Struct MemoryTiers 

pub struct MemoryTiers {
    pub host_ram_bytes: u64,
    pub device_vram_bytes: u64,
}

Per-tier memory footprint of a candidate’s placement, in bytes.

Tracked per tier (not collapsed) because which tier binds depends on the target machine — a fast-but-VRAM-heavy path and a slow-but- host-light path are both legitimately Pareto-optimal. Disk is reserved for when the Layer-1 cost model grows a disk-footprint axis; today’s CostEstimate has none, so it stays zero.

Fields

host_ram_bytes: u64

Host-RAM footprint (bytes) attributed to a CPU-tier placement.

device_vram_bytes: u64

Device-VRAM footprint (bytes) attributed to a GPU-tier placement (CUDA / Vulkan / Metal).

Implementations

impl MemoryTiers

pub fn for_placement(device: DeviceLocation, bytes: u64) -> Self

Attribute bytes to the tier implied by device: GPU devices land on VRAM, the CPU on host RAM. The Layer-1 cost model has one footprint number (bytes_moved); placement decides which tier it loads.

Trait Implementations

impl Clone for MemoryTiers

fn clone(&self) -> Self

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Copy for MemoryTiers

impl Debug for MemoryTiers

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

Formats the value using the given formatter.

impl Default for MemoryTiers

fn default() -> Self

Returns the “default value” for a type.

impl Eq for MemoryTiers

impl PartialEq for MemoryTiers

fn eq(&self, other: &Self) -> bool

Equality operator ==.

fn ne(&self, other: &Rhs) -> bool

Inequality operator !=.

impl StructuralPartialEq for MemoryTiers