[flang][OpenMP] Add -f[no]-openmp-simd

clang/include/clang/Driver/Options.td

@@ -3731,14 +3731,20 @@ def fopenmp_relocatable_target : Flag<["-"], "fopenmp-relocatable-target">,
 def fnoopenmp_relocatable_target : Flag<["-"], "fnoopenmp-relocatable-target">,
   Group<f_Group>, Flags<[NoArgumentUnused, HelpHidden]>,
   Visibility<[ClangOption, CC1Option]>;
-def fopenmp_simd : Flag<["-"], "fopenmp-simd">, Group<f_Group>,
-  Flags<[NoArgumentUnused]>, Visibility<[ClangOption, CC1Option]>,
-  HelpText<"Emit OpenMP code only for SIMD-based constructs.">;
+def fopenmp_simd : Flag<["-"], "fopenmp-simd">,
+                   Group<f_Group>,
+                   Flags<[NoArgumentUnused]>,
+                   Visibility<[ClangOption, CC1Option, FlangOption, FC1Option]>,
+                   HelpText<"Emit OpenMP code only for SIMD-based constructs.">;
 def fopenmp_enable_irbuilder : Flag<["-"], "fopenmp-enable-irbuilder">, Group<f_Group>,
   Flags<[NoArgumentUnused, HelpHidden]>, Visibility<[ClangOption, CC1Option]>,
   HelpText<"Use the experimental OpenMP-IR-Builder codegen path.">;
-def fno_openmp_simd : Flag<["-"], "fno-openmp-simd">, Group<f_Group>,
-  Flags<[NoArgumentUnused]>, Visibility<[ClangOption, CC1Option]>;
+def fno_openmp_simd
+    : Flag<["-"], "fno-openmp-simd">,
+      Group<f_Group>,
+      Flags<[NoArgumentUnused]>,
+      Visibility<[ClangOption, CC1Option, FlangOption, FC1Option]>,
+      HelpText<"Do not emit code for any OpenMP constructs.">;
 def fopenmp_cuda_mode : Flag<["-"], "fopenmp-cuda-mode">, Group<f_Group>,
   Flags<[NoArgumentUnused, HelpHidden]>, Visibility<[ClangOption, CC1Option]>;
 def fno_openmp_cuda_mode : Flag<["-"], "fno-openmp-cuda-mode">, Group<f_Group>,

clang/lib/Driver/ToolChains/Flang.cpp

@@ -937,6 +937,8 @@ void Flang::ConstructJob(Compilation &C, const JobAction &JA,
 
       if (Args.hasArg(options::OPT_fopenmp_force_usm))
         CmdArgs.push_back("-fopenmp-force-usm");
+      Args.AddLastArg(CmdArgs, options::OPT_fopenmp_simd,
+                      options::OPT_fno_openmp_simd);
 
       // FIXME: Clang supports a whole bunch more flags here.
       break;
@@ -952,6 +954,9 @@ void Flang::ConstructJob(Compilation &C, const JobAction &JA,
           << A->getSpelling() << A->getValue();
       break;
     }
+  } else {
+    Args.AddLastArg(CmdArgs, options::OPT_fopenmp_simd,
+                    options::OPT_fno_openmp_simd);
   }
 
   // Pass the path to compiler resource files.

flang/include/flang/Optimizer/OpenMP/Passes.td

@@ -112,4 +112,9 @@ def GenericLoopConversionPass
   ];
 }
 
+def SimdOnlyPass : Pass<"omp-simd-only", "mlir::ModuleOp"> {
+  let summary = "Filters out non-simd OpenMP constructs";
+  let dependentDialects = ["mlir::omp::OpenMPDialect"];
+}
+
 #endif //FORTRAN_OPTIMIZER_OPENMP_PASSES

flang/include/flang/Optimizer/Passes/Pipelines.h

@@ -119,13 +119,16 @@ void registerDefaultInlinerPass(MLIRToLLVMPassPipelineConfig &config);
 void createDefaultFIROptimizerPassPipeline(mlir::PassManager &pm,
                                            MLIRToLLVMPassPipelineConfig &pc);
 
+/// Select which mode to enable OpenMP support in.
+enum class EnableOpenMP { None, Simd, Full };
+
 /// Create a pass pipeline for lowering from HLFIR to FIR
 ///
 /// \param pm - MLIR pass manager that will hold the pipeline definition
 /// \param optLevel - optimization level used for creating FIR optimization
 ///   passes pipeline
 void createHLFIRToFIRPassPipeline(
-    mlir::PassManager &pm, bool enableOpenMP,
+    mlir::PassManager &pm, EnableOpenMP enableOpenMP,
     llvm::OptimizationLevel optLevel = defaultOptLevel);
 
 struct OpenMPFIRPassPipelineOpts {

flang/include/flang/Semantics/openmp-directive-sets.h

@@ -401,6 +401,22 @@ static const OmpDirectiveSet nestedWorkshareErrSet{
         Directive::OMPD_taskloop,
     } | workShareSet,
 };
+
+//===----------------------------------------------------------------------===//
+// Misc directive sets
+//===----------------------------------------------------------------------===//
+
+// Simple standalone directives than can be erased by -fopenmp-simd.
+static const OmpDirectiveSet simpleStandaloneNonSimdOnlySet{
+    Directive::OMPD_taskyield,
+    Directive::OMPD_barrier,
+    Directive::OMPD_ordered,
+    Directive::OMPD_target_enter_data,
+    Directive::OMPD_target_exit_data,
+    Directive::OMPD_target_update,
+    Directive::OMPD_taskwait,
+};
+
 } // namespace llvm::omp
 
 #endif // FORTRAN_SEMANTICS_OPENMP_DIRECTIVE_SETS_H_

flang/include/flang/Support/LangOptions.def

@@ -58,6 +58,8 @@ LANGOPT(OpenMPTeamSubscription, 1, 0)
 LANGOPT(OpenMPNoThreadState, 1, 0)
 /// Assume that no thread in a parallel region will encounter a parallel region
 LANGOPT(OpenMPNoNestedParallelism, 1, 0)
+/// Use SIMD only OpenMP support.
+LANGOPT(OpenMPSimd, 1, false)
 
 LANGOPT(VScaleMin, 32, 0)  ///< Minimum vscale range value
 LANGOPT(VScaleMax, 32, 0)  ///< Maximum vscale range value

flang/include/flang/Tools/CrossToolHelpers.h

@@ -134,6 +134,7 @@ struct MLIRToLLVMPassPipelineConfig : public FlangEPCallBacks {
                                       ///< functions.
   bool NSWOnLoopVarInc = true; ///< Add nsw flag to loop variable increments.
   bool EnableOpenMP = false; ///< Enable OpenMP lowering.
+  bool EnableOpenMPSimd = false; ///< Enable OpenMP simd-only mode.
   std::string InstrumentFunctionEntry =
       ""; ///< Name of the instrument-function that is called on each
           ///< function-entry

flang/lib/Frontend/CompilerInvocation.cpp

@@ -1162,8 +1162,15 @@ static bool parseOpenMPArgs(CompilerInvocation &res, llvm::opt::ArgList &args,
                             clang::DiagnosticsEngine &diags) {
   llvm::opt::Arg *arg = args.getLastArg(clang::driver::options::OPT_fopenmp,
                                         clang::driver::options::OPT_fno_openmp);
-  if (!arg || arg->getOption().matches(clang::driver::options::OPT_fno_openmp))
-    return true;
+  if (!arg ||
+      arg->getOption().matches(clang::driver::options::OPT_fno_openmp)) {
+    bool isSimdSpecified = args.hasFlag(
+        clang::driver::options::OPT_fopenmp_simd,
+        clang::driver::options::OPT_fno_openmp_simd, /*Default=*/false);
+    if (!isSimdSpecified)
+      return true;
+    res.getLangOpts().OpenMPSimd = 1;
+  }
 
   unsigned numErrorsBefore = diags.getNumErrors();
   llvm::Triple t(res.getTargetOpts().triple);

flang/lib/Frontend/FrontendActions.cpp

@@ -298,6 +298,7 @@ bool CodeGenAction::beginSourceFileAction() {
   bool isOpenMPEnabled =
       ci.getInvocation().getFrontendOpts().features.IsEnabled(
           Fortran::common::LanguageFeature::OpenMP);
+  bool isOpenMPSimd = ci.getInvocation().getLangOpts().OpenMPSimd;
 
   fir::OpenMPFIRPassPipelineOpts opts;
 
@@ -329,12 +330,13 @@ bool CodeGenAction::beginSourceFileAction() {
     if (auto offloadMod = llvm::dyn_cast<mlir::omp::OffloadModuleInterface>(
             mlirModule->getOperation()))
       opts.isTargetDevice = offloadMod.getIsTargetDevice();
+  }
 
-    // WARNING: This pipeline must be run immediately after the lowering to
-    // ensure that the FIR is correct with respect to OpenMP operations/
-    // attributes.
+  // WARNING: This pipeline must be run immediately after the lowering to
+  // ensure that the FIR is correct with respect to OpenMP operations/
+  // attributes.
+  if (isOpenMPEnabled || isOpenMPSimd)
     fir::createOpenMPFIRPassPipeline(pm, opts);
-  }
 
   pm.enableVerifier(/*verifyPasses=*/true);
   pm.addPass(std::make_unique<Fortran::lower::VerifierPass>());
@@ -617,12 +619,14 @@ void CodeGenAction::lowerHLFIRToFIR() {
   pm.addPass(std::make_unique<Fortran::lower::VerifierPass>());
   pm.enableVerifier(/*verifyPasses=*/true);
 
+  fir::EnableOpenMP enableOpenMP = fir::EnableOpenMP::None;
+  if (ci.getInvocation().getFrontendOpts().features.IsEnabled(
+          Fortran::common::LanguageFeature::OpenMP))
+    enableOpenMP = fir::EnableOpenMP::Full;
+  if (ci.getInvocation().getLangOpts().OpenMPSimd)
+    enableOpenMP = fir::EnableOpenMP::Simd;
   // Create the pass pipeline
-  fir::createHLFIRToFIRPassPipeline(
-      pm,
-      ci.getInvocation().getFrontendOpts().features.IsEnabled(
-          Fortran::common::LanguageFeature::OpenMP),
-      level);
+  fir::createHLFIRToFIRPassPipeline(pm, enableOpenMP, level);
   (void)mlir::applyPassManagerCLOptions(pm);
 
   mlir::TimingScope timingScopeMLIRPasses = timingScopeRoot.nest(
@@ -748,6 +752,9 @@ void CodeGenAction::generateLLVMIR() {
           Fortran::common::LanguageFeature::OpenMP))
     config.EnableOpenMP = true;
 
+  if (ci.getInvocation().getLangOpts().OpenMPSimd)
+    config.EnableOpenMPSimd = true;
+
   if (ci.getInvocation().getLoweringOpts().getIntegerWrapAround())
     config.NSWOnLoopVarInc = false;
 

flang/lib/Lower/OpenMP/ClauseProcessor.h

@@ -208,11 +208,15 @@ void ClauseProcessor::processTODO(mlir::Location currentLocation,
     if (!x)
       return;
     unsigned version = semaCtx.langOptions().OpenMPVersion;
-    TODO(currentLocation,
-         "Unhandled clause " + llvm::omp::getOpenMPClauseName(id).upper() +
-             " in " +
-             llvm::omp::getOpenMPDirectiveName(directive, version).upper() +
-             " construct");
+    bool isSimdDirective = llvm::omp::getOpenMPDirectiveName(directive, version)
+                               .upper()
+                               .find("SIMD") != llvm::StringRef::npos;
+    if (!semaCtx.langOptions().OpenMPSimd || isSimdDirective)
+      TODO(currentLocation,
+           "Unhandled clause " + llvm::omp::getOpenMPClauseName(id).upper() +
+               " in " +
+               llvm::omp::getOpenMPDirectiveName(directive, version).upper() +
+               " construct");
   };
 
   for (ClauseIterator it = clauses.begin(); it != clauses.end(); ++it)

flang/lib/Lower/OpenMP/OpenMP.cpp

@@ -2262,7 +2262,8 @@ genOrderedOp(lower::AbstractConverter &converter, lower::SymMap &symTable,
              semantics::SemanticsContext &semaCtx, lower::pft::Evaluation &eval,
              mlir::Location loc, const ConstructQueue &queue,
              ConstructQueue::const_iterator item) {
-  TODO(loc, "OMPD_ordered");
+  if (!semaCtx.langOptions().OpenMPSimd)
+    TODO(loc, "OMPD_ordered");
   return nullptr;
 }
 
@@ -2449,7 +2450,8 @@ genScopeOp(lower::AbstractConverter &converter, lower::SymMap &symTable,
            semantics::SemanticsContext &semaCtx, lower::pft::Evaluation &eval,
            mlir::Location loc, const ConstructQueue &queue,
            ConstructQueue::const_iterator item) {
-  TODO(loc, "Scope construct");
+  if (!semaCtx.langOptions().OpenMPSimd)
+    TODO(loc, "Scope construct");
   return nullptr;
 }
 
@@ -3276,7 +3278,8 @@ static mlir::omp::TaskloopOp genCompositeTaskloopSimd(
     lower::pft::Evaluation &eval, mlir::Location loc,
     const ConstructQueue &queue, ConstructQueue::const_iterator item) {
   assert(std::distance(item, queue.end()) == 2 && "Invalid leaf constructs");
-  TODO(loc, "Composite TASKLOOP SIMD");
+  if (!semaCtx.langOptions().OpenMPSimd)
+    TODO(loc, "Composite TASKLOOP SIMD");
   return nullptr;
 }
 
@@ -3448,8 +3451,10 @@ static void genOMPDispatch(lower::AbstractConverter &converter,
     break;
   case llvm::omp::Directive::OMPD_tile: {
     unsigned version = semaCtx.langOptions().OpenMPVersion;
-    TODO(loc, "Unhandled loop directive (" +
-                  llvm::omp::getOpenMPDirectiveName(dir, version) + ")");
+    if (!semaCtx.langOptions().OpenMPSimd)
+      TODO(loc, "Unhandled loop directive (" +
+                    llvm::omp::getOpenMPDirectiveName(dir, version) + ")");
+    break;
   }
   case llvm::omp::Directive::OMPD_unroll:
     genUnrollOp(converter, symTable, stmtCtx, semaCtx, eval, loc, queue, item);
@@ -3484,35 +3489,40 @@ static void
 genOMP(lower::AbstractConverter &converter, lower::SymMap &symTable,
        semantics::SemanticsContext &semaCtx, lower::pft::Evaluation &eval,
        const parser::OpenMPDeclarativeAllocate &declarativeAllocate) {
-  TODO(converter.getCurrentLocation(), "OpenMPDeclarativeAllocate");
+  if (!semaCtx.langOptions().OpenMPSimd)
+    TODO(converter.getCurrentLocation(), "OpenMPDeclarativeAllocate");
 }
 
 static void genOMP(lower::AbstractConverter &converter, lower::SymMap &symTable,
                    semantics::SemanticsContext &semaCtx,
                    lower::pft::Evaluation &eval,
                    const parser::OpenMPDeclarativeAssumes &assumesConstruct) {
-  TODO(converter.getCurrentLocation(), "OpenMP ASSUMES declaration");
+  if (!semaCtx.langOptions().OpenMPSimd)
+    TODO(converter.getCurrentLocation(), "OpenMP ASSUMES declaration");
 }
 
 static void
 genOMP(lower::AbstractConverter &converter, lower::SymMap &symTable,
        semantics::SemanticsContext &semaCtx, lower::pft::Evaluation &eval,
        const parser::OmpDeclareVariantDirective &declareVariantDirective) {
-  TODO(converter.getCurrentLocation(), "OmpDeclareVariantDirective");
+  if (!semaCtx.langOptions().OpenMPSimd)
+    TODO(converter.getCurrentLocation(), "OmpDeclareVariantDirective");
 }
 
 static void genOMP(
     lower::AbstractConverter &converter, lower::SymMap &symTable,
     semantics::SemanticsContext &semaCtx, lower::pft::Evaluation &eval,
     const parser::OpenMPDeclareReductionConstruct &declareReductionConstruct) {
-  TODO(converter.getCurrentLocation(), "OpenMPDeclareReductionConstruct");
+  if (!semaCtx.langOptions().OpenMPSimd)
+    TODO(converter.getCurrentLocation(), "OpenMPDeclareReductionConstruct");
 }
 
 static void
 genOMP(lower::AbstractConverter &converter, lower::SymMap &symTable,
        semantics::SemanticsContext &semaCtx, lower::pft::Evaluation &eval,
        const parser::OpenMPDeclareSimdConstruct &declareSimdConstruct) {
-  TODO(converter.getCurrentLocation(), "OpenMPDeclareSimdConstruct");
+  if (!semaCtx.langOptions().OpenMPSimd)
+    TODO(converter.getCurrentLocation(), "OpenMPDeclareSimdConstruct");
 }
 
 static void
@@ -3706,14 +3716,16 @@ static void genOMP(lower::AbstractConverter &converter, lower::SymMap &symTable,
   (void)objects;
   (void)clauses;
 
-  TODO(converter.getCurrentLocation(), "OpenMPDepobjConstruct");
+  if (!semaCtx.langOptions().OpenMPSimd)
+    TODO(converter.getCurrentLocation(), "OpenMPDepobjConstruct");
 }
 
 static void genOMP(lower::AbstractConverter &converter, lower::SymMap &symTable,
                    semantics::SemanticsContext &semaCtx,
                    lower::pft::Evaluation &eval,
                    const parser::OpenMPInteropConstruct &interopConstruct) {
-  TODO(converter.getCurrentLocation(), "OpenMPInteropConstruct");
+  if (!semaCtx.langOptions().OpenMPSimd)
+    TODO(converter.getCurrentLocation(), "OpenMPInteropConstruct");
 }
 
 static void
@@ -3729,7 +3741,8 @@ static void genOMP(lower::AbstractConverter &converter, lower::SymMap &symTable,
                    semantics::SemanticsContext &semaCtx,
                    lower::pft::Evaluation &eval,
                    const parser::OpenMPAllocatorsConstruct &allocsConstruct) {
-  TODO(converter.getCurrentLocation(), "OpenMPAllocatorsConstruct");
+  if (!semaCtx.langOptions().OpenMPSimd)
+    TODO(converter.getCurrentLocation(), "OpenMPAllocatorsConstruct");
 }
 
 //===----------------------------------------------------------------------===//
@@ -3795,7 +3808,8 @@ static void genOMP(lower::AbstractConverter &converter, lower::SymMap &symTable,
         !std::holds_alternative<clause::Detach>(clause.u)) {
       std::string name =
           parser::ToUpperCaseLetters(llvm::omp::getOpenMPClauseName(clause.id));
-      TODO(clauseLocation, name + " clause is not implemented yet");
+      if (!semaCtx.langOptions().OpenMPSimd)
+        TODO(clauseLocation, name + " clause is not implemented yet");
     }
   }
 
@@ -3811,7 +3825,8 @@ static void genOMP(lower::AbstractConverter &converter, lower::SymMap &symTable,
                    lower::pft::Evaluation &eval,
                    const parser::OpenMPAssumeConstruct &assumeConstruct) {
   mlir::Location clauseLocation = converter.genLocation(assumeConstruct.source);
-  TODO(clauseLocation, "OpenMP ASSUME construct");
+  if (!semaCtx.langOptions().OpenMPSimd)
+    TODO(clauseLocation, "OpenMP ASSUME construct");
 }
 
 static void genOMP(lower::AbstractConverter &converter, lower::SymMap &symTable,
@@ -3847,21 +3862,24 @@ static void genOMP(lower::AbstractConverter &converter, lower::SymMap &symTable,
                    semantics::SemanticsContext &semaCtx,
                    lower::pft::Evaluation &eval,
                    const parser::OpenMPUtilityConstruct &) {
-  TODO(converter.getCurrentLocation(), "OpenMPUtilityConstruct");
+  if (!semaCtx.langOptions().OpenMPSimd)
+    TODO(converter.getCurrentLocation(), "OpenMPUtilityConstruct");
 }
 
 static void genOMP(lower::AbstractConverter &converter, lower::SymMap &symTable,
                    semantics::SemanticsContext &semaCtx,
                    lower::pft::Evaluation &eval,
                    const parser::OpenMPDispatchConstruct &) {
-  TODO(converter.getCurrentLocation(), "OpenMPDispatchConstruct");
+  if (!semaCtx.langOptions().OpenMPSimd)
+    TODO(converter.getCurrentLocation(), "OpenMPDispatchConstruct");
 }
 
 static void genOMP(lower::AbstractConverter &converter, lower::SymMap &symTable,
                    semantics::SemanticsContext &semaCtx,
                    lower::pft::Evaluation &eval,
                    const parser::OpenMPExecutableAllocate &execAllocConstruct) {
-  TODO(converter.getCurrentLocation(), "OpenMPExecutableAllocate");
+  if (!semaCtx.langOptions().OpenMPSimd)
+    TODO(converter.getCurrentLocation(), "OpenMPExecutableAllocate");
 }
 
 static void genOMP(lower::AbstractConverter &converter, lower::SymMap &symTable,

flang/lib/Optimizer/OpenMP/CMakeLists.txt

@@ -9,6 +9,7 @@ add_flang_library(FlangOpenMPTransforms
   MarkDeclareTarget.cpp
   LowerWorkshare.cpp
   LowerNontemporal.cpp
+  SimdOnly.cpp
 
   DEPENDS
   FIRDialect