[openmp] Default to new rtl for amdgpu

Reverts D114965 as the compiler backend appears to be working again

Reviewed By: jhuber6

Differential Revision: https://reviews.llvm.org/D115157

clang/include/clang/Driver/Options.td

@@ -2442,7 +2442,7 @@ def fno_openmp_assume_teams_oversubscription : Flag<["-"], "fno-openmp-assume-te
 def fno_openmp_assume_threads_oversubscription : Flag<["-"], "fno-openmp-assume-threads-oversubscription">, 
   Group<f_Group>, Flags<[CC1Option, NoArgumentUnused, HelpHidden]>;
 defm openmp_target_new_runtime: BoolFOption<"openmp-target-new-runtime",
-  LangOpts<"OpenMPTargetNewRuntime">, DefaultFalse,
+  LangOpts<"OpenMPTargetNewRuntime">, DefaultTrue,
   PosFlag<SetTrue, [CC1Option], "Use the new bitcode library for OpenMP offloading">,
   NegFlag<SetFalse>>;
 defm openmp_optimistic_collapse : BoolFOption<"openmp-optimistic-collapse",

clang/lib/Driver/ToolChains/AMDGPUOpenMP.cpp

@@ -267,7 +267,7 @@ void AMDGPUOpenMPToolChain::addClangTargetOptions(
 
   std::string BitcodeSuffix;
   if (DriverArgs.hasFlag(options::OPT_fopenmp_target_new_runtime,
-                         options::OPT_fno_openmp_target_new_runtime, false))
+                         options::OPT_fno_openmp_target_new_runtime, true))
     BitcodeSuffix = "new-amdgpu-" + GPUArch;
   else
     BitcodeSuffix = "amdgcn-" + GPUArch;

clang/lib/Driver/ToolChains/Clang.cpp

@@ -5904,7 +5904,7 @@ void Clang::ConstructJob(Compilation &C, const JobAction &JA,
       // runtime.
       if (Args.hasFlag(options::OPT_fopenmp_target_new_runtime,
                        options::OPT_fno_openmp_target_new_runtime,
-                       /*Default=*/!getToolChain().getTriple().isAMDGCN()))
+                       /*Default=*/true))
         CmdArgs.push_back("-fopenmp-target-new-runtime");
 
       // When in OpenMP offloading mode, enable debugging on the device.