Struct RouteEnv 

pub struct RouteEnv {
    pub keep_nvfp4: bool,
    pub dispatch: GemmDispatch,
    pub attn_w4a4: bool,
    pub attn_prefill_q_t: bool,
}

2026-09-25: The environment half of DenseFp8Inputs. load_layers resolves it once, so every layer of one load is planned from the same values.

Fields

keep_nvfp4: bool

dispatch: GemmDispatch

attn_w4a4: bool

2026-09-25: The METRALE_ATTN_W4A4 check of prefill/paged_oproj.rs and prefill/paged_qkv.rs, which read the variable on every call.

attn_prefill_q_t: bool

2026-09-25: The METRALE_ATTN_PREFILL_Q_T check of prefill/cache_skip_qkv.rs, also read on every call.

Implementations

impl RouteEnv

pub fn from_env() -> Self

pub fn plan( &self, ffn_fp8: bool, attn_fp8: bool, w8a8_kernels: bool, ) -> DenseFp8Plan

2026-09-25: One layer’s plan. w8a8_kernels comes from the layer’s resolved kernel handles, so it is an argument, not a field.

pub fn attn_nvfp4(&self, attn_fp8: bool) -> bool

2026-09-25: The attention NVFP4 decision. The loader asks it before the layer exists and passes w8a8_kernels = true, which is sound because attn_nvfp4 does not depend on it (attn_nvfp4_does_not_depend_on_the_w8a8_kernels).

pub fn attn_fp8_twins(&self, attn_fp8: bool, w8a8_kernels: bool) -> Fp8TwinSet

2026-09-25: Which FP8 prefill twins this layer needs. The loader passes the constructed layer’s has_w8a8_prefill_kernels().

Trait Implementations

impl Clone for RouteEnv

fn clone(&self) -> RouteEnv

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for RouteEnv

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

Formats the value using the given formatter.

impl Copy for RouteEnv