diff --git a/lib/Analysis/InstructionSimplify.cpp b/lib/Analysis/InstructionSimplify.cpp
index ea83c83ed0c..aa7993b875c 100644
--- a/lib/Analysis/InstructionSimplify.cpp
+++ b/lib/Analysis/InstructionSimplify.cpp
@@ -3884,8 +3884,11 @@ static Value *SimplifyGEPInst(Type *SrcTy, ArrayRef<Value *> Ops,
   if (!all_of(Ops, [](Value *V) { return isa<Constant>(V); }))
     return nullptr;
 
-  return ConstantExpr::getGetElementPtr(SrcTy, cast<Constant>(Ops[0]),
-                                        Ops.slice(1));
+  auto *CE = ConstantExpr::getGetElementPtr(SrcTy, cast<Constant>(Ops[0]),
+                                            Ops.slice(1));
+  if (auto *CEFolded = ConstantFoldConstant(CE, Q.DL))
+    return CEFolded;
+  return CE;
 }
 
 Value *llvm::SimplifyGEPInst(Type *SrcTy, ArrayRef<Value *> Ops,
diff --git a/test/Analysis/ConstantFolding/gep-constanfolding-error.ll b/test/Analysis/ConstantFolding/gep-constanfolding-error.ll
index 50ad61a8f10..16bc8a983e4 100644
--- a/test/Analysis/ConstantFolding/gep-constanfolding-error.ll
+++ b/test/Analysis/ConstantFolding/gep-constanfolding-error.ll
@@ -44,7 +44,7 @@ entry:
   %9 = add i32 %f.promoted, %smax
   %10 = add i32 %9, 2
   call void @llvm.memset.p0i8.i32(i8* %scevgep, i8 %conv6, i32 %10, i32 1, i1 false)
-; CHECK:  call void @llvm.memset.p0i8.i32(i8* getelementptr inbounds ([6 x [6 x [7 x i8]]], [6 x [6 x [7 x i8]]]* @j, i32 0, i64 5, i64 4, i32 1), i8 %conv6, i32 1, i32 1, i1 false)
+; CHECK:  call void @llvm.memset.p0i8.i32(i8* getelementptr inbounds ([6 x [6 x [7 x i8]]], [6 x [6 x [7 x i8]]]* @j, i32 0, i{{32|64}} 5, i{{32|64}} 4, i32 1), i8 %conv6, i32 1, i32 1, i1 false)
 ; CHECK-NOT: call void @llvm.memset.p0i8.i32(i8* getelementptr ([6 x [6 x [7 x i8]]], [6 x [6 x [7 x i8]]]* @j, i64 1, i64 4, i64 4, i32 1)
   ret i32 0
 }
diff --git a/test/Transforms/InstSimplify/compare.ll b/test/Transforms/InstSimplify/compare.ll
index 20ebd36991a..2fe07901916 100644
--- a/test/Transforms/InstSimplify/compare.ll
+++ b/test/Transforms/InstSimplify/compare.ll
@@ -69,7 +69,7 @@ define i1 @gep4() {
 
 define i1 @PR31262() {
 ; CHECK-LABEL: @PR31262(
-; CHECK-NEXT:    ret i1 icmp uge (i32* getelementptr ([1 x i32], [1 x i32]* @a, i64 0, i64 undef), i32* getelementptr inbounds ([1 x i32], [1 x i32]* @a, i32 0, i32 0))
+; CHECK-NEXT:    ret i1 icmp uge (i32* getelementptr ([1 x i32], [1 x i32]* @a, i32 0, i32 undef), i32* getelementptr inbounds ([1 x i32], [1 x i32]* @a, i32 0, i32 0))
 ;
   %idx = getelementptr inbounds [1 x i32], [1 x i32]* @a, i64 0, i64 undef
   %cmp = icmp uge i32* %idx, getelementptr inbounds ([1 x i32], [1 x i32]* @a, i32 0, i32 0)
diff --git a/test/Transforms/InstSimplify/simplify-nested-bitcast.ll b/test/Transforms/InstSimplify/simplify-nested-bitcast.ll
new file mode 100644
index 00000000000..b7ee79415a2
--- /dev/null
+++ b/test/Transforms/InstSimplify/simplify-nested-bitcast.ll
@@ -0,0 +1,54 @@
+; RUN: opt -always-inline -S %s | FileCheck %s
+%0 = type { i64, i64, i8 addrspace(1)*, i8 addrspace(1)* }
+%__aaa_struct = type { { i8**, i32, i32, i8*, %struct.__block_descriptor addrspace(1)* }, %0, [17 x i8], { i8**, i32, i32, i8*, %struct.__block_descriptor addrspace(1)* }, %0, [18 x i8] }
+%struct.__block_descriptor = type { i64, i64 }
+%struct.__block_literal_generic = type { i8*, i32, i32, i8*, %struct.__block_descriptor addrspace(1)* }
+
+@__aaa_struct_ptr = external addrspace(1) global %__aaa_struct
+@__aaa_const_init = constant %__aaa_struct { { i8**, i32, i32, i8*, %struct.__block_descriptor addrspace(1)* } { i8** null, i32 1342177280, i32 0, i8* bitcast (i32 (i8 addrspace(4)*, i32 addrspace(1)*)* @bl0_block_invoke to i8*), %struct.__block_descriptor addrspace(1)* bitcast (%0 addrspace(1)* getelementptr inbounds (%__aaa_struct, %__aaa_struct addrspace(1)* @__aaa_struct_ptr, i32 0, i32 1) to %struct.__block_descriptor addrspace(1)*) }, %0 { i64 0, i64 32, i8 addrspace(1)* getelementptr inbounds (%__aaa_struct, %__aaa_struct addrspace(1)* @__aaa_struct_ptr, i32 0, i32 2, i32 0), i8 addrspace(1)* null }, [17 x i8] c"bl0_block_invoke\00", { i8**, i32, i32, i8*, %struct.__block_descriptor addrspace(1)* } { i8** null, i32 1342177280, i32 0, i8* bitcast (i32 (i8 addrspace(4)*, i32 addrspace(1)*)* @__f1_block_invoke to i8*), %struct.__block_descriptor addrspace(1)* bitcast (%0 addrspace(1)* getelementptr inbounds (%__aaa_struct, %__aaa_struct addrspace(1)* @__aaa_struct_ptr, i32 0, i32 4) to %struct.__block_descriptor addrspace(1)*) }, %0 { i64 0, i64 32, i8 addrspace(1)* getelementptr inbounds (%__aaa_struct, %__aaa_struct addrspace(1)* @__aaa_struct_ptr, i32 0, i32 5, i32 0), i8 addrspace(1)* null }, [18 x i8] c"__f1_block_invoke\00" }
+
+; Function Attrs: alwaysinline norecurse nounwind readonly
+define i32 @bl0_block_invoke(i8 addrspace(4)* nocapture readnone, i32 addrspace(1)* nocapture readonly) #0 {
+entry:
+  %2 = load i32, i32 addrspace(1)* %1, align 4
+  %mul = shl nsw i32 %2, 1
+  ret i32 %mul
+}
+
+; Function Attrs: alwaysinline nounwind
+define i32 @f0(i32 addrspace(1)*, i32 (i32 addrspace(1)*) addrspace(4)*) #1 {
+entry:
+  %block.literal = bitcast i32 (i32 addrspace(1)*) addrspace(4)* %1 to %struct.__block_literal_generic addrspace(4)*
+  %2 = getelementptr inbounds %struct.__block_literal_generic, %struct.__block_literal_generic addrspace(4)* %block.literal, i64 0, i32 3
+  %3 = bitcast i32 (i32 addrspace(1)*) addrspace(4)* %1 to i8 addrspace(4)*
+  %4 = bitcast i8* addrspace(4)* %2 to i32 (i8 addrspace(4)*, i32 addrspace(1)*)* addrspace(4)*
+  %5 = load i32 (i8 addrspace(4)*, i32 addrspace(1)*)*, i32 (i8 addrspace(4)*, i32 addrspace(1)*)* addrspace(4)* %4, align 8
+  %call = tail call i32 %5(i8 addrspace(4)* %3, i32 addrspace(1)* %0) #2
+  ret i32 %call
+}
+
+; CHECK-LABEL: define void @f1
+; CHECK: %1 = load i32 (i8 addrspace(4)*, i32 addrspace(1)*)*, i32 (i8 addrspace(4)*, i32 addrspace(1)*)* addrspace(4)* bitcast (i8* addrspace(4)* getelementptr (%__aaa_struct, %__aaa_struct addrspace(4)* addrspacecast (%__aaa_struct addrspace(1)* @__aaa_struct_ptr to %__aaa_struct addrspace(4)*), i64 0, i32 0, i32 3) to i32 (i8 addrspace(4)*, i32 addrspace(1)*)* addrspace(4)*), align 8
+
+; Function Attrs: alwaysinline nounwind
+define void @f1(i32 addrspace(1)*) #1 {
+entry:
+  %call = tail call i32 @f0(i32 addrspace(1)* %0, i32 (i32 addrspace(1)*) addrspace(4)* addrspacecast (i32 (i32 addrspace(1)*) addrspace(1)* bitcast (%__aaa_struct addrspace(1)* @__aaa_struct_ptr to i32 (i32 addrspace(1)*) addrspace(1)*) to i32 (i32 addrspace(1)*) addrspace(4)*)) #3
+  store i32 %call, i32 addrspace(1)* %0, align 4
+  %call1 = tail call i32 @f0(i32 addrspace(1)* %0, i32 (i32 addrspace(1)*) addrspace(4)* addrspacecast (i32 (i32 addrspace(1)*) addrspace(1)* bitcast ({ i8**, i32, i32, i8*, %struct.__block_descriptor addrspace(1)* } addrspace(1)* getelementptr inbounds (%__aaa_struct, %__aaa_struct addrspace(1)* @__aaa_struct_ptr, i32 0, i32 3) to i32 (i32 addrspace(1)*) addrspace(1)*) to i32 (i32 addrspace(1)*) addrspace(4)*)) #3
+  store i32 %call1, i32 addrspace(1)* %0, align 4
+  ret void
+}
+
+; Function Attrs: alwaysinline norecurse nounwind readonly
+define i32 @__f1_block_invoke(i8 addrspace(4)* nocapture readnone, i32 addrspace(1)* nocapture readonly) #0 {
+entry:
+  %2 = load i32, i32 addrspace(1)* %1, align 4
+  %add = add nsw i32 %2, 1
+  ret i32 %add
+}
+
+attributes #0 = { alwaysinline norecurse nounwind readonly }
+attributes #1 = { alwaysinline nounwind }
+attributes #2 = { nobuiltin nounwind }
+attributes #3 = { nobuiltin }
diff --git a/test/Transforms/InstSimplify/vector_gep.ll b/test/Transforms/InstSimplify/vector_gep.ll
index b8e61a05cc0..cdf4732d4b5 100644
--- a/test/Transforms/InstSimplify/vector_gep.ll
+++ b/test/Transforms/InstSimplify/vector_gep.ll
@@ -51,7 +51,7 @@ define <4 x i8*> @test5() {
   ret <4 x i8*> %gep
 
 ; CHECK-LABEL: @test5
-; CHECK-NEXT: ret <4 x i8*> getelementptr (i8, <4 x i8*> <i8* inttoptr (i64 1 to i8*), i8* inttoptr (i64 2 to i8*), i8* inttoptr (i64 3 to i8*), i8* inttoptr (i64 4 to i8*)>, <4 x i32> <i32 1, i32 1, i32 1, i32 1>)
+; CHECK-NEXT: ret <4 x i8*> getelementptr (i8, <4 x i8*> <i8* inttoptr (i64 1 to i8*), i8* inttoptr (i64 2 to i8*), i8* inttoptr (i64 3 to i8*), i8* inttoptr (i64 4 to i8*)>, <4 x i64> <i64 1, i64 1, i64 1, i64 1>)
 }
 
 @v = global [24 x [42 x [3 x i32]]] zeroinitializer, align 16
diff --git a/test/Transforms/NewGVN/completeness.ll b/test/Transforms/NewGVN/completeness.ll
index 2b28f12df9d..2c35871e9ae 100644
--- a/test/Transforms/NewGVN/completeness.ll
+++ b/test/Transforms/NewGVN/completeness.ll
@@ -395,7 +395,7 @@ define void @test10() {
 ; CHECK:       g:
 ; CHECK-NEXT:    [[N:%.*]] = phi i32* [ [[H:%.*]], [[I:%.*]] ], [ null, [[B:%.*]] ]
 ; CHECK-NEXT:    [[H]] = getelementptr i32, i32* [[N]], i64 1
-; CHECK-NEXT:    [[J:%.*]] = icmp eq i32* [[H]], getelementptr (i32, i32* null, i64 8)
+; CHECK-NEXT:    [[J:%.*]] = icmp eq i32* %h, inttoptr (i64 32 to i32*)
 ; CHECK-NEXT:    br i1 [[J]], label [[C:%.*]], label [[I]]
 ; CHECK:       i:
 ; CHECK-NEXT:    br i1 undef, label [[K:%.*]], label [[G]]
diff --git a/test/Transforms/NewGVN/loadforward.ll b/test/Transforms/NewGVN/loadforward.ll
index d66b5332601..b4cbcc6b0f4 100644
--- a/test/Transforms/NewGVN/loadforward.ll
+++ b/test/Transforms/NewGVN/loadforward.ll
@@ -9,8 +9,8 @@ target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128"
 ;; Test that we forward the first store to the second load
 define i16 @bazinga() {
 ; CHECK-LABEL: @bazinga(
-; CHECK-NEXT:    [[_TMP10:%.*]] = load i16, i16* getelementptr inbounds (%rec11, %rec11* @str, i16 0, i32 1)
-; CHECK-NEXT:    store i16 [[_TMP10]], i16* getelementptr inbounds (%rec11, %rec11* @str, i16 0, i32 0)
+; CHECK-NEXT:    [[_TMP10:%.*]] = load i16, i16* getelementptr inbounds (%rec11, %rec11* @str, i64 0, i32 1)
+; CHECK-NEXT:    store i16 [[_TMP10]], i16* getelementptr inbounds (%rec11, %rec11* @str, i64 0, i32 0)
 ; CHECK-NEXT:    [[_TMP15:%.*]] = icmp eq i16 [[_TMP10]], 3
 ; CHECK-NEXT:    [[_TMP16:%.*]] = select i1 [[_TMP15]], i16 1, i16 0
 ; CHECK-NEXT:    br label [[BB1:%.*]]