Struct M16TcLevers 

pub struct M16TcLevers {
    pub ffn: bool,
    pub attn: bool,
    pub ffn_n_tile: u32,
}

Which projection families the tensor-core tier serves, and at what CTA width. SSOT for the whole lever grammar; every call site resolves it ONCE at construction into a field.

Fields

ffn: bool

The dense-FFN arm (ATLAS_FFN_M16_TC).

attn: bool

The multi-seq FP8 QKV tier and the FP8 o_proj tier (ATLAS_ATTN_M16_TC).

ffn_n_tile: u32

CTA N width for the FFN arm: 32 (default) or 64 (ATLAS_FFN_M16_TC_NTILE=64). Attention always runs 32 — the wide tile has no strided twin and its N is already CTA-starved.

Trait Implementations

impl Clone for M16TcLevers

fn clone(&self) -> M16TcLevers

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for M16TcLevers

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

Formats the value using the given formatter.

impl PartialEq for M16TcLevers

fn eq(&self, other: &M16TcLevers) -> bool

Tests for self and other values to be equal, and is used by ==.

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

Tests for !=. The default implementation is almost always sufficient, and should not be overridden without very good reason.

impl Copy for M16TcLevers

impl Eq for M16TcLevers

impl StructuralPartialEq for M16TcLevers