[ConstantFold] Get rid of special cases for sizeof etc.

[efriedma]

Summary

Constant folding will fold these down to simple constants (or at least, expressions that don't involve a GEP). We don't need heroics to try to optimize the form of the expression before that happens.

Fixes https://bugs.llvm.org/show_bug.cgi?id=51232 .

[mstorsjo]

I can’t say much about the code change itself, but it does indeed fix the build failure I was seeing, thanks!

[nikic]

LGTM -- the target-dependent folder is going to reduce these to constants, I don't see any value in trying very hard to optimize them in the target-indepenedent folder.

Note that you need to adjust clang codegen tests as well.

llvm/test/Other/constant-fold-gep.ll

50

Outdated comment.

[mstorsjo]

Thanks! Do you think this would be suitable for the 13.x release branch too?

[efriedma]

Just committed 6eb2ffba to fix a couple regression tests.

Seems low-risk to cherry-pick to 13.x.

clang/test/CodeGen/catch-nullptr-and-nonzero-offset-in-offsetof-idiom.c

+// NOTE: Assertions have been autogenerated by utils/update_cc_test_checks.py
 // RUN: %clang_cc1 -x c   -emit-llvm %s -o - -triple x86_64-linux-gnu | FileCheck %s
 // RUN: %clang_cc1 -x c   -fsanitize=pointer-overflow -fno-sanitize-recover=pointer-overflow -emit-llvm %s -o - -triple x86_64-linux-gnu | FileCheck %s
 
 // RUN: %clang_cc1 -x c++ -emit-llvm %s -o - -triple x86_64-linux-gnu | FileCheck %s
 // RUN: %clang_cc1 -x c++ -fsanitize=pointer-overflow -fno-sanitize-recover=pointer-overflow -emit-llvm %s -o - -triple x86_64-linux-gnu | FileCheck %s
 
 #include <stdint.h>
 
 #ifdef __cplusplus
 extern "C" {
 #endif
 
 struct S {
   int x, y;
 };
 
-// CHECK-LABEL: define{{.*}} i64 @{{.*}}get_offset_of_y_naively{{.*}}(
+// CHECK-LABEL: @get_offset_of_y_naively(
+// CHECK-NEXT:  entry:
+// CHECK-NEXT:    ret i64 ptrtoint (i32* getelementptr inbounds ([[STRUCT_S:%.*]], %struct.S* null, i32 0, i32 1) to i64)
+//
 uintptr_t get_offset_of_y_naively() {
-  // CHECK: [[ENTRY:.*]]:
-  // CHECK-NEXT:   ret i64 ptrtoint (i32* getelementptr (i32, i32* null, i32 1) to i64)
-  // CHECK-NEXT: }
   return ((uintptr_t)(&(((struct S *)0)->y)));
 }
 
-// CHECK-LABEL: define{{.*}} i64 @{{.*}}get_offset_of_y_via_builtin{{.*}}(
-uintptr_t get_offset_of_y_via_builtin() {
-  // CHECK: [[ENTRY:.*]]:
-  // CHECK-NEXT:   ret i64 4
-  // CHECK-NEXT: }
+// CHECK-LABEL: @get_offset_of_y_via_builtin(
+// CHECK-NEXT:  entry:
+// CHECK-NEXT:    ret i64 4
+//
+uintptr_t get_offset_of_y_via_builtin() {
   return __builtin_offsetof(struct S, y);
 }
 
 #ifdef __cplusplus
 }
 #endif

clang/test/CodeGen/catch-nullptr-and-nonzero-offset.c

 }
 
 char *nullptr_allones_BAD() {
   // CHECK:                           define{{.*}} i8* @nullptr_allones_BAD()
   // CHECK-NEXT:                      [[ENTRY:.*]]:
   // CHECK-SANITIZE-C-NEXT:             br i1 false, label %[[CONT:.*]], label %[[HANDLER_POINTER_OVERFLOW:[^,]+]],{{.*}} !nosanitize
   // CHECK-SANITIZE-CPP-NEXT:           br i1 icmp eq (i64 mul (i64 ptrtoint (i8* getelementptr (i8, i8* null, i32 1) to i64), i64 -1), i64 0), label %[[CONT:.*]], label %[[HANDLER_POINTER_OVERFLOW:[^,]+]],{{.*}} !nosanitize
   // CHECK-SANITIZE:                  [[HANDLER_POINTER_OVERFLOW]]:
-  // CHECK-SANITIZE-NORECOVER-NEXT:     call void @__ubsan_handle_pointer_overflow_abort(i8* bitcast ({ {{{.*}}} }* @[[LINE_800]] to i8*), i64 0, i64 mul (i64 ptrtoint (i8* getelementptr (i8, i8* null, i32 1) to i64), i64 -1))
+  // CHECK-SANITIZE-NORECOVER-NEXT:     call void @__ubsan_handle_pointer_overflow_abort(i8* bitcast ({ {{{.*}}} }* @[[LINE_800]] to i8*), i64 0, i64 ptrtoint (i8* getelementptr inbounds (i8, i8* null, i64 -1) to i64))
   // CHECK-SANITIZE-RECOVER-NEXT:       call void @__ubsan_handle_pointer_overflow(i8* bitcast ({ {{{.*}}} }* @[[LINE_800]] to i8*), i64 0, i64 mul (i64 ptrtoint (i8* getelementptr (i8, i8* null, i32 1) to i64), i64 -1))
   // CHECK-SANITIZE-TRAP-NEXT:          call void @llvm.ubsantrap(i8 19){{.*}}, !nosanitize
   // CHECK-SANITIZE-UNREACHABLE-NEXT:   unreachable, !nosanitize
   // CHECK-SANITIZE:                  [[CONT]]:
   // CHECK-NEXT:                        ret i8* getelementptr inbounds (i8, i8* null, i64 -1)
   static char *const base = (char *)0;
   static const unsigned long offset = -1;
 #line 800

clang/test/CodeGenCXX/catch-nullptr-and-nonzero-offset-in-offsetof-idiom.cpp

+// NOTE: Assertions have been autogenerated by utils/update_cc_test_checks.py
 // RUN: %clang_cc1 -x c++ -emit-llvm %s -o - -triple x86_64-linux-gnu | FileCheck %s
 // RUN: %clang_cc1 -x c++ -fsanitize=pointer-overflow -fno-sanitize-recover=pointer-overflow -emit-llvm %s -o - -triple x86_64-linux-gnu | FileCheck %s
 
 #include <stdint.h>
 
 struct S {
   int x, y;
 };
 
-// CHECK-LABEL: define{{.*}} i64 @{{.*}}get_offset_of_y_naively{{.*}}(
+// CHECK-LABEL: @_Z23get_offset_of_y_naivelyv(
+// CHECK-NEXT:  entry:
+// CHECK-NEXT:    ret i64 ptrtoint (i32* getelementptr inbounds ([[STRUCT_S:%.*]], %struct.S* null, i32 0, i32 1) to i64)
+//
 uintptr_t get_offset_of_y_naively() {
-  // CHECK: [[ENTRY:.*]]:
-  // CHECK-NEXT:   ret i64 ptrtoint (i32* getelementptr (i32, i32* null, i32 1) to i64)
-  // CHECK-NEXT: }
   return ((uintptr_t)(&(((S *)nullptr)->y)));
 }
 
-// CHECK-LABEL: define{{.*}} i64 @{{.*}}get_offset_of_y_via_builtin{{.*}}(
-uintptr_t get_offset_of_y_via_builtin() {
-  // CHECK: [[ENTRY:.*]]:
-  // CHECK-NEXT:   ret i64 4
-  // CHECK-NEXT: }
+// CHECK-LABEL: @_Z27get_offset_of_y_via_builtinv(
+// CHECK-NEXT:  entry:
+// CHECK-NEXT:    ret i64 4
+//
+uintptr_t get_offset_of_y_via_builtin() {
   return __builtin_offsetof(S, y);
 }

clang/test/OpenMP/for_reduction_codegen.cpp

 // CHECK1-NEXT:    [[TMP23:%.*]] = sext i32 [[TMP22]] to i64
 // CHECK1-NEXT:    [[LB_ADD_LEN11:%.*]] = add nsw i64 0, [[TMP23]]
 // CHECK1-NEXT:    [[ARRAYIDX12:%.*]] = getelementptr inbounds [10 x [4 x %struct.S]], [10 x [4 x %struct.S]]* [[TMP4]], i64 0, i64 [[LB_ADD_LEN11]]
 // CHECK1-NEXT:    [[ARRAYDECAY13:%.*]] = getelementptr inbounds [4 x %struct.S], [4 x %struct.S]* [[ARRAYIDX12]], i64 0, i64 0
 // CHECK1-NEXT:    [[ARRAYIDX14:%.*]] = getelementptr inbounds [[STRUCT_S]], %struct.S* [[ARRAYDECAY13]], i64 2
 // CHECK1-NEXT:    [[TMP24:%.*]] = ptrtoint %struct.S* [[ARRAYIDX14]] to i64
 // CHECK1-NEXT:    [[TMP25:%.*]] = ptrtoint %struct.S* [[ARRAYIDX9]] to i64
 // CHECK1-NEXT:    [[TMP26:%.*]] = sub i64 [[TMP24]], [[TMP25]]
-// CHECK1-NEXT:    [[TMP27:%.*]] = sdiv exact i64 [[TMP26]], ptrtoint (float* getelementptr (float, float* null, i32 1) to i64)
+// CHECK1-NEXT:    [[TMP27:%.*]] = sdiv exact i64 [[TMP26]], ptrtoint (%struct.S* getelementptr ([[STRUCT_S]], %struct.S* null, i32 1) to i64)
 // CHECK1-NEXT:    [[TMP28:%.*]] = add nuw i64 [[TMP27]], 1
-// CHECK1-NEXT:    [[TMP29:%.*]] = mul nuw i64 [[TMP28]], ptrtoint (float* getelementptr (float, float* null, i32 1) to i64)
+// CHECK1-NEXT:    [[TMP29:%.*]] = mul nuw i64 [[TMP28]], ptrtoint (%struct.S* getelementptr ([[STRUCT_S]], %struct.S* null, i32 1) to i64)
 // CHECK1-NEXT:    [[VLA15:%.*]] = alloca [[STRUCT_S]], i64 [[TMP28]], align 16
 // CHECK1-NEXT:    store i64 [[TMP28]], i64* [[__VLA_EXPR1]], align 8
 // CHECK1-NEXT:    [[TMP30:%.*]] = getelementptr [[STRUCT_S]], %struct.S* [[VLA15]], i64 [[TMP28]]
 // CHECK1-NEXT:    [[OMP_ARRAYINIT_ISEMPTY16:%.*]] = icmp eq %struct.S* [[VLA15]], [[TMP30]]
 // CHECK1-NEXT:    br i1 [[OMP_ARRAYINIT_ISEMPTY16]], label [[OMP_ARRAYINIT_DONE21:%.*]], label [[OMP_ARRAYINIT_BODY17:%.*]]
 // CHECK1:       omp.arrayinit.body17:
 // CHECK1-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST18:%.*]] = phi %struct.S* [ [[VLA15]], [[OMP_ARRAYINIT_DONE]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT19:%.*]], [[OMP_ARRAYINIT_BODY17]] ]
 // CHECK1-NEXT:    call void @_ZN1SIfEC1Ev(%struct.S* nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_DESTELEMENTPAST18]])
 // CHECK1-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT19]] = getelementptr [[STRUCT_S]], %struct.S* [[OMP_ARRAYCPY_DESTELEMENTPAST18]], i32 1
 // CHECK1-NEXT:    [[OMP_ARRAYCPY_DONE20:%.*]] = icmp eq %struct.S* [[OMP_ARRAYCPY_DEST_ELEMENT19]], [[TMP30]]
 // CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_DONE20]], label [[OMP_ARRAYINIT_DONE21]], label [[OMP_ARRAYINIT_BODY17]]
 // CHECK1:       omp.arrayinit.done21:
 // CHECK1-NEXT:    [[TMP31:%.*]] = bitcast [10 x [4 x %struct.S]]* [[TMP4]] to %struct.S*
 // CHECK1-NEXT:    [[TMP32:%.*]] = ptrtoint %struct.S* [[TMP31]] to i64
 // CHECK1-NEXT:    [[TMP33:%.*]] = ptrtoint %struct.S* [[ARRAYIDX9]] to i64
 // CHECK1-NEXT:    [[TMP34:%.*]] = sub i64 [[TMP32]], [[TMP33]]
-// CHECK1-NEXT:    [[TMP35:%.*]] = sdiv exact i64 [[TMP34]], ptrtoint (float* getelementptr (float, float* null, i32 1) to i64)
+// CHECK1-NEXT:    [[TMP35:%.*]] = sdiv exact i64 [[TMP34]], ptrtoint (%struct.S* getelementptr ([[STRUCT_S]], %struct.S* null, i32 1) to i64)
 // CHECK1-NEXT:    [[TMP36:%.*]] = getelementptr [[STRUCT_S]], %struct.S* [[VLA15]], i64 [[TMP35]]
 // CHECK1-NEXT:    [[TMP37:%.*]] = bitcast %struct.S* [[TMP36]] to [10 x [4 x %struct.S]]*
 // CHECK1-NEXT:    [[TMP38:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK1-NEXT:    [[TMP39:%.*]] = load i32, i32* [[TMP38]], align 4
 // CHECK1-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP39]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
 // CHECK1-NEXT:    [[TMP40:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP40]], 9
 // CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]

 // CHECK1-NEXT:    [[ARRAYIDX1:%.*]] = getelementptr inbounds [[STRUCT_S:%.*]], %struct.S* [[TMP2]], i64 1
 // CHECK1-NEXT:    [[TMP3:%.*]] = load %struct.S**, %struct.S*** [[TMP0]], align 8
 // CHECK1-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds %struct.S*, %struct.S** [[TMP3]], i64 4
 // CHECK1-NEXT:    [[TMP4:%.*]] = load %struct.S*, %struct.S** [[ARRAYIDX2]], align 8
 // CHECK1-NEXT:    [[ARRAYIDX3:%.*]] = getelementptr inbounds [[STRUCT_S]], %struct.S* [[TMP4]], i64 6
 // CHECK1-NEXT:    [[TMP5:%.*]] = ptrtoint %struct.S* [[ARRAYIDX3]] to i64
 // CHECK1-NEXT:    [[TMP6:%.*]] = ptrtoint %struct.S* [[ARRAYIDX1]] to i64
 // CHECK1-NEXT:    [[TMP7:%.*]] = sub i64 [[TMP5]], [[TMP6]]
-// CHECK1-NEXT:    [[TMP8:%.*]] = sdiv exact i64 [[TMP7]], ptrtoint (float* getelementptr (float, float* null, i32 1) to i64)
+// CHECK1-NEXT:    [[TMP8:%.*]] = sdiv exact i64 [[TMP7]], ptrtoint (%struct.S* getelementptr ([[STRUCT_S]], %struct.S* null, i32 1) to i64)
 // CHECK1-NEXT:    [[TMP9:%.*]] = add nuw i64 [[TMP8]], 1
-// CHECK1-NEXT:    [[TMP10:%.*]] = mul nuw i64 [[TMP9]], ptrtoint (float* getelementptr (float, float* null, i32 1) to i64)
+// CHECK1-NEXT:    [[TMP10:%.*]] = mul nuw i64 [[TMP9]], ptrtoint (%struct.S* getelementptr ([[STRUCT_S]], %struct.S* null, i32 1) to i64)
 // CHECK1-NEXT:    [[TMP11:%.*]] = call i8* @llvm.stacksave()
 // CHECK1-NEXT:    store i8* [[TMP11]], i8** [[SAVED_STACK]], align 8
 // CHECK1-NEXT:    [[VLA:%.*]] = alloca [[STRUCT_S]], i64 [[TMP9]], align 16
 // CHECK1-NEXT:    store i64 [[TMP9]], i64* [[__VLA_EXPR0]], align 8
 // CHECK1-NEXT:    [[TMP12:%.*]] = getelementptr [[STRUCT_S]], %struct.S* [[VLA]], i64 [[TMP9]]
 // CHECK1-NEXT:    [[OMP_ARRAYINIT_ISEMPTY:%.*]] = icmp eq %struct.S* [[VLA]], [[TMP12]]
 // CHECK1-NEXT:    br i1 [[OMP_ARRAYINIT_ISEMPTY]], label [[OMP_ARRAYINIT_DONE:%.*]], label [[OMP_ARRAYINIT_BODY:%.*]]
 // CHECK1:       omp.arrayinit.body:
 // CHECK1-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi %struct.S* [ [[VLA]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYINIT_BODY]] ]
 // CHECK1-NEXT:    call void @_ZN1SIfEC1Ev(%struct.S* nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_DESTELEMENTPAST]])
 // CHECK1-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr [[STRUCT_S]], %struct.S* [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
 // CHECK1-NEXT:    [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq %struct.S* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP12]]
 // CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYINIT_DONE]], label [[OMP_ARRAYINIT_BODY]]
 // CHECK1:       omp.arrayinit.done:
 // CHECK1-NEXT:    [[TMP13:%.*]] = load %struct.S**, %struct.S*** [[TMP0]], align 8
 // CHECK1-NEXT:    [[TMP14:%.*]] = load %struct.S*, %struct.S** [[TMP13]], align 8
 // CHECK1-NEXT:    [[TMP15:%.*]] = ptrtoint %struct.S* [[TMP14]] to i64
 // CHECK1-NEXT:    [[TMP16:%.*]] = ptrtoint %struct.S* [[ARRAYIDX1]] to i64
 // CHECK1-NEXT:    [[TMP17:%.*]] = sub i64 [[TMP15]], [[TMP16]]
-// CHECK1-NEXT:    [[TMP18:%.*]] = sdiv exact i64 [[TMP17]], ptrtoint (float* getelementptr (float, float* null, i32 1) to i64)
+// CHECK1-NEXT:    [[TMP18:%.*]] = sdiv exact i64 [[TMP17]], ptrtoint (%struct.S* getelementptr ([[STRUCT_S]], %struct.S* null, i32 1) to i64)
 // CHECK1-NEXT:    [[TMP19:%.*]] = getelementptr [[STRUCT_S]], %struct.S* [[VLA]], i64 [[TMP18]]
 // CHECK1-NEXT:    store %struct.S** [[_TMP5]], %struct.S*** [[_TMP4]], align 8
 // CHECK1-NEXT:    store %struct.S* [[TMP19]], %struct.S** [[_TMP5]], align 8
 // CHECK1-NEXT:    [[TMP20:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK1-NEXT:    [[TMP21:%.*]] = load i32, i32* [[TMP20]], align 4
 // CHECK1-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP21]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
 // CHECK1-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP22]], 9

 // CHECK1-NEXT:    [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq %struct.S* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP5]]
 // CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYINIT_DONE]], label [[OMP_ARRAYINIT_BODY]]
 // CHECK1:       omp.arrayinit.done:
 // CHECK1-NEXT:    [[TMP6:%.*]] = load %struct.S**, %struct.S*** [[TMP0]], align 8
 // CHECK1-NEXT:    [[TMP7:%.*]] = load %struct.S*, %struct.S** [[TMP6]], align 8
 // CHECK1-NEXT:    [[TMP8:%.*]] = ptrtoint %struct.S* [[TMP7]] to i64
 // CHECK1-NEXT:    [[TMP9:%.*]] = ptrtoint %struct.S* [[ARRAYIDX1]] to i64
 // CHECK1-NEXT:    [[TMP10:%.*]] = sub i64 [[TMP8]], [[TMP9]]
-// CHECK1-NEXT:    [[TMP11:%.*]] = sdiv exact i64 [[TMP10]], ptrtoint (float* getelementptr (float, float* null, i32 1) to i64)
+// CHECK1-NEXT:    [[TMP11:%.*]] = sdiv exact i64 [[TMP10]], ptrtoint (%struct.S* getelementptr ([[STRUCT_S]], %struct.S* null, i32 1) to i64)
 // CHECK1-NEXT:    [[TMP12:%.*]] = bitcast [1 x [6 x %struct.S]]* [[VAR24]] to %struct.S*
 // CHECK1-NEXT:    [[TMP13:%.*]] = getelementptr [[STRUCT_S]], %struct.S* [[TMP12]], i64 [[TMP11]]
 // CHECK1-NEXT:    store %struct.S** [[_TMP6]], %struct.S*** [[_TMP5]], align 8
 // CHECK1-NEXT:    store %struct.S* [[TMP13]], %struct.S** [[_TMP6]], align 8
 // CHECK1-NEXT:    [[RHS_BEGIN:%.*]] = bitcast [1 x [6 x %struct.S]]* [[VAR24]] to %struct.S*
 // CHECK1-NEXT:    [[TMP14:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK1-NEXT:    [[TMP15:%.*]] = load i32, i32* [[TMP14]], align 4
 // CHECK1-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP15]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)

 // CHECK1-NEXT:    [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq %struct.S* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP5]]
 // CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYINIT_DONE]], label [[OMP_ARRAYINIT_BODY]]
 // CHECK1:       omp.arrayinit.done:
 // CHECK1-NEXT:    [[TMP6:%.*]] = load %struct.S**, %struct.S*** [[TMP0]], align 8
 // CHECK1-NEXT:    [[TMP7:%.*]] = load %struct.S*, %struct.S** [[TMP6]], align 8
 // CHECK1-NEXT:    [[TMP8:%.*]] = ptrtoint %struct.S* [[TMP7]] to i64
 // CHECK1-NEXT:    [[TMP9:%.*]] = ptrtoint %struct.S* [[ARRAYIDX1]] to i64
 // CHECK1-NEXT:    [[TMP10:%.*]] = sub i64 [[TMP8]], [[TMP9]]
-// CHECK1-NEXT:    [[TMP11:%.*]] = sdiv exact i64 [[TMP10]], ptrtoint (float* getelementptr (float, float* null, i32 1) to i64)
+// CHECK1-NEXT:    [[TMP11:%.*]] = sdiv exact i64 [[TMP10]], ptrtoint (%struct.S* getelementptr ([[STRUCT_S]], %struct.S* null, i32 1) to i64)
 // CHECK1-NEXT:    [[TMP12:%.*]] = bitcast [1 x [6 x %struct.S]]* [[VAR24]] to %struct.S*
 // CHECK1-NEXT:    [[TMP13:%.*]] = getelementptr [[STRUCT_S]], %struct.S* [[TMP12]], i64 [[TMP11]]
 // CHECK1-NEXT:    store %struct.S** [[_TMP6]], %struct.S*** [[_TMP5]], align 8
 // CHECK1-NEXT:    store %struct.S* [[TMP13]], %struct.S** [[_TMP6]], align 8
 // CHECK1-NEXT:    [[RHS_BEGIN:%.*]] = bitcast [1 x [6 x %struct.S]]* [[VAR24]] to %struct.S*
 // CHECK1-NEXT:    [[TMP14:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK1-NEXT:    [[TMP15:%.*]] = load i32, i32* [[TMP14]], align 4
 // CHECK1-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP15]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)

 // CHECK1-NEXT:    [[TMP4:%.*]] = load %struct.S*, %struct.S** [[ARRAYIDX2]], align 8
 // CHECK1-NEXT:    [[ARRAYIDX3:%.*]] = getelementptr inbounds [[STRUCT_S]], %struct.S* [[TMP4]], i64 1
 // CHECK1-NEXT:    call void @_ZN1SIfEC1Ev(%struct.S* nonnull align 4 dereferenceable(4) [[VAR24]])
 // CHECK1-NEXT:    [[TMP5:%.*]] = load %struct.S**, %struct.S*** [[TMP0]], align 8
 // CHECK1-NEXT:    [[TMP6:%.*]] = load %struct.S*, %struct.S** [[TMP5]], align 8
 // CHECK1-NEXT:    [[TMP7:%.*]] = ptrtoint %struct.S* [[TMP6]] to i64
 // CHECK1-NEXT:    [[TMP8:%.*]] = ptrtoint %struct.S* [[ARRAYIDX1]] to i64
 // CHECK1-NEXT:    [[TMP9:%.*]] = sub i64 [[TMP7]], [[TMP8]]
-// CHECK1-NEXT:    [[TMP10:%.*]] = sdiv exact i64 [[TMP9]], ptrtoint (float* getelementptr (float, float* null, i32 1) to i64)
+// CHECK1-NEXT:    [[TMP10:%.*]] = sdiv exact i64 [[TMP9]], ptrtoint (%struct.S* getelementptr ([[STRUCT_S]], %struct.S* null, i32 1) to i64)
 // CHECK1-NEXT:    [[TMP11:%.*]] = getelementptr [[STRUCT_S]], %struct.S* [[VAR24]], i64 [[TMP10]]
 // CHECK1-NEXT:    store %struct.S** [[_TMP6]], %struct.S*** [[_TMP5]], align 8
 // CHECK1-NEXT:    store %struct.S* [[TMP11]], %struct.S** [[_TMP6]], align 8
 // CHECK1-NEXT:    [[TMP12:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK1-NEXT:    [[TMP13:%.*]] = load i32, i32* [[TMP12]], align 4
 // CHECK1-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP13]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
 // CHECK1-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP14]], 9

 // CHECK1-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr [[STRUCT_S]], %struct.S* [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
 // CHECK1-NEXT:    [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq %struct.S* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP1]]
 // CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYINIT_DONE]], label [[OMP_ARRAYINIT_BODY]]
 // CHECK1:       omp.arrayinit.done:
 // CHECK1-NEXT:    [[TMP2:%.*]] = bitcast [5 x %struct.S]* [[TMP0]] to %struct.S*
 // CHECK1-NEXT:    [[TMP3:%.*]] = ptrtoint %struct.S* [[TMP2]] to i64
 // CHECK1-NEXT:    [[TMP4:%.*]] = ptrtoint %struct.S* [[ARRAYIDX]] to i64
 // CHECK1-NEXT:    [[TMP5:%.*]] = sub i64 [[TMP3]], [[TMP4]]
-// CHECK1-NEXT:    [[TMP6:%.*]] = sdiv exact i64 [[TMP5]], ptrtoint (float* getelementptr (float, float* null, i32 1) to i64)
+// CHECK1-NEXT:    [[TMP6:%.*]] = sdiv exact i64 [[TMP5]], ptrtoint (%struct.S* getelementptr ([[STRUCT_S]], %struct.S* null, i32 1) to i64)
 // CHECK1-NEXT:    [[TMP7:%.*]] = bitcast [5 x %struct.S]* [[VVAR22]] to %struct.S*
 // CHECK1-NEXT:    [[TMP8:%.*]] = getelementptr [[STRUCT_S]], %struct.S* [[TMP7]], i64 [[TMP6]]
 // CHECK1-NEXT:    [[TMP9:%.*]] = bitcast %struct.S* [[TMP8]] to [5 x %struct.S]*
 // CHECK1-NEXT:    [[RHS_BEGIN:%.*]] = bitcast [5 x %struct.S]* [[VVAR22]] to %struct.S*
 // CHECK1-NEXT:    [[TMP10:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK1-NEXT:    [[TMP11:%.*]] = load i32, i32* [[TMP10]], align 4
 // CHECK1-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP11]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
 // CHECK1-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4

 // CHECK1-NEXT:    [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq %struct.S* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP4]]
 // CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYINIT_DONE]], label [[OMP_ARRAYINIT_BODY]]
 // CHECK1:       omp.arrayinit.done:
 // CHECK1-NEXT:    [[TMP5:%.*]] = load [4 x %struct.S]*, [4 x %struct.S]** [[_TMP1]], align 8
 // CHECK1-NEXT:    [[TMP6:%.*]] = bitcast [4 x %struct.S]* [[TMP5]] to %struct.S*
 // CHECK1-NEXT:    [[TMP7:%.*]] = ptrtoint %struct.S* [[TMP6]] to i64
 // CHECK1-NEXT:    [[TMP8:%.*]] = ptrtoint %struct.S* [[ARRAYIDX]] to i64
 // CHECK1-NEXT:    [[TMP9:%.*]] = sub i64 [[TMP7]], [[TMP8]]
-// CHECK1-NEXT:    [[TMP10:%.*]] = sdiv exact i64 [[TMP9]], ptrtoint (float* getelementptr (float, float* null, i32 1) to i64)
+// CHECK1-NEXT:    [[TMP10:%.*]] = sdiv exact i64 [[TMP9]], ptrtoint (%struct.S* getelementptr ([[STRUCT_S]], %struct.S* null, i32 1) to i64)
 // CHECK1-NEXT:    [[TMP11:%.*]] = bitcast [2 x %struct.S]* [[VAR34]] to %struct.S*
 // CHECK1-NEXT:    [[TMP12:%.*]] = getelementptr [[STRUCT_S]], %struct.S* [[TMP11]], i64 [[TMP10]]
 // CHECK1-NEXT:    [[TMP13:%.*]] = bitcast %struct.S* [[TMP12]] to [4 x %struct.S]*
 // CHECK1-NEXT:    store [4 x %struct.S]* [[TMP13]], [4 x %struct.S]** [[_TMP5]], align 8
 // CHECK1-NEXT:    [[RHS_BEGIN:%.*]] = bitcast [2 x %struct.S]* [[VAR34]] to %struct.S*
 // CHECK1-NEXT:    [[TMP14:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK1-NEXT:    [[TMP15:%.*]] = load i32, i32* [[TMP14]], align 4
 // CHECK1-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP15]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)

 // CHECK1-NEXT:    [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq %struct.S* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP4]]
 // CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYINIT_DONE]], label [[OMP_ARRAYINIT_BODY]]
 // CHECK1:       omp.arrayinit.done:
 // CHECK1-NEXT:    [[TMP5:%.*]] = load [4 x %struct.S]*, [4 x %struct.S]** [[_TMP1]], align 8
 // CHECK1-NEXT:    [[TMP6:%.*]] = bitcast [4 x %struct.S]* [[TMP5]] to %struct.S*
 // CHECK1-NEXT:    [[TMP7:%.*]] = ptrtoint %struct.S* [[TMP6]] to i64
 // CHECK1-NEXT:    [[TMP8:%.*]] = ptrtoint %struct.S* [[ARRAYIDX]] to i64
 // CHECK1-NEXT:    [[TMP9:%.*]] = sub i64 [[TMP7]], [[TMP8]]
-// CHECK1-NEXT:    [[TMP10:%.*]] = sdiv exact i64 [[TMP9]], ptrtoint (float* getelementptr (float, float* null, i32 1) to i64)
+// CHECK1-NEXT:    [[TMP10:%.*]] = sdiv exact i64 [[TMP9]], ptrtoint (%struct.S* getelementptr ([[STRUCT_S]], %struct.S* null, i32 1) to i64)
 // CHECK1-NEXT:    [[TMP11:%.*]] = bitcast [2 x %struct.S]* [[VAR34]] to %struct.S*
 // CHECK1-NEXT:    [[TMP12:%.*]] = getelementptr [[STRUCT_S]], %struct.S* [[TMP11]], i64 [[TMP10]]
 // CHECK1-NEXT:    [[TMP13:%.*]] = bitcast %struct.S* [[TMP12]] to [4 x %struct.S]*
 // CHECK1-NEXT:    store [4 x %struct.S]* [[TMP13]], [4 x %struct.S]** [[_TMP5]], align 8
 // CHECK1-NEXT:    [[RHS_BEGIN:%.*]] = bitcast [2 x %struct.S]* [[VAR34]] to %struct.S*
 // CHECK1-NEXT:    [[TMP14:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK1-NEXT:    [[TMP15:%.*]] = load i32, i32* [[TMP14]], align 4
 // CHECK1-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP15]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)

 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
 // CHECK1-NEXT:    [[TMP2:%.*]] = load [4 x %struct.S]*, [4 x %struct.S]** [[_TMP1]], align 8
 // CHECK1-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [4 x %struct.S], [4 x %struct.S]* [[TMP2]], i64 0, i64 2
 // CHECK1-NEXT:    [[TMP3:%.*]] = load [4 x %struct.S]*, [4 x %struct.S]** [[_TMP1]], align 8
 // CHECK1-NEXT:    [[ARRAYIDX3:%.*]] = getelementptr inbounds [4 x %struct.S], [4 x %struct.S]* [[TMP3]], i64 0, i64 3
 // CHECK1-NEXT:    [[TMP4:%.*]] = ptrtoint %struct.S* [[ARRAYIDX3]] to i64
 // CHECK1-NEXT:    [[TMP5:%.*]] = ptrtoint %struct.S* [[ARRAYIDX]] to i64
 // CHECK1-NEXT:    [[TMP6:%.*]] = sub i64 [[TMP4]], [[TMP5]]
-// CHECK1-NEXT:    [[TMP7:%.*]] = sdiv exact i64 [[TMP6]], ptrtoint (float* getelementptr (float, float* null, i32 1) to i64)
+// CHECK1-NEXT:    [[TMP7:%.*]] = sdiv exact i64 [[TMP6]], ptrtoint (%struct.S* getelementptr ([[STRUCT_S:%.*]], %struct.S* null, i32 1) to i64)
 // CHECK1-NEXT:    [[TMP8:%.*]] = add nuw i64 [[TMP7]], 1
-// CHECK1-NEXT:    [[TMP9:%.*]] = mul nuw i64 [[TMP8]], ptrtoint (float* getelementptr (float, float* null, i32 1) to i64)
+// CHECK1-NEXT:    [[TMP9:%.*]] = mul nuw i64 [[TMP8]], ptrtoint (%struct.S* getelementptr ([[STRUCT_S]], %struct.S* null, i32 1) to i64)
 // CHECK1-NEXT:    [[TMP10:%.*]] = call i8* @llvm.stacksave()
 // CHECK1-NEXT:    store i8* [[TMP10]], i8** [[SAVED_STACK]], align 8
-// CHECK1-NEXT:    [[VLA:%.*]] = alloca [[STRUCT_S:%.*]], i64 [[TMP8]], align 16
+// CHECK1-NEXT:    [[VLA:%.*]] = alloca [[STRUCT_S]], i64 [[TMP8]], align 16
 // CHECK1-NEXT:    store i64 [[TMP8]], i64* [[__VLA_EXPR0]], align 8
 // CHECK1-NEXT:    [[TMP11:%.*]] = getelementptr [[STRUCT_S]], %struct.S* [[VLA]], i64 [[TMP8]]
 // CHECK1-NEXT:    [[OMP_ARRAYINIT_ISEMPTY:%.*]] = icmp eq %struct.S* [[VLA]], [[TMP11]]
 // CHECK1-NEXT:    br i1 [[OMP_ARRAYINIT_ISEMPTY]], label [[OMP_ARRAYINIT_DONE:%.*]], label [[OMP_ARRAYINIT_BODY:%.*]]
 // CHECK1:       omp.arrayinit.body:
 // CHECK1-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi %struct.S* [ [[VLA]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYINIT_BODY]] ]
 // CHECK1-NEXT:    call void @_ZN1SIfEC1Ev(%struct.S* nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_DESTELEMENTPAST]])
 // CHECK1-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr [[STRUCT_S]], %struct.S* [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
 // CHECK1-NEXT:    [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq %struct.S* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP11]]
 // CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYINIT_DONE]], label [[OMP_ARRAYINIT_BODY]]
 // CHECK1:       omp.arrayinit.done:
 // CHECK1-NEXT:    [[TMP12:%.*]] = load [4 x %struct.S]*, [4 x %struct.S]** [[_TMP1]], align 8
 // CHECK1-NEXT:    [[TMP13:%.*]] = bitcast [4 x %struct.S]* [[TMP12]] to %struct.S*
 // CHECK1-NEXT:    [[TMP14:%.*]] = ptrtoint %struct.S* [[TMP13]] to i64
 // CHECK1-NEXT:    [[TMP15:%.*]] = ptrtoint %struct.S* [[ARRAYIDX]] to i64
 // CHECK1-NEXT:    [[TMP16:%.*]] = sub i64 [[TMP14]], [[TMP15]]
-// CHECK1-NEXT:    [[TMP17:%.*]] = sdiv exact i64 [[TMP16]], ptrtoint (float* getelementptr (float, float* null, i32 1) to i64)
+// CHECK1-NEXT:    [[TMP17:%.*]] = sdiv exact i64 [[TMP16]], ptrtoint (%struct.S* getelementptr ([[STRUCT_S]], %struct.S* null, i32 1) to i64)
 // CHECK1-NEXT:    [[TMP18:%.*]] = getelementptr [[STRUCT_S]], %struct.S* [[VLA]], i64 [[TMP17]]
 // CHECK1-NEXT:    [[TMP19:%.*]] = bitcast %struct.S* [[TMP18]] to [4 x %struct.S]*
 // CHECK1-NEXT:    store [4 x %struct.S]* [[TMP19]], [4 x %struct.S]** [[_TMP4]], align 8
 // CHECK1-NEXT:    [[TMP20:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK1-NEXT:    [[TMP21:%.*]] = load i32, i32* [[TMP20]], align 4
 // CHECK1-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP21]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
 // CHECK1-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP22]], 9

 // CHECK1-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr [[STRUCT_S_0]], %struct.S.0* [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
 // CHECK1-NEXT:    [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq %struct.S.0* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP6]]
 // CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYINIT_DONE]], label [[OMP_ARRAYINIT_BODY]]
 // CHECK1:       omp.arrayinit.done:
 // CHECK1-NEXT:    [[TMP7:%.*]] = bitcast [42 x %struct.S.0]* [[TMP0]] to %struct.S.0*
 // CHECK1-NEXT:    [[TMP8:%.*]] = ptrtoint %struct.S.0* [[TMP7]] to i64
 // CHECK1-NEXT:    [[TMP9:%.*]] = ptrtoint %struct.S.0* [[ARRAYIDX]] to i64
 // CHECK1-NEXT:    [[TMP10:%.*]] = sub i64 [[TMP8]], [[TMP9]]
-// CHECK1-NEXT:    [[TMP11:%.*]] = sdiv exact i64 [[TMP10]], ptrtoint (i32* getelementptr (i32, i32* null, i32 1) to i64)
+// CHECK1-NEXT:    [[TMP11:%.*]] = sdiv exact i64 [[TMP10]], ptrtoint (%struct.S.0* getelementptr ([[STRUCT_S_0]], %struct.S.0* null, i32 1) to i64)
 // CHECK1-NEXT:    [[TMP12:%.*]] = bitcast [40 x %struct.S.0]* [[ARR4]] to %struct.S.0*
 // CHECK1-NEXT:    [[TMP13:%.*]] = getelementptr [[STRUCT_S_0]], %struct.S.0* [[TMP12]], i64 [[TMP11]]
 // CHECK1-NEXT:    [[TMP14:%.*]] = bitcast %struct.S.0* [[TMP13]] to [42 x %struct.S.0]*
 // CHECK1-NEXT:    [[RHS_BEGIN:%.*]] = bitcast [40 x %struct.S.0]* [[ARR4]] to %struct.S.0*
 // CHECK1-NEXT:    [[TMP15:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK1-NEXT:    [[TMP16:%.*]] = load i32, i32* [[TMP15]], align 4
 // CHECK1-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP16]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
 // CHECK1-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4

 // CHECK2-NEXT:    [[TMP23:%.*]] = sext i32 [[TMP22]] to i64
 // CHECK2-NEXT:    [[LB_ADD_LEN11:%.*]] = add nsw i64 0, [[TMP23]]
 // CHECK2-NEXT:    [[ARRAYIDX12:%.*]] = getelementptr inbounds [10 x [4 x %struct.S]], [10 x [4 x %struct.S]]* [[TMP4]], i64 0, i64 [[LB_ADD_LEN11]]
 // CHECK2-NEXT:    [[ARRAYDECAY13:%.*]] = getelementptr inbounds [4 x %struct.S], [4 x %struct.S]* [[ARRAYIDX12]], i64 0, i64 0
 // CHECK2-NEXT:    [[ARRAYIDX14:%.*]] = getelementptr inbounds [[STRUCT_S]], %struct.S* [[ARRAYDECAY13]], i64 2
 // CHECK2-NEXT:    [[TMP24:%.*]] = ptrtoint %struct.S* [[ARRAYIDX14]] to i64
 // CHECK2-NEXT:    [[TMP25:%.*]] = ptrtoint %struct.S* [[ARRAYIDX9]] to i64
 // CHECK2-NEXT:    [[TMP26:%.*]] = sub i64 [[TMP24]], [[TMP25]]
-// CHECK2-NEXT:    [[TMP27:%.*]] = sdiv exact i64 [[TMP26]], ptrtoint (float* getelementptr (float, float* null, i32 1) to i64)
+// CHECK2-NEXT:    [[TMP27:%.*]] = sdiv exact i64 [[TMP26]], ptrtoint (%struct.S* getelementptr ([[STRUCT_S]], %struct.S* null, i32 1) to i64)
 // CHECK2-NEXT:    [[TMP28:%.*]] = add nuw i64 [[TMP27]], 1
-// CHECK2-NEXT:    [[TMP29:%.*]] = mul nuw i64 [[TMP28]], ptrtoint (float* getelementptr (float, float* null, i32 1) to i64)
+// CHECK2-NEXT:    [[TMP29:%.*]] = mul nuw i64 [[TMP28]], ptrtoint (%struct.S* getelementptr ([[STRUCT_S]], %struct.S* null, i32 1) to i64)
 // CHECK2-NEXT:    [[VLA15:%.*]] = alloca [[STRUCT_S]], i64 [[TMP28]], align 16
 // CHECK2-NEXT:    store i64 [[TMP28]], i64* [[__VLA_EXPR1]], align 8
 // CHECK2-NEXT:    [[TMP30:%.*]] = getelementptr [[STRUCT_S]], %struct.S* [[VLA15]], i64 [[TMP28]]
 // CHECK2-NEXT:    [[OMP_ARRAYINIT_ISEMPTY16:%.*]] = icmp eq %struct.S* [[VLA15]], [[TMP30]]
 // CHECK2-NEXT:    br i1 [[OMP_ARRAYINIT_ISEMPTY16]], label [[OMP_ARRAYINIT_DONE21:%.*]], label [[OMP_ARRAYINIT_BODY17:%.*]]
 // CHECK2:       omp.arrayinit.body17:
 // CHECK2-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST18:%.*]] = phi %struct.S* [ [[VLA15]], [[OMP_ARRAYINIT_DONE]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT19:%.*]], [[OMP_ARRAYINIT_BODY17]] ]
 // CHECK2-NEXT:    call void @_ZN1SIfEC1Ev(%struct.S* nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_DESTELEMENTPAST18]])
 // CHECK2-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT19]] = getelementptr [[STRUCT_S]], %struct.S* [[OMP_ARRAYCPY_DESTELEMENTPAST18]], i32 1
 // CHECK2-NEXT:    [[OMP_ARRAYCPY_DONE20:%.*]] = icmp eq %struct.S* [[OMP_ARRAYCPY_DEST_ELEMENT19]], [[TMP30]]
 // CHECK2-NEXT:    br i1 [[OMP_ARRAYCPY_DONE20]], label [[OMP_ARRAYINIT_DONE21]], label [[OMP_ARRAYINIT_BODY17]]
 // CHECK2:       omp.arrayinit.done21:
 // CHECK2-NEXT:    [[TMP31:%.*]] = bitcast [10 x [4 x %struct.S]]* [[TMP4]] to %struct.S*
 // CHECK2-NEXT:    [[TMP32:%.*]] = ptrtoint %struct.S* [[TMP31]] to i64
 // CHECK2-NEXT:    [[TMP33:%.*]] = ptrtoint %struct.S* [[ARRAYIDX9]] to i64
 // CHECK2-NEXT:    [[TMP34:%.*]] = sub i64 [[TMP32]], [[TMP33]]
-// CHECK2-NEXT:    [[TMP35:%.*]] = sdiv exact i64 [[TMP34]], ptrtoint (float* getelementptr (float, float* null, i32 1) to i64)
+// CHECK2-NEXT:    [[TMP35:%.*]] = sdiv exact i64 [[TMP34]], ptrtoint (%struct.S* getelementptr ([[STRUCT_S]], %struct.S* null, i32 1) to i64)
 // CHECK2-NEXT:    [[TMP36:%.*]] = getelementptr [[STRUCT_S]], %struct.S* [[VLA15]], i64 [[TMP35]]
 // CHECK2-NEXT:    [[TMP37:%.*]] = bitcast %struct.S* [[TMP36]] to [10 x [4 x %struct.S]]*
 // CHECK2-NEXT:    [[TMP38:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK2-NEXT:    [[TMP39:%.*]] = load i32, i32* [[TMP38]], align 4
 // CHECK2-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP39]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
 // CHECK2-NEXT:    [[TMP40:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK2-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP40]], 9
 // CHECK2-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]

 // CHECK2-NEXT:    [[ARRAYIDX1:%.*]] = getelementptr inbounds [[STRUCT_S:%.*]], %struct.S* [[TMP2]], i64 1
 // CHECK2-NEXT:    [[TMP3:%.*]] = load %struct.S**, %struct.S*** [[TMP0]], align 8
 // CHECK2-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds %struct.S*, %struct.S** [[TMP3]], i64 4
 // CHECK2-NEXT:    [[TMP4:%.*]] = load %struct.S*, %struct.S** [[ARRAYIDX2]], align 8
 // CHECK2-NEXT:    [[ARRAYIDX3:%.*]] = getelementptr inbounds [[STRUCT_S]], %struct.S* [[TMP4]], i64 6
 // CHECK2-NEXT:    [[TMP5:%.*]] = ptrtoint %struct.S* [[ARRAYIDX3]] to i64
 // CHECK2-NEXT:    [[TMP6:%.*]] = ptrtoint %struct.S* [[ARRAYIDX1]] to i64
 // CHECK2-NEXT:    [[TMP7:%.*]] = sub i64 [[TMP5]], [[TMP6]]
-// CHECK2-NEXT:    [[TMP8:%.*]] = sdiv exact i64 [[TMP7]], ptrtoint (float* getelementptr (float, float* null, i32 1) to i64)
+// CHECK2-NEXT:    [[TMP8:%.*]] = sdiv exact i64 [[TMP7]], ptrtoint (%struct.S* getelementptr ([[STRUCT_S]], %struct.S* null, i32 1) to i64)
 // CHECK2-NEXT:    [[TMP9:%.*]] = add nuw i64 [[TMP8]], 1
-// CHECK2-NEXT:    [[TMP10:%.*]] = mul nuw i64 [[TMP9]], ptrtoint (float* getelementptr (float, float* null, i32 1) to i64)
+// CHECK2-NEXT:    [[TMP10:%.*]] = mul nuw i64 [[TMP9]], ptrtoint (%struct.S* getelementptr ([[STRUCT_S]], %struct.S* null, i32 1) to i64)
 // CHECK2-NEXT:    [[TMP11:%.*]] = call i8* @llvm.stacksave()
 // CHECK2-NEXT:    store i8* [[TMP11]], i8** [[SAVED_STACK]], align 8
 // CHECK2-NEXT:    [[VLA:%.*]] = alloca [[STRUCT_S]], i64 [[TMP9]], align 16
 // CHECK2-NEXT:    store i64 [[TMP9]], i64* [[__VLA_EXPR0]], align 8
 // CHECK2-NEXT:    [[TMP12:%.*]] = getelementptr [[STRUCT_S]], %struct.S* [[VLA]], i64 [[TMP9]]
 // CHECK2-NEXT:    [[OMP_ARRAYINIT_ISEMPTY:%.*]] = icmp eq %struct.S* [[VLA]], [[TMP12]]
 // CHECK2-NEXT:    br i1 [[OMP_ARRAYINIT_ISEMPTY]], label [[OMP_ARRAYINIT_DONE:%.*]], label [[OMP_ARRAYINIT_BODY:%.*]]
 // CHECK2:       omp.arrayinit.body:
 // CHECK2-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi %struct.S* [ [[VLA]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYINIT_BODY]] ]
 // CHECK2-NEXT:    call void @_ZN1SIfEC1Ev(%struct.S* nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_DESTELEMENTPAST]])
 // CHECK2-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr [[STRUCT_S]], %struct.S* [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
 // CHECK2-NEXT:    [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq %struct.S* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP12]]
 // CHECK2-NEXT:    br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYINIT_DONE]], label [[OMP_ARRAYINIT_BODY]]
 // CHECK2:       omp.arrayinit.done:
 // CHECK2-NEXT:    [[TMP13:%.*]] = load %struct.S**, %struct.S*** [[TMP0]], align 8
 // CHECK2-NEXT:    [[TMP14:%.*]] = load %struct.S*, %struct.S** [[TMP13]], align 8
 // CHECK2-NEXT:    [[TMP15:%.*]] = ptrtoint %struct.S* [[TMP14]] to i64
 // CHECK2-NEXT:    [[TMP16:%.*]] = ptrtoint %struct.S* [[ARRAYIDX1]] to i64
 // CHECK2-NEXT:    [[TMP17:%.*]] = sub i64 [[TMP15]], [[TMP16]]
-// CHECK2-NEXT:    [[TMP18:%.*]] = sdiv exact i64 [[TMP17]], ptrtoint (float* getelementptr (float, float* null, i32 1) to i64)
+// CHECK2-NEXT:    [[TMP18:%.*]] = sdiv exact i64 [[TMP17]], ptrtoint (%struct.S* getelementptr ([[STRUCT_S]], %struct.S* null, i32 1) to i64)
 // CHECK2-NEXT:    [[TMP19:%.*]] = getelementptr [[STRUCT_S]], %struct.S* [[VLA]], i64 [[TMP18]]
 // CHECK2-NEXT:    store %struct.S** [[_TMP5]], %struct.S*** [[_TMP4]], align 8
 // CHECK2-NEXT:    store %struct.S* [[TMP19]], %struct.S** [[_TMP5]], align 8
 // CHECK2-NEXT:    [[TMP20:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK2-NEXT:    [[TMP21:%.*]] = load i32, i32* [[TMP20]], align 4
 // CHECK2-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP21]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
 // CHECK2-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK2-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP22]], 9

 // CHECK2-NEXT:    [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq %struct.S* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP5]]
 // CHECK2-NEXT:    br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYINIT_DONE]], label [[OMP_ARRAYINIT_BODY]]
 // CHECK2:       omp.arrayinit.done:
 // CHECK2-NEXT:    [[TMP6:%.*]] = load %struct.S**, %struct.S*** [[TMP0]], align 8
 // CHECK2-NEXT:    [[TMP7:%.*]] = load %struct.S*, %struct.S** [[TMP6]], align 8
 // CHECK2-NEXT:    [[TMP8:%.*]] = ptrtoint %struct.S* [[TMP7]] to i64
 // CHECK2-NEXT:    [[TMP9:%.*]] = ptrtoint %struct.S* [[ARRAYIDX1]] to i64
 // CHECK2-NEXT:    [[TMP10:%.*]] = sub i64 [[TMP8]], [[TMP9]]
-// CHECK2-NEXT:    [[TMP11:%.*]] = sdiv exact i64 [[TMP10]], ptrtoint (float* getelementptr (float, float* null, i32 1) to i64)
+// CHECK2-NEXT:    [[TMP11:%.*]] = sdiv exact i64 [[TMP10]], ptrtoint (%struct.S* getelementptr ([[STRUCT_S]], %struct.S* null, i32 1) to i64)
 // CHECK2-NEXT:    [[TMP12:%.*]] = bitcast [1 x [6 x %struct.S]]* [[VAR24]] to %struct.S*
 // CHECK2-NEXT:    [[TMP13:%.*]] = getelementptr [[STRUCT_S]], %struct.S* [[TMP12]], i64 [[TMP11]]
 // CHECK2-NEXT:    store %struct.S** [[_TMP6]], %struct.S*** [[_TMP5]], align 8
 // CHECK2-NEXT:    store %struct.S* [[TMP13]], %struct.S** [[_TMP6]], align 8
 // CHECK2-NEXT:    [[RHS_BEGIN:%.*]] = bitcast [1 x [6 x %struct.S]]* [[VAR24]] to %struct.S*
 // CHECK2-NEXT:    [[TMP14:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK2-NEXT:    [[TMP15:%.*]] = load i32, i32* [[TMP14]], align 4
 // CHECK2-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP15]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)

 // CHECK2-NEXT:    [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq %struct.S* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP5]]
 // CHECK2-NEXT:    br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYINIT_DONE]], label [[OMP_ARRAYINIT_BODY]]
 // CHECK2:       omp.arrayinit.done:
 // CHECK2-NEXT:    [[TMP6:%.*]] = load %struct.S**, %struct.S*** [[TMP0]], align 8
 // CHECK2-NEXT:    [[TMP7:%.*]] = load %struct.S*, %struct.S** [[TMP6]], align 8
 // CHECK2-NEXT:    [[TMP8:%.*]] = ptrtoint %struct.S* [[TMP7]] to i64
 // CHECK2-NEXT:    [[TMP9:%.*]] = ptrtoint %struct.S* [[ARRAYIDX1]] to i64
 // CHECK2-NEXT:    [[TMP10:%.*]] = sub i64 [[TMP8]], [[TMP9]]
-// CHECK2-NEXT:    [[TMP11:%.*]] = sdiv exact i64 [[TMP10]], ptrtoint (float* getelementptr (float, float* null, i32 1) to i64)
+// CHECK2-NEXT:    [[TMP11:%.*]] = sdiv exact i64 [[TMP10]], ptrtoint (%struct.S* getelementptr ([[STRUCT_S]], %struct.S* null, i32 1) to i64)
 // CHECK2-NEXT:    [[TMP12:%.*]] = bitcast [1 x [6 x %struct.S]]* [[VAR24]] to %struct.S*
 // CHECK2-NEXT:    [[TMP13:%.*]] = getelementptr [[STRUCT_S]], %struct.S* [[TMP12]], i64 [[TMP11]]
 // CHECK2-NEXT:    store %struct.S** [[_TMP6]], %struct.S*** [[_TMP5]], align 8
 // CHECK2-NEXT:    store %struct.S* [[TMP13]], %struct.S** [[_TMP6]], align 8
 // CHECK2-NEXT:    [[RHS_BEGIN:%.*]] = bitcast [1 x [6 x %struct.S]]* [[VAR24]] to %struct.S*
 // CHECK2-NEXT:    [[TMP14:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK2-NEXT:    [[TMP15:%.*]] = load i32, i32* [[TMP14]], align 4
 // CHECK2-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP15]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)

 // CHECK2-NEXT:    [[TMP4:%.*]] = load %struct.S*, %struct.S** [[ARRAYIDX2]], align 8
 // CHECK2-NEXT:    [[ARRAYIDX3:%.*]] = getelementptr inbounds [[STRUCT_S]], %struct.S* [[TMP4]], i64 1
 // CHECK2-NEXT:    call void @_ZN1SIfEC1Ev(%struct.S* nonnull align 4 dereferenceable(4) [[VAR24]])
 // CHECK2-NEXT:    [[TMP5:%.*]] = load %struct.S**, %struct.S*** [[TMP0]], align 8
 // CHECK2-NEXT:    [[TMP6:%.*]] = load %struct.S*, %struct.S** [[TMP5]], align 8
 // CHECK2-NEXT:    [[TMP7:%.*]] = ptrtoint %struct.S* [[TMP6]] to i64
 // CHECK2-NEXT:    [[TMP8:%.*]] = ptrtoint %struct.S* [[ARRAYIDX1]] to i64
 // CHECK2-NEXT:    [[TMP9:%.*]] = sub i64 [[TMP7]], [[TMP8]]
-// CHECK2-NEXT:    [[TMP10:%.*]] = sdiv exact i64 [[TMP9]], ptrtoint (float* getelementptr (float, float* null, i32 1) to i64)
+// CHECK2-NEXT:    [[TMP10:%.*]] = sdiv exact i64 [[TMP9]], ptrtoint (%struct.S* getelementptr ([[STRUCT_S]], %struct.S* null, i32 1) to i64)
 // CHECK2-NEXT:    [[TMP11:%.*]] = getelementptr [[STRUCT_S]], %struct.S* [[VAR24]], i64 [[TMP10]]
 // CHECK2-NEXT:    store %struct.S** [[_TMP6]], %struct.S*** [[_TMP5]], align 8
 // CHECK2-NEXT:    store %struct.S* [[TMP11]], %struct.S** [[_TMP6]], align 8
 // CHECK2-NEXT:    [[TMP12:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK2-NEXT:    [[TMP13:%.*]] = load i32, i32* [[TMP12]], align 4
 // CHECK2-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP13]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
 // CHECK2-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK2-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP14]], 9

 // CHECK2-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr [[STRUCT_S]], %struct.S* [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
 // CHECK2-NEXT:    [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq %struct.S* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP1]]
 // CHECK2-NEXT:    br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYINIT_DONE]], label [[OMP_ARRAYINIT_BODY]]
 // CHECK2:       omp.arrayinit.done:
 // CHECK2-NEXT:    [[TMP2:%.*]] = bitcast [5 x %struct.S]* [[TMP0]] to %struct.S*
 // CHECK2-NEXT:    [[TMP3:%.*]] = ptrtoint %struct.S* [[TMP2]] to i64
 // CHECK2-NEXT:    [[TMP4:%.*]] = ptrtoint %struct.S* [[ARRAYIDX]] to i64
 // CHECK2-NEXT:    [[TMP5:%.*]] = sub i64 [[TMP3]], [[TMP4]]
-// CHECK2-NEXT:    [[TMP6:%.*]] = sdiv exact i64 [[TMP5]], ptrtoint (float* getelementptr (float, float* null, i32 1) to i64)
+// CHECK2-NEXT:    [[TMP6:%.*]] = sdiv exact i64 [[TMP5]], ptrtoint (%struct.S* getelementptr ([[STRUCT_S]], %struct.S* null, i32 1) to i64)
 // CHECK2-NEXT:    [[TMP7:%.*]] = bitcast [5 x %struct.S]* [[VVAR22]] to %struct.S*
 // CHECK2-NEXT:    [[TMP8:%.*]] = getelementptr [[STRUCT_S]], %struct.S* [[TMP7]], i64 [[TMP6]]
 // CHECK2-NEXT:    [[TMP9:%.*]] = bitcast %struct.S* [[TMP8]] to [5 x %struct.S]*
 // CHECK2-NEXT:    [[RHS_BEGIN:%.*]] = bitcast [5 x %struct.S]* [[VVAR22]] to %struct.S*
 // CHECK2-NEXT:    [[TMP10:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK2-NEXT:    [[TMP11:%.*]] = load i32, i32* [[TMP10]], align 4
 // CHECK2-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP11]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
 // CHECK2-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4

 // CHECK2-NEXT:    [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq %struct.S* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP4]]
 // CHECK2-NEXT:    br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYINIT_DONE]], label [[OMP_ARRAYINIT_BODY]]
 // CHECK2:       omp.arrayinit.done:
 // CHECK2-NEXT:    [[TMP5:%.*]] = load [4 x %struct.S]*, [4 x %struct.S]** [[_TMP1]], align 8
 // CHECK2-NEXT:    [[TMP6:%.*]] = bitcast [4 x %struct.S]* [[TMP5]] to %struct.S*
 // CHECK2-NEXT:    [[TMP7:%.*]] = ptrtoint %struct.S* [[TMP6]] to i64
 // CHECK2-NEXT:    [[TMP8:%.*]] = ptrtoint %struct.S* [[ARRAYIDX]] to i64
 // CHECK2-NEXT:    [[TMP9:%.*]] = sub i64 [[TMP7]], [[TMP8]]
-// CHECK2-NEXT:    [[TMP10:%.*]] = sdiv exact i64 [[TMP9]], ptrtoint (float* getelementptr (float, float* null, i32 1) to i64)
+// CHECK2-NEXT:    [[TMP10:%.*]] = sdiv exact i64 [[TMP9]], ptrtoint (%struct.S* getelementptr ([[STRUCT_S]], %struct.S* null, i32 1) to i64)
 // CHECK2-NEXT:    [[TMP11:%.*]] = bitcast [2 x %struct.S]* [[VAR34]] to %struct.S*
 // CHECK2-NEXT:    [[TMP12:%.*]] = getelementptr [[STRUCT_S]], %struct.S* [[TMP11]], i64 [[TMP10]]
 // CHECK2-NEXT:    [[TMP13:%.*]] = bitcast %struct.S* [[TMP12]] to [4 x %struct.S]*
 // CHECK2-NEXT:    store [4 x %struct.S]* [[TMP13]], [4 x %struct.S]** [[_TMP5]], align 8
 // CHECK2-NEXT:    [[RHS_BEGIN:%.*]] = bitcast [2 x %struct.S]* [[VAR34]] to %struct.S*
 // CHECK2-NEXT:    [[TMP14:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK2-NEXT:    [[TMP15:%.*]] = load i32, i32* [[TMP14]], align 4
 // CHECK2-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP15]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)

 // CHECK2-NEXT:    [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq %struct.S* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP4]]
 // CHECK2-NEXT:    br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYINIT_DONE]], label [[OMP_ARRAYINIT_BODY]]
 // CHECK2:       omp.arrayinit.done:
 // CHECK2-NEXT:    [[TMP5:%.*]] = load [4 x %struct.S]*, [4 x %struct.S]** [[_TMP1]], align 8
 // CHECK2-NEXT:    [[TMP6:%.*]] = bitcast [4 x %struct.S]* [[TMP5]] to %struct.S*
 // CHECK2-NEXT:    [[TMP7:%.*]] = ptrtoint %struct.S* [[TMP6]] to i64
 // CHECK2-NEXT:    [[TMP8:%.*]] = ptrtoint %struct.S* [[ARRAYIDX]] to i64
 // CHECK2-NEXT:    [[TMP9:%.*]] = sub i64 [[TMP7]], [[TMP8]]
-// CHECK2-NEXT:    [[TMP10:%.*]] = sdiv exact i64 [[TMP9]], ptrtoint (float* getelementptr (float, float* null, i32 1) to i64)
+// CHECK2-NEXT:    [[TMP10:%.*]] = sdiv exact i64 [[TMP9]], ptrtoint (%struct.S* getelementptr ([[STRUCT_S]], %struct.S* null, i32 1) to i64)
 // CHECK2-NEXT:    [[TMP11:%.*]] = bitcast [2 x %struct.S]* [[VAR34]] to %struct.S*
 // CHECK2-NEXT:    [[TMP12:%.*]] = getelementptr [[STRUCT_S]], %struct.S* [[TMP11]], i64 [[TMP10]]
 // CHECK2-NEXT:    [[TMP13:%.*]] = bitcast %struct.S* [[TMP12]] to [4 x %struct.S]*
 // CHECK2-NEXT:    store [4 x %struct.S]* [[TMP13]], [4 x %struct.S]** [[_TMP5]], align 8
 // CHECK2-NEXT:    [[RHS_BEGIN:%.*]] = bitcast [2 x %struct.S]* [[VAR34]] to %struct.S*
 // CHECK2-NEXT:    [[TMP14:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK2-NEXT:    [[TMP15:%.*]] = load i32, i32* [[TMP14]], align 4
 // CHECK2-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP15]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)

 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
 // CHECK2-NEXT:    [[TMP2:%.*]] = load [4 x %struct.S]*, [4 x %struct.S]** [[_TMP1]], align 8
 // CHECK2-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [4 x %struct.S], [4 x %struct.S]* [[TMP2]], i64 0, i64 2
 // CHECK2-NEXT:    [[TMP3:%.*]] = load [4 x %struct.S]*, [4 x %struct.S]** [[_TMP1]], align 8
 // CHECK2-NEXT:    [[ARRAYIDX3:%.*]] = getelementptr inbounds [4 x %struct.S], [4 x %struct.S]* [[TMP3]], i64 0, i64 3
 // CHECK2-NEXT:    [[TMP4:%.*]] = ptrtoint %struct.S* [[ARRAYIDX3]] to i64
 // CHECK2-NEXT:    [[TMP5:%.*]] = ptrtoint %struct.S* [[ARRAYIDX]] to i64
 // CHECK2-NEXT:    [[TMP6:%.*]] = sub i64 [[TMP4]], [[TMP5]]
-// CHECK2-NEXT:    [[TMP7:%.*]] = sdiv exact i64 [[TMP6]], ptrtoint (float* getelementptr (float, float* null, i32 1) to i64)
+// CHECK2-NEXT:    [[TMP7:%.*]] = sdiv exact i64 [[TMP6]], ptrtoint (%struct.S* getelementptr ([[STRUCT_S:%.*]], %struct.S* null, i32 1) to i64)
 // CHECK2-NEXT:    [[TMP8:%.*]] = add nuw i64 [[TMP7]], 1
-// CHECK2-NEXT:    [[TMP9:%.*]] = mul nuw i64 [[TMP8]], ptrtoint (float* getelementptr (float, float* null, i32 1) to i64)
+// CHECK2-NEXT:    [[TMP9:%.*]] = mul nuw i64 [[TMP8]], ptrtoint (%struct.S* getelementptr ([[STRUCT_S]], %struct.S* null, i32 1) to i64)
 // CHECK2-NEXT:    [[TMP10:%.*]] = call i8* @llvm.stacksave()
 // CHECK2-NEXT:    store i8* [[TMP10]], i8** [[SAVED_STACK]], align 8
-// CHECK2-NEXT:    [[VLA:%.*]] = alloca [[STRUCT_S:%.*]], i64 [[TMP8]], align 16
+// CHECK2-NEXT:    [[VLA:%.*]] = alloca [[STRUCT_S]], i64 [[TMP8]], align 16
 // CHECK2-NEXT:    store i64 [[TMP8]], i64* [[__VLA_EXPR0]], align 8
 // CHECK2-NEXT:    [[TMP11:%.*]] = getelementptr [[STRUCT_S]], %struct.S* [[VLA]], i64 [[TMP8]]
 // CHECK2-NEXT:    [[OMP_ARRAYINIT_ISEMPTY:%.*]] = icmp eq %struct.S* [[VLA]], [[TMP11]]
 // CHECK2-NEXT:    br i1 [[OMP_ARRAYINIT_ISEMPTY]], label [[OMP_ARRAYINIT_DONE:%.*]], label [[OMP_ARRAYINIT_BODY:%.*]]
 // CHECK2:       omp.arrayinit.body:
 // CHECK2-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi %struct.S* [ [[VLA]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYINIT_BODY]] ]
 // CHECK2-NEXT:    call void @_ZN1SIfEC1Ev(%struct.S* nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_DESTELEMENTPAST]])
 // CHECK2-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr [[STRUCT_S]], %struct.S* [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
 // CHECK2-NEXT:    [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq %struct.S* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP11]]
 // CHECK2-NEXT:    br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYINIT_DONE]], label [[OMP_ARRAYINIT_BODY]]
 // CHECK2:       omp.arrayinit.done:
 // CHECK2-NEXT:    [[TMP12:%.*]] = load [4 x %struct.S]*, [4 x %struct.S]** [[_TMP1]], align 8
 // CHECK2-NEXT:    [[TMP13:%.*]] = bitcast [4 x %struct.S]* [[TMP12]] to %struct.S*
 // CHECK2-NEXT:    [[TMP14:%.*]] = ptrtoint %struct.S* [[TMP13]] to i64
 // CHECK2-NEXT:    [[TMP15:%.*]] = ptrtoint %struct.S* [[ARRAYIDX]] to i64
 // CHECK2-NEXT:    [[TMP16:%.*]] = sub i64 [[TMP14]], [[TMP15]]
-// CHECK2-NEXT:    [[TMP17:%.*]] = sdiv exact i64 [[TMP16]], ptrtoint (float* getelementptr (float, float* null, i32 1) to i64)
+// CHECK2-NEXT:    [[TMP17:%.*]] = sdiv exact i64 [[TMP16]], ptrtoint (%struct.S* getelementptr ([[STRUCT_S]], %struct.S* null, i32 1) to i64)
 // CHECK2-NEXT:    [[TMP18:%.*]] = getelementptr [[STRUCT_S]], %struct.S* [[VLA]], i64 [[TMP17]]
 // CHECK2-NEXT:    [[TMP19:%.*]] = bitcast %struct.S* [[TMP18]] to [4 x %struct.S]*
 // CHECK2-NEXT:    store [4 x %struct.S]* [[TMP19]], [4 x %struct.S]** [[_TMP4]], align 8
 // CHECK2-NEXT:    [[TMP20:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK2-NEXT:    [[TMP21:%.*]] = load i32, i32* [[TMP20]], align 4
 // CHECK2-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP21]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
 // CHECK2-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
 // CHECK2-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP22]], 9

 // CHECK2-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr [[STRUCT_S_0]], %struct.S.0* [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
 // CHECK2-NEXT:    [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq %struct.S.0* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP6]]
 // CHECK2-NEXT:    br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYINIT_DONE]], label [[OMP_ARRAYINIT_BODY]]
 // CHECK2:       omp.arrayinit.done:
 // CHECK2-NEXT:    [[TMP7:%.*]] = bitcast [42 x %struct.S.0]* [[TMP0]] to %struct.S.0*
 // CHECK2-NEXT:    [[TMP8:%.*]] = ptrtoint %struct.S.0* [[TMP7]] to i64
 // CHECK2-NEXT:    [[TMP9:%.*]] = ptrtoint %struct.S.0* [[ARRAYIDX]] to i64
 // CHECK2-NEXT:    [[TMP10:%.*]] = sub i64 [[TMP8]], [[TMP9]]
-// CHECK2-NEXT:    [[TMP11:%.*]] = sdiv exact i64 [[TMP10]], ptrtoint (i32* getelementptr (i32, i32* null, i32 1) to i64)
+// CHECK2-NEXT:    [[TMP11:%.*]] = sdiv exact i64 [[TMP10]], ptrtoint (%struct.S.0* getelementptr ([[STRUCT_S_0]], %struct.S.0* null, i32 1) to i64)
 // CHECK2-NEXT:    [[TMP12:%.*]] = bitcast [40 x %struct.S.0]* [[ARR4]] to %struct.S.0*
 // CHECK2-NEXT:    [[TMP13:%.*]] = getelementptr [[STRUCT_S_0]], %struct.S.0* [[TMP12]], i64 [[TMP11]]
 // CHECK2-NEXT:    [[TMP14:%.*]] = bitcast %struct.S.0* [[TMP13]] to [42 x %struct.S.0]*
 // CHECK2-NEXT:    [[RHS_BEGIN:%.*]] = bitcast [40 x %struct.S.0]* [[ARR4]] to %struct.S.0*
 // CHECK2-NEXT:    [[TMP15:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK2-NEXT:    [[TMP16:%.*]] = load i32, i32* [[TMP15]], align 4
 // CHECK2-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP16]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
 // CHECK2-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4

 // CHECK4-NEXT:    [[ADD:%.*]] = fadd double [[TMP18]], [[TMP19]]
 // CHECK4-NEXT:    store double [[ADD]], double* [[TMP11]], align 8
 // CHECK4-NEXT:    [[TMP20:%.*]] = load double, double* [[TMP17]], align 8
 // CHECK4-NEXT:    [[TMP21:%.*]] = load double, double* [[TMP14]], align 8
 // CHECK4-NEXT:    [[ADD2:%.*]] = fadd double [[TMP20]], [[TMP21]]
 // CHECK4-NEXT:    store double [[ADD2]], double* [[TMP17]], align 8
 // CHECK4-NEXT:    ret void
 //
-//
 

clang/test/OpenMP/master_taskloop_reduction_codegen.cpp

 // CHECK-DAG:    store i8* [[TMP31]], i8** [[TMP28:%[^,]+]],
 // CHECK-DAG:    [[TMP28]] = getelementptr inbounds %struct.kmp_taskred_input_t, %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_4:%.+]], i32 0, i32 0
 // CHECK-DAG:    [[TMP31:%.*]] = bitcast %struct.S* [[ARRAYIDX5]] to i8*
 // CHECK-DAG:    store i8* [[TMP31]], i8** [[TMP28:%[^,]+]],
 // CHECK-DAG:    [[TMP28]] = getelementptr inbounds %struct.kmp_taskred_input_t, %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_4]], i32 0, i32 1
 // CHECK-DAG:    [[TMP32:%.*]] = ptrtoint %struct.S* [[ARRAYIDX6]] to i64
 // CHECK-DAG:    [[TMP33:%.*]] = ptrtoint %struct.S* [[ARRAYIDX5]] to i64
 // CHECK-DAG:    [[TMP34:%.*]] = sub i64 [[TMP32]], [[TMP33]]
-// CHECK-DAG:    [[TMP35:%.*]] = sdiv exact i64 [[TMP34]], ptrtoint (float* getelementptr (float, float* null, i32 1) to i64)
+// CHECK-DAG:    [[TMP35:%.*]] = sdiv exact i64 [[TMP34]], ptrtoint (%struct.S* getelementptr (%struct.S, %struct.S* null, i32 1) to i64)
 // CHECK-DAG:    [[TMP36:%.*]] = add nuw i64 [[TMP35]], 1
-// CHECK-DAG:    [[TMP37:%.*]] = mul nuw i64 [[TMP36]], ptrtoint (float* getelementptr (float, float* null, i32 1) to i64)
+// CHECK-DAG:    [[TMP37:%.*]] = mul nuw i64 [[TMP36]], ptrtoint (%struct.S* getelementptr (%struct.S, %struct.S* null, i32 1) to i64)
 // CHECK-DAG:    store i64 [[TMP37]], i64* [[TMP38:%[^,]+]],
 // CHECK-DAG:    [[TMP38]] = getelementptr inbounds %struct.kmp_taskred_input_t, %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_4]], i32 0, i32 2
 // CHECK-DAG:    [[TMP39:%.*]] = getelementptr inbounds %struct.kmp_taskred_input_t, %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_4]], i32 0, i32 3
 // CHECK-DAG:    store i8* bitcast (void (i8*, i8*)* @[[RED_INIT2:.+]] to i8*), i8** [[TMP39]],
 // CHECK-DAG:    [[TMP40:%.*]] = getelementptr inbounds %struct.kmp_taskred_input_t, %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_4]], i32 0, i32 4
 // CHECK-DAG:    store i8* bitcast (void (i8*)* @[[RED_FINI2:.+]] to i8*), i8** [[TMP40]],
 // CHECK-DAG:    [[TMP41:%.*]] = getelementptr inbounds %struct.kmp_taskred_input_t, %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_4]], i32 0, i32 5
 // CHECK-DAG:    store i8* bitcast (void (i8*, i8*)* @[[RED_COMB2:.+]] to i8*), i8** [[TMP41]],

clang/test/OpenMP/master_taskloop_simd_reduction_codegen.cpp

 // CHECK-DAG:    store i8* [[TMP31]], i8** [[TMP28:%[^,]+]],
 // CHECK-DAG:    [[TMP28]] = getelementptr inbounds %struct.kmp_taskred_input_t, %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_4:%.+]], i32 0, i32 0
 // CHECK-DAG:    [[TMP31:%.*]] = bitcast %struct.S* [[ARRAYIDX5]] to i8*
 // CHECK-DAG:    store i8* [[TMP31]], i8** [[TMP28:%[^,]+]],
 // CHECK-DAG:    [[TMP28]] = getelementptr inbounds %struct.kmp_taskred_input_t, %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_4:%.+]], i32 0, i32 1
 // CHECK-DAG:    [[TMP32:%.*]] = ptrtoint %struct.S* [[ARRAYIDX6]] to i64
 // CHECK-DAG:    [[TMP33:%.*]] = ptrtoint %struct.S* [[ARRAYIDX5]] to i64
 // CHECK-DAG:    [[TMP34:%.*]] = sub i64 [[TMP32]], [[TMP33]]
-// CHECK-DAG:    [[TMP35:%.*]] = sdiv exact i64 [[TMP34]], ptrtoint (float* getelementptr (float, float* null, i32 1) to i64)
+// CHECK-DAG:    [[TMP35:%.*]] = sdiv exact i64 [[TMP34]], ptrtoint (%struct.S* getelementptr (%struct.S, %struct.S* null, i32 1) to i64)
 // CHECK-DAG:    [[TMP36:%.*]] = add nuw i64 [[TMP35]], 1
-// CHECK-DAG:    [[TMP37:%.*]] = mul nuw i64 [[TMP36]], ptrtoint (float* getelementptr (float, float* null, i32 1) to i64)
+// CHECK-DAG:    [[TMP37:%.*]] = mul nuw i64 [[TMP36]], ptrtoint (%struct.S* getelementptr (%struct.S, %struct.S* null, i32 1) to i64)
 // CHECK-DAG:    store i64 [[TMP37]], i64* [[TMP38:%[^,]+]],
 // CHECK-DAG:    [[TMP38]] = getelementptr inbounds %struct.kmp_taskred_input_t, %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_4]], i32 0, i32 2
 // CHECK-DAG:    [[TMP39:%.*]] = getelementptr inbounds %struct.kmp_taskred_input_t, %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_4]], i32 0, i32 3
 // CHECK-DAG:    store i8* bitcast (void (i8*, i8*)* @[[RED_INIT2:.+]] to i8*), i8** [[TMP39]],
 // CHECK-DAG:    [[TMP40:%.*]] = getelementptr inbounds %struct.kmp_taskred_input_t, %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_4]], i32 0, i32 4
 // CHECK-DAG:    store i8* bitcast (void (i8*)* @[[RED_FINI2:.+]] to i8*), i8** [[TMP40]],
 // CHECK-DAG:    [[TMP41:%.*]] = getelementptr inbounds %struct.kmp_taskred_input_t, %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_4]], i32 0, i32 5
 // CHECK-DAG:    store i8* bitcast (void (i8*, i8*)* @[[RED_COMB2:.+]] to i8*), i8** [[TMP41]],

clang/test/OpenMP/parallel_master_taskloop_reduction_codegen.cpp

 // CHECK-DAG:    store i8* [[TMP31]], i8** [[TMP28:%[^,]+]],
 // CHECK-DAG:    [[TMP28]] = getelementptr inbounds %struct.kmp_taskred_input_t, %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_4:%.+]], i32 0, i32 0
 // CHECK-DAG:    [[TMP31:%.*]] = bitcast %struct.S* [[ARRAYIDX5]] to i8*
 // CHECK-DAG:    store i8* [[TMP31]], i8** [[TMP28:%[^,]+]],
 // CHECK-DAG:    [[TMP28]] = getelementptr inbounds %struct.kmp_taskred_input_t, %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_4:%.+]], i32 0, i32 1
 // CHECK-DAG:    [[TMP32:%.*]] = ptrtoint %struct.S* [[ARRAYIDX6]] to i64
 // CHECK-DAG:    [[TMP33:%.*]] = ptrtoint %struct.S* [[ARRAYIDX5]] to i64
 // CHECK-DAG:    [[TMP34:%.*]] = sub i64 [[TMP32]], [[TMP33]]
-// CHECK-DAG:    [[TMP35:%.*]] = sdiv exact i64 [[TMP34]], ptrtoint (float* getelementptr (float, float* null, i32 1) to i64)
+// CHECK-DAG:    [[TMP35:%.*]] = sdiv exact i64 [[TMP34]], ptrtoint (%struct.S* getelementptr (%struct.S, %struct.S* null, i32 1) to i64)
 // CHECK-DAG:    [[TMP36:%.*]] = add nuw i64 [[TMP35]], 1
-// CHECK-DAG:    [[TMP37:%.*]] = mul nuw i64 [[TMP36]], ptrtoint (float* getelementptr (float, float* null, i32 1) to i64)
+// CHECK-DAG:    [[TMP37:%.*]] = mul nuw i64 [[TMP36]], ptrtoint (%struct.S* getelementptr (%struct.S, %struct.S* null, i32 1) to i64)
 // CHECK-DAG:    store i64 [[TMP37]], i64* [[TMP38:%[^,]+]],
 // CHECK-DAG:    [[TMP38]] = getelementptr inbounds %struct.kmp_taskred_input_t, %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_4]], i32 0, i32 2
 // CHECK-DAG:    [[TMP39:%.*]] = getelementptr inbounds %struct.kmp_taskred_input_t, %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_4]], i32 0, i32 3
 // CHECK-DAG:    store i8* bitcast (void (i8*, i8*)* @[[RED_INIT2:.+]] to i8*), i8** [[TMP39]],
 // CHECK-DAG:    [[TMP40:%.*]] = getelementptr inbounds %struct.kmp_taskred_input_t, %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_4]], i32 0, i32 4
 // CHECK-DAG:    store i8* bitcast (void (i8*)* @[[RED_FINI2:.+]] to i8*), i8** [[TMP40]],
 // CHECK-DAG:    [[TMP41:%.*]] = getelementptr inbounds %struct.kmp_taskred_input_t, %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_4]], i32 0, i32 5
 // CHECK-DAG:    store i8* bitcast (void (i8*, i8*)* @[[RED_COMB2:.+]] to i8*), i8** [[TMP41]],

clang/test/OpenMP/parallel_master_taskloop_simd_reduction_codegen.cpp

 // CHECK-DAG:    store i8* [[TMP31]], i8** [[TMP28:%[^,]+]],
 // CHECK-DAG:    [[TMP28]] = getelementptr inbounds %struct.kmp_taskred_input_t, %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_4:%.+]], i32 0, i32 0
 // CHECK-DAG:    [[TMP31:%.*]] = bitcast %struct.S* [[ARRAYIDX5]] to i8*
 // CHECK-DAG:    store i8* [[TMP31]], i8** [[TMP28:%[^,]+]],
 // CHECK-DAG:    [[TMP28]] = getelementptr inbounds %struct.kmp_taskred_input_t, %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_4]], i32 0, i32 1
 // CHECK-DAG:    [[TMP32:%.*]] = ptrtoint %struct.S* [[ARRAYIDX6]] to i64
 // CHECK-DAG:    [[TMP33:%.*]] = ptrtoint %struct.S* [[ARRAYIDX5]] to i64
 // CHECK-DAG:    [[TMP34:%.*]] = sub i64 [[TMP32]], [[TMP33]]
-// CHECK-DAG:    [[TMP35:%.*]] = sdiv exact i64 [[TMP34]], ptrtoint (float* getelementptr (float, float* null, i32 1) to i64)
+// CHECK-DAG:    [[TMP35:%.*]] = sdiv exact i64 [[TMP34]], ptrtoint (%struct.S* getelementptr (%struct.S, %struct.S* null, i32 1) to i64)
 // CHECK-DAG:    [[TMP36:%.*]] = add nuw i64 [[TMP35]], 1
-// CHECK-DAG:    [[TMP37:%.*]] = mul nuw i64 [[TMP36]], ptrtoint (float* getelementptr (float, float* null, i32 1) to i64)
+// CHECK-DAG:    [[TMP37:%.*]] = mul nuw i64 [[TMP36]], ptrtoint (%struct.S* getelementptr (%struct.S, %struct.S* null, i32 1) to i64)
 // CHECK-DAG:    store i64 [[TMP37]], i64* [[TMP38:%[^,]+]],
 // CHECK-DAG:    [[TMP38]] = getelementptr inbounds %struct.kmp_taskred_input_t, %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_4]], i32 0, i32 2
 // CHECK-DAG:    [[TMP39:%.*]] = getelementptr inbounds %struct.kmp_taskred_input_t, %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_4]], i32 0, i32 3
 // CHECK-DAG:    store i8* bitcast (void (i8*, i8*)* @[[RED_INIT2:.+]] to i8*), i8** [[TMP39]],
 // CHECK-DAG:    [[TMP40:%.*]] = getelementptr inbounds %struct.kmp_taskred_input_t, %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_4]], i32 0, i32 4
 // CHECK-DAG:    store i8* bitcast (void (i8*)* @[[RED_FINI2:.+]] to i8*), i8** [[TMP40]],
 // CHECK-DAG:    [[TMP41:%.*]] = getelementptr inbounds %struct.kmp_taskred_input_t, %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_4]], i32 0, i32 5
 // CHECK-DAG:    store i8* bitcast (void (i8*, i8*)* @[[RED_COMB2:.+]] to i8*), i8** [[TMP41]],

clang/test/OpenMP/taskloop_reduction_codegen.cpp

 // CHECK-DAG:    store i8* [[TMP31]], i8** [[TMP28:%[^,]+]],
 // CHECK-DAG:    [[TMP28]] = getelementptr inbounds %struct.kmp_taskred_input_t, %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_4:%.+]], i32 0, i32 0
 // CHECK-DAG:    [[TMP31:%.*]] = bitcast %struct.S* [[ARRAYIDX5]] to i8*
 // CHECK-DAG:    store i8* [[TMP31]], i8** [[TMP28:%[^,]+]],
 // CHECK-DAG:    [[TMP28]] = getelementptr inbounds %struct.kmp_taskred_input_t, %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_4]], i32 0, i32 1
 // CHECK-DAG:    [[TMP32:%.*]] = ptrtoint %struct.S* [[ARRAYIDX6]] to i64
 // CHECK-DAG:    [[TMP33:%.*]] = ptrtoint %struct.S* [[ARRAYIDX5]] to i64
 // CHECK-DAG:    [[TMP34:%.*]] = sub i64 [[TMP32]], [[TMP33]]
-// CHECK-DAG:    [[TMP35:%.*]] = sdiv exact i64 [[TMP34]], ptrtoint (float* getelementptr (float, float* null, i32 1) to i64)
+// CHECK-DAG:    [[TMP35:%.*]] = sdiv exact i64 [[TMP34]], ptrtoint (%struct.S* getelementptr (%struct.S, %struct.S* null, i32 1) to i64)
 // CHECK-DAG:    [[TMP36:%.*]] = add nuw i64 [[TMP35]], 1
-// CHECK-DAG:    [[TMP37:%.*]] = mul nuw i64 [[TMP36]], ptrtoint (float* getelementptr (float, float* null, i32 1) to i64)
+// CHECK-DAG:    [[TMP37:%.*]] = mul nuw i64 [[TMP36]], ptrtoint (%struct.S* getelementptr (%struct.S, %struct.S* null, i32 1) to i64)
 // CHECK-DAG:    store i64 [[TMP37]], i64* [[TMP38:%[^,]+]],
 // CHECK-DAG:    [[TMP38]] = getelementptr inbounds %struct.kmp_taskred_input_t, %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_4]], i32 0, i32 2
 // CHECK-DAG:    [[TMP39:%.*]] = getelementptr inbounds %struct.kmp_taskred_input_t, %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_4]], i32 0, i32 3
 // CHECK-DAG:    store i8* bitcast (void (i8*, i8*)* @[[RED_INIT2:.+]] to i8*), i8** [[TMP39]],
 // CHECK-DAG:    [[TMP40:%.*]] = getelementptr inbounds %struct.kmp_taskred_input_t, %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_4]], i32 0, i32 4
 // CHECK-DAG:    store i8* bitcast (void (i8*)* @[[RED_FINI2:.+]] to i8*), i8** [[TMP40]],
 // CHECK-DAG:    [[TMP41:%.*]] = getelementptr inbounds %struct.kmp_taskred_input_t, %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_4]], i32 0, i32 5
 // CHECK-DAG:    store i8* bitcast (void (i8*, i8*)* @[[RED_COMB2:.+]] to i8*), i8** [[TMP41]],

clang/test/OpenMP/taskloop_simd_reduction_codegen.cpp

 // CHECK-DAG:    store i8* [[TMP31]], i8** [[TMP28:%[^,]+]],
 // CHECK-DAG:    [[TMP28]] = getelementptr inbounds %struct.kmp_taskred_input_t, %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_4:%.+]], i32 0, i32 0
 // CHECK-DAG:    [[TMP31:%.*]] = bitcast %struct.S* [[ARRAYIDX5]] to i8*
 // CHECK-DAG:    store i8* [[TMP31]], i8** [[TMP28:%[^,]+]],
 // CHECK-DAG:    [[TMP28]] = getelementptr inbounds %struct.kmp_taskred_input_t, %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_4]], i32 0, i32 1
 // CHECK-DAG:    [[TMP32:%.*]] = ptrtoint %struct.S* [[ARRAYIDX6]] to i64
 // CHECK-DAG:    [[TMP33:%.*]] = ptrtoint %struct.S* [[ARRAYIDX5]] to i64
 // CHECK-DAG:    [[TMP34:%.*]] = sub i64 [[TMP32]], [[TMP33]]
-// CHECK-DAG:    [[TMP35:%.*]] = sdiv exact i64 [[TMP34]], ptrtoint (float* getelementptr (float, float* null, i32 1) to i64)
+// CHECK-DAG:    [[TMP35:%.*]] = sdiv exact i64 [[TMP34]], ptrtoint (%struct.S* getelementptr (%struct.S, %struct.S* null, i32 1) to i64)
 // CHECK-DAG:    [[TMP36:%.*]] = add nuw i64 [[TMP35]], 1
-// CHECK-DAG:    [[TMP37:%.*]] = mul nuw i64 [[TMP36]], ptrtoint (float* getelementptr (float, float* null, i32 1) to i64)
+// CHECK-DAG:    [[TMP37:%.*]] = mul nuw i64 [[TMP36]], ptrtoint (%struct.S* getelementptr (%struct.S, %struct.S* null, i32 1) to i64)
 // CHECK-DAG:    store i64 [[TMP37]], i64* [[TMP38:%[^,]+]],
 // CHECK-DAG:    [[TMP38]] = getelementptr inbounds %struct.kmp_taskred_input_t, %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_4]], i32 0, i32 2
 // CHECK-DAG:    [[TMP39:%.*]] = getelementptr inbounds %struct.kmp_taskred_input_t, %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_4]], i32 0, i32 3
 // CHECK-DAG:    store i8* bitcast (void (i8*, i8*)* @[[RED_INIT2:.+]] to i8*), i8** [[TMP39]],
 // CHECK-DAG:    [[TMP40:%.*]] = getelementptr inbounds %struct.kmp_taskred_input_t, %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_4]], i32 0, i32 4
 // CHECK-DAG:    store i8* bitcast (void (i8*)* @[[RED_FINI2:.+]] to i8*), i8** [[TMP40]],
 // CHECK-DAG:    [[TMP41:%.*]] = getelementptr inbounds %struct.kmp_taskred_input_t, %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_4]], i32 0, i32 5
 // CHECK-DAG:    store i8* bitcast (void (i8*, i8*)* @[[RED_COMB2:.+]] to i8*), i8** [[TMP41]],

llvm/lib/IR/ConstantFold.cpp

case Instruction::ZExt: {
     }
 
     // TODO: Handle the 'partially zero' case.
     return nullptr;
   }
   }
 }
 
-/// Wrapper around getFoldedSizeOfImpl() that adds caching.
-static Constant *getFoldedSizeOf(Type *Ty, Type *DestTy, bool Folded,
-                                 DenseMap<Type *, Constant *> &Cache);
-
-/// Return a ConstantExpr with type DestTy for sizeof on Ty, with any known
-/// factors factored out. If Folded is false, return null if no factoring was
-/// possible, to avoid endlessly bouncing an unfoldable expression back into the
-/// top-level folder.
-static Constant *getFoldedSizeOfImpl(Type *Ty, Type *DestTy, bool Folded,
-                                     DenseMap<Type *, Constant *> &Cache) {
-  // This is the actual implementation of getFoldedSizeOf(). To get the caching
-  // behavior, we need to call getFoldedSizeOf() when we recurse.
-
-  if (ArrayType *ATy = dyn_cast<ArrayType>(Ty)) {
-    Constant *N = ConstantInt::get(DestTy, ATy->getNumElements());
-    Constant *E = getFoldedSizeOf(ATy->getElementType(), DestTy, true, Cache);
-    return ConstantExpr::getNUWMul(E, N);
-  }
-
-  if (StructType *STy = dyn_cast<StructType>(Ty))
-    if (!STy->isPacked()) {
-      unsigned NumElems = STy->getNumElements();
-      // An empty struct has size zero.
-      if (NumElems == 0)
-        return ConstantExpr::getNullValue(DestTy);
-      // Check for a struct with all members having the same size.
-      Constant *MemberSize =
-          getFoldedSizeOf(STy->getElementType(0), DestTy, true, Cache);
-      bool AllSame = true;
-      for (unsigned i = 1; i != NumElems; ++i)
-        if (MemberSize !=
-            getFoldedSizeOf(STy->getElementType(i), DestTy, true, Cache)) {
-          AllSame = false;
-          break;
-        }
-      if (AllSame) {
-        Constant *N = ConstantInt::get(DestTy, NumElems);
-        return ConstantExpr::getNUWMul(MemberSize, N);
-      }
-    }
-
-  // Pointer size doesn't depend on the pointee type, so canonicalize them
-  // to an arbitrary pointee.
-  if (PointerType *PTy = dyn_cast<PointerType>(Ty))
-    if (!PTy->getElementType()->isIntegerTy(1))
-      return getFoldedSizeOf(
-          PointerType::get(IntegerType::get(PTy->getContext(), 1),
-                           PTy->getAddressSpace()),
-          DestTy, true, Cache);
-
-  // If there's no interesting folding happening, bail so that we don't create
-  // a constant that looks like it needs folding but really doesn't.
-  if (!Folded)
-    return nullptr;
-
-  // Base case: Get a regular sizeof expression.
-  Constant *C = ConstantExpr::getSizeOf(Ty);
-  C = ConstantExpr::getCast(CastInst::getCastOpcode(C, false,
-                                                    DestTy, false),
-                            C, DestTy);
-  return C;
-}
-
-static Constant *getFoldedSizeOf(Type *Ty, Type *DestTy, bool Folded,
-                                 DenseMap<Type *, Constant *> &Cache) {
-  // Check for previously generated folded size constant.
-  auto It = Cache.find(Ty);
-  if (It != Cache.end())
-    return It->second;
-  return Cache[Ty] = getFoldedSizeOfImpl(Ty, DestTy, Folded, Cache);
-}
-
-static Constant *getFoldedSizeOf(Type *Ty, Type *DestTy, bool Folded) {
-  DenseMap<Type *, Constant *> Cache;
-  return getFoldedSizeOf(Ty, DestTy, Folded, Cache);
-}
-
-/// Return a ConstantExpr with type DestTy for alignof on Ty, with any known
-/// factors factored out. If Folded is false, return null if no factoring was
-/// possible, to avoid endlessly bouncing an unfoldable expression back into the
-/// top-level folder.
-static Constant *getFoldedAlignOf(Type *Ty, Type *DestTy, bool Folded) {
-  // The alignment of an array is equal to the alignment of the
-  // array element. Note that this is not always true for vectors.
-  if (ArrayType *ATy = dyn_cast<ArrayType>(Ty)) {
-    Constant *C = ConstantExpr::getAlignOf(ATy->getElementType());
-    C = ConstantExpr::getCast(CastInst::getCastOpcode(C, false,
-                                                      DestTy,
-                                                      false),
-                              C, DestTy);
-    return C;
-  }
-
-  if (StructType *STy = dyn_cast<StructType>(Ty)) {
-    // Packed structs always have an alignment of 1.
-    if (STy->isPacked())
-      return ConstantInt::get(DestTy, 1);
-
-    // Otherwise, struct alignment is the maximum alignment of any member.
-    // Without target data, we can't compare much, but we can check to see
-    // if all the members have the same alignment.
-    unsigned NumElems = STy->getNumElements();
-    // An empty struct has minimal alignment.
-    if (NumElems == 0)
-      return ConstantInt::get(DestTy, 1);
-    // Check for a struct with all members having the same alignment.
-    Constant *MemberAlign =
-      getFoldedAlignOf(STy->getElementType(0), DestTy, true);
-    bool AllSame = true;
-    for (unsigned i = 1; i != NumElems; ++i)
-      if (MemberAlign != getFoldedAlignOf(STy->getElementType(i), DestTy, true)) {
-        AllSame = false;
-        break;
-      }
-    if (AllSame)
-      return MemberAlign;
-  }
-
-  // Pointer alignment doesn't depend on the pointee type, so canonicalize them
-  // to an arbitrary pointee.
-  if (PointerType *PTy = dyn_cast<PointerType>(Ty))
-    if (!PTy->getElementType()->isIntegerTy(1))
-      return
-        getFoldedAlignOf(PointerType::get(IntegerType::get(PTy->getContext(),
-                                                           1),
-                                          PTy->getAddressSpace()),
-                         DestTy, true);
-
-  // If there's no interesting folding happening, bail so that we don't create
-  // a constant that looks like it needs folding but really doesn't.
-  if (!Folded)
-    return nullptr;
-
-  // Base case: Get a regular alignof expression.
-  Constant *C = ConstantExpr::getAlignOf(Ty);
-  C = ConstantExpr::getCast(CastInst::getCastOpcode(C, false,
-                                                    DestTy, false),
-                            C, DestTy);
-  return C;
-}
-
-/// Return a ConstantExpr with type DestTy for offsetof on Ty and FieldNo, with
-/// any known factors factored out. If Folded is false, return null if no
-/// factoring was possible, to avoid endlessly bouncing an unfoldable expression
-/// back into the top-level folder.
-static Constant *getFoldedOffsetOf(Type *Ty, Constant *FieldNo, Type *DestTy,
-                                   bool Folded) {
-  if (ArrayType *ATy = dyn_cast<ArrayType>(Ty)) {
-    Constant *N = ConstantExpr::getCast(CastInst::getCastOpcode(FieldNo, false,
-                                                                DestTy, false),
-                                        FieldNo, DestTy);
-    Constant *E = getFoldedSizeOf(ATy->getElementType(), DestTy, true);
-    return ConstantExpr::getNUWMul(E, N);
-  }
-
-  if (StructType *STy = dyn_cast<StructType>(Ty))
-    if (!STy->isPacked()) {
-      unsigned NumElems = STy->getNumElements();
-      // An empty struct has no members.
-      if (NumElems == 0)
-        return nullptr;
-      // Check for a struct with all members having the same size.
-      Constant *MemberSize =
-        getFoldedSizeOf(STy->getElementType(0), DestTy, true);
-      bool AllSame = true;
-      for (unsigned i = 1; i != NumElems; ++i)
-        if (MemberSize !=
-            getFoldedSizeOf(STy->getElementType(i), DestTy, true)) {
-          AllSame = false;
-          break;
-        }
-      if (AllSame) {
-        Constant *N = ConstantExpr::getCast(CastInst::getCastOpcode(FieldNo,
-                                                                    false,
-                                                                    DestTy,
-                                                                    false),
-                                            FieldNo, DestTy);
-        return ConstantExpr::getNUWMul(MemberSize, N);
-      }
-    }
-
-  // If there's no interesting folding happening, bail so that we don't create
-  // a constant that looks like it needs folding but really doesn't.
-  if (!Folded)
-    return nullptr;
-
-  // Base case: Get a regular offsetof expression.
-  Constant *C = ConstantExpr::getOffsetOf(Ty, FieldNo);
-  C = ConstantExpr::getCast(CastInst::getCastOpcode(C, false,
-                                                    DestTy, false),
-                            C, DestTy);
-  return C;
-}
-
 Constant *llvm::ConstantFoldCastInstruction(unsigned opc, Constant *V,
                                             Type *DestTy) {
   if (isa<PoisonValue>(V))
     return PoisonValue::get(DestTy);
 
   if (isa<UndefValue>(V)) {
     // zext(undef) = 0, because the top bits will be zero.
     // sext(undef) = 0, because the top bits will all be the same.
Constant *llvm::ConstantFoldCastInstruction(unsigned opc, Constant *V,
   case Instruction::IntToPtr:   //always treated as unsigned
     if (V->isNullValue())       // Is it an integral null value?
       return ConstantPointerNull::get(cast<PointerType>(DestTy));
     return nullptr;                   // Other pointer types cannot be casted
   case Instruction::PtrToInt:   // always treated as unsigned
     // Is it a null pointer value?
     if (V->isNullValue())
       return ConstantInt::get(DestTy, 0);
-    // If this is a sizeof-like expression, pull out multiplications by
-    // known factors to expose them to subsequent folding. If it's an
-    // alignof-like expression, factor out known factors.
-    if (ConstantExpr *CE = dyn_cast<ConstantExpr>(V))
-      if (CE->getOpcode() == Instruction::GetElementPtr &&
-          CE->getOperand(0)->isNullValue()) {
-        // FIXME: Looks like getFoldedSizeOf(), getFoldedOffsetOf() and
-        // getFoldedAlignOf() don't handle the case when DestTy is a vector of
-        // pointers yet. We end up in asserts in CastInst::getCastOpcode (see
-        // test/Analysis/ConstantFolding/cast-vector.ll). I've only seen this
-        // happen in one "real" C-code test case, so it does not seem to be an
-        // important optimization to handle vectors here. For now, simply bail
-        // out.
-        if (DestTy->isVectorTy())
-          return nullptr;
-        GEPOperator *GEPO = cast<GEPOperator>(CE);
-        Type *Ty = GEPO->getSourceElementType();
-        if (CE->getNumOperands() == 2) {
-          // Handle a sizeof-like expression.
-          Constant *Idx = CE->getOperand(1);
-          bool isOne = isa<ConstantInt>(Idx) && cast<ConstantInt>(Idx)->isOne();
-          if (Constant *C = getFoldedSizeOf(Ty, DestTy, !isOne)) {
-            Idx = ConstantExpr::getCast(CastInst::getCastOpcode(Idx, true,
-                                                                DestTy, false),
-                                        Idx, DestTy);
-            return ConstantExpr::getMul(C, Idx);
-          }
-        } else if (CE->getNumOperands() == 3 &&
-                   CE->getOperand(1)->isNullValue()) {
-          // Handle an alignof-like expression.
-          if (StructType *STy = dyn_cast<StructType>(Ty))
-            if (!STy->isPacked()) {
-              ConstantInt *CI = cast<ConstantInt>(CE->getOperand(2));
-              if (CI->isOne() &&
-                  STy->getNumElements() == 2 &&
-                  STy->getElementType(0)->isIntegerTy(1)) {
-                return getFoldedAlignOf(STy->getElementType(1), DestTy, false);
-              }
-            }
-          // Handle an offsetof-like expression.
-          if (Ty->isStructTy() || Ty->isArrayTy()) {
-            if (Constant *C = getFoldedOffsetOf(Ty, CE->getOperand(2),
-                                                DestTy, false))
-              return C;
-          }
-        }
-      }
     // Other pointer types cannot be casted
     return nullptr;
   case Instruction::UIToFP:
   case Instruction::SIToFP:
     if (ConstantInt *CI = dyn_cast<ConstantInt>(V)) {
       const APInt &api = CI->getValue();
       APFloat apf(DestTy->getFltSemantics(),
                   APInt::getNullValue(DestTy->getPrimitiveSizeInBits()));

llvm/test/Other/constant-fold-gep.ll

 @G1 = global i1* getelementptr (i1, i1* inttoptr (i32 1 to i1*), i32 -1)
 @F8 = global i8* getelementptr (i8, i8* inttoptr (i32 1 to i8*), i32 -2)
 @F1 = global i1* getelementptr (i1, i1* inttoptr (i32 1 to i1*), i32 -2)
 @H8 = global i8* getelementptr (i8, i8* inttoptr (i32 0 to i8*), i32 -1)
 @H1 = global i1* getelementptr (i1, i1* inttoptr (i32 0 to i1*), i32 -1)
 
 ; The target-independent folder should be able to do some clever
 ; simplifications on sizeof, alignof, and offsetof expressions. The
 ; target-dependent folder should fold these down to constants.
 
-; PLAIN: @a = constant i64 mul (i64 ptrtoint (double* getelementptr (double, double* null, i32 1) to i64), i64 2310)
-; PLAIN: @b = constant i64 ptrtoint (double* getelementptr ({ i1, double }, { i1, double }* null, i64 0, i32 1) to i64)
-; PLAIN: @c = constant i64 mul nuw (i64 ptrtoint (double* getelementptr (double, double* null, i32 1) to i64), i64 2)
-; PLAIN: @d = constant i64 mul nuw (i64 ptrtoint (double* getelementptr (double, double* null, i32 1) to i64), i64 11)
+; PLAIN: @a = constant i64 mul (i64 ptrtoint ({ [7 x double], [7 x double] }* getelementptr ({ [7 x double], [7 x double] }, { [7 x double], [7 x double] }* null, i64 11) to i64), i64 15)
+; PLAIN: @b = constant i64 ptrtoint ([13 x double]* getelementptr ({ i1, [13 x double] }, { i1, [13 x double] }* null, i64 0, i32 1) to i64)
+; PLAIN: @c = constant i64 ptrtoint (double* getelementptr ({ double, double, double, double }, { double, double, double, double }* null, i64 0, i32 2) to i64)
+; PLAIN: @d = constant i64 ptrtoint (double* getelementptr ([13 x double], [13 x double]* null, i64 0, i32 11) to i64)
 ; PLAIN: @e = constant i64 ptrtoint (double* getelementptr ({ double, float, double, double }, { double, float, double, double }* null, i64 0, i32 2) to i64)
-; PLAIN: @f = constant i64 1
-; PLAIN: @g = constant i64 ptrtoint (double* getelementptr ({ i1, double }, { i1, double }* null, i64 0, i32 1) to i64)
-; PLAIN: @h = constant i64 ptrtoint (i1** getelementptr (i1*, i1** null, i32 1) to i64)
-; PLAIN: @i = constant i64 ptrtoint (i1** getelementptr ({ i1, i1* }, { i1, i1* }* null, i64 0, i32 1) to i64)
+; PLAIN: @f = constant i64 ptrtoint (<{ i16, i128 }>* getelementptr ({ i1, <{ i16, i128 }> }, { i1, <{ i16, i128 }> }* null, i64 0, i32 1) to i64)
+; PLAIN: @g = constant i64 ptrtoint ({ double, double }* getelementptr ({ i1, { double, double } }, { i1, { double, double } }* null, i64 0, i32 1) to i64)
+; PLAIN: @h = constant i64 ptrtoint (double** getelementptr (double*, double** null, i64 1) to i64)
+; PLAIN: @i = constant i64 ptrtoint (double** getelementptr ({ i1, double* }, { i1, double* }* null, i64 0, i32 1) to i64)
 ; OPT: @a = local_unnamed_addr constant i64 18480
 ; OPT: @b = local_unnamed_addr constant i64 8
 ; OPT: @c = local_unnamed_addr constant i64 16
 ; OPT: @d = local_unnamed_addr constant i64 88
 ; OPT: @e = local_unnamed_addr constant i64 16
 ; OPT: @f = local_unnamed_addr constant i64 1
 ; OPT: @g = local_unnamed_addr constant i64 8
 ; OPT: @h = local_unnamed_addr constant i64 8
define i8* @hoo8() nounwind {
   ret i8* %t
 }
 define i1* @hoo1() nounwind {
   %t = bitcast i1* getelementptr (i1, i1* inttoptr (i32 0 to i1*), i32 -1) to i1*
   ret i1* %t
 }
 
 ; PLAIN: define i64 @fa() #0 {
-; PLAIN:   %t = bitcast i64 mul (i64 ptrtoint (double* getelementptr (double, double* null, i32 1) to i64), i64 2310) to i64
+; PLAIN:   %t = bitcast i64 mul (i64 ptrtoint ({ [7 x double], [7 x double] }* getelementptr ({ [7 x double], [7 x double] }, { [7 x double], [7 x double] }* null, i64 11) to i64), i64 15) to i64
 ; PLAIN:   ret i64 %t
 ; PLAIN: }
 ; PLAIN: define i64 @fb() #0 {
-; PLAIN:   %t = bitcast i64 ptrtoint (double* getelementptr ({ i1, double }, { i1, double }* null, i64 0, i32 1) to i64) to i64
+; PLAIN:   %t = bitcast i64 ptrtoint ([13 x double]* getelementptr ({ i1, [13 x double] }, { i1, [13 x double] }* null, i64 0, i32 1) to i64) to i64
 ; PLAIN:   ret i64 %t
 ; PLAIN: }
 ; PLAIN: define i64 @fc() #0 {
-; PLAIN:   %t = bitcast i64 mul nuw (i64 ptrtoint (double* getelementptr (double, double* null, i32 1) to i64), i64 2) to i64
+; PLAIN:   %t = bitcast i64 ptrtoint (double* getelementptr ({ double, double, double, double }, { double, double, double, double }* null, i64 0, i32 2) to i64) to i64
 ; PLAIN:   ret i64 %t
 ; PLAIN: }
 ; PLAIN: define i64 @fd() #0 {
-; PLAIN:   %t = bitcast i64 mul nuw (i64 ptrtoint (double* getelementptr (double, double* null, i32 1) to i64), i64 11) to i64
+; PLAIN:   %t = bitcast i64 ptrtoint (double* getelementptr ([13 x double], [13 x double]* null, i64 0, i32 11) to i64) to i64
 ; PLAIN:   ret i64 %t
 ; PLAIN: }
 ; PLAIN: define i64 @fe() #0 {
 ; PLAIN:   %t = bitcast i64 ptrtoint (double* getelementptr ({ double, float, double, double }, { double, float, double, double }* null, i64 0, i32 2) to i64) to i64
 ; PLAIN:   ret i64 %t
 ; PLAIN: }
 ; PLAIN: define i64 @ff() #0 {
-; PLAIN:   %t = bitcast i64 1 to i64
+; PLAIN:   %t = bitcast i64 ptrtoint (<{ i16, i128 }>* getelementptr ({ i1, <{ i16, i128 }> }, { i1, <{ i16, i128 }> }* null, i64 0, i32 1) to i64) to i64
 ; PLAIN:   ret i64 %t
 ; PLAIN: }
 ; PLAIN: define i64 @fg() #0 {
-; PLAIN:   %t = bitcast i64 ptrtoint (double* getelementptr ({ i1, double }, { i1, double }* null, i64 0, i32 1) to i64) to i64
+; PLAIN:   %t = bitcast i64 ptrtoint ({ double, double }* getelementptr ({ i1, { double, double } }, { i1, { double, double } }* null, i64 0, i32 1) to i64) to i64
 ; PLAIN:   ret i64 %t
 ; PLAIN: }
 ; PLAIN: define i64 @fh() #0 {
-; PLAIN:   %t = bitcast i64 ptrtoint (i1** getelementptr (i1*, i1** null, i32 1) to i64) to i64
+; PLAIN:   %t = bitcast i64 ptrtoint (double** getelementptr (double*, double** null, i32 1) to i64) to i64
 ; PLAIN:   ret i64 %t
 ; PLAIN: }
 ; PLAIN: define i64 @fi() #0 {
-; PLAIN:   %t = bitcast i64 ptrtoint (i1** getelementptr ({ i1, i1* }, { i1, i1* }* null, i64 0, i32 1) to i64) to i64
+; PLAIN:   %t = bitcast i64 ptrtoint (double** getelementptr ({ i1, double* }, { i1, double* }* null, i64 0, i32 1) to i64) to i64
 ; PLAIN:   ret i64 %t
 ; PLAIN: }
 ; OPT: define i64 @fa() local_unnamed_addr #0 {
 ; OPT:   ret i64 18480
 ; OPT: }
 ; OPT: define i64 @fb() local_unnamed_addr #0 {
 ; OPT:   ret i64 8
 ; OPT: }

 ; TO:   ret i64 8
 ; TO: }
 ; TO: define i64 @fh() local_unnamed_addr #0 {
 ; TO:   ret i64 8
 ; TO: }
 ; TO: define i64 @fi() local_unnamed_addr #0 {
 ; TO:   ret i64 8
 ; TO: }
-; SCEV: Classifying expressions for: @fa
-; SCEV:   %t = bitcast i64 mul (i64 ptrtoint (double* getelementptr (double, double* null, i32 1) to i64), i64 2310) to i64
+; SCEV-LABEL: Classifying expressions for: @fa
+; SCEV:   %t = bitcast i64 mul (i64 ptrtoint ({ [7 x double], [7 x double] }* getelementptr ({ [7 x double], [7 x double] }, { [7 x double], [7 x double] }* null, i64 11) to i64), i64 15) to i64
 ; SCEV:   -->  18480
-; SCEV: Classifying expressions for: @fb
-; SCEV:   %t = bitcast i64 ptrtoint (double* getelementptr ({ i1, double }, { i1, double }* null, i64 0, i32 1) to i64) to i64
+; SCEV-LABEL: Classifying expressions for: @fb
+; SCEV:  %t = bitcast i64 ptrtoint ([13 x double]* getelementptr ({ i1, [13 x double] }, { i1, [13 x double] }* null, i64 0, i32 1) to i64) to i64
 ; SCEV:   -->  8
-; SCEV: Classifying expressions for: @fc
-; SCEV:   %t = bitcast i64 mul nuw (i64 ptrtoint (double* getelementptr (double, double* null, i32 1) to i64), i64 2) to i64
+; SCEV-LABEL: Classifying expressions for: @fc
+; SCEV:  %t = bitcast i64 ptrtoint (double* getelementptr ({ double, double, double, double }, { double, double, double, double }* null, i64 0, i32 2) to i64) to i64
 ; SCEV:   -->  16
-; SCEV: Classifying expressions for: @fd
-; SCEV:   %t = bitcast i64 mul nuw (i64 ptrtoint (double* getelementptr (double, double* null, i32 1) to i64), i64 11) to i64
+; SCEV-LABEL: Classifying expressions for: @fd
+; SCEV:   %t = bitcast i64 ptrtoint (double* getelementptr ([13 x double], [13 x double]* null, i64 0, i32 11) to i64) to i64
 ; SCEV:   -->  88
-; SCEV: Classifying expressions for: @fe
+; SCEV-LABEL: Classifying expressions for: @fe
 ; SCEV:   %t = bitcast i64 ptrtoint (double* getelementptr ({ double, float, double, double }, { double, float, double, double }* null, i64 0, i32 2) to i64) to i64
 ; SCEV:   -->  16
-; SCEV: Classifying expressions for: @ff
-; SCEV:   %t = bitcast i64 1 to i64
+; SCEV-LABEL: Classifying expressions for: @ff
+; SCEV:   %t = bitcast i64 ptrtoint (<{ i16, i128 }>* getelementptr ({ i1, <{ i16, i128 }> }, { i1, <{ i16, i128 }> }* null, i64 0, i32 1) to i64) to i64
 ; SCEV:   -->  1
-; SCEV: Classifying expressions for: @fg
-; SCEV:   %t = bitcast i64 ptrtoint (double* getelementptr ({ i1, double }, { i1, double }* null, i64 0, i32 1) to i64) to i64
+; SCEV-LABEL: Classifying expressions for: @fg
+; SCEV:   %t = bitcast i64 ptrtoint ({ double, double }* getelementptr ({ i1, { double, double } }, { i1, { double, double } }* null, i64 0, i32 1) to i64) to i64
 ; SCEV:   -->  8
-; SCEV: Classifying expressions for: @fh
-; SCEV:   %t = bitcast i64 ptrtoint (i1** getelementptr (i1*, i1** null, i32 1) to i64) to i64
+; SCEV-LABEL: Classifying expressions for: @fh
+; SCEV:   %t = bitcast i64 ptrtoint (double** getelementptr (double*, double** null, i32 1) to i64) to i64
 ; SCEV:   --> 8
-; SCEV: Classifying expressions for: @fi
-; SCEV:   %t = bitcast i64 ptrtoint (i1** getelementptr ({ i1, i1* }, { i1, i1* }* null, i64 0, i32 1) to i64) to i64
+; SCEV-LABEL: Classifying expressions for: @fi
+; SCEV:   %t = bitcast i64 ptrtoint (double** getelementptr ({ i1, double* }, { i1, double* }* null, i64 0, i32 1) to i64) to i64
 ; SCEV:   --> 8
 
 define i64 @fa() nounwind {
   %t = bitcast i64 mul (i64 3, i64 mul (i64 ptrtoint ({[7 x double], [7 x double]}* getelementptr ({[7 x double], [7 x double]}, {[7 x double], [7 x double]}* null, i64 11) to i64), i64 5)) to i64
   ret i64 %t
 }
 define i64 @fb() nounwind {
   %t = bitcast i64 ptrtoint ([13 x double]* getelementptr ({i1, [13 x double]}, {i1, [13 x double]}* null, i64 0, i32 1) to i64) to i64

llvm/test/Transforms/LowerTypeTests/function-disjoint.ll

 
 declare i1 @llvm.type.test(i8* %ptr, metadata %bitset) nounwind readnone
 
 define i1 @foo(i8* %p) {
   ; X64: icmp eq i64 {{.*}}, ptrtoint (void ()* @[[JT0]] to i64)
   ; WASM32: icmp eq i64 {{.*}}, ptrtoint (i8* getelementptr (i8, i8* null, i64 1) to i64)
   %x = call i1 @llvm.type.test(i8* %p, metadata !"typeid1")
   ; X64: icmp eq i64 {{.*}}, ptrtoint (void ()* @[[JT1]] to i64)
-  ; WASM32: icmp eq i64 {{.*}}, mul (i64 ptrtoint (i8* getelementptr (i8, i8* null, i32 1) to i64), i64 2)
+  ; WASM32: icmp eq i64 {{.*}}, ptrtoint (i8* getelementptr (i8, i8* null, i64 2) to i64)
   %y = call i1 @llvm.type.test(i8* %p, metadata !"typeid2")
   %z = add i1 %x, %y
   ret i1 %z
 }
 
 ; X64: define private void @[[JT0]]() #{{.*}} align 8 {
 ; X64:   call void asm sideeffect "jmp ${0:c}@plt\0Aint3\0Aint3\0Aint3\0A", "s"(void ()* @f.cfi)
 
 ; X64: define private void @[[JT1]]() #{{.*}} align 8 {
 ; X64:   call void asm sideeffect "jmp ${0:c}@plt\0Aint3\0Aint3\0Aint3\0A", "s"(void ()* @g.cfi)
 
 ; WASM32: ![[I0]] = !{i64 1}
 ; WASM32: ![[I1]] = !{i64 2}

llvm/test/tools/llvm-as/slow-ptrtoint.ll

 %11 = type { %12, %12, %12, %12, %12, %12, %12, %12 }
 %12 = type { %13, %13, %13, %13, %13, %13, %13, %13 }
 %13 = type { i32, i32 }
 
 ; it would take a naive recursive implementation ~4 days
 ; to constant fold the size of %0
 define i64 @f_i64() {
 ; CHECK-LABEL: @f_i64
-; CHECK:         ret i64 mul (i64 ptrtoint (i32* getelementptr (i32, i32* null, i32 1) to i64), i64 1099511627776)
+; CHECK:         ret i64 ptrtoint (%0* getelementptr (%0, %0* null, i32 1) to i64)
   ret i64 ptrtoint (%0* getelementptr (%0, %0* null, i32 1) to i64)
 }
 
 define i32 @f_i32() {
 ; CHECK-LABEL: @f_i32
-; CHECK:         ret i32 mul (i32 ptrtoint (i32* getelementptr (i32, i32* null, i32 1) to i32), i32 -2147483648)
+; CHECK:         ret i32 ptrtoint (%3* getelementptr (%3, %3* null, i32 1) to i32)
   ret i32 ptrtoint (%3* getelementptr (%3, %3* null, i32 1) to i32)
 }