diff --git a/llvm/lib/Target/X86/X86FastISel.cpp b/llvm/lib/Target/X86/X86FastISel.cpp
index f29d127c85d1..9d42ac2e470c 100644
--- a/llvm/lib/Target/X86/X86FastISel.cpp
+++ b/llvm/lib/Target/X86/X86FastISel.cpp
@@ -1933,7 +1933,7 @@ bool X86FastISel::TryToFoldLoad(MachineInstr *MI, unsigned OpNo,
     XII.foldMemoryOperandImpl(*FuncInfo.MF, MI, OpNo, AddrOps, Size, Alignment);
   if (Result == 0) return false;
 
-  MI->getParent()->insert(MI, Result);
+  FuncInfo.MBB->insert(FuncInfo.InsertPt, Result);
   MI->eraseFromParent();
   return true;
 }
diff --git a/llvm/test/CodeGen/X86/fast-isel-gep.ll b/llvm/test/CodeGen/X86/fast-isel-gep.ll
index 577dd7223a4d..622a1ff831d0 100644
--- a/llvm/test/CodeGen/X86/fast-isel-gep.ll
+++ b/llvm/test/CodeGen/X86/fast-isel-gep.ll
@@ -70,3 +70,20 @@ entry:
 ; X64: test4:
 ; X64: 128(%r{{.*}},%r{{.*}},8)
 }
+
+; PR8961 - Make sure the sext for the GEP addressing comes before the load that
+; is folded.
+define i64 @test5(i8* %A, i32 %I, i64 %B) nounwind {
+  %v8 = getelementptr i8* %A, i32 %I
+  %v9 = bitcast i8* %v8 to i64*
+  %v10 = load i64* %v9
+  %v11 = add i64 %B, %v10
+  ret i64 %v11
+; X64: test5:
+; X64: movslq	%esi, %rax
+; X64-NEXT: movq	(%rdi,%rax), %rax
+; X64-NEXT: addq	%rdx, %rax
+; X64-NEXT: ret
+}
+
+