[Mlir-commits] [mlir] [mlir][arith] Add `exact` to `index_cast{, ui}` (PR #183395)

[Slava Zakharin]

================
@@ -1579,15 +1580,34 @@ def IndexCastTypeConstraint : TypeConstraint<Or<[
 
 def Arith_IndexCastOp
   : Arith_CastOp<"index_cast", IndexCastTypeConstraint, IndexCastTypeConstraint,
-                 [DeclareOpInterfaceMethods<InferIntRangeInterface, ["inferResultRanges"]>]> {
+                 [DeclareOpInterfaceMethods<InferIntRangeInterface, ["inferResultRanges"]>,
+                  DeclareOpInterfaceMethods<ArithExactFlagInterface>]> {
   let summary = "cast between index and integer types";
   let description = [{
     Casts between scalar or vector integers and corresponding 'index' scalar or
     vectors. Index is an integer of platform-specific bit width. If casting to
     a wider integer, the value is sign-extended. If casting to a narrower
     integer, the value is truncated.
+
+    If the `exact` attribute is present, it is assumed that the operand
+    contains a value that fits in the destination's representation, therefore
+    the cast does not lose any information. When this assumption is violated,
+    the result is poison.
+
+    Example:
+
+    ```mlir
+      %0 = arith.index_cast %a : index to i64
+      %1 = arith.index_cast %a exact : index to i64
+      %2 = arith.index_cast %b exact : i32 to index
+    ```
   }];
 
+  let arguments = (ins IndexCastTypeConstraint:$in, UnitAttr:$isExact);
----------------
vzakhari wrote:

Okay, you may ignore my comment. Since it does not trigger the immediate UB, it should be okay. Thanks!

https://github.com/llvm/llvm-project/pull/183395