--- /dev/null
+; RUN: opt %loadPolly -polly-scops -analyze < %s | FileCheck %s
+;
+; The load access to A has a pointer-bitcast to another elements size before the
+; GetElementPtr. Verify that we do not the GEP delinearization because it
+; mismatches with the size of the loaded element type.
+;
+; void f(short A[][4], int N, int P) {
+; short(*B)[4] = &A[P][0];
+; for (int i = 0; i < N; i++)
+; *((<4 x short> *)&A[7 * i][0]) = *((<4 x short>)&B[7 * i][0]);
+; }
+;
+define void @f([4 x i16]* %A, i32 %N, i32 %P) {
+entry:
+ %arrayidx1 = getelementptr inbounds [4 x i16], [4 x i16]* %A, i32 %P, i64 0
+ %tmp = bitcast i16* %arrayidx1 to [4 x i16]*
+ br label %for.cond
+
+for.cond:
+ %indvars.iv = phi i32 [ %indvars.iv.next, %for.inc ], [ 0, %entry ]
+ %cmp = icmp slt i32 %indvars.iv, %N
+ br i1 %cmp, label %for.body, label %for.end
+
+for.body:
+ %mul = mul nsw i32 %indvars.iv, 7
+ %arrayidx4 = getelementptr inbounds [4 x i16], [4 x i16]* %tmp, i32 %mul, i64 0
+ %bc4 = bitcast i16* %arrayidx4 to <4 x i16>*
+ %tmp3 = load <4 x i16>, <4 x i16>* %bc4
+ %arrayidx8 = getelementptr inbounds [4 x i16], [4 x i16]* %A, i32 %mul, i64 0
+ %bc8 = bitcast i16* %arrayidx8 to <4 x i16>*
+ store <4 x i16> %tmp3, <4 x i16>* %bc8
+ br label %for.inc
+
+for.inc:
+ %indvars.iv.next = add nuw nsw i32 %indvars.iv, 1
+ br label %for.cond
+
+for.end:
+ ret void
+}
+
+
+; CHECK: Arrays {
+; CHECK-NEXT: <4 x i16> MemRef_A[*]; // Element size 8
+; CHECK-NEXT: }
+; CHECK: Statements {
+; CHECK-NEXT: Stmt_for_body
+; CHECK-NEXT: Domain :=
+; CHECK-NEXT: [N, P] -> { Stmt_for_body[i0] : 0 <= i0 < N };
+; CHECK-NEXT: Schedule :=
+; CHECK-NEXT: [N, P] -> { Stmt_for_body[i0] -> [i0] };
+; CHECK-NEXT: ReadAccess := [Reduction Type: NONE] [Scalar: 0]
+; CHECK-NEXT: [N, P] -> { Stmt_for_body[i0] -> MemRef_A[P + 7i0] };
+; CHECK-NEXT: MustWriteAccess := [Reduction Type: NONE] [Scalar: 0]
+; CHECK-NEXT: [N, P] -> { Stmt_for_body[i0] -> MemRef_A[7i0] };
+; CHECK-NEXT: }