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 module

structured

Provides AMD GPU building blocks for structured matrix multiply-accumulate kernels.

Defines thread roles, pipeline-stage shared memory layouts, workgroup and warp barriers, and the MMA configuration and tile operator used to drive AMD tensor core matrix multiplication.

Structs

• ​AMDSharedMemoryBarrier: Implements a workgroup-level shared memory barrier for AMD GPUs.
• ​AmdTileOperator: Manages tensor core operations for matrix multiplication on AMD GPUs.
• ​AMDWarpSharedMemoryBarrier: Implements a per-warp shared memory barrier for AMD GPUs.
• ​MMAConfig: Configures matrix multiply-accumulate parameters for AMD tensor cores.
• ​SMemBuffer: Manages shared memory and returns 2D tile slices of the buffer.
• ​ThreadRole: Represents the role a thread plays in a producer-consumer pipeline.

Traits

• ​Enum: Defines a comparable enum-like trait exposing an integer value and equality.

Functions

• ​pipeline_layout: Builds a 2D layout extended with a pipeline-stage dimension.