Struct DiscoveredWorker 

pub struct DiscoveredWorker {
    pub name: String,
    pub host: String,
    pub port: u16,
    pub gpus: Vec<GpuInfo>,
    pub backend: String,
    pub hostname: String,
    pub os: String,
}

A worker discovered via broadcast.

Fields

name: String

host: String

port: u16

gpus: Vec<GpuInfo>

backend: String

hostname: String

os: String

Implementations

impl DiscoveredWorker

pub fn total_vram(&self) -> u64

Total VRAM across all GPUs.

pub fn max_layers_for_size(&self, layer_size_bytes: u64) -> usize

Maximum number of layers this worker can fit, based on per-GPU VRAM.

For dedicated GPUs (CUDA), reserves ~5% for driver/runtime overhead (typically 200–600 MiB for CUDA context + cuBLAS workspace). For unified-memory devices (Apple Silicon), reserves 28% of total (minimum 6 GiB) for macOS + inference working memory, since model weights compete with the OS for the same physical RAM and insufficient headroom causes catastrophic memory-compressor thrashing.

pub fn total_tflops(&self) -> f64

Total estimated TFLOPS across all GPUs. Falls back to a VRAM-based estimate when workers report 0 (old binaries).

Trait Implementations

impl Clone for DiscoveredWorker

fn clone(&self) -> DiscoveredWorker

Returns a duplicate of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.

impl Debug for DiscoveredWorker

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.