[OpenMP] Add RTL function to externalization RAII

This patch adds the '_kmpc_get_hardware_num_threads_in_block'
OpenMP RTL function to the externalization RAII struct. This was getting
optimized out and then being replaced with an undefined value once added
back in, causing bugs for complex reductions.

Fixes #53909.

Reviewed By: jdoerfert

Differential Revision: https://reviews.llvm.org/D120076

llvm/lib/Transforms/IPO/OpenMPOpt.cpp

@@ -2119,6 +2119,8 @@ struct OpenMPOpt {
                                        OMPRTL___kmpc_barrier_simple_generic);
     ExternalizationRAII ThreadId(OMPInfoCache,
                                  OMPRTL___kmpc_get_hardware_thread_id_in_block);
+    ExternalizationRAII NumThreads(
+        OMPInfoCache, OMPRTL___kmpc_get_hardware_num_threads_in_block);
     ExternalizationRAII WarpSize(OMPInfoCache, OMPRTL___kmpc_get_warp_size);
 
     registerAAs(IsModulePass);

llvm/test/Transforms/OpenMP/get_hardware_num_threads_in_block_fold.ll

@@ -178,7 +178,16 @@ entry:
   ret void
 }
 
-declare i32 @__kmpc_get_hardware_num_threads_in_block()
+define internal i32 @__kmpc_get_hardware_num_threads_in_block() {
+; CHECK-LABEL: define {{[^@]+}}@__kmpc_get_hardware_num_threads_in_block
+; CHECK-SAME: () #[[ATTR1]] {
+; CHECK-NEXT:    [[RET:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block_dummy()
+; CHECK-NEXT:    ret i32 [[RET]]
+;
+  %ret = call i32 @__kmpc_get_hardware_num_threads_in_block_dummy()
+  ret i32 %ret
+}
+declare i32 @__kmpc_get_hardware_num_threads_in_block_dummy()
 declare i32 @__kmpc_target_init(%struct.ident_t*, i8, i1 zeroext, i1 zeroext) #1
 declare void @__kmpc_target_deinit(%struct.ident_t* nocapture readnone, i8, i1 zeroext) #1
 declare void @__kmpc_parallel_51(%struct.ident_t*, i32, i32, i32, i32, i8*, i8*, i8**, i64)