pub fn fp8_quant_hopper_lane(tid: u32, span: u32) -> Option<(u32, u32, u32)>Expand description
Which 8 elements of a 128-element group thread tid of a Hopper CTA owns,
as a half-open [start, end) within the group, plus the group slot it is
working on. Mirrors sub/lane in the .cu. Returns None for a thread
whose group slot is past the CTA’s span.