[OPENMP] Fix codegen for lastprivate loop counters.

Patch fixes bug with codegen for lastprivate loop counters. Also it may
improve performance for lastprivates calculations in some cases.

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@261209 91177308-0d34-0410-b5e6-96231b3b80d8

lib/CodeGen/CGStmtOpenMP.cpp

@@ -653,28 +653,15 @@ void CodeGenFunction::EmitOMPLastprivateClauseFinal(
     EmitBlock(ThenBB);
   }
   llvm::DenseMap<const Decl *, const Expr *> LoopCountersAndUpdates;
-  const Expr *LastIterVal = nullptr;
-  const Expr *IVExpr = nullptr;
-  const Expr *IncExpr = nullptr;
   if (auto *LoopDirective = dyn_cast<OMPLoopDirective>(&D)) {
-    if (isOpenMPWorksharingDirective(D.getDirectiveKind())) {
-      LastIterVal = cast<VarDecl>(cast<DeclRefExpr>(
-                                      LoopDirective->getUpperBoundVariable())
-                                      ->getDecl())
-                        ->getAnyInitializer();
-      IVExpr = LoopDirective->getIterationVariable();
-      IncExpr = LoopDirective->getInc();
-      auto IUpdate = LoopDirective->updates().begin();
-      for (auto *E : LoopDirective->counters()) {
-        auto *D = cast<DeclRefExpr>(E)->getDecl()->getCanonicalDecl();
-        LoopCountersAndUpdates[D] = *IUpdate;
-        ++IUpdate;
+    auto IC = LoopDirective->counters().begin();
+    for (auto F : LoopDirective->finals()) {
+      auto *D = cast<DeclRefExpr>(*IC)->getDecl()->getCanonicalDecl();
+      LoopCountersAndUpdates[D] = F;
+      ++IC;
     }
   }
-  }
-  {
   llvm::DenseSet<const VarDecl *> AlreadyEmittedVars;
-    bool FirstLCV = true;
   for (const auto *C : D.getClausesOfKind<OMPLastprivateClause>()) {
     auto IRef = C->varlist_begin();
     auto ISrcRef = C->source_exprs().begin();
@@ -687,16 +674,8 @@ void CodeGenFunction::EmitOMPLastprivateClauseFinal(
         // If lastprivate variable is a loop control variable for loop-based
         // directive, update its value before copyin back to original
         // variable.
-          if (auto *UpExpr = LoopCountersAndUpdates.lookup(CanonicalVD)) {
-            if (FirstLCV && LastIterVal) {
-              EmitAnyExprToMem(LastIterVal, EmitLValue(IVExpr).getAddress(),
-                               IVExpr->getType().getQualifiers(),
-                               /*IsInitializer=*/false);
-              EmitIgnoredExpr(IncExpr);
-              FirstLCV = false;
-            }
+        if (auto *UpExpr = LoopCountersAndUpdates.lookup(CanonicalVD))
           EmitIgnoredExpr(UpExpr);
-          }
         auto *SrcVD = cast<VarDecl>(cast<DeclRefExpr>(*ISrcRef)->getDecl());
         auto *DestVD = cast<VarDecl>(cast<DeclRefExpr>(*IDestRef)->getDecl());
         // Get the address of the original variable.
@@ -714,10 +693,8 @@ void CodeGenFunction::EmitOMPLastprivateClauseFinal(
       ++IDestRef;
     }
   }
-  }
-  if (IsLastIterCond) {
+  if (IsLastIterCond)
     EmitBlock(DoneBB, /*IsFinished=*/true);
-  }
 }
 
 static Address castToBase(CodeGenFunction &CGF, QualType BaseTy, QualType ElTy,
@@ -1052,10 +1029,6 @@ void CodeGenFunction::EmitOMPLoopBody(const OMPLoopDirective &D,
   // The end (updates/cleanups).
   EmitBlock(Continue.getBlock());
   BreakContinueStack.pop_back();
-    // TODO: Update lastprivates if the SeparateIter flag is true.
-    // This will be implemented in a follow-up OMPLastprivateClause patch, but
-    // result should be still correct without it, as we do not make these
-    // variables private yet.
 }
 
 void CodeGenFunction::EmitOMPInnerLoop(

test/OpenMP/for_lastprivate_codegen.cpp

@@ -396,20 +396,9 @@ int main() {
 // CHECK: br i1 [[IS_LAST_ITER:%.+]], label %[[LAST_THEN:.+]], label %[[LAST_DONE:.+]]
 // CHECK: [[LAST_THEN]]
 
-// Calculate last iter count
-// CHECK: store i32 1, i32* [[OMP_IV]]
-// CHECK: [[IV1_1:%.+]] = load i32, i32* [[OMP_IV]]
-// CHECK-NEXT: [[CALC_I_2:%.+]] = add nsw i32 [[IV1_1]], 1
-// CHECK-NEXT: store i32 [[CALC_I_2]], i32* [[OMP_IV]]
-// Actual copying.
-
-// original cnt=private_cnt;
 // Calculate private cnt value.
-// CHECK: [[IV1_1:%.+]] = load i32, i32* [[OMP_IV]]
-// CHECK: [[MUL:%.+]] = mul nsw i32 [[IV1_1]], 1
-// CHECK: [[ADD:%.+]] = add nsw i32 0, [[MUL]]
-// CHECK: [[CONV:%.+]] = trunc i32 [[ADD]] to i8
-// CHECK: store i8 [[CONV]], i8* [[CNT_PRIV]]
+// CHECK: store i8 2, i8* [[CNT_PRIV]]
+// original cnt=private_cnt;
 // CHECK: [[CNT_VAL:%.+]] = load i8, i8* [[CNT_PRIV]],
 // CHECK: store i8 [[CNT_VAL]], i8* [[CNT]],