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

host

Implements the gpu host package.

Packages

compile: Implements GPU compilation operations.
constant_memory_mapping: This module provides functionality for mapping constant memory between host and device.
device_attribute: This module defines GPU device attributes that can be queried from CUDA-compatible devices.
device_context: This module provides functionality for interacting with accelerators. In particular the DeviceContext struct, which represents a single stream of execution on a given accelerator. You can use this struct to allocate accelerator memory, copy data to and from the accelerator, and compile and execute functions on the accelerator.
device_graph: This module provides functionality for building and replaying device graphs. A device graph captures a sequence of GPU operations (such as kernel launches, memory copies, and memsets) as a reusable graph that can be replayed at a lower overhead than re-enqueueing each operation individually. The main entry point is DeviceGraph.create(), which hands a DeviceGraphBuilder to a scoped callback.
dim: This module implements the dim type.
func_attribute: GPU Kernel Function Attributes Module.
info: Re-exports accelerator target information from std.gpu.host.info.
launch_attribute: GPU Launch Attributes for Kernel Execution Control.