_async: Tensor Core Async Module.
• ​tensor_storage: Defines storage abstractions for tile-backed tensor views.
• ​tile_io: Trait and utilities for copying data between TileTensors.
• ​tile_layout: Provides a mixed compile-time/runtime layout system for tensor memory mapping.
• ​tile_tensor: TileTensor type for structured memory access with compile-time layout information.
• ​tma_async: Tensor Memory Accelerator (TMA) Asynchronous Operations Module.