[OpenMP] Make default parallel for schedule in NVPTX target regions in SPMD mode achieve coalescing

Summary: Set default schedule for parallel for loops to schedule(static, 1) when using SPMD mode on the NVPTX device offloading toolchain to ensure coalescing.

Reviewers: ABataev, Hahnfeld, caomhin

Reviewed By: ABataev

Subscribers: jholewinski, guansong, cfe-commits

Differential Revision: https://reviews.llvm.org/D52629

llvm-svn: 343260
diff --git a/clang/lib/CodeGen/CGOpenMPRuntime.h b/clang/lib/CodeGen/CGOpenMPRuntime.h
index 982aeb3..35f75a9 100644
--- a/clang/lib/CodeGen/CGOpenMPRuntime.h
+++ b/clang/lib/CodeGen/CGOpenMPRuntime.h
@@ -1496,6 +1496,12 @@
       const OMPLoopDirective &S, OpenMPDistScheduleClauseKind &ScheduleKind,
       llvm::Value *&Chunk) const {}
 
+  /// Choose default schedule type and chunk value for the
+  /// schedule clause.
+  virtual void getDefaultScheduleAndChunk(CodeGenFunction &CGF,
+      const OMPLoopDirective &S, OpenMPScheduleClauseKind &ScheduleKind,
+      llvm::Value *&Chunk) const {}
+
   /// Emits call of the outlined function with the provided arguments,
   /// translating these arguments to correct target-specific arguments.
   virtual void
diff --git a/clang/lib/CodeGen/CGOpenMPRuntimeNVPTX.cpp b/clang/lib/CodeGen/CGOpenMPRuntimeNVPTX.cpp
index 56b244d..14fd4a3 100644
--- a/clang/lib/CodeGen/CGOpenMPRuntimeNVPTX.cpp
+++ b/clang/lib/CodeGen/CGOpenMPRuntimeNVPTX.cpp
@@ -4093,3 +4093,14 @@
         S.getIterationVariable()->getType(), S.getBeginLoc());
   }
 }
+
+void CGOpenMPRuntimeNVPTX::getDefaultScheduleAndChunk(
+    CodeGenFunction &CGF, const OMPLoopDirective &S,
+    OpenMPScheduleClauseKind &ScheduleKind,
+    llvm::Value *&Chunk) const {
+  if (getExecutionMode() == CGOpenMPRuntimeNVPTX::EM_SPMD) {
+    ScheduleKind = OMPC_SCHEDULE_static;
+    Chunk = CGF.Builder.getIntN(CGF.getContext().getTypeSize(
+        S.getIterationVariable()->getType()), 1);
+  }
+}
diff --git a/clang/lib/CodeGen/CGOpenMPRuntimeNVPTX.h b/clang/lib/CodeGen/CGOpenMPRuntimeNVPTX.h
index 76343df..fc8cd24 100644
--- a/clang/lib/CodeGen/CGOpenMPRuntimeNVPTX.h
+++ b/clang/lib/CodeGen/CGOpenMPRuntimeNVPTX.h
@@ -340,11 +340,16 @@
   ///
   void functionFinished(CodeGenFunction &CGF) override;
 
-  /// Choose a default value for the schedule clause.
+  /// Choose a default value for the dist_schedule clause.
   void getDefaultDistScheduleAndChunk(CodeGenFunction &CGF,
       const OMPLoopDirective &S, OpenMPDistScheduleClauseKind &ScheduleKind,
       llvm::Value *&Chunk) const override;
 
+  /// Choose a default value for the schedule clause.
+  void getDefaultScheduleAndChunk(CodeGenFunction &CGF,
+      const OMPLoopDirective &S, OpenMPScheduleClauseKind &ScheduleKind,
+      llvm::Value *&Chunk) const override;
+
 private:
   /// Track the execution mode when codegening directives within a target
   /// region. The appropriate mode (SPMD/NON-SPMD) is set on entry to the
diff --git a/clang/lib/CodeGen/CGStmtOpenMP.cpp b/clang/lib/CodeGen/CGStmtOpenMP.cpp
index 4bafb8b..79ffa7c 100644
--- a/clang/lib/CodeGen/CGStmtOpenMP.cpp
+++ b/clang/lib/CodeGen/CGStmtOpenMP.cpp
@@ -2310,6 +2310,10 @@
                                        S.getIterationVariable()->getType(),
                                        S.getBeginLoc());
         }
+      } else {
+        // Default behaviour for schedule clause.
+        CGM.getOpenMPRuntime().getDefaultScheduleAndChunk(
+            *this, S, ScheduleKind.Schedule, Chunk);
       }
       const unsigned IVSize = getContext().getTypeSize(IVExpr->getType());
       const bool IVSigned = IVExpr->getType()->hasSignedIntegerRepresentation();