[libomptarget-nvptx] reduction: Determine if runtime uninitialized

[Hahnfeld]

Summary

Pass in the correct value of isRuntimeUninitialized() which solves
parallel reductions as reported on the mailing list.
For reference: r333285 did the same for loop scheduling.

[ABataev]

LG

[grokos]

Looks good. Thanks for taking care of this issue!

libomptarget/deviceRTLs/nvptx/src/reduction.cu

 }
 
 INLINE
 int32_t nvptx_parallel_reduce_nowait(int32_t global_tid, int32_t num_vars,
                                      size_t reduce_size, void *reduce_data,
                                      kmp_ShuffleReductFctPtr shflFct,
                                      kmp_InterWarpCopyFctPtr cpyFct,
                                      bool isSPMDExecutionMode,
-                                     bool isRuntimeUninitialized = false) {
+                                     bool isRuntimeUninitialized) {
   uint32_t BlockThreadId = GetLogicalThreadIdInBlock();
   uint32_t NumThreads = GetNumberOfOmpThreads(
       BlockThreadId, isSPMDExecutionMode, isRuntimeUninitialized);
   if (NumThreads == 1)
     return 1;
   /*
    * This reduce function handles reduction within a team. It handles
    * parallel regions in both L1 and L2 parallelism levels. It also

 }
 
 EXTERN
 int32_t __kmpc_nvptx_parallel_reduce_nowait(
     int32_t global_tid, int32_t num_vars, size_t reduce_size, void *reduce_data,
     kmp_ShuffleReductFctPtr shflFct, kmp_InterWarpCopyFctPtr cpyFct) {
   return nvptx_parallel_reduce_nowait(global_tid, num_vars, reduce_size,
                                       reduce_data, shflFct, cpyFct,
-                                      /*isSPMDExecutionMode=*/isSPMDMode());
+                                      /*isSPMDExecutionMode=*/isSPMDMode(),
+                                      /*isRuntimeUninitialized=*/isRuntimeUninitialized());
 }
 
 EXTERN
 int32_t __kmpc_nvptx_parallel_reduce_nowait_simple_spmd(
     int32_t global_tid, int32_t num_vars, size_t reduce_size, void *reduce_data,
     kmp_ShuffleReductFctPtr shflFct, kmp_InterWarpCopyFctPtr cpyFct) {
   return nvptx_parallel_reduce_nowait(global_tid, num_vars, reduce_size,
                                       reduce_data, shflFct, cpyFct,
return nvptx_parallel_reduce_nowait(global_tid, num_vars, reduce_size,
                                       /*isRuntimeUninitialized=*/true);
 }
 
 INLINE
 int32_t nvptx_teams_reduce_nowait(
     int32_t global_tid, int32_t num_vars, size_t reduce_size, void *reduce_data,
     kmp_ShuffleReductFctPtr shflFct, kmp_InterWarpCopyFctPtr cpyFct,
     kmp_CopyToScratchpadFctPtr scratchFct, kmp_LoadReduceFctPtr ldFct,
-    bool isSPMDExecutionMode, bool isRuntimeUninitialized = false) {
+    bool isSPMDExecutionMode, bool isRuntimeUninitialized) {
   uint32_t ThreadId = GetLogicalThreadIdInBlock();
   // In non-generic mode all workers participate in the teams reduction.
   // In generic mode only the team master participates in the teams
   // reduction because the workers are waiting for parallel work.
   uint32_t NumThreads =
       isSPMDExecutionMode
           ? GetNumberOfOmpThreads(ThreadId, /*isSPMDExecutionMode=*/true,
                                   isRuntimeUninitialized)

 int32_t __kmpc_nvptx_teams_reduce_nowait(int32_t global_tid, int32_t num_vars,
                                          size_t reduce_size, void *reduce_data,
                                          kmp_ShuffleReductFctPtr shflFct,
                                          kmp_InterWarpCopyFctPtr cpyFct,
                                          kmp_CopyToScratchpadFctPtr scratchFct,
                                          kmp_LoadReduceFctPtr ldFct) {
   return nvptx_teams_reduce_nowait(global_tid, num_vars, reduce_size,
                                    reduce_data, shflFct, cpyFct, scratchFct,
-                                   ldFct, /*isSPMDExecutionMode=*/isSPMDMode());
+                                   ldFct, /*isSPMDExecutionMode=*/isSPMDMode(),
+                                   /*isRuntimeUninitialized=*/isRuntimeUninitialized());
 }
 
 EXTERN
 int32_t __kmpc_nvptx_teams_reduce_nowait_simple_spmd(
     int32_t global_tid, int32_t num_vars, size_t reduce_size, void *reduce_data,
     kmp_ShuffleReductFctPtr shflFct, kmp_InterWarpCopyFctPtr cpyFct,
     kmp_CopyToScratchpadFctPtr scratchFct, kmp_LoadReduceFctPtr ldFct) {
   return nvptx_teams_reduce_nowait(global_tid, num_vars, reduce_size,