[Mlir-commits] [mlir] [mlir][vector] Don't fold in_bounds for negative constant indices (PR #219681)

[Federico Bruzzone]

================
@@ -1435,6 +1435,53 @@ func.func @no_fold_transfer_write_in_bounds_scalable(%m: memref<4xf32>, %v: vect
 
 // -----
 
+// `in_bounds` promises that the transfer stays within the source *including its
+// starting point*, so a negative index is not in bounds even when the end of
+// the transfer would land inside the source. Here `-1 + 4 <= 8` holds, but
+// element `-1` is still read from outside `%m`.
+
+// CHECK-LABEL: func @no_fold_transfer_read_in_bounds_negative_index
+//       CHECK:   vector.transfer_read
+//   CHECK-NOT:   in_bounds
+//       CHECK:   : memref<8xf32>, vector<4xf32>
+func.func @no_fold_transfer_read_in_bounds_negative_index(%m: memref<8xf32>, %p: f32) -> vector<4xf32> {
+  %c-1 = arith.constant -1 : index
+  %v = vector.transfer_read %m[%c-1], %p : memref<8xf32>, vector<4xf32>
+  return %v : vector<4xf32>
+}
+
+// -----
+
+// Same for the write path, where the fold would permit a store to element -1.
+
+// CHECK-LABEL: func @no_fold_transfer_write_in_bounds_negative_index
+//       CHECK:   vector.transfer_write
+//   CHECK-NOT:   in_bounds
+//       CHECK:   : vector<4xf32>, memref<8xf32>
+func.func @no_fold_transfer_write_in_bounds_negative_index(%m: memref<8xf32>, %v: vector<4xf32>) {
+  %c-1 = arith.constant -1 : index
+  vector.transfer_write %v, %m[%c-1] : vector<4xf32>, memref<8xf32>
+  return
+}
+
+// -----
+
+// The bound is computed as `sourceSize - vectorSize` rather than
+// `index + vectorSize`, so an index large enough to overflow the addition is
+// still rejected.
----------------
FedericoBruzzone wrote:

Same as above. Probably we could integrate the overflow behavior into the name of the test :D

https://github.com/llvm/llvm-project/pull/219681