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

ring_buffer

Ring Buffer implementation for producer-consumer synchronization in GPU kernels.

This ring buffer coordinates data transfer between producer warps (loading from global memory) and consumer warps (performing computation) through shared memory tiles.

Key features:

• Configurable synchronization strategies via the SyncStrategy trait
• Pipeline stages for overlapping data transfer and computation
• Context managers for automatic acquire/release of tiles
• Phase-based synchronization to prevent data races

Structs

• ​ConsumerTile: Context manager for consumer access to a single ring buffer tile.
• ​ConsumerView: Consumer view of the unified ring buffer.
• ​ProducerTile: Context manager for producer access to a single ring buffer tile.
• ​ProducerView: Producer view of the unified ring buffer.
• ​RingBuffer: Ring buffer for coordinating producer-consumer warps in matrix multiplication.