[OPENMP][NVPTX]Basic support for reductions across the teams.

alexey-bataev · alexey-bataev · commit d4de439cf45a · 2018-11-27T21:06:09.000Z
Summary: Added functions __kmpc_nvptx_teams_reduce_nowait_simple and __kmpc_nvptx_teams_end_reduce_nowait_simple to implement basic support for reductions across the teams. Reviewers: gtbercea, kkwli0 Subscribers: guansong, jfb, caomhin, openmp-commits Differential Revision: https://reviews.llvm.org/D54967 llvm-svn: 347710
diff --git a/openmp/libomptarget/deviceRTLs/nvptx/src/interface.h b/openmp/libomptarget/deviceRTLs/nvptx/src/interface.h
@@ -419,6 +419,12 @@ EXTERN int32_t __kmpc_nvptx_teams_reduce_nowait_simple_generic(
     int32_t global_tid, int32_t num_vars, size_t reduce_size, void *reduce_data,
     kmp_ShuffleReductFctPtr shflFct, kmp_InterWarpCopyFctPtr cpyFct,
     kmp_CopyToScratchpadFctPtr sratchFct, kmp_LoadReduceFctPtr ldFct);
+EXTERN int32_t __kmpc_nvptx_teams_reduce_nowait_simple(kmp_Ident *loc,
+                                                       int32_t global_tid,
+                                                       kmp_CriticalName *crit);
+EXTERN void __kmpc_nvptx_teams_end_reduce_nowait_simple(kmp_Ident *loc,
+                                                        int32_t global_tid,
+                                                        kmp_CriticalName *crit);
 EXTERN int32_t __kmpc_shuffle_int32(int32_t val, int16_t delta, int16_t size);
 EXTERN int64_t __kmpc_shuffle_int64(int64_t val, int16_t delta, int16_t size);
 
diff --git a/openmp/libomptarget/deviceRTLs/nvptx/src/reduction.cu b/openmp/libomptarget/deviceRTLs/nvptx/src/reduction.cu
@@ -430,3 +430,21 @@ int32_t __kmpc_nvptx_teams_reduce_nowait_simple_generic(
                                    /*isSPMDExecutionMode=*/false,
                                    /*isRuntimeUninitialized=*/true);
 }
+
+EXTERN int32_t __kmpc_nvptx_teams_reduce_nowait_simple(kmp_Ident *loc,
+                                                       int32_t global_tid,
+                                                       kmp_CriticalName *crit) {
+  if (checkSPMDMode(loc) && GetThreadIdInBlock() != 0)
+    return 0;
+  // The master thread of the team actually does the reduction.
+  while (atomicCAS((uint32_t *)crit, 0, 1))
+    ;
+  return 1;
+}
+
+EXTERN void
+__kmpc_nvptx_teams_end_reduce_nowait_simple(kmp_Ident *loc, int32_t global_tid,
+                                            kmp_CriticalName *crit) {
+  (void)atomicExch((uint32_t *)crit, 0);
+}
+
