Skip to main content

optimize_workgroup_size

Function optimize_workgroup_size 

Source
pub fn optimize_workgroup_size(
    hw: &WorkgroupHwParams,
    shared_mem_bytes: u32,
    regs_per_thread: u32,
) -> [u32; 3]
Expand description

Find the workgroup size that maximizes occupancy.

Explores candidate sizes that are multiples of warp_size, within hardware dimension limits.