[OpenMP][FIX] Enforce a function boundary for a new data environment

Whenever we enter a new OpenMP data environment we want to enter a function to simplify reasoning. Later we probably want to remove the entire specialization wrt. the if clause and pass the result to the runtime, for now this should fix PR48686. Reviewed By: ABataev Differential Revision: https://reviews.llvm.org/D94315
llvm · Jan 26, 2021 · bd75628 · bd75628
1 parent f19849a
commit bd75628
Show file tree

Hide file tree

Showing 3 changed files with 29 additions and 15 deletions.
diff --git a/clang/lib/CodeGen/CGOpenMPRuntime.cpp b/clang/lib/CodeGen/CGOpenMPRuntime.cpp
@@ -2098,6 +2098,14 @@ void CGOpenMPRuntime::emitParallelCall(CodeGenFunction &CGF, SourceLocation Loc,
     OutlinedFnArgs.push_back(ThreadIDAddr.getPointer());
     OutlinedFnArgs.push_back(ZeroAddrBound.getPointer());
     OutlinedFnArgs.append(CapturedVars.begin(), CapturedVars.end());
+
+    // Ensure we do not inline the function. This is trivially true for the ones
+    // passed to __kmpc_fork_call but the ones calles in serialized regions
+    // could be inlined. This is not a perfect but it is closer to the invariant
+    // we want, namely, every data environment starts with a new function.
+    // TODO: We should pass the if condition to the runtime function and do the
+    //       handling there. Much cleaner code.
+    OutlinedFn->addFnAttr(llvm::Attribute::NoInline);
     RT.emitOutlinedFunctionCall(CGF, Loc, OutlinedFn, OutlinedFnArgs);
 
     // __kmpc_end_serialized_parallel(&Loc, GTid);

diff --git a/clang/lib/CodeGen/CGOpenMPRuntimeGPU.cpp b/clang/lib/CodeGen/CGOpenMPRuntimeGPU.cpp
@@ -1576,11 +1576,6 @@ llvm::Function *CGOpenMPRuntimeGPU::emitParallelOutlinedFunction(
   auto *OutlinedFun =
       cast<llvm::Function>(CGOpenMPRuntime::emitParallelOutlinedFunction(
           D, ThreadIDVar, InnermostKind, CodeGen));
-  if (CGM.getLangOpts().Optimize) {
-    OutlinedFun->removeFnAttr(llvm::Attribute::NoInline);
-    OutlinedFun->removeFnAttr(llvm::Attribute::OptimizeNone);
-    OutlinedFun->addFnAttr(llvm::Attribute::AlwaysInline);
-  }
   IsInTargetMasterThreadRegion = PrevIsInTargetMasterThreadRegion;
   IsInTTDRegion = PrevIsInTTDRegion;
   if (getExecutionMode() != CGOpenMPRuntimeGPU::EM_SPMD &&
@@ -1698,11 +1693,6 @@ llvm::Function *CGOpenMPRuntimeGPU::emitTeamsOutlinedFunction(
   CodeGen.setAction(Action);
   llvm::Function *OutlinedFun = CGOpenMPRuntime::emitTeamsOutlinedFunction(
       D, ThreadIDVar, InnermostKind, CodeGen);
-  if (CGM.getLangOpts().Optimize) {
-    OutlinedFun->removeFnAttr(llvm::Attribute::NoInline);
-    OutlinedFun->removeFnAttr(llvm::Attribute::OptimizeNone);
-    OutlinedFun->addFnAttr(llvm::Attribute::AlwaysInline);
-  }
 
   return OutlinedFun;
 }
@@ -2102,6 +2092,14 @@ void CGOpenMPRuntimeGPU::emitNonSPMDParallelCall(
   // Force inline this outlined function at its call site.
   Fn->setLinkage(llvm::GlobalValue::InternalLinkage);
 
+  // Ensure we do not inline the function. This is trivially true for the ones
+  // passed to __kmpc_fork_call but the ones calles in serialized regions
+  // could be inlined. This is not a perfect but it is closer to the invariant
+  // we want, namely, every data environment starts with a new function.
+  // TODO: We should pass the if condition to the runtime function and do the
+  //       handling there. Much cleaner code.
+  cast<llvm::Function>(OutlinedFn)->addFnAttr(llvm::Attribute::NoInline);
+
   Address ZeroAddr = CGF.CreateDefaultAlignTempAlloca(CGF.Int32Ty,
                                                       /*Name=*/".zero.addr");
   CGF.InitTempAlloca(ZeroAddr, CGF.Builder.getInt32(/*C*/ 0));
@@ -3134,11 +3132,6 @@ static llvm::Function *emitShuffleAndReduceFunction(
       "_omp_reduction_shuffle_and_reduce_func", &CGM.getModule());
   CGM.SetInternalFunctionAttributes(GlobalDecl(), Fn, CGFI);
   Fn->setDoesNotRecurse();
-  if (CGM.getLangOpts().Optimize) {
-    Fn->removeFnAttr(llvm::Attribute::NoInline);
-    Fn->removeFnAttr(llvm::Attribute::OptimizeNone);
-    Fn->addFnAttr(llvm::Attribute::AlwaysInline);
-  }
 
   CodeGenFunction CGF(CGM);
   CGF.StartFunction(GlobalDecl(), C.VoidTy, Fn, CGFI, Args, Loc, Loc);

diff --git a/clang/test/OpenMP/parallel_if_codegen.cpp b/clang/test/OpenMP/parallel_if_codegen.cpp
@@ -7,6 +7,7 @@
 // RUN: %clang_cc1 -fopenmp-simd -fopenmp-version=45 -x c++ -triple %itanium_abi_triple -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck --check-prefix SIMD-ONLY0 %s
 // SIMD-ONLY0-NOT: {{__kmpc|__tgt}}
 
+// RUN: %clang_cc1 -verify -fopenmp -x c++ -triple %itanium_abi_triple -emit-llvm %s -disable-O0-optnone -o - | FileCheck %s --check-prefix=WOPT
 // RUN: %clang_cc1 -verify -fopenmp -x c++ -triple %itanium_abi_triple -emit-llvm %s -o - | FileCheck %s
 // RUN: %clang_cc1 -fopenmp -x c++ -std=c++11 -triple %itanium_abi_triple -emit-pch -o %t %s
 // RUN: %clang_cc1 -fopenmp -x c++ -triple %itanium_abi_triple -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s
@@ -96,14 +97,20 @@ int main() {
   return tmain(Arg);
 }
 
+// WOPT:     noinline
+// WOPT-NOT: optnone
 // CHECK: define internal {{.*}}void [[CAP_FN4]]
 // CHECK: call {{.*}}void @{{.+}}fn4
 // CHECK: ret void
 
+// WOPT:     noinline
+// WOPT-NOT: optnone
 // CHECK: define internal {{.*}}void [[CAP_FN5]]
 // CHECK: call {{.*}}void @{{.+}}fn5
 // CHECK: ret void
 
+// WOPT:     noinline
+// WOPT-NOT: optnone
 // CHECK: define internal {{.*}}void [[CAP_FN6]]
 // CHECK: call {{.*}}void @{{.+}}fn6
 // CHECK: ret void
@@ -129,14 +136,20 @@ int main() {
 // CHECK: br label %[[OMP_END]]
 // CHECK: [[OMP_END]]
 
+// WOPT:     noinline
+// WOPT-NOT: optnone
 // CHECK: define internal {{.*}}void [[CAP_FN1]]
 // CHECK: call {{.*}}void @{{.+}}fn1
 // CHECK: ret void
 
+// WOPT:     noinline
+// WOPT-NOT: optnone
 // CHECK: define internal {{.*}}void [[CAP_FN2]]
 // CHECK: call {{.*}}void @{{.+}}fn2
 // CHECK: ret void
 
+// WOPT:     noinline
+// WOPT-NOT: optnone
 // CHECK: define internal {{.*}}void [[CAP_FN3]]
 // CHECK: call {{.*}}void @{{.+}}fn3
 // CHECK: ret void