[flang] Updating drivers to create data layout before semantics (#73301)

Preliminary patch to change lowering/code generation to use
llvm::DataLayout information instead of generating "sizeof" GEP (see
https://github.com/llvm/llvm-project/issues/71507).

Fortran Semantic analysis needs to know about the target type size and
alignment to deal with common blocks, and intrinsics like
C_SIZEOF/TRANSFER. This information should be obtained from the
llvm::DataLayout so that it is consistent during the whole compilation
flow.

This change is changing flang-new and bbc drivers to:
1. Create the llvm::TargetMachine so that the data layout of the target
can be obtained before semantics.
2. Sharing bbc/flang-new set-up of the
SemanticConstext.targetCharateristics from the llvm::TargetMachine. For
now, the actual part that set-up the Fortran type size and alignment
from the llvm::DataLayout is left TODO so that this change is mostly an
NFC impacting the drivers.
3. Let the lowering bridge set-up the mlir::Module datalayout attributes
since it is doing it for the target attribute, and that allows the llvm
data layout information to be available during lowering.

For flang-new, the changes are code shuffling: the `llvm::TargetMachine`
instance is moved to `CompilerInvocation` class so that it can be used
to set-up the semantic contexts. `setMLIRDataLayout` is moved to
`flang/Optimizer/Support/DataLayout.h` (it will need to be used from
codegen pass for fir-opt target independent testing.)), and the code
setting-up semantics targetCharacteristics is moved to
`Tools/TargetSetup.h` so that it can be shared with bbc.

As a consequence, LLVM targets must be registered when running
semantics, and it is not possible to run semantics for a target that is
not registered with the -triple option (hence the power pc specific
modules can only be built if the PowerPC target is available.

6 files changed, 212 insertions, 177 deletions

diff --git a/flang/lib/Frontend/CompilerInstance.cpp b/flang/lib/Frontend/CompilerInstance.cpp
index 328b377..a6b8f1a 100644
--- a/flang/lib/Frontend/CompilerInstance.cpp
+++ b/flang/lib/Frontend/CompilerInstance.cpp
@@ -17,11 +17,15 @@
 #include "flang/Parser/parsing.h"
 #include "flang/Parser/provenance.h"
 #include "flang/Semantics/semantics.h"
+#include "clang/Basic/DiagnosticFrontend.h"
+#include "llvm/ADT/StringExtras.h"
+#include "llvm/MC/TargetRegistry.h"
 #include "llvm/Support/Errc.h"
 #include "llvm/Support/Error.h"
 #include "llvm/Support/FileSystem.h"
 #include "llvm/Support/Path.h"
 #include "llvm/Support/raw_ostream.h"
+#include "llvm/TargetParser/TargetParser.h"
 #include "llvm/TargetParser/Triple.h"
 
 using namespace Fortran::frontend;
@@ -156,8 +160,10 @@ bool CompilerInstance::executeAction(FrontendAction &act) {
   invoc.setFortranOpts();
   // Set the encoding to read all input files in based on user input.
   allSources->set_encoding(invoc.getFortranOpts().encoding);
+  if (!setUpTargetMachine())
+    return false;
   // Create the semantics context
-  semaContext = invoc.getSemanticsCtx(*allCookedSources);
+  semaContext = invoc.getSemanticsCtx(*allCookedSources, getTargetMachine());
   // Set options controlling lowering to FIR.
   invoc.setLoweringOptions();
 
@@ -197,3 +203,129 @@ CompilerInstance::createDiagnostics(clang::DiagnosticOptions *opts,
   }
   return diags;
 }
+
+// Get feature string which represents combined explicit target features
+// for AMD GPU and the target features specified by the user
+static std::string
+getExplicitAndImplicitAMDGPUTargetFeatures(clang::DiagnosticsEngine &diags,
+                                           const TargetOptions &targetOpts,
+                                           const llvm::Triple triple) {
+  llvm::StringRef cpu = targetOpts.cpu;
+  llvm::StringMap<bool> implicitFeaturesMap;
+  std::string errorMsg;
+  // Get the set of implicit target features
+  llvm::AMDGPU::fillAMDGPUFeatureMap(cpu, triple, implicitFeaturesMap);
+
+  // Add target features specified by the user
+  for (auto &userFeature : targetOpts.featuresAsWritten) {
+    std::string userKeyString = userFeature.substr(1);
+    implicitFeaturesMap[userKeyString] = (userFeature[0] == '+');
+  }
+
+  if (!llvm::AMDGPU::insertWaveSizeFeature(cpu, triple, implicitFeaturesMap,
+                                           errorMsg)) {
+    unsigned diagID = diags.getCustomDiagID(clang::DiagnosticsEngine::Error,
+                                            "Unsupported feature ID: %0");
+    diags.Report(diagID) << errorMsg.data();
+    return std::string();
+  }
+
+  llvm::SmallVector<std::string> featuresVec;
+  for (auto &implicitFeatureItem : implicitFeaturesMap) {
+    featuresVec.push_back((llvm::Twine(implicitFeatureItem.second ? "+" : "-") +
+                           implicitFeatureItem.first().str())
+                              .str());
+  }
+  llvm::sort(featuresVec);
+  return llvm::join(featuresVec, ",");
+}
+
+// Get feature string which represents combined explicit target features
+// for NVPTX and the target features specified by the user/
+// TODO: Have a more robust target conf like `clang/lib/Basic/Targets/NVPTX.cpp`
+static std::string
+getExplicitAndImplicitNVPTXTargetFeatures(clang::DiagnosticsEngine &diags,
+                                          const TargetOptions &targetOpts,
+                                          const llvm::Triple triple) {
+  llvm::StringRef cpu = targetOpts.cpu;
+  llvm::StringMap<bool> implicitFeaturesMap;
+  std::string errorMsg;
+  bool ptxVer = false;
+
+  // Add target features specified by the user
+  for (auto &userFeature : targetOpts.featuresAsWritten) {
+    llvm::StringRef userKeyString(llvm::StringRef(userFeature).drop_front(1));
+    implicitFeaturesMap[userKeyString.str()] = (userFeature[0] == '+');
+    // Check if the user provided a PTX version
+    if (userKeyString.startswith("ptx"))
+      ptxVer = true;
+  }
+
+  // Set the default PTX version to `ptx61` if none was provided.
+  // TODO: set the default PTX version based on the chip.
+  if (!ptxVer)
+    implicitFeaturesMap["ptx61"] = true;
+
+  // Set the compute capability.
+  implicitFeaturesMap[cpu.str()] = true;
+
+  llvm::SmallVector<std::string> featuresVec;
+  for (auto &implicitFeatureItem : implicitFeaturesMap) {
+    featuresVec.push_back((llvm::Twine(implicitFeatureItem.second ? "+" : "-") +
+                           implicitFeatureItem.first().str())
+                              .str());
+  }
+  llvm::sort(featuresVec);
+  return llvm::join(featuresVec, ",");
+}
+
+std::string CompilerInstance::getTargetFeatures() {
+  const TargetOptions &targetOpts = getInvocation().getTargetOpts();
+  const llvm::Triple triple(targetOpts.triple);
+
+  // Clang does not append all target features to the clang -cc1 invocation.
+  // Some target features are parsed implicitly by clang::TargetInfo child
+  // class. Clang::TargetInfo classes are the basic clang classes and
+  // they cannot be reused by Flang.
+  // That's why we need to extract implicit target features and add
+  // them to the target features specified by the user
+  if (triple.isAMDGPU()) {
+    return getExplicitAndImplicitAMDGPUTargetFeatures(getDiagnostics(),
+                                                      targetOpts, triple);
+  } else if (triple.isNVPTX()) {
+    return getExplicitAndImplicitNVPTXTargetFeatures(getDiagnostics(),
+                                                     targetOpts, triple);
+  }
+  return llvm::join(targetOpts.featuresAsWritten.begin(),
+                    targetOpts.featuresAsWritten.end(), ",");
+}
+
+bool CompilerInstance::setUpTargetMachine() {
+  const TargetOptions &targetOpts = getInvocation().getTargetOpts();
+  const std::string &theTriple = targetOpts.triple;
+
+  // Create `Target`
+  std::string error;
+  const llvm::Target *theTarget =
+      llvm::TargetRegistry::lookupTarget(theTriple, error);
+  if (!theTarget) {
+    getDiagnostics().Report(clang::diag::err_fe_unable_to_create_target)
+        << error;
+    return false;
+  }
+
+  // Create `TargetMachine`
+  const auto &CGOpts = getInvocation().getCodeGenOpts();
+  std::optional<llvm::CodeGenOptLevel> OptLevelOrNone =
+      llvm::CodeGenOpt::getLevel(CGOpts.OptimizationLevel);
+  assert(OptLevelOrNone && "Invalid optimization level!");
+  llvm::CodeGenOptLevel OptLevel = *OptLevelOrNone;
+  std::string featuresStr = getTargetFeatures();
+  targetMachine.reset(theTarget->createTargetMachine(
+      theTriple, /*CPU=*/targetOpts.cpu,
+      /*Features=*/featuresStr, llvm::TargetOptions(),
+      /*Reloc::Model=*/CGOpts.getRelocationModel(),
+      /*CodeModel::Model=*/std::nullopt, OptLevel));
+  assert(targetMachine && "Failed to create TargetMachine");
+  return true;
+}
diff --git a/flang/lib/Frontend/CompilerInvocation.cpp b/flang/lib/Frontend/CompilerInvocation.cpp
index c623969..b3f32bb 100644
--- a/flang/lib/Frontend/CompilerInvocation.cpp
+++ b/flang/lib/Frontend/CompilerInvocation.cpp
@@ -18,6 +18,7 @@
 #include "flang/Frontend/PreprocessorOptions.h"
 #include "flang/Frontend/TargetOptions.h"
 #include "flang/Semantics/semantics.h"
+#include "flang/Tools/TargetSetup.h"
 #include "flang/Version.inc"
 #include "clang/Basic/AllDiagnostics.h"
 #include "clang/Basic/DiagnosticDriver.h"
@@ -1347,7 +1348,8 @@ void CompilerInvocation::setFortranOpts() {
 
 std::unique_ptr<Fortran::semantics::SemanticsContext>
 CompilerInvocation::getSemanticsCtx(
-    Fortran::parser::AllCookedSources &allCookedSources) {
+    Fortran::parser::AllCookedSources &allCookedSources,
+    const llvm::TargetMachine &targetMachine) {
   auto &fortranOptions = getFortranOpts();
 
   auto semanticsContext = std::make_unique<semantics::SemanticsContext>(
@@ -1360,21 +1362,10 @@ CompilerInvocation::getSemanticsCtx(
       .set_moduleFileSuffix(getModuleFileSuffix())
       .set_underscoring(getCodeGenOpts().Underscoring);
 
-  llvm::Triple targetTriple{llvm::Triple(this->targetOpts.triple)};
-  // FIXME: Handle real(3) ?
-  if (targetTriple.getArch() != llvm::Triple::ArchType::x86_64) {
-    semanticsContext->targetCharacteristics().DisableType(
-        Fortran::common::TypeCategory::Real, /*kind=*/10);
-  }
-
-  std::string version = Fortran::common::getFlangFullVersion();
-  semanticsContext->targetCharacteristics()
-      .set_compilerOptionsString(allCompilerInvocOpts)
-      .set_compilerVersionString(version);
-
-  if (targetTriple.isPPC())
-    semanticsContext->targetCharacteristics().set_isPPC(true);
-
+  std::string compilerVersion = Fortran::common::getFlangFullVersion();
+  Fortran::tools::setUpTargetCharacteristics(
+      semanticsContext->targetCharacteristics(), targetMachine, compilerVersion,
+      allCompilerInvocOpts);
   return semanticsContext;
 }
 
diff --git a/flang/lib/Frontend/FrontendActions.cpp b/flang/lib/Frontend/FrontendActions.cpp
index 1be95cc..b114c55 100644
--- a/flang/lib/Frontend/FrontendActions.cpp
+++ b/flang/lib/Frontend/FrontendActions.cpp
@@ -21,6 +21,7 @@
 #include "flang/Lower/Support/Verifier.h"
 #include "flang/Optimizer/Dialect/Support/FIRContext.h"
 #include "flang/Optimizer/Dialect/Support/KindMapping.h"
+#include "flang/Optimizer/Support/DataLayout.h"
 #include "flang/Optimizer/Support/InitFIR.h"
 #include "flang/Optimizer/Support/Utils.h"
 #include "flang/Optimizer/Transforms/Passes.h"
@@ -53,7 +54,6 @@
 #include "llvm/IR/LegacyPassManager.h"
 #include "llvm/IR/Verifier.h"
 #include "llvm/IRReader/IRReader.h"
-#include "llvm/MC/TargetRegistry.h"
 #include "llvm/Object/OffloadBinary.h"
 #include "llvm/Passes/PassBuilder.h"
 #include "llvm/Passes/PassPlugin.h"
@@ -65,7 +65,6 @@
 #include "llvm/Support/SourceMgr.h"
 #include "llvm/Support/ToolOutputFile.h"
 #include "llvm/Target/TargetMachine.h"
-#include "llvm/TargetParser/TargetParser.h"
 #include "llvm/Transforms/Utils/ModuleUtils.h"
 #include <memory>
 #include <system_error>
@@ -139,111 +138,6 @@ bool PrescanAndSemaDebugAction::beginSourceFileAction() {
          (generateRtTypeTables() || true);
 }
 
-// Get feature string which represents combined explicit target features
-// for AMD GPU and the target features specified by the user
-static std::string
-getExplicitAndImplicitAMDGPUTargetFeatures(CompilerInstance &ci,
-                                           const TargetOptions &targetOpts,
-                                           const llvm::Triple triple) {
-  llvm::StringRef cpu = targetOpts.cpu;
-  llvm::StringMap<bool> implicitFeaturesMap;
-  std::string errorMsg;
-  // Get the set of implicit target features
-  llvm::AMDGPU::fillAMDGPUFeatureMap(cpu, triple, implicitFeaturesMap);
-
-  // Add target features specified by the user
-  for (auto &userFeature : targetOpts.featuresAsWritten) {
-    std::string userKeyString = userFeature.substr(1);
-    implicitFeaturesMap[userKeyString] = (userFeature[0] == '+');
-  }
-
-  if (!llvm::AMDGPU::insertWaveSizeFeature(cpu, triple, implicitFeaturesMap,
-                                           errorMsg)) {
-    unsigned diagID = ci.getDiagnostics().getCustomDiagID(
-        clang::DiagnosticsEngine::Error, "Unsupported feature ID: %0");
-    ci.getDiagnostics().Report(diagID) << errorMsg.data();
-    return std::string();
-  }
-
-  llvm::SmallVector<std::string> featuresVec;
-  for (auto &implicitFeatureItem : implicitFeaturesMap) {
-    featuresVec.push_back((llvm::Twine(implicitFeatureItem.second ? "+" : "-") +
-                           implicitFeatureItem.first().str())
-                              .str());
-  }
-  llvm::sort(featuresVec);
-  return llvm::join(featuresVec, ",");
-}
-
-// Get feature string which represents combined explicit target features
-// for NVPTX and the target features specified by the user/
-// TODO: Have a more robust target conf like `clang/lib/Basic/Targets/NVPTX.cpp`
-static std::string
-getExplicitAndImplicitNVPTXTargetFeatures(CompilerInstance &ci,
-                                          const TargetOptions &targetOpts,
-                                          const llvm::Triple triple) {
-  llvm::StringRef cpu = targetOpts.cpu;
-  llvm::StringMap<bool> implicitFeaturesMap;
-  std::string errorMsg;
-  bool ptxVer = false;
-
-  // Add target features specified by the user
-  for (auto &userFeature : targetOpts.featuresAsWritten) {
-    llvm::StringRef userKeyString(llvm::StringRef(userFeature).drop_front(1));
-    implicitFeaturesMap[userKeyString.str()] = (userFeature[0] == '+');
-    // Check if the user provided a PTX version
-    if (userKeyString.startswith("ptx"))
-      ptxVer = true;
-  }
-
-  // Set the default PTX version to `ptx61` if none was provided.
-  // TODO: set the default PTX version based on the chip.
-  if (!ptxVer)
-    implicitFeaturesMap["ptx61"] = true;
-
-  // Set the compute capability.
-  implicitFeaturesMap[cpu.str()] = true;
-
-  llvm::SmallVector<std::string> featuresVec;
-  for (auto &implicitFeatureItem : implicitFeaturesMap) {
-    featuresVec.push_back((llvm::Twine(implicitFeatureItem.second ? "+" : "-") +
-                           implicitFeatureItem.first().str())
-                              .str());
-  }
-  llvm::sort(featuresVec);
-  return llvm::join(featuresVec, ",");
-}
-
-// Produces the string which represents target feature
-static std::string getTargetFeatures(CompilerInstance &ci) {
-  const TargetOptions &targetOpts = ci.getInvocation().getTargetOpts();
-  const llvm::Triple triple(targetOpts.triple);
-
-  // Clang does not append all target features to the clang -cc1 invocation.
-  // Some target features are parsed implicitly by clang::TargetInfo child
-  // class. Clang::TargetInfo classes are the basic clang classes and
-  // they cannot be reused by Flang.
-  // That's why we need to extract implicit target features and add
-  // them to the target features specified by the user
-  if (triple.isAMDGPU()) {
-    return getExplicitAndImplicitAMDGPUTargetFeatures(ci, targetOpts, triple);
-  } else if (triple.isNVPTX()) {
-    return getExplicitAndImplicitNVPTXTargetFeatures(ci, targetOpts, triple);
-  }
-  return llvm::join(targetOpts.featuresAsWritten.begin(),
-                    targetOpts.featuresAsWritten.end(), ",");
-}
-
-static void setMLIRDataLayout(mlir::ModuleOp &mlirModule,
-                              const llvm::DataLayout &dl) {
-  mlir::MLIRContext *context = mlirModule.getContext();
-  mlirModule->setAttr(
-      mlir::LLVM::LLVMDialect::getDataLayoutAttrName(),
-      mlir::StringAttr::get(context, dl.getStringRepresentation()));
-  mlir::DataLayoutSpecInterface dlSpec = mlir::translateDataLayout(dl, context);
-  mlirModule->setAttr(mlir::DLTIDialect::kDataLayoutAttrName, dlSpec);
-}
-
 static void addDependentLibs(mlir::ModuleOp &mlirModule, CompilerInstance &ci) {
   const std::vector<std::string> &libs =
       ci.getInvocation().getCodeGenOpts().DependentLibs;
@@ -352,6 +246,8 @@ bool CodeGenAction::beginSourceFileAction() {
   fir::support::loadDialects(*mlirCtx);
   fir::support::registerLLVMTranslation(*mlirCtx);
 
+  const llvm::TargetMachine &targetMachine = ci.getTargetMachine();
+
   // If the input is an MLIR file, just parse it and return.
   if (this->getCurrentInput().getKind().getLanguage() == Language::MLIR) {
     llvm::SourceMgr sourceMgr;
@@ -369,10 +265,8 @@ bool CodeGenAction::beginSourceFileAction() {
     }
 
     mlirModule = std::make_unique<mlir::ModuleOp>(module.release());
-    if (!setUpTargetMachine())
-      return false;
-    const llvm::DataLayout &dl = tm->createDataLayout();
-    setMLIRDataLayout(*mlirModule, dl);
+    const llvm::DataLayout &dl = targetMachine.createDataLayout();
+    fir::support::setMLIRDataLayout(*mlirModule, dl);
     return true;
   }
 
@@ -394,6 +288,8 @@ bool CodeGenAction::beginSourceFileAction() {
       ci.getSemanticsContext().defaultKinds();
   fir::KindMapping kindMap(mlirCtx.get(), llvm::ArrayRef<fir::KindTy>{
                                               fir::fromDefaultKinds(defKinds)});
+  const llvm::DataLayout &dl = targetMachine.createDataLayout();
+
   lower::LoweringBridge lb = Fortran::lower::LoweringBridge::create(
       *mlirCtx, ci.getSemanticsContext(), defKinds,
       ci.getSemanticsContext().intrinsics(),
@@ -401,27 +297,22 @@ bool CodeGenAction::beginSourceFileAction() {
       ci.getParsing().allCooked(), ci.getInvocation().getTargetOpts().triple,
       kindMap, ci.getInvocation().getLoweringOpts(),
       ci.getInvocation().getFrontendOpts().envDefaults,
-      ci.getInvocation().getFrontendOpts().features);
+      ci.getInvocation().getFrontendOpts().features, &dl);
 
   // Fetch module from lb, so we can set
   mlirModule = std::make_unique<mlir::ModuleOp>(lb.getModule());
 
-  if (!setUpTargetMachine())
-    return false;
-
   if (ci.getInvocation().getFrontendOpts().features.IsEnabled(
           Fortran::common::LanguageFeature::OpenMP)) {
     setOffloadModuleInterfaceAttributes(*mlirModule,
                                         ci.getInvocation().getLangOpts());
-    setOffloadModuleInterfaceTargetAttribute(*mlirModule, tm->getTargetCPU(),
-                                             tm->getTargetFeatureString());
+    setOffloadModuleInterfaceTargetAttribute(
+        *mlirModule, targetMachine.getTargetCPU(),
+        targetMachine.getTargetFeatureString());
     setOpenMPVersionAttribute(*mlirModule,
                               ci.getInvocation().getLangOpts().OpenMPVersion);
   }
 
-  const llvm::DataLayout &dl = tm->createDataLayout();
-  setMLIRDataLayout(*mlirModule, dl);
-
   // Create a parse tree and lower it to FIR
   Fortran::parser::Program &parseTree{*ci.getParsing().parseTree()};
   lb.lower(parseTree, ci.getSemanticsContext());
@@ -830,7 +721,7 @@ getVScaleRange(CompilerInstance &ci,
     return std::pair<unsigned, unsigned>(
         langOpts.VScaleMin ? langOpts.VScaleMin : 1, langOpts.VScaleMax);
 
-  std::string featuresStr = getTargetFeatures(ci);
+  std::string featuresStr = ci.getTargetFeatures();
   if (featuresStr.find("+sve") != std::string::npos)
     return std::pair<unsigned, unsigned>(1, 16);
 
@@ -911,38 +802,6 @@ void CodeGenAction::generateLLVMIR() {
   }
 }
 
-bool CodeGenAction::setUpTargetMachine() {
-  CompilerInstance &ci = this->getInstance();
-
-  const TargetOptions &targetOpts = ci.getInvocation().getTargetOpts();
-  const std::string &theTriple = targetOpts.triple;
-
-  // Create `Target`
-  std::string error;
-  const llvm::Target *theTarget =
-      llvm::TargetRegistry::lookupTarget(theTriple, error);
-  if (!theTarget) {
-    ci.getDiagnostics().Report(clang::diag::err_fe_unable_to_create_target)
-        << error;
-    return false;
-  }
-
-  // Create `TargetMachine`
-  const auto &CGOpts = ci.getInvocation().getCodeGenOpts();
-  std::optional<llvm::CodeGenOptLevel> OptLevelOrNone =
-      llvm::CodeGenOpt::getLevel(CGOpts.OptimizationLevel);
-  assert(OptLevelOrNone && "Invalid optimization level!");
-  llvm::CodeGenOptLevel OptLevel = *OptLevelOrNone;
-  std::string featuresStr = getTargetFeatures(ci);
-  tm.reset(theTarget->createTargetMachine(
-      theTriple, /*CPU=*/targetOpts.cpu,
-      /*Features=*/featuresStr, llvm::TargetOptions(),
-      /*Reloc::Model=*/CGOpts.getRelocationModel(),
-      /*CodeModel::Model=*/std::nullopt, OptLevel));
-  assert(tm && "Failed to create TargetMachine");
-  return true;
-}
-
 static std::unique_ptr<llvm::raw_pwrite_stream>
 getOutputStream(CompilerInstance &ci, llvm::StringRef inFile,
                 BackendActionTy action) {
@@ -1020,6 +879,7 @@ void CodeGenAction::runOptimizationPipeline(llvm::raw_pwrite_stream &os) {
   auto &diags = getInstance().getDiagnostics();
   llvm::OptimizationLevel level = mapToLevel(opts);
 
+  llvm::TargetMachine *targetMachine = &getInstance().getTargetMachine();
   // Create the analysis managers.
   llvm::LoopAnalysisManager lam;
   llvm::FunctionAnalysisManager fam;
@@ -1033,7 +893,7 @@ void CodeGenAction::runOptimizationPipeline(llvm::raw_pwrite_stream &os) {
   llvm::StandardInstrumentations si(llvmModule->getContext(),
                                     opts.DebugPassManager);
   si.registerCallbacks(pic, &mam);
-  llvm::PassBuilder pb(tm.get(), pto, pgoOpt, &pic);
+  llvm::PassBuilder pb(targetMachine, pto, pgoOpt, &pic);
 
   // Attempt to load pass plugins and register their callbacks with PB.
   for (auto &pluginFile : opts.LLVMPassPlugins) {
@@ -1299,9 +1159,8 @@ void CodeGenAction::executeAction() {
   // Set the triple based on the targetmachine (this comes compiler invocation
   // and the command-line target option if specified, or the default if not
   // given on the command-line).
-  if (!setUpTargetMachine())
-    return;
-  const std::string &theTriple = tm->getTargetTriple().str();
+  llvm::TargetMachine &targetMachine = ci.getTargetMachine();
+  const std::string &theTriple = targetMachine.getTargetTriple().str();
 
   if (llvmModule->getTargetTriple() != theTriple) {
     diags.Report(clang::diag::warn_fe_override_module) << theTriple;
@@ -1311,7 +1170,7 @@ void CodeGenAction::executeAction() {
   // Note that this overwrites any datalayout stored in the LLVM-IR. This avoids
   // an assert for incompatible data layout when the code-generation happens.
   llvmModule->setTargetTriple(theTriple);
-  llvmModule->setDataLayout(tm->createDataLayout());
+  llvmModule->setDataLayout(targetMachine.createDataLayout());
 
   // Embed offload objects specified with -fembed-offload-object
   if (!codeGenOpts.OffloadObjects.empty())
@@ -1361,7 +1220,7 @@ void CodeGenAction::executeAction() {
   if (action == BackendActionTy::Backend_EmitAssembly ||
       action == BackendActionTy::Backend_EmitObj) {
     generateMachineCodeOrAssemblyImpl(
-        diags, *tm, action, *llvmModule, codeGenOpts,
+        diags, targetMachine, action, *llvmModule, codeGenOpts,
         ci.isOutputStreamNull() ? *os : ci.getOutputStream());
     return;
   }
diff --git a/flang/lib/Lower/Bridge.cpp b/flang/lib/Lower/Bridge.cpp
index 1258e3b..7e64adc 100644
--- a/flang/lib/Lower/Bridge.cpp
+++ b/flang/lib/Lower/Bridge.cpp
@@ -44,6 +44,7 @@
 #include "flang/Optimizer/Dialect/FIROps.h"
 #include "flang/Optimizer/Dialect/Support/FIRContext.h"
 #include "flang/Optimizer/HLFIR/HLFIROps.h"
+#include "flang/Optimizer/Support/DataLayout.h"
 #include "flang/Optimizer/Support/FatalError.h"
 #include "flang/Optimizer/Support/InternalNames.h"
 #include "flang/Optimizer/Transforms/Passes.h"
@@ -5089,7 +5090,8 @@ Fortran::lower::LoweringBridge::LoweringBridge(
     fir::KindMapping &kindMap,
     const Fortran::lower::LoweringOptions &loweringOptions,
     const std::vector<Fortran::lower::EnvironmentDefault> &envDefaults,
-    const Fortran::common::LanguageFeatureControl &languageFeatures)
+    const Fortran::common::LanguageFeatureControl &languageFeatures,
+    const llvm::DataLayout *dataLayout)
     : semanticsContext{semanticsContext}, defaultKinds{defaultKinds},
       intrinsics{intrinsics}, targetCharacteristics{targetCharacteristics},
       cooked{&cooked}, context{context}, kindMap{kindMap},
@@ -5145,4 +5147,6 @@ Fortran::lower::LoweringBridge::LoweringBridge(
   assert(module.get() && "module was not created");
   fir::setTargetTriple(*module.get(), triple);
   fir::setKindMapping(*module.get(), kindMap);
+  if (dataLayout)
+    fir::support::setMLIRDataLayout(*module.get(), *dataLayout);
 }
diff --git a/flang/lib/Optimizer/Support/CMakeLists.txt b/flang/lib/Optimizer/Support/CMakeLists.txt
index 2d03bc5..55f5718 100644
--- a/flang/lib/Optimizer/Support/CMakeLists.txt
+++ b/flang/lib/Optimizer/Support/CMakeLists.txt
@@ -2,6 +2,7 @@ get_property(dialect_libs GLOBAL PROPERTY MLIR_DIALECT_LIBS)
 get_property(extension_libs GLOBAL PROPERTY MLIR_EXTENSION_LIBS)
 
 add_flang_library(FIRSupport
+  DataLayout.cpp
   InitFIR.cpp
   InternalNames.cpp
 
@@ -20,6 +21,7 @@ add_flang_library(FIRSupport
   MLIROpenMPToLLVMIRTranslation
   MLIRLLVMToLLVMIRTranslation
   MLIRTargetLLVMIRExport
+  MLIRTargetLLVMIRImport
 
   LINK_COMPONENTS
   TargetParser
diff --git a/flang/lib/Optimizer/Support/DataLayout.cpp b/flang/lib/Optimizer/Support/DataLayout.cpp
new file mode 100644
index 0000000..5cd9c01
--- /dev/null
+++ b/flang/lib/Optimizer/Support/DataLayout.cpp
@@ -0,0 +1,47 @@
+//===-- Optimizer/Support/DataLayout.cpp ----------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#include "flang/Optimizer/Support/DataLayout.h"
+#include "flang/Optimizer/Dialect/Support/FIRContext.h"
+#include "flang/Optimizer/Support/FatalError.h"
+#include "mlir/Dialect/DLTI/DLTI.h"
+#include "mlir/Dialect/LLVMIR/LLVMDialect.h"
+#include "mlir/IR/BuiltinOps.h"
+#include "mlir/Interfaces/DataLayoutInterfaces.h"
+#include "mlir/Support/LLVM.h"
+#include "mlir/Target/LLVMIR/Import.h"
+#include "llvm/IR/DataLayout.h"
+#include "llvm/MC/TargetRegistry.h"
+#include "llvm/Support/TargetSelect.h"
+#include "llvm/Target/TargetMachine.h"
+
+void fir::support::setMLIRDataLayout(mlir::ModuleOp mlirModule,
+                                     const llvm::DataLayout &dl) {
+  mlir::MLIRContext *context = mlirModule.getContext();
+  mlirModule->setAttr(
+      mlir::LLVM::LLVMDialect::getDataLayoutAttrName(),
+      mlir::StringAttr::get(context, dl.getStringRepresentation()));
+  mlir::DataLayoutSpecInterface dlSpec = mlir::translateDataLayout(dl, context);
+  mlirModule->setAttr(mlir::DLTIDialect::kDataLayoutAttrName, dlSpec);
+}
+
+void fir::support::setMLIRDataLayoutFromAttributes(mlir::ModuleOp mlirModule,
+                                                   bool allowDefaultLayout) {
+  if (mlirModule.getDataLayoutSpec())
+    return; // Already set.
+  if (auto dataLayoutString = mlirModule->getAttrOfType<mlir::StringAttr>(
+          mlir::LLVM::LLVMDialect::getDataLayoutAttrName())) {
+    llvm::DataLayout llvmDataLayout(dataLayoutString);
+    fir::support::setMLIRDataLayout(mlirModule, llvmDataLayout);
+    return;
+  }
+  if (!allowDefaultLayout)
+    return;
+  llvm::DataLayout llvmDataLayout("");
+  fir::support::setMLIRDataLayout(mlirModule, llvmDataLayout);
+}