Overview

One compiled register config: tr/tc thread tile, stages, the forms instantiated, and whether NN takes the unpredicated leg when the layout allows (derived per call, never a gate).

Definition at line 54 of file choice.hh.

#include </home/runner/work/BatchLAS/BatchLAS/src/ops/gemm/choice.hh>

Functions

◆ threads()

int batchlas::ops::gemm::RegCfg::threads ( ) const

inlineconstexpr

Definition at line 58 of file choice.hh.

◆ slm_bytes()

std::int64_t batchlas::ops::gemm::RegCfg::slm_bytes ( ) const

inlineconstexpr

Float local memory per work-group (gemm_kernels.cc static_asserts it against each kernel).

Definition at line 60 of file choice.hh.