Expand description
B2 — nearest NCHW resize via Forge on the auxiliary GPU circuit when available.
Uses existing OpNode::Resize2d / WGSL resize2d_main (certified against
the same numerical contract as ops::resize_nearest_nchw_f32). Fail closed
to caller (CPU path) on any adapter / forge error.
Device routing: vision SR runs on the auxiliary circuit (the iGPU when present) to keep the
discrete GPU free for the LLM. The device comes from
crate::gpu_context::device_registry::try_auxiliary_gpu, whose fallback chain is
auxiliary → primary → CPU: on a single-GPU box it transparently uses the primary device, and on
a GPU-less box it returns None, so this function fails closed and the caller degrades to the
CPU oracle. It never panics.
Functions§
- try_
resize_ nearest_ shared_ gpu - Attempt GPU resize on the process-wide shared device.