[OpenMP][DeviceRTL] Fix the support for tasking on the device

[tianshilei1992]

Summary

This patch fixes the support for tasking on the device.

Note: AMDGPU doesn't support it yet because of no support for malloc and free.

Fix #59946.

➜  ./test_parallel_master_device
[OMPVV_RESULT: test_parallel_master_device.c] Test passed on the device.

[jdoerfert]

LG

openmp/libomptarget/DeviceRTL/include/Interface.h

 /// TODO
 uint16_t __kmpc_parallel_level(IdentTy *Loc, uint32_t);
 
 ///}
 
 /// Tasking
 ///
 ///{
-TaskDescriptorTy *__kmpc_omp_task_alloc(IdentTy *, uint32_t, int32_t,
-                                        uint32_t TaskSizeInclPrivateValues,
-                                        uint32_t SharedValuesSize,
+TaskDescriptorTy *__kmpc_omp_task_alloc(IdentTy *, int32_t, int32_t,
+                                        size_t TaskSizeInclPrivateValues,
+                                        size_t SharedValuesSize,
                                         TaskFnTy TaskFn);
 
 int32_t __kmpc_omp_task(IdentTy *Loc, uint32_t TId,
                         TaskDescriptorTy *TaskDescriptor);
 
 int32_t __kmpc_omp_task_with_deps(IdentTy *Loc, uint32_t TId,
                                   TaskDescriptorTy *TaskDescriptor, int32_t,
                                   void *, int32_t, void *);

openmp/libomptarget/DeviceRTL/src/Tasking.cpp

 #include "State.h"
 #include "Types.h"
 #include "Utils.h"
 
 using namespace ompx;
 
 #pragma omp begin declare target device_type(nohost)
 
-TaskDescriptorTy *__kmpc_omp_task_alloc(IdentTy *, uint32_t, int32_t,
-                                        uint64_t TaskSizeInclPrivateValues,
-                                        uint64_t SharedValuesSize,
+extern "C" {
+
+TaskDescriptorTy *__kmpc_omp_task_alloc(IdentTy *, int32_t, int32_t,
+                                        size_t TaskSizeInclPrivateValues,
+                                        size_t SharedValuesSize,
                                         TaskFnTy TaskFn) {
   FunctionTracingRAII();
   auto TaskSizeInclPrivateValuesPadded =
       utils::roundUp(TaskSizeInclPrivateValues, uint64_t(sizeof(void *)));
   auto TaskSizeTotal = TaskSizeInclPrivateValuesPadded + SharedValuesSize;
   TaskDescriptorTy *TaskDescriptor = (TaskDescriptorTy *)memory::allocGlobal(
       TaskSizeTotal, "explicit task descriptor");
   TaskDescriptor->Payload =
int omp_in_final(void) {
   // treat all tasks as final... Specs may expect runtime to keep
   // track more precisely if a task was actively set by users... This
   // is not explicitly specified; will treat as if runtime can
   // actively decide to put a non-final task into a final one.
   return 1;
 }
 
 int omp_get_max_task_priority(void) { return 0; }
+}
 
 #pragma omp end declare target