[llvm] [AMDGPU] Introduce iglp_opt(2): Generalized exp/mfma interleaving for select kernels (PR #81342)
Austin Kerbow via llvm-commits
llvm-commits at lists.llvm.org
Fri Feb 23 15:49:27 PST 2024
================
@@ -1726,6 +1749,8 @@ bool MFMAExpInterleaveOpt::applyIGLPStrategy(
// In loop MFMAs
auto MFMAInLoop = MFMAPipeCount - (MFMAEnablement * 2);
auto MFMALoopCount = MFMAInLoop / MFMARatio;
+ auto VALUOps =
----------------
kerbowa wrote:
Maybe just double-check whether any of these divs need a check for div-by-zero.
https://github.com/llvm/llvm-project/pull/81342
More information about the llvm-commits
mailing list