[llvm] ConstraintElimination: use DynamicAPInt (PR #99019)

Florian Hahn via llvm-commits llvm-commits at lists.llvm.org
Tue Jul 16 06:16:12 PDT 2024


fhahn wrote:

> To automatically handle overflows correctly

IIUC this should be fine, as we only allow expressions that won't wrap in IR semantics, so the only source of wrapping should be multiplications during solving. Unfortunately it comes at a cost of compile-time, presumably not due to overflows happening frequently but extra overhead from DynamicAPInt?

https://github.com/llvm/llvm-project/pull/99019


More information about the llvm-commits mailing list