[OpenMP] Avoid calling isSPMDMode during RT initialization

Until we hit the first barrier we should not call `mapping::isSPMDMode`
with all threads. Instead, we now have (and use during initialization) a
`mapping::isMainThreadInGenericMode` overload that takes the known
SPMD-mode state and one that queries it.

Reviewed By: tianshilei1992

Differential Revision: https://reviews.llvm.org/D111381

Author: jhuber6

openmp/libomptarget/DeviceRTL/include/Mapping.h

@@ -35,6 +35,7 @@ bool isGenericMode();
 
 /// Return true if the executing thread is the main thread in generic mode.
 bool isMainThreadInGenericMode();
+bool isMainThreadInGenericMode(bool IsSPMD);
 
 /// Return true if the executing thread has the lowest Id of the active threads
 /// in the warp.

openmp/libomptarget/DeviceRTL/src/Kernel.cpp

@@ -81,7 +81,7 @@ int32_t __kmpc_target_init(IdentTy *Ident, int8_t Mode,
     return -1;
   }
 
-  if (mapping::isMainThreadInGenericMode())
+  if (mapping::isMainThreadInGenericMode(IsSPMD))
     return -1;
 
   if (UseGenericStateMachine)

openmp/libomptarget/DeviceRTL/src/Mapping.cpp

@@ -165,8 +165,8 @@ uint32_t getWarpSize() { return getGridValue().GV_Warp_Size; }
 } // namespace impl
 } // namespace _OMP
 
-bool mapping::isMainThreadInGenericMode() {
-  if (mapping::isSPMDMode() || icv::Level)
+bool mapping::isMainThreadInGenericMode(bool IsSPMD) {
+  if (IsSPMD || icv::Level)
     return false;
 
   // Check if this is the last warp in the block.
@@ -175,6 +175,10 @@ bool mapping::isMainThreadInGenericMode() {
   return mapping::getThreadIdInBlock() == MainTId;
 }
 
+bool mapping::isMainThreadInGenericMode() {
+  return mapping::isMainThreadInGenericMode(mapping::isSPMDMode());
+}
+
 bool mapping::isLeaderInWarp() {
   __kmpc_impl_lanemask_t Active = mapping::activemask();
   __kmpc_impl_lanemask_t LaneMaskLT = mapping::lanemaskLT();

openmp/libomptarget/DeviceRTL/src/Reduction.cpp

@@ -72,7 +72,7 @@ static int32_t nvptx_parallel_reduce_nowait(int32_t TId, int32_t num_vars,
                                             InterWarpCopyFnTy cpyFct,
                                             bool isSPMDExecutionMode, bool) {
   uint32_t BlockThreadId = mapping::getThreadIdInBlock();
-  if (mapping::isMainThreadInGenericMode())
+  if (mapping::isMainThreadInGenericMode(/* IsSPMD */ false))
     BlockThreadId = 0;
   uint32_t NumThreads = omp_get_num_threads();
   if (NumThreads == 1)