vyre_driver_wgpu/engine.rs
1//! Layer 3 complete compute engines.
2//!
3//! Each engine is a self-contained GPU compute pipeline: structured input
4//! in, compute passes on a real GPU backend, typed output back.
5//!
6//! This module owns backend execution helpers only: command graphs, work
7//! partitioning policy, persistent queues, host-ingress compatibility, and
8//! shared record/readback. Domain algorithms live in `vyre-libs` or
9//! wrappers that consume vyre.
10
11/// Per-thread scratch arenas for record/readback hot-path vectors.
12pub(crate) mod dispatch_scratch;
13/// GPU-resident command graph execution.
14pub mod graph;
15/// Mockable multi-GPU work partitioning.
16pub mod multi_gpu;
17/// Resident persistent-kernel queue engine.
18pub mod persistent;
19/// Shared command recording and readback for vyre IR dispatch paths.
20pub(crate) mod record_and_readback;
21/// Host-ingress chunk bridge for callers that still receive bytes through CPU memory.
22pub mod streaming;