diff --git a/clang/test/OpenMP/distribute_parallel_for_reduction_task_codegen.cpp b/clang/test/OpenMP/distribute_parallel_for_reduction_task_codegen.cpp
--- a/clang/test/OpenMP/distribute_parallel_for_reduction_task_codegen.cpp
+++ b/clang/test/OpenMP/distribute_parallel_for_reduction_task_codegen.cpp
@@ -448,7 +448,7 @@
 // CHECK1-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR]], align 8
 // CHECK1-NEXT:    store ptr [[TMP1]], ptr [[DOTADDR1]], align 8
 // CHECK1-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTADDR]], align 8
-// CHECK1-NEXT:    [[TMP3:%.*]] = call ptr @llvm.threadlocal.address.p0(ptr @{{reduction_size[.].+[.]}})
+// CHECK1-NEXT:    [[TMP3:%.*]] = call align 8 ptr @llvm.threadlocal.address.p0(ptr align 8 @{{reduction_size[.].+[.]}})
 // CHECK1-NEXT:    [[TMP4:%.*]] = load i64, ptr [[TMP3]], align 8
 // CHECK1-NEXT:    [[TMP5:%.*]] = getelementptr i8, ptr [[TMP2]], i64 [[TMP4]]
 // CHECK1-NEXT:    [[OMP_ARRAYINIT_ISEMPTY:%.*]] = icmp eq ptr [[TMP2]], [[TMP5]]
@@ -470,7 +470,7 @@
 // CHECK1-NEXT:    [[DOTADDR1:%.*]] = alloca ptr, align 8
 // CHECK1-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR]], align 8
 // CHECK1-NEXT:    store ptr [[TMP1]], ptr [[DOTADDR1]], align 8
-// CHECK1-NEXT:    [[TMP2:%.*]] = call ptr @llvm.threadlocal.address.p0(ptr @{{reduction_size[.].+[.]}})
+// CHECK1-NEXT:    [[TMP2:%.*]] = call align 8 ptr @llvm.threadlocal.address.p0(ptr align 8 @{{reduction_size[.].+[.]}})
 // CHECK1-NEXT:    [[TMP3:%.*]] = load i64, ptr [[TMP2]], align 8
 // CHECK1-NEXT:    [[TMP4:%.*]] = load ptr, ptr [[DOTADDR]], align 8
 // CHECK1-NEXT:    [[TMP5:%.*]] = load ptr, ptr [[DOTADDR1]], align 8
diff --git a/clang/test/OpenMP/for_reduction_task_codegen.cpp b/clang/test/OpenMP/for_reduction_task_codegen.cpp
--- a/clang/test/OpenMP/for_reduction_task_codegen.cpp
+++ b/clang/test/OpenMP/for_reduction_task_codegen.cpp
@@ -19,22 +19,6 @@
   }
 }
 
-
-
-// Init firstprivate copy of argc
-
-// Init firstprivate copy of argv[0:10][0:argc]
-
-// Register task reduction.
-
-
-
-
-
-
-
-
-
 #endif
 // CHECK1-LABEL: define {{[^@]+}}@main
 // CHECK1-SAME: (i32 noundef [[ARGC:%.*]], ptr noundef [[ARGV:%.*]]) #[[ATTR0:[0-9]+]] {
@@ -361,7 +345,7 @@
 // CHECK1-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR]], align 8
 // CHECK1-NEXT:    store ptr [[TMP1]], ptr [[DOTADDR1]], align 8
 // CHECK1-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTADDR]], align 8
-// CHECK1-NEXT:    [[TMP3:%.*]] = call ptr @llvm.threadlocal.address.p0(ptr @{{reduction_size[.].+[.]}})
+// CHECK1-NEXT:    [[TMP3:%.*]] = call align 8 ptr @llvm.threadlocal.address.p0(ptr align 8 @{{reduction_size[.].+[.]}})
 // CHECK1-NEXT:    [[TMP4:%.*]] = load i64, ptr [[TMP3]], align 8
 // CHECK1-NEXT:    [[TMP5:%.*]] = getelementptr i8, ptr [[TMP2]], i64 [[TMP4]]
 // CHECK1-NEXT:    [[OMP_ARRAYINIT_ISEMPTY:%.*]] = icmp eq ptr [[TMP2]], [[TMP5]]
@@ -383,7 +367,7 @@
 // CHECK1-NEXT:    [[DOTADDR1:%.*]] = alloca ptr, align 8
 // CHECK1-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR]], align 8
 // CHECK1-NEXT:    store ptr [[TMP1]], ptr [[DOTADDR1]], align 8
-// CHECK1-NEXT:    [[TMP2:%.*]] = call ptr @llvm.threadlocal.address.p0(ptr @{{reduction_size[.].+[.]}})
+// CHECK1-NEXT:    [[TMP2:%.*]] = call align 8 ptr @llvm.threadlocal.address.p0(ptr align 8 @{{reduction_size[.].+[.]}})
 // CHECK1-NEXT:    [[TMP3:%.*]] = load i64, ptr [[TMP2]], align 8
 // CHECK1-NEXT:    [[TMP4:%.*]] = load ptr, ptr [[DOTADDR]], align 8
 // CHECK1-NEXT:    [[TMP5:%.*]] = load ptr, ptr [[DOTADDR1]], align 8
diff --git a/clang/test/OpenMP/parallel_for_reduction_task_codegen.cpp b/clang/test/OpenMP/parallel_for_reduction_task_codegen.cpp
--- a/clang/test/OpenMP/parallel_for_reduction_task_codegen.cpp
+++ b/clang/test/OpenMP/parallel_for_reduction_task_codegen.cpp
@@ -356,7 +356,7 @@
 // CHECK1-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR]], align 8
 // CHECK1-NEXT:    store ptr [[TMP1]], ptr [[DOTADDR1]], align 8
 // CHECK1-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTADDR]], align 8
-// CHECK1-NEXT:    [[TMP3:%.*]] = call ptr @llvm.threadlocal.address.p0(ptr @{{reduction_size[.].+[.]}})
+// CHECK1-NEXT:    [[TMP3:%.*]] = call align 8 ptr @llvm.threadlocal.address.p0(ptr align 8 @{{reduction_size[.].+[.]}})
 // CHECK1-NEXT:    [[TMP4:%.*]] = load i64, ptr [[TMP3]], align 8
 // CHECK1-NEXT:    [[TMP5:%.*]] = getelementptr i8, ptr [[TMP2]], i64 [[TMP4]]
 // CHECK1-NEXT:    [[OMP_ARRAYINIT_ISEMPTY:%.*]] = icmp eq ptr [[TMP2]], [[TMP5]]
@@ -378,7 +378,7 @@
 // CHECK1-NEXT:    [[DOTADDR1:%.*]] = alloca ptr, align 8
 // CHECK1-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR]], align 8
 // CHECK1-NEXT:    store ptr [[TMP1]], ptr [[DOTADDR1]], align 8
-// CHECK1-NEXT:    [[TMP2:%.*]] = call ptr @llvm.threadlocal.address.p0(ptr @{{reduction_size[.].+[.]}})
+// CHECK1-NEXT:    [[TMP2:%.*]] = call align 8 ptr @llvm.threadlocal.address.p0(ptr align 8 @{{reduction_size[.].+[.]}})
 // CHECK1-NEXT:    [[TMP3:%.*]] = load i64, ptr [[TMP2]], align 8
 // CHECK1-NEXT:    [[TMP4:%.*]] = load ptr, ptr [[DOTADDR]], align 8
 // CHECK1-NEXT:    [[TMP5:%.*]] = load ptr, ptr [[DOTADDR1]], align 8
diff --git a/clang/test/OpenMP/parallel_master_reduction_task_codegen.cpp b/clang/test/OpenMP/parallel_master_reduction_task_codegen.cpp
--- a/clang/test/OpenMP/parallel_master_reduction_task_codegen.cpp
+++ b/clang/test/OpenMP/parallel_master_reduction_task_codegen.cpp
@@ -311,7 +311,7 @@
 // CHECK1-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR]], align 8
 // CHECK1-NEXT:    store ptr [[TMP1]], ptr [[DOTADDR1]], align 8
 // CHECK1-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTADDR]], align 8
-// CHECK1-NEXT:    [[TMP3:%.*]] = call ptr @llvm.threadlocal.address.p0(ptr @{{reduction_size[.].+[.]}})
+// CHECK1-NEXT:    [[TMP3:%.*]] = call align 8 ptr @llvm.threadlocal.address.p0(ptr align 8 @{{reduction_size[.].+[.]}})
 // CHECK1-NEXT:    [[TMP4:%.*]] = load i64, ptr [[TMP3]], align 8
 // CHECK1-NEXT:    [[TMP5:%.*]] = getelementptr i8, ptr [[TMP2]], i64 [[TMP4]]
 // CHECK1-NEXT:    [[OMP_ARRAYINIT_ISEMPTY:%.*]] = icmp eq ptr [[TMP2]], [[TMP5]]
@@ -333,7 +333,7 @@
 // CHECK1-NEXT:    [[DOTADDR1:%.*]] = alloca ptr, align 8
 // CHECK1-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR]], align 8
 // CHECK1-NEXT:    store ptr [[TMP1]], ptr [[DOTADDR1]], align 8
-// CHECK1-NEXT:    [[TMP2:%.*]] = call ptr @llvm.threadlocal.address.p0(ptr @{{reduction_size[.].+[.]}})
+// CHECK1-NEXT:    [[TMP2:%.*]] = call align 8 ptr @llvm.threadlocal.address.p0(ptr align 8 @{{reduction_size[.].+[.]}})
 // CHECK1-NEXT:    [[TMP3:%.*]] = load i64, ptr [[TMP2]], align 8
 // CHECK1-NEXT:    [[TMP4:%.*]] = load ptr, ptr [[DOTADDR]], align 8
 // CHECK1-NEXT:    [[TMP5:%.*]] = load ptr, ptr [[DOTADDR1]], align 8
diff --git a/clang/test/OpenMP/parallel_reduction_task_codegen.cpp b/clang/test/OpenMP/parallel_reduction_task_codegen.cpp
--- a/clang/test/OpenMP/parallel_reduction_task_codegen.cpp
+++ b/clang/test/OpenMP/parallel_reduction_task_codegen.cpp
@@ -302,7 +302,7 @@
 // CHECK1-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR]], align 8
 // CHECK1-NEXT:    store ptr [[TMP1]], ptr [[DOTADDR1]], align 8
 // CHECK1-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTADDR]], align 8
-// CHECK1-NEXT:    [[TMP3:%.*]] = call ptr @llvm.threadlocal.address.p0(ptr @{{reduction_size[.].+[.]}})
+// CHECK1-NEXT:    [[TMP3:%.*]] = call align 8 ptr @llvm.threadlocal.address.p0(ptr align 8 @{{reduction_size[.].+[.]}})
 // CHECK1-NEXT:    [[TMP4:%.*]] = load i64, ptr [[TMP3]], align 8
 // CHECK1-NEXT:    [[TMP5:%.*]] = getelementptr i8, ptr [[TMP2]], i64 [[TMP4]]
 // CHECK1-NEXT:    [[OMP_ARRAYINIT_ISEMPTY:%.*]] = icmp eq ptr [[TMP2]], [[TMP5]]
@@ -324,7 +324,7 @@
 // CHECK1-NEXT:    [[DOTADDR1:%.*]] = alloca ptr, align 8
 // CHECK1-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR]], align 8
 // CHECK1-NEXT:    store ptr [[TMP1]], ptr [[DOTADDR1]], align 8
-// CHECK1-NEXT:    [[TMP2:%.*]] = call ptr @llvm.threadlocal.address.p0(ptr @{{reduction_size[.].+[.]}})
+// CHECK1-NEXT:    [[TMP2:%.*]] = call align 8 ptr @llvm.threadlocal.address.p0(ptr align 8 @{{reduction_size[.].+[.]}})
 // CHECK1-NEXT:    [[TMP3:%.*]] = load i64, ptr [[TMP2]], align 8
 // CHECK1-NEXT:    [[TMP4:%.*]] = load ptr, ptr [[DOTADDR]], align 8
 // CHECK1-NEXT:    [[TMP5:%.*]] = load ptr, ptr [[DOTADDR1]], align 8
diff --git a/clang/test/OpenMP/parallel_sections_reduction_task_codegen.cpp b/clang/test/OpenMP/parallel_sections_reduction_task_codegen.cpp
--- a/clang/test/OpenMP/parallel_sections_reduction_task_codegen.cpp
+++ b/clang/test/OpenMP/parallel_sections_reduction_task_codegen.cpp
@@ -344,7 +344,7 @@
 // CHECK1-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR]], align 8
 // CHECK1-NEXT:    store ptr [[TMP1]], ptr [[DOTADDR1]], align 8
 // CHECK1-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTADDR]], align 8
-// CHECK1-NEXT:    [[TMP3:%.*]] = call ptr @llvm.threadlocal.address.p0(ptr @{{reduction_size[.].+[.]}})
+// CHECK1-NEXT:    [[TMP3:%.*]] = call align 8 ptr @llvm.threadlocal.address.p0(ptr align 8 @{{reduction_size[.].+[.]}})
 // CHECK1-NEXT:    [[TMP4:%.*]] = load i64, ptr [[TMP3]], align 8
 // CHECK1-NEXT:    [[TMP5:%.*]] = getelementptr i8, ptr [[TMP2]], i64 [[TMP4]]
 // CHECK1-NEXT:    [[OMP_ARRAYINIT_ISEMPTY:%.*]] = icmp eq ptr [[TMP2]], [[TMP5]]
@@ -366,7 +366,7 @@
 // CHECK1-NEXT:    [[DOTADDR1:%.*]] = alloca ptr, align 8
 // CHECK1-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR]], align 8
 // CHECK1-NEXT:    store ptr [[TMP1]], ptr [[DOTADDR1]], align 8
-// CHECK1-NEXT:    [[TMP2:%.*]] = call ptr @llvm.threadlocal.address.p0(ptr @{{reduction_size[.].+[.]}})
+// CHECK1-NEXT:    [[TMP2:%.*]] = call align 8 ptr @llvm.threadlocal.address.p0(ptr align 8 @{{reduction_size[.].+[.]}})
 // CHECK1-NEXT:    [[TMP3:%.*]] = load i64, ptr [[TMP2]], align 8
 // CHECK1-NEXT:    [[TMP4:%.*]] = load ptr, ptr [[DOTADDR]], align 8
 // CHECK1-NEXT:    [[TMP5:%.*]] = load ptr, ptr [[DOTADDR1]], align 8
diff --git a/clang/test/OpenMP/reduction_implicit_map.cpp b/clang/test/OpenMP/reduction_implicit_map.cpp
--- a/clang/test/OpenMP/reduction_implicit_map.cpp
+++ b/clang/test/OpenMP/reduction_implicit_map.cpp
@@ -704,7 +704,7 @@
 // CHECK1-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR]], align 8
 // CHECK1-NEXT:    store ptr [[TMP1]], ptr [[DOTADDR1]], align 8
 // CHECK1-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTADDR]], align 8
-// CHECK1-NEXT:    [[TMP3:%.*]] = call ptr @llvm.threadlocal.address.p0(ptr @{{reduction_size[.].+[.]}})
+// CHECK1-NEXT:    [[TMP3:%.*]] = call align 8 ptr @llvm.threadlocal.address.p0(ptr align 8 @{{reduction_size[.].+[.]}})
 // CHECK1-NEXT:    [[TMP4:%.*]] = load i64, ptr [[TMP3]], align 8
 // CHECK1-NEXT:    [[TMP5:%.*]] = getelementptr double, ptr [[TMP2]], i64 [[TMP4]]
 // CHECK1-NEXT:    [[OMP_ARRAYINIT_ISEMPTY:%.*]] = icmp eq ptr [[TMP2]], [[TMP5]]
@@ -726,7 +726,7 @@
 // CHECK1-NEXT:    [[DOTADDR1:%.*]] = alloca ptr, align 8
 // CHECK1-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR]], align 8
 // CHECK1-NEXT:    store ptr [[TMP1]], ptr [[DOTADDR1]], align 8
-// CHECK1-NEXT:    [[TMP2:%.*]] = call ptr @llvm.threadlocal.address.p0(ptr @{{reduction_size[.].+[.]}})
+// CHECK1-NEXT:    [[TMP2:%.*]] = call align 8 ptr @llvm.threadlocal.address.p0(ptr align 8 @{{reduction_size[.].+[.]}})
 // CHECK1-NEXT:    [[TMP3:%.*]] = load i64, ptr [[TMP2]], align 8
 // CHECK1-NEXT:    [[TMP4:%.*]] = load ptr, ptr [[DOTADDR]], align 8
 // CHECK1-NEXT:    [[TMP5:%.*]] = load ptr, ptr [[DOTADDR1]], align 8
diff --git a/clang/test/OpenMP/sections_reduction_task_codegen.cpp b/clang/test/OpenMP/sections_reduction_task_codegen.cpp
--- a/clang/test/OpenMP/sections_reduction_task_codegen.cpp
+++ b/clang/test/OpenMP/sections_reduction_task_codegen.cpp
@@ -349,7 +349,7 @@
 // CHECK1-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR]], align 8
 // CHECK1-NEXT:    store ptr [[TMP1]], ptr [[DOTADDR1]], align 8
 // CHECK1-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTADDR]], align 8
-// CHECK1-NEXT:    [[TMP3:%.*]] = call ptr @llvm.threadlocal.address.p0(ptr @{{reduction_size[.].+[.]}})
+// CHECK1-NEXT:    [[TMP3:%.*]] = call align 8 ptr @llvm.threadlocal.address.p0(ptr align 8 @{{reduction_size[.].+[.]}})
 // CHECK1-NEXT:    [[TMP4:%.*]] = load i64, ptr [[TMP3]], align 8
 // CHECK1-NEXT:    [[TMP5:%.*]] = getelementptr i8, ptr [[TMP2]], i64 [[TMP4]]
 // CHECK1-NEXT:    [[OMP_ARRAYINIT_ISEMPTY:%.*]] = icmp eq ptr [[TMP2]], [[TMP5]]
@@ -371,7 +371,7 @@
 // CHECK1-NEXT:    [[DOTADDR1:%.*]] = alloca ptr, align 8
 // CHECK1-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR]], align 8
 // CHECK1-NEXT:    store ptr [[TMP1]], ptr [[DOTADDR1]], align 8
-// CHECK1-NEXT:    [[TMP2:%.*]] = call ptr @llvm.threadlocal.address.p0(ptr @{{reduction_size[.].+[.]}})
+// CHECK1-NEXT:    [[TMP2:%.*]] = call align 8 ptr @llvm.threadlocal.address.p0(ptr align 8 @{{reduction_size[.].+[.]}})
 // CHECK1-NEXT:    [[TMP3:%.*]] = load i64, ptr [[TMP2]], align 8
 // CHECK1-NEXT:    [[TMP4:%.*]] = load ptr, ptr [[DOTADDR]], align 8
 // CHECK1-NEXT:    [[TMP5:%.*]] = load ptr, ptr [[DOTADDR1]], align 8
diff --git a/clang/test/OpenMP/target_parallel_for_reduction_task_codegen.cpp b/clang/test/OpenMP/target_parallel_for_reduction_task_codegen.cpp
--- a/clang/test/OpenMP/target_parallel_for_reduction_task_codegen.cpp
+++ b/clang/test/OpenMP/target_parallel_for_reduction_task_codegen.cpp
@@ -369,7 +369,7 @@
 // CHECK1-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR]], align 8
 // CHECK1-NEXT:    store ptr [[TMP1]], ptr [[DOTADDR1]], align 8
 // CHECK1-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTADDR]], align 8
-// CHECK1-NEXT:    [[TMP3:%.*]] = call ptr @llvm.threadlocal.address.p0(ptr @{{reduction_size[.].+[.]}})
+// CHECK1-NEXT:    [[TMP3:%.*]] = call align 8 ptr @llvm.threadlocal.address.p0(ptr align 8 @{{reduction_size[.].+[.]}})
 // CHECK1-NEXT:    [[TMP4:%.*]] = load i64, ptr [[TMP3]], align 8
 // CHECK1-NEXT:    [[TMP5:%.*]] = getelementptr i8, ptr [[TMP2]], i64 [[TMP4]]
 // CHECK1-NEXT:    [[OMP_ARRAYINIT_ISEMPTY:%.*]] = icmp eq ptr [[TMP2]], [[TMP5]]
@@ -391,7 +391,7 @@
 // CHECK1-NEXT:    [[DOTADDR1:%.*]] = alloca ptr, align 8
 // CHECK1-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR]], align 8
 // CHECK1-NEXT:    store ptr [[TMP1]], ptr [[DOTADDR1]], align 8
-// CHECK1-NEXT:    [[TMP2:%.*]] = call ptr @llvm.threadlocal.address.p0(ptr @{{reduction_size[.].+[.]}})
+// CHECK1-NEXT:    [[TMP2:%.*]] = call align 8 ptr @llvm.threadlocal.address.p0(ptr align 8 @{{reduction_size[.].+[.]}})
 // CHECK1-NEXT:    [[TMP3:%.*]] = load i64, ptr [[TMP2]], align 8
 // CHECK1-NEXT:    [[TMP4:%.*]] = load ptr, ptr [[DOTADDR]], align 8
 // CHECK1-NEXT:    [[TMP5:%.*]] = load ptr, ptr [[DOTADDR1]], align 8
diff --git a/clang/test/OpenMP/target_parallel_reduction_task_codegen.cpp b/clang/test/OpenMP/target_parallel_reduction_task_codegen.cpp
--- a/clang/test/OpenMP/target_parallel_reduction_task_codegen.cpp
+++ b/clang/test/OpenMP/target_parallel_reduction_task_codegen.cpp
@@ -315,7 +315,7 @@
 // CHECK1-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR]], align 8
 // CHECK1-NEXT:    store ptr [[TMP1]], ptr [[DOTADDR1]], align 8
 // CHECK1-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTADDR]], align 8
-// CHECK1-NEXT:    [[TMP3:%.*]] = call ptr @llvm.threadlocal.address.p0(ptr @{{reduction_size[.].+[.]}})
+// CHECK1-NEXT:    [[TMP3:%.*]] = call align 8 ptr @llvm.threadlocal.address.p0(ptr align 8 @{{reduction_size[.].+[.]}})
 // CHECK1-NEXT:    [[TMP4:%.*]] = load i64, ptr [[TMP3]], align 8
 // CHECK1-NEXT:    [[TMP5:%.*]] = getelementptr i8, ptr [[TMP2]], i64 [[TMP4]]
 // CHECK1-NEXT:    [[OMP_ARRAYINIT_ISEMPTY:%.*]] = icmp eq ptr [[TMP2]], [[TMP5]]
@@ -337,7 +337,7 @@
 // CHECK1-NEXT:    [[DOTADDR1:%.*]] = alloca ptr, align 8
 // CHECK1-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR]], align 8
 // CHECK1-NEXT:    store ptr [[TMP1]], ptr [[DOTADDR1]], align 8
-// CHECK1-NEXT:    [[TMP2:%.*]] = call ptr @llvm.threadlocal.address.p0(ptr @{{reduction_size[.].+[.]}})
+// CHECK1-NEXT:    [[TMP2:%.*]] = call align 8 ptr @llvm.threadlocal.address.p0(ptr align 8 @{{reduction_size[.].+[.]}})
 // CHECK1-NEXT:    [[TMP3:%.*]] = load i64, ptr [[TMP2]], align 8
 // CHECK1-NEXT:    [[TMP4:%.*]] = load ptr, ptr [[DOTADDR]], align 8
 // CHECK1-NEXT:    [[TMP5:%.*]] = load ptr, ptr [[DOTADDR1]], align 8
diff --git a/clang/test/OpenMP/target_teams_distribute_parallel_for_reduction_task_codegen.cpp b/clang/test/OpenMP/target_teams_distribute_parallel_for_reduction_task_codegen.cpp
--- a/clang/test/OpenMP/target_teams_distribute_parallel_for_reduction_task_codegen.cpp
+++ b/clang/test/OpenMP/target_teams_distribute_parallel_for_reduction_task_codegen.cpp
@@ -344,7 +344,7 @@
 // CHECK1-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR]], align 8
 // CHECK1-NEXT:    store ptr [[TMP1]], ptr [[DOTADDR1]], align 8
 // CHECK1-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTADDR]], align 8
-// CHECK1-NEXT:    [[TMP3:%.*]] = call ptr @llvm.threadlocal.address.p0(ptr @{{reduction_size[.].+[.]}})
+// CHECK1-NEXT:    [[TMP3:%.*]] = call align 8 ptr @llvm.threadlocal.address.p0(ptr align 8 @{{reduction_size[.].+[.]}})
 // CHECK1-NEXT:    [[TMP4:%.*]] = load i64, ptr [[TMP3]], align 8
 // CHECK1-NEXT:    [[TMP5:%.*]] = getelementptr i8, ptr [[TMP2]], i64 [[TMP4]]
 // CHECK1-NEXT:    [[OMP_ARRAYINIT_ISEMPTY:%.*]] = icmp eq ptr [[TMP2]], [[TMP5]]
@@ -366,7 +366,7 @@
 // CHECK1-NEXT:    [[DOTADDR1:%.*]] = alloca ptr, align 8
 // CHECK1-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR]], align 8
 // CHECK1-NEXT:    store ptr [[TMP1]], ptr [[DOTADDR1]], align 8
-// CHECK1-NEXT:    [[TMP2:%.*]] = call ptr @llvm.threadlocal.address.p0(ptr @{{reduction_size[.].+[.]}})
+// CHECK1-NEXT:    [[TMP2:%.*]] = call align 8 ptr @llvm.threadlocal.address.p0(ptr align 8 @{{reduction_size[.].+[.]}})
 // CHECK1-NEXT:    [[TMP3:%.*]] = load i64, ptr [[TMP2]], align 8
 // CHECK1-NEXT:    [[TMP4:%.*]] = load ptr, ptr [[DOTADDR]], align 8
 // CHECK1-NEXT:    [[TMP5:%.*]] = load ptr, ptr [[DOTADDR1]], align 8
@@ -708,7 +708,7 @@
 // CHECK1-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR]], align 8
 // CHECK1-NEXT:    store ptr [[TMP1]], ptr [[DOTADDR1]], align 8
 // CHECK1-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTADDR]], align 8
-// CHECK1-NEXT:    [[TMP3:%.*]] = call ptr @llvm.threadlocal.address.p0(ptr @{{reduction_size[.].+[.]}})
+// CHECK1-NEXT:    [[TMP3:%.*]] = call align 8 ptr @llvm.threadlocal.address.p0(ptr align 8 @{{reduction_size[.].+[.]}})
 // CHECK1-NEXT:    [[TMP4:%.*]] = load i64, ptr [[TMP3]], align 8
 // CHECK1-NEXT:    [[TMP5:%.*]] = getelementptr i8, ptr [[TMP2]], i64 [[TMP4]]
 // CHECK1-NEXT:    [[OMP_ARRAYINIT_ISEMPTY:%.*]] = icmp eq ptr [[TMP2]], [[TMP5]]
@@ -730,7 +730,7 @@
 // CHECK1-NEXT:    [[DOTADDR1:%.*]] = alloca ptr, align 8
 // CHECK1-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR]], align 8
 // CHECK1-NEXT:    store ptr [[TMP1]], ptr [[DOTADDR1]], align 8
-// CHECK1-NEXT:    [[TMP2:%.*]] = call ptr @llvm.threadlocal.address.p0(ptr @{{reduction_size[.].+[.]}})
+// CHECK1-NEXT:    [[TMP2:%.*]] = call align 8 ptr @llvm.threadlocal.address.p0(ptr align 8 @{{reduction_size[.].+[.]}})
 // CHECK1-NEXT:    [[TMP3:%.*]] = load i64, ptr [[TMP2]], align 8
 // CHECK1-NEXT:    [[TMP4:%.*]] = load ptr, ptr [[DOTADDR]], align 8
 // CHECK1-NEXT:    [[TMP5:%.*]] = load ptr, ptr [[DOTADDR1]], align 8
diff --git a/clang/test/OpenMP/taskloop_reduction_codegen.cpp b/clang/test/OpenMP/taskloop_reduction_codegen.cpp
--- a/clang/test/OpenMP/taskloop_reduction_codegen.cpp
+++ b/clang/test/OpenMP/taskloop_reduction_codegen.cpp
@@ -175,13 +175,13 @@
 // CHECK: call void @llvm.memcpy.p0.p0.i64(
 
 // CHECK: define internal void @[[RED_FINI2]](ptr noundef %0)
-// CHECK: [[RED_SIZE1_ADDR:%.+]] = call ptr @llvm.threadlocal.address.p0(ptr [[RED_SIZE1]]
+// CHECK: [[RED_SIZE1_ADDR:%.+]] = call align 8 ptr @llvm.threadlocal.address.p0(ptr align 8 [[RED_SIZE1]]
 // CHECK: load i64, ptr [[RED_SIZE1_ADDR]]
 // CHECK: call void @
 // CHECK: ret void
 
 // CHECK: define internal void @[[RED_COMB2]](ptr noundef %0, ptr noundef %1)
-// CHECK: [[RED_SIZE1_ADDR2:%.+]] = call ptr @llvm.threadlocal.address.p0(ptr [[RED_SIZE1]]
+// CHECK: [[RED_SIZE1_ADDR2:%.+]] = call align 8 ptr @llvm.threadlocal.address.p0(ptr align 8 [[RED_SIZE1]]
 // CHECK: load i64, ptr [[RED_SIZE1_ADDR2]]
 // CHECK: call void [[OMP_COMB1]](
 // CHECK: ret void
@@ -196,13 +196,13 @@
 // CHECK: ret void
 
 // CHECK: define internal void @[[RED_INIT4]](ptr noalias noundef %{{.+}}, ptr noalias noundef %{{.+}})
-// CHECK: [[RED_SIZE2_ADDR:%.+]] = call ptr @llvm.threadlocal.address.p0(ptr [[RED_SIZE2]]
+// CHECK: [[RED_SIZE2_ADDR:%.+]] = call align 8 ptr @llvm.threadlocal.address.p0(ptr align 8 [[RED_SIZE2]]
 // CHECK: load i64, ptr [[RED_SIZE2_ADDR]]
 // CHECK: store float 0.000000e+00, ptr %
 // CHECK: ret void
 
 // CHECK: define internal void @[[RED_COMB4]](ptr noundef %0, ptr noundef %1)
-// CHECK: [[RED_SIZE2_ADDR2:%.+]] = call ptr @llvm.threadlocal.address.p0(ptr [[RED_SIZE2]]
+// CHECK: [[RED_SIZE2_ADDR2:%.+]] = call align 8 ptr @llvm.threadlocal.address.p0(ptr align 8 [[RED_SIZE2]]
 // CHECK: load i64, ptr [[RED_SIZE2_ADDR2]]
 // CHECK: fadd float %
 // CHECK: store float %{{.+}}, ptr %
diff --git a/clang/test/OpenMP/teams_distribute_parallel_for_reduction_task_codegen.cpp b/clang/test/OpenMP/teams_distribute_parallel_for_reduction_task_codegen.cpp
--- a/clang/test/OpenMP/teams_distribute_parallel_for_reduction_task_codegen.cpp
+++ b/clang/test/OpenMP/teams_distribute_parallel_for_reduction_task_codegen.cpp
@@ -348,7 +348,7 @@
 // CHECK1-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR]], align 8
 // CHECK1-NEXT:    store ptr [[TMP1]], ptr [[DOTADDR1]], align 8
 // CHECK1-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTADDR]], align 8
-// CHECK1-NEXT:    [[TMP3:%.*]] = call ptr @llvm.threadlocal.address.p0(ptr @{{reduction_size[.].+[.]}})
+// CHECK1-NEXT:    [[TMP3:%.*]] = call align 8 ptr @llvm.threadlocal.address.p0(ptr align 8 @{{reduction_size[.].+[.]}})
 // CHECK1-NEXT:    [[TMP4:%.*]] = load i64, ptr [[TMP3]], align 8
 // CHECK1-NEXT:    [[TMP5:%.*]] = getelementptr i8, ptr [[TMP2]], i64 [[TMP4]]
 // CHECK1-NEXT:    [[OMP_ARRAYINIT_ISEMPTY:%.*]] = icmp eq ptr [[TMP2]], [[TMP5]]
@@ -370,7 +370,7 @@
 // CHECK1-NEXT:    [[DOTADDR1:%.*]] = alloca ptr, align 8
 // CHECK1-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR]], align 8
 // CHECK1-NEXT:    store ptr [[TMP1]], ptr [[DOTADDR1]], align 8
-// CHECK1-NEXT:    [[TMP2:%.*]] = call ptr @llvm.threadlocal.address.p0(ptr @{{reduction_size[.].+[.]}})
+// CHECK1-NEXT:    [[TMP2:%.*]] = call align 8 ptr @llvm.threadlocal.address.p0(ptr align 8 @{{reduction_size[.].+[.]}})
 // CHECK1-NEXT:    [[TMP3:%.*]] = load i64, ptr [[TMP2]], align 8
 // CHECK1-NEXT:    [[TMP4:%.*]] = load ptr, ptr [[DOTADDR]], align 8
 // CHECK1-NEXT:    [[TMP5:%.*]] = load ptr, ptr [[DOTADDR1]], align 8
@@ -712,7 +712,7 @@
 // CHECK1-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR]], align 8
 // CHECK1-NEXT:    store ptr [[TMP1]], ptr [[DOTADDR1]], align 8
 // CHECK1-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTADDR]], align 8
-// CHECK1-NEXT:    [[TMP3:%.*]] = call ptr @llvm.threadlocal.address.p0(ptr @{{reduction_size[.].+[.]}})
+// CHECK1-NEXT:    [[TMP3:%.*]] = call align 8 ptr @llvm.threadlocal.address.p0(ptr align 8 @{{reduction_size[.].+[.]}})
 // CHECK1-NEXT:    [[TMP4:%.*]] = load i64, ptr [[TMP3]], align 8
 // CHECK1-NEXT:    [[TMP5:%.*]] = getelementptr i8, ptr [[TMP2]], i64 [[TMP4]]
 // CHECK1-NEXT:    [[OMP_ARRAYINIT_ISEMPTY:%.*]] = icmp eq ptr [[TMP2]], [[TMP5]]
@@ -734,7 +734,7 @@
 // CHECK1-NEXT:    [[DOTADDR1:%.*]] = alloca ptr, align 8
 // CHECK1-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR]], align 8
 // CHECK1-NEXT:    store ptr [[TMP1]], ptr [[DOTADDR1]], align 8
-// CHECK1-NEXT:    [[TMP2:%.*]] = call ptr @llvm.threadlocal.address.p0(ptr @{{reduction_size[.].+[.]}})
+// CHECK1-NEXT:    [[TMP2:%.*]] = call align 8 ptr @llvm.threadlocal.address.p0(ptr align 8 @{{reduction_size[.].+[.]}})
 // CHECK1-NEXT:    [[TMP3:%.*]] = load i64, ptr [[TMP2]], align 8
 // CHECK1-NEXT:    [[TMP4:%.*]] = load ptr, ptr [[DOTADDR]], align 8
 // CHECK1-NEXT:    [[TMP5:%.*]] = load ptr, ptr [[DOTADDR1]], align 8
diff --git a/llvm/lib/Frontend/OpenMP/OMPIRBuilder.cpp b/llvm/lib/Frontend/OpenMP/OMPIRBuilder.cpp
--- a/llvm/lib/Frontend/OpenMP/OMPIRBuilder.cpp
+++ b/llvm/lib/Frontend/OpenMP/OMPIRBuilder.cpp
@@ -4225,10 +4225,12 @@
     // variable for possibly changing that to internal or private, or maybe
     // create different versions of the function for different OMP internal
     // variables.
-    Elem.second = new GlobalVariable(
+    auto *GV = new GlobalVariable(
         M, Ty, /*IsConstant=*/false, GlobalValue::CommonLinkage,
         Constant::getNullValue(Ty), Elem.first(),
         /*InsertBefore=*/nullptr, GlobalValue::NotThreadLocal, AddressSpace);
+    GV->setAlignment(M.getDataLayout().getABITypeAlign(Ty));
+    Elem.second = GV;
   }
 
   return cast<GlobalVariable>(&*Elem.second);