For the complete documentation index, see llms.txt. Markdown versions of all pages are available by appending .md to any URL (e.g. /max/get-started.md).
Mojo struct
TargetProfile
struct TargetProfile
Unified hardware target description.
Bundles everything the framework needs to know about a specific GPU target into a single struct:
- cost_model: per-op costs (resource, latency, role)
- pipeline: pipeline structure (depth, MMA grid, buffer strategy)
The algorithm declares WHAT ops exist (logical op table). The TargetProfile describes HOW the hardware executes them. Platform- specific factories (e.g., mi355x_target() in amd_target.mojo) provide both from a single call — no redundant configuration.
Usage: # Platform-specific factory (from amd_target): comptime target = mi355x_target() var annotated = annotate_ops(logical_ops, target.cost_model) var config = target.pipeline
Fields
- cost_model (
TargetCostModel): - pipeline (
PipelineConfig):
Implemented traits
AnyType,
Copyable,
ImplicitlyCopyable,
ImplicitlyDestructible,
Movable
Methods
__init__
__init__(out self, cost_model: TargetCostModel, pipeline: PipelineConfig)
Was this page helpful?
Thank you! We'll create more content like this.
Thank you for helping us improve!