Skip to main content

AMPERE_SM_REGS

logicaffeine_compile::codegen_cuda::gemm_tile

Constant AMPERE_SM_REGS 

Source 
pub const AMPERE_SM_REGS: usize = 65536;
Expand description

SM-level register file on Ampere (sm_86): 65,536 32-bit registers per SM. A tile needs regs_per_thread · threads ≤ this to place at least one block per SM.