[llvm-branch-commits] [clang] [flang] [flang] Add -f[no-]version-loops-for-trip-count and enable it at -O3 (2/2) (PR #231207)

[via llvm-branch-commits]

================
@@ -0,0 +1,36 @@
+! Check that loop trip-count versioning fires on real lowering output and is
+! controlled by -f[no-]version-loops-for-trip-count. The pass matches the
+! canonical form of array-section loops (arith.maxsi bound, shared slice and
+! loop bound), so this test guards against lowering or canonicalization
+! changes that would silently stop it from matching.
+!
+! RUN: %flang_fc1 -emit-llvm -O3 -fversion-loops-for-trip-count \
----------------
MattPD wrote:

Confirmed, the new runs catch ignored overrides.

https://github.com/llvm/llvm-project/pull/231207