[flang-commits] [flang] [llvm] [flang] Support static-unit array slices in FIR LoopVersioning (PR #222723)

[Sergey Shcherbinin via flang-commits]

================
@@ -285,6 +920,458 @@ static mlir::Value getIndex(fir::FirOpBuilder &builder, mlir::Operation *op,
   return mlir::arith::SubIOp::create(builder, coop.getLoc(), index, lb);
 }
 
+/// Convert one cloned slice operand to the index type before arithmetic.
+/// Preflight proved that this conversion preserves the effective index-domain
+/// bit pattern used by generic lowering.
+static mlir::Value materializeIndex(fir::FirOpBuilder &builder,
+                                    mlir::Location loc, mlir::Value value) {
+  mlir::Type indexType = builder.getIndexType();
+  if (value.getType() == indexType)
+    return value;
+  return builder.createConvert(loc, indexType, value);
+}
----------------
SergeyShch01 wrote:

The `materializeIndex` helper has been removed, and the rewritten code calls `createConvert` directly where no special integer-cast semantics are required.



https://github.com/llvm/llvm-project/pull/222723