[mlir] Introduce dialect interfaces for translation to LLVM IR

[ftynse]

Summary

The existing approach to translation to the LLVM IR relies on a single
translation supporting the base LLVM dialect, extensible through inheritance to
support intrinsic-based dialects also derived from LLVM IR such as NVVM and
AVX512. This approach does not scale well as it requires additional
translations to be created for each new intrinsic-based dialect and does not
allow them to mix in the same module, contrary to the rest of the MLIR
infrastructure. Furthermore, OpenMP translation ingrained itself into the main
translation mechanism.

Start refactoring the translation to LLVM IR to operate using dialect
interfaces. Each dialect that contains ops translatable to LLVM IR can
implement the interface for translating them, and the top-level translation
driver can operate on interfaces without knowing about specific dialects.
Furthermore, the delayed dialect registration mechanism allows one to avoid a
dependency on LLVM IR in the dialect that is translated to it by implementing
the translation as a separate library and only registering it at the client
level.

This change introduces the new mechanism and factors out the translation of the
"main" LLVM dialect. The remaining dialects will follow suit.

[mehdi_amini]

Look great! I am only concerned about registerLLVMDialectTranslation(MLIRContext&) right now, the rest LGTM.

mlir/lib/Target/LLVMIR/ConvertToLLVMIR.cpp

46

The fact that this API won't do anything if the dialect isn't loaded does not seem right to me.

What prevents us from always doing unconditionally the addition like this?

DialectRegistry registry;
registry.insert<LLVM::LLVMDialect>();
registry.addDialectInterface<LLVM::LLVMDialect,
                             LLVMDialectLLVMIRTranslationInterface>();
context.appendDialectRegistry(registry);

[ftynse]

Address review

mlir/lib/Target/LLVMIR/ConvertToLLVMIR.cpp

46

Good catch!

We cannot register unconditionally because if the interface is already registered on the dialect (that is, the dialect is already loaded), an attempt to register it for the second time will trigger a double-registration assertion. We should probably fix this and just ignore repeated registrations of the same interface on the same dialect, but that deserves a separate patch.

Changed the condition in this patch to add the interface if the dialect doesn't have it yet _or_ if it has not been loaded.

[ftynse]

mlir/lib/Target/LLVMIR/ConvertToLLVMIR.cpp

46

Also see https://reviews.llvm.org/D96606.

[nicolasvasilache]

Thanks!

mlir/examples/toy/Ch6/toyc.cpp

int dumpAST() {
   if (!moduleAST)
     return 1;
 
   dump(*moduleAST);
   return 0;
 }
 
 int dumpLLVMIR(mlir::ModuleOp module) {
+  // Register the translation to LLVM IR with the MLIR context.
+  mlir::registerLLVMDialectTranslation(*module->getContext());
+
   // Convert the module to LLVM IR in a new LLVM IR context.
   llvm::LLVMContext llvmContext;
   auto llvmModule = mlir::translateModuleToLLVMIR(module, llvmContext);
   if (!llvmModule) {
     llvm::errs() << "Failed to emit LLVM IR\n";
     return -1;
   }
 
int dumpLLVMIR(mlir::ModuleOp module) {
   return 0;
 }
 
 int runJit(mlir::ModuleOp module) {
   // Initialize LLVM targets.
   llvm::InitializeNativeTarget();
   llvm::InitializeNativeTargetAsmPrinter();
 
+  // Register the translation from MLIR to LLVM IR, which must happen before we
+  // can JIT-compile.
+  mlir::registerLLVMDialectTranslation(*module->getContext());
+
   // An optimization pipeline to use within the execution engine.
   auto optPipeline = mlir::makeOptimizingTransformer(
       /*optLevel=*/enableOpt ? 3 : 0, /*sizeLevel=*/0,
       /*targetMachine=*/nullptr);
 
   // Create an MLIR execution engine. The execution engine eagerly JIT-compiles
   // the module.
   auto maybeEngine = mlir::ExecutionEngine::create(

mlir/examples/toy/Ch7/toyc.cpp

int dumpAST() {
   if (!moduleAST)
     return 1;
 
   dump(*moduleAST);
   return 0;
 }
 
 int dumpLLVMIR(mlir::ModuleOp module) {
+  // Register the translation to LLVM IR with the MLIR context.
+  mlir::registerLLVMDialectTranslation(*module->getContext());
+
   // Convert the module to LLVM IR in a new LLVM IR context.
   llvm::LLVMContext llvmContext;
   auto llvmModule = mlir::translateModuleToLLVMIR(module, llvmContext);
   if (!llvmModule) {
     llvm::errs() << "Failed to emit LLVM IR\n";
     return -1;
   }
 
int dumpLLVMIR(mlir::ModuleOp module) {
   return 0;
 }
 
 int runJit(mlir::ModuleOp module) {
   // Initialize LLVM targets.
   llvm::InitializeNativeTarget();
   llvm::InitializeNativeTargetAsmPrinter();
 
+  // Register the translation from MLIR to LLVM IR, which must happen before we
+  // can JIT-compile.
+  mlir::registerLLVMDialectTranslation(*module->getContext());
+
   // An optimization pipeline to use within the execution engine.
   auto optPipeline = mlir::makeOptimizingTransformer(
       /*optLevel=*/enableOpt ? 3 : 0, /*sizeLevel=*/0,
       /*targetMachine=*/nullptr);
 
   // Create an MLIR execution engine. The execution engine eagerly JIT-compiles
   // the module.
   auto maybeEngine = mlir::ExecutionEngine::create(

mlir/include/mlir/Target/LLVMIR.h

 // Forward-declare LLVM classes.
 namespace llvm {
 class LLVMContext;
 class Module;
 } // namespace llvm
 
 namespace mlir {
 
+class DialectRegistry;
 class OwningModuleRef;
 class MLIRContext;
 class ModuleOp;
 
 /// Convert the given MLIR module into LLVM IR.  The LLVM context is extracted
 /// from the registered LLVM IR dialect.  In case of error, report it
 /// to the error handler registered with the MLIR context, if any (obtained from
 /// the MLIR module), and return `nullptr`.
 std::unique_ptr<llvm::Module>
 translateModuleToLLVMIR(ModuleOp m, llvm::LLVMContext &llvmContext,
                         StringRef name = "LLVMDialectModule");
 
 /// Convert the given LLVM module into MLIR's LLVM dialect.  The LLVM context is
 /// extracted from the registered LLVM IR dialect. In case of error, report it
 /// to the error handler registered with the MLIR context, if any (obtained from
 /// the MLIR module), and return `{}`.
 OwningModuleRef
 translateLLVMIRToModule(std::unique_ptr<llvm::Module> llvmModule,
                         MLIRContext *context);
 
+/// Register the LLVM dialect and the translation from it to the LLVM IR in the
+/// given registry;
+void registerLLVMDialectTranslation(DialectRegistry &registry);
+
+/// Register the LLVM dialect and the translation from it in the registry
+/// associated with the given context. This checks if the interface is already
+/// registered and avoids double registation.
+void registerLLVMDialectTranslation(MLIRContext &context);
+
 } // namespace mlir
 
 #endif // MLIR_TARGET_LLVMIR_H

mlir/include/mlir/Target/LLVMIR/Dialect/LLVMIR/LLVMToLLVMIRTranslation.h

+//===- LLVMToLLVMIRTranslation.h - LLVM Dialect to LLVM IR-------*- C++ -*-===//
+//
+// 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
+//
+//===----------------------------------------------------------------------===//
+//
+// This file implements the dialect interface for translating the LLVM dialect
+// to LLVM IR.
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef MLIR_TARGET_LLVMIR_DIALECT_LLVMIR_LLVMTOLLVMIRTRANSLATION_H
+#define MLIR_TARGET_LLVMIR_DIALECT_LLVMIR_LLVMTOLLVMIRTRANSLATION_H
+
+#include "mlir/Target/LLVMIR/LLVMTranslationInterface.h"
+
+namespace mlir {
+
+/// Implementation of the dialect interface that converts operations beloning to
+/// the LLVM dialect to LLVM IR.
+class LLVMDialectLLVMIRTranslationInterface
+    : public LLVMTranslationDialectInterface {
+public:
+  using LLVMTranslationDialectInterface::LLVMTranslationDialectInterface;
+
+  /// Translates the given operation to LLVM IR using the provided IR builder
+  /// and saving the state in `moduleTranslation`.
+  LogicalResult
+  convertOperation(Operation *op, llvm::IRBuilderBase &builder,
+                   LLVM::ModuleTranslation &moduleTranslation) const final;
+};
+
+} // namespace mlir
+
+#endif // MLIR_TARGET_LLVMIR_DIALECT_LLVMIR_LLVMTOLLVMIRTRANSLATION_H

mlir/include/mlir/Target/LLVMIR/LLVMTranslationInterface.h

+//===- LLVMTranslationInterface.h - Translation to LLVM iface ---*- C++ -*-===//
+//
+// 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
+//
+//===----------------------------------------------------------------------===//
+//
+// This header file defines dialect interfaces for translation to LLVM IR.
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef MLIR_TARGET_LLVMIR_LLVMTRANSLATIONINTERFACE_H
+#define MLIR_TARGET_LLVMIR_LLVMTRANSLATIONINTERFACE_H
+
+#include "mlir/IR/DialectInterface.h"
+#include "mlir/Support/LogicalResult.h"
+
+namespace llvm {
+class IRBuilderBase;
+}
+
+namespace mlir {
+namespace LLVM {
+class ModuleTranslation;
+} // namespace LLVM
+
+/// Base class for dialect interfaces providing translation to LLVM IR.
+/// Dialects that can be translated should provide an implementation of this
+/// interface for the supported operations. The interface may be implemented in
+/// a separate library to avoid the "main" dialect library depending on LLVM IR.
+/// The interface can be attached using the delayed registration mechanism
+/// available in DialectRegistry.
+class LLVMTranslationDialectInterface
+    : public DialectInterface::Base<LLVMTranslationDialectInterface> {
+public:
+  LLVMTranslationDialectInterface(Dialect *dialect) : Base(dialect) {}
+
+  /// Hook for derived dialect interface to provide translation of the
+  /// operations to LLVM IR.
+  virtual LogicalResult
+  convertOperation(Operation *op, llvm::IRBuilderBase &builder,
+                   LLVM::ModuleTranslation &moduleTranslation) const {
+    return failure();
+  }
+};
+
+/// Interface collection for translation to LLVM IR, dispatches to a concrete
+/// interface implementation based on the dialect to which the given op belongs.
+class LLVMTranslationInterface
+    : public DialectInterfaceCollection<LLVMTranslationDialectInterface> {
+public:
+  using Base::Base;
+
+  /// Translates the given operation to LLVM IR using the interface implemented
+  /// by the op's dialect.
+  virtual LogicalResult
+  convertOperation(Operation *op, llvm::IRBuilderBase &builder,
+                   LLVM::ModuleTranslation &moduleTranslation) const {
+    if (const LLVMTranslationDialectInterface *iface = getInterfaceFor(op))
+      return iface->convertOperation(op, builder, moduleTranslation);
+    return failure();
+  }
+};
+
+} // namespace mlir
+
+#endif // MLIR_TARGET_LLVMIR_LLVMTRANSLATIONINTERFACE_H

mlir/include/mlir/Target/LLVMIR/ModuleTranslation.h

 #ifndef MLIR_TARGET_LLVMIR_MODULETRANSLATION_H
 #define MLIR_TARGET_LLVMIR_MODULETRANSLATION_H
 
 #include "mlir/Dialect/LLVMIR/LLVMDialect.h"
 #include "mlir/Dialect/LLVMIR/Transforms/LegalizeForExport.h"
 #include "mlir/IR/Block.h"
 #include "mlir/IR/BuiltinOps.h"
 #include "mlir/IR/Value.h"
+#include "mlir/Target/LLVMIR/LLVMTranslationInterface.h"
 #include "mlir/Target/LLVMIR/TypeTranslation.h"
 
 #include "llvm/Frontend/OpenMP/OMPIRBuilder.h"
 #include "llvm/IR/BasicBlock.h"
 #include "llvm/IR/Function.h"
 #include "llvm/IR/IRBuilder.h"
 #include "llvm/IR/MatrixBuilder.h"
 #include "llvm/IR/Value.h"
public:
   }
 
   /// Finds an LLVM IR instruction that corresponds to the given MLIR operation
   /// with successors.
   llvm::Instruction *lookupBranch(Operation *op) const {
     return branchMapping.lookup(op);
   }
 
+  /// Converts the type from MLIR LLVM dialect to LLVM.
+  llvm::Type *convertType(Type type);
+
+  /// Looks up remapped a list of remapped values.
+  SmallVector<llvm::Value *, 8> lookupValues(ValueRange values);
+
+  /// Create an LLVM IR constant of `llvmType` from the MLIR attribute `attr`.
+  /// This currently supports integer, floating point, splat and dense element
+  /// attributes and combinations thereof.  In case of error, report it to `loc`
+  /// and return nullptr.
+  llvm::Constant *getLLVMConstant(llvm::Type *llvmType, Attribute attr,
+                                  Location loc);
+
+  /// Returns the MLIR context of the module being translated.
+  MLIRContext &getContext() { return *mlirModule->getContext(); }
+
+  /// Returns the LLVM context in which the IR is being constructed.
+  llvm::LLVMContext &getLLVMContext() { return llvmModule->getContext(); }
+
+  /// Finds an LLVM IR global value that corresponds to the given MLIR operation
+  /// defining a global value.
+  llvm::GlobalValue *lookupGlobal(Operation *op) {
+    return globalsMapping.lookup(op);
+  }
+
 protected:
   /// Translate the given MLIR module expressed in MLIR LLVM IR dialect into an
   /// LLVM IR module. The MLIR LLVM IR dialect holds a pointer to an
   /// LLVMContext, the LLVM IR module will be created in that context.
   ModuleTranslation(Operation *module,
                     std::unique_ptr<llvm::Module> llvmModule);
   virtual ~ModuleTranslation();
 
   virtual LogicalResult convertOperation(Operation &op,
                                          llvm::IRBuilder<> &builder);
   virtual LogicalResult convertOmpOperation(Operation &op,
                                             llvm::IRBuilder<> &builder);
   virtual LogicalResult convertOmpParallel(Operation &op,
                                            llvm::IRBuilder<> &builder);
   virtual LogicalResult convertOmpMaster(Operation &op,
                                          llvm::IRBuilder<> &builder);
   void convertOmpOpRegions(Region &region, StringRef blockName,
                            llvm::BasicBlock &sourceBlock,
                            llvm::BasicBlock &continuationBlock,
                            llvm::IRBuilder<> &builder,
                            LogicalResult &bodyGenStatus);
   virtual LogicalResult convertOmpWsLoop(Operation &opInst,
                                          llvm::IRBuilder<> &builder);
 
-  /// Converts the type from MLIR LLVM dialect to LLVM.
-  llvm::Type *convertType(Type type);
-
   static std::unique_ptr<llvm::Module>
   prepareLLVMModule(Operation *m, llvm::LLVMContext &llvmContext,
                     StringRef name);
 
-  /// A helper to look up remapped operands in the value remapping table.
-  SmallVector<llvm::Value *, 8> lookupValues(ValueRange values);
-
 private:
   /// Check whether the module contains only supported ops directly in its body.
   static LogicalResult checkSupportedModuleOps(Operation *m);
 
   LogicalResult convertFunctionSignatures();
   LogicalResult convertFunctions();
   LogicalResult convertGlobals();
   LogicalResult convertOneFunction(LLVMFuncOp func);
   LogicalResult convertBlock(Block &bb, bool ignoreArguments,
                              llvm::IRBuilder<> &builder);
 
-  llvm::Constant *getLLVMConstant(llvm::Type *llvmType, Attribute attr,
-                                  Location loc);
-
   /// Original and translated module.
   Operation *mlirModule;
   std::unique_ptr<llvm::Module> llvmModule;
   /// A converter for translating debug information.
   std::unique_ptr<detail::DebugTranslation> debugTranslation;
 
   /// Builder for LLVM IR generation of OpenMP constructs.
   std::unique_ptr<llvm::OpenMPIRBuilder> ompBuilder;
 
   /// Precomputed pointer to OpenMP dialect. Note this can be nullptr if the
   /// OpenMP dialect hasn't been loaded (it is always loaded if there are OpenMP
   /// operations in the module though).
   const Dialect *ompDialect;
 
   /// Mappings between llvm.mlir.global definitions and corresponding globals.
   DenseMap<Operation *, llvm::GlobalValue *> globalsMapping;
 
   /// A stateful object used to translate types.
   TypeToLLVMIRTranslator typeTranslator;
 
-private:
+  LLVMTranslationInterface iface;
+
   /// Mappings between original and translated values, used for lookups.
   llvm::StringMap<llvm::Function *> functionMapping;
   DenseMap<Value, llvm::Value *> valueMapping;
   DenseMap<Block *, llvm::BasicBlock *> blockMapping;
 
   /// A mapping between MLIR LLVM dialect terminators and LLVM IR terminators
   /// they are converted to. This allows for connecting PHI nodes to the source
   /// values after all operations are converted.
   DenseMap<Operation *, llvm::Instruction *> branchMapping;
 };
 
 } // namespace LLVM
 } // namespace mlir
 
 #endif // MLIR_TARGET_LLVMIR_MODULETRANSLATION_H

mlir/lib/Target/CMakeLists.txt

 add_subdirectory(SPIRV)
+add_subdirectory(LLVMIR)
 
 add_mlir_translation_library(MLIRTargetLLVMIRModuleTranslation
   LLVMIR/DebugTranslation.cpp
   LLVMIR/ModuleTranslation.cpp
   LLVMIR/TypeTranslation.cpp
 
   ADDITIONAL_HEADER_DIRS
   ${MLIR_MAIN_INCLUDE_DIR}/mlir/Target/LLVMIR
add_mlir_translation_library(MLIRTargetAVX512
 
   LINK_COMPONENTS
   Core
 
   LINK_LIBS PUBLIC
   MLIRIR
   MLIRLLVMAVX512
   MLIRLLVMIR
+  MLIRTargetLLVMIR
   MLIRTargetLLVMIRModuleTranslation
   )
 
 add_mlir_translation_library(MLIRTargetLLVMIR
   LLVMIR/ConvertFromLLVMIR.cpp
   LLVMIR/ConvertToLLVMIR.cpp
 
   ADDITIONAL_HEADER_DIRS
   ${MLIR_MAIN_INCLUDE_DIR}/mlir/Target/LLVMIR
 
   LINK_COMPONENTS
   Core
   IRReader
 
   LINK_LIBS PUBLIC
+  MLIRLLVMToLLVMIRTranslation
   MLIRTargetLLVMIRModuleTranslation
   )
 
 add_mlir_translation_library(MLIRTargetArmNeon
   LLVMIR/LLVMArmNeonIntr.cpp
 
   ADDITIONAL_HEADER_DIRS
   ${MLIR_MAIN_INCLUDE_DIR}/mlir/Target/LLVMIR
 
   DEPENDS
   MLIRLLVMArmNeonConversionsIncGen
 
   LINK_COMPONENTS
   Core
 
   LINK_LIBS PUBLIC
   MLIRIR
   MLIRLLVMArmNeon
   MLIRLLVMIR
+  MLIRTargetLLVMIR
   MLIRTargetLLVMIRModuleTranslation
   )
 
 add_mlir_translation_library(MLIRTargetArmSVE
   LLVMIR/LLVMArmSVEIntr.cpp
 
   ADDITIONAL_HEADER_DIRS
   ${MLIR_MAIN_INCLUDE_DIR}/mlir/Target/LLVMIR
 
   DEPENDS
   MLIRLLVMArmSVEConversionsIncGen
 
   LINK_COMPONENTS
   Core
 
   LINK_LIBS PUBLIC
   MLIRIR
   MLIRLLVMArmSVE
   MLIRLLVMIR
+  MLIRTargetLLVMIR
   MLIRTargetLLVMIRModuleTranslation
   )
 
 add_mlir_translation_library(MLIRTargetNVVMIR
   LLVMIR/ConvertToNVVMIR.cpp
 
   ADDITIONAL_HEADER_DIRS
   ${MLIR_MAIN_INCLUDE_DIR}/mlir/Target/LLVMIR
 
   DEPENDS
   intrinsics_gen
 
   LINK_COMPONENTS
   Core
 
   LINK_LIBS PUBLIC
   MLIRGPU
   MLIRIR
   MLIRLLVMIR
   MLIRNVVMIR
+  MLIRTargetLLVMIR
   MLIRTargetLLVMIRModuleTranslation
   )
 
 add_mlir_translation_library(MLIRTargetROCDLIR
   LLVMIR/ConvertToROCDLIR.cpp
 
   ADDITIONAL_HEADER_DIRS
   ${MLIR_MAIN_INCLUDE_DIR}/mlir/Target/LLVMIR
 
   DEPENDS
   intrinsics_gen
 
   LINK_COMPONENTS
   Core
 
   LINK_LIBS PUBLIC
   MLIRGPU
   MLIRIR
   MLIRLLVMIR
   MLIRROCDLIR
+  MLIRTargetLLVMIR
   MLIRTargetLLVMIRModuleTranslation
   )

mlir/lib/Target/LLVMIR/CMakeLists.txt

+add_subdirectory(Dialect)

mlir/lib/Target/LLVMIR/ConvertToLLVMIR.cpp

 //===- ConvertToLLVMIR.cpp - MLIR to LLVM IR conversion -------------------===//
 //
 // 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
 //
 //===----------------------------------------------------------------------===//
 //
 // This file implements a translation between the MLIR LLVM dialect and LLVM IR.
 //
 //===----------------------------------------------------------------------===//
 
 #include "mlir/Target/LLVMIR.h"
 
 #include "mlir/Dialect/OpenMP/OpenMPDialect.h"
+#include "mlir/Target/LLVMIR/Dialect/LLVMIR/LLVMToLLVMIRTranslation.h"
 #include "mlir/Target/LLVMIR/ModuleTranslation.h"
 #include "mlir/Translation.h"
 
 #include "llvm/ADT/StringRef.h"
 #include "llvm/IR/Module.h"
 #include "llvm/IR/Verifier.h"
 #include "llvm/Support/ToolOutputFile.h"
 
 using namespace mlir;
 
 std::unique_ptr<llvm::Module>
 mlir::translateModuleToLLVMIR(ModuleOp m, llvm::LLVMContext &llvmContext,
                               StringRef name) {
   auto llvmModule =
       LLVM::ModuleTranslation::translateModule<>(m, llvmContext, name);
   if (!llvmModule)
     emitError(m.getLoc(), "Fail to convert MLIR to LLVM IR");
   else if (verifyModule(*llvmModule))
     emitError(m.getLoc(), "LLVM IR fails to verify");
   return llvmModule;
 }
 
+void mlir::registerLLVMDialectTranslation(DialectRegistry &registry) {
+  registry.insert<LLVM::LLVMDialect>();
+  registry.addDialectInterface<LLVM::LLVMDialect,
+                               LLVMDialectLLVMIRTranslationInterface>();
+}
+
+void mlir::registerLLVMDialectTranslation(MLIRContext &context) {
+  if (auto *dialect = context.getLoadedDialect<LLVM::LLVMDialect>()) {
+    if (!dialect->getRegisteredInterface<
+            LLVMDialectLLVMIRTranslationInterface>()) {
+      DialectRegistry registry;
+      registry.insert<LLVM::LLVMDialect>();
+      registry.addDialectInterface<LLVM::LLVMDialect,
+                                   LLVMDialectLLVMIRTranslationInterface>();
+      context.appendDialectRegistry(registry);
+    }
+  }
+}
+
 namespace mlir {
 void registerToLLVMIRTranslation() {
   TranslateFromMLIRRegistration registration(
       "mlir-to-llvmir",
       [](ModuleOp module, raw_ostream &output) {
         llvm::LLVMContext llvmContext;
         auto llvmModule = LLVM::ModuleTranslation::translateModule<>(
             module, llvmContext, "LLVMDialectModule");
         if (!llvmModule)
           return failure();
 
         llvmModule->print(output, nullptr);
         return success();
       },
       [](DialectRegistry &registry) {
-        registry.insert<LLVM::LLVMDialect, omp::OpenMPDialect>();
+        registry.insert<omp::OpenMPDialect>();
+        registerLLVMDialectTranslation(registry);
       });
 }
 } // namespace mlir

mlir/lib/Target/LLVMIR/ConvertToNVVMIR.cpp

 //===----------------------------------------------------------------------===//
 
 #include "mlir/Target/NVVMIR.h"
 
 #include "mlir/Dialect/GPU/GPUDialect.h"
 #include "mlir/Dialect/LLVMIR/LLVMDialect.h"
 #include "mlir/Dialect/LLVMIR/NVVMDialect.h"
 #include "mlir/IR/BuiltinOps.h"
+#include "mlir/Target/LLVMIR.h"
 #include "mlir/Target/LLVMIR/ModuleTranslation.h"
 #include "mlir/Translation.h"
 
 #include "llvm/ADT/StringRef.h"
 #include "llvm/IR/IntrinsicsNVPTX.h"
 #include "llvm/IR/Module.h"
 #include "llvm/Support/ToolOutputFile.h"
 
#include "mlir/Dialect/LLVMIR/NVVMConversions.inc"
   /// Allow access to the constructor.
   friend LLVM::ModuleTranslation;
 };
 } // namespace
 
 std::unique_ptr<llvm::Module>
 mlir::translateModuleToNVVMIR(Operation *m, llvm::LLVMContext &llvmContext,
                               StringRef name) {
+  // Register the translation to LLVM IR if nobody else did before. This may
+  // happen if this translation is called inside a pass pipeline that converts
+  // GPU dialects to binary blobs without translating the rest of the code.
+  registerLLVMDialectTranslation(*m->getContext());
+
   auto llvmModule = LLVM::ModuleTranslation::translateModule<ModuleTranslation>(
       m, llvmContext, name);
   if (!llvmModule)
     return llvmModule;
 
   // Insert the nvvm.annotations kernel so that the NVVM backend recognizes the
   // function as a kernel.
   for (auto func :
TranslateFromMLIRRegistration registration(
         auto llvmModule = mlir::translateModuleToNVVMIR(module, llvmContext);
         if (!llvmModule)
           return failure();
 
         llvmModule->print(output, nullptr);
         return success();
       },
       [](DialectRegistry &registry) {
-        registry.insert<LLVM::LLVMDialect, NVVM::NVVMDialect>();
+        registry.insert<NVVM::NVVMDialect>();
+        registerLLVMDialectTranslation(registry);
       });
 }
 } // namespace mlir

mlir/lib/Target/LLVMIR/ConvertToROCDLIR.cpp

 //===----------------------------------------------------------------------===//
 
 #include "mlir/Target/ROCDLIR.h"
 
 #include "mlir/Dialect/GPU/GPUDialect.h"
 #include "mlir/Dialect/LLVMIR/LLVMDialect.h"
 #include "mlir/Dialect/LLVMIR/ROCDLDialect.h"
 #include "mlir/IR/BuiltinOps.h"
+#include "mlir/Target/LLVMIR.h"
 #include "mlir/Target/LLVMIR/ModuleTranslation.h"
 #include "mlir/Translation.h"
 
 #include "llvm/ADT/StringRef.h"
 #include "llvm/IR/IntrinsicsAMDGPU.h"
 #include "llvm/IR/Module.h"
 #include "llvm/Support/ToolOutputFile.h"
 
#include "mlir/Dialect/LLVMIR/ROCDLConversions.inc"
   /// Allow access to the constructor.
   friend LLVM::ModuleTranslation;
 };
 } // namespace
 
 std::unique_ptr<llvm::Module>
 mlir::translateModuleToROCDLIR(Operation *m, llvm::LLVMContext &llvmContext,
                                StringRef name) {
-  // lower MLIR (with RODL Dialect) to LLVM IR (with ROCDL intrinsics)
+  // Register the translation to LLVM IR if nobody else did before. This may
+  // happen if this translation is called inside a pass pipeline that converts
+  // GPU dialects to binary blobs without translating the rest of the code.
+  registerLLVMDialectTranslation(*m->getContext());
+
+  // Lower MLIR (with RODL Dialect) to LLVM IR (with ROCDL intrinsics).
   auto llvmModule = LLVM::ModuleTranslation::translateModule<ModuleTranslation>(
       m, llvmContext, name);
 
-  // foreach GPU kernel
+  // Foreach GPU kernel:
   // 1. Insert AMDGPU_KERNEL calling convention.
   // 2. Insert amdgpu-flat-workgroup-size(1, 1024) attribute.
   for (auto func :
        ModuleTranslation::getModuleBody(m).getOps<LLVM::LLVMFuncOp>()) {
     if (!func->getAttrOfType<UnitAttr>(
             gpu::GPUDialect::getKernelFuncAttrName()))
       continue;
 
TranslateFromMLIRRegistration registration(
         auto llvmModule = mlir::translateModuleToROCDLIR(module, llvmContext);
         if (!llvmModule)
           return failure();
 
         llvmModule->print(output, nullptr);
         return success();
       },
       [](DialectRegistry &registry) {
-        registry.insert<ROCDL::ROCDLDialect, LLVM::LLVMDialect>();
+        registry.insert<ROCDL::ROCDLDialect>();
+        registerLLVMDialectTranslation(registry);
       });
 }
 } // namespace mlir

mlir/lib/Target/LLVMIR/Dialect/CMakeLists.txt

+add_subdirectory(LLVMIR)

mlir/lib/Target/LLVMIR/Dialect/LLVMIR/CMakeLists.txt

+add_mlir_translation_library(MLIRLLVMToLLVMIRTranslation
+  LLVMToLLVMIRTranslation.cpp
+
+  LINK_COMPONENTS
+  Core
+
+  LINK_LIBS PUBLIC
+  MLIRIR
+  MLIRLLVMIR
+  MLIRSupport
+  MLIRTargetLLVMIRModuleTranslation
+  )

mlir/lib/Target/LLVMIR/Dialect/LLVMIR/LLVMToLLVMIRTranslation.cpp

+//===- LLVMToLLVMIRTranslation.cpp - Translate LLVM dialect to LLVM IR ----===//
+//
+// 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
+//
+//===----------------------------------------------------------------------===//
+//
+// This file implements a translation between the MLIR LLVM dialect and LLVM IR.
+//
+//===----------------------------------------------------------------------===//
+
+#include "mlir/Target/LLVMIR/Dialect/LLVMIR/LLVMToLLVMIRTranslation.h"
+#include "mlir/Dialect/LLVMIR/LLVMDialect.h"
+#include "mlir/IR/Operation.h"
+#include "mlir/Support/LLVM.h"
+#include "mlir/Target/LLVMIR/ModuleTranslation.h"
+
+#include "llvm/IR/IRBuilder.h"
+#include "llvm/IR/InlineAsm.h"
+#include "llvm/IR/MDBuilder.h"
+#include "llvm/IR/Operator.h"
+
+using namespace mlir;
+using namespace mlir::LLVM;
+
+#include "mlir/Dialect/LLVMIR/LLVMConversionEnumsToLLVM.inc"
+
+/// Convert MLIR integer comparison predicate to LLVM IR comparison predicate.
+static llvm::CmpInst::Predicate getLLVMCmpPredicate(ICmpPredicate p) {
+  switch (p) {
+  case LLVM::ICmpPredicate::eq:
+    return llvm::CmpInst::Predicate::ICMP_EQ;
+  case LLVM::ICmpPredicate::ne:
+    return llvm::CmpInst::Predicate::ICMP_NE;
+  case LLVM::ICmpPredicate::slt:
+    return llvm::CmpInst::Predicate::ICMP_SLT;
+  case LLVM::ICmpPredicate::sle:
+    return llvm::CmpInst::Predicate::ICMP_SLE;
+  case LLVM::ICmpPredicate::sgt:
+    return llvm::CmpInst::Predicate::ICMP_SGT;
+  case LLVM::ICmpPredicate::sge:
+    return llvm::CmpInst::Predicate::ICMP_SGE;
+  case LLVM::ICmpPredicate::ult:
+    return llvm::CmpInst::Predicate::ICMP_ULT;
+  case LLVM::ICmpPredicate::ule:
+    return llvm::CmpInst::Predicate::ICMP_ULE;
+  case LLVM::ICmpPredicate::ugt:
+    return llvm::CmpInst::Predicate::ICMP_UGT;
+  case LLVM::ICmpPredicate::uge:
+    return llvm::CmpInst::Predicate::ICMP_UGE;
+  }
+  llvm_unreachable("incorrect comparison predicate");
+}
+
+static llvm::CmpInst::Predicate getLLVMCmpPredicate(FCmpPredicate p) {
+  switch (p) {
+  case LLVM::FCmpPredicate::_false:
+    return llvm::CmpInst::Predicate::FCMP_FALSE;
+  case LLVM::FCmpPredicate::oeq:
+    return llvm::CmpInst::Predicate::FCMP_OEQ;
+  case LLVM::FCmpPredicate::ogt:
+    return llvm::CmpInst::Predicate::FCMP_OGT;
+  case LLVM::FCmpPredicate::oge:
+    return llvm::CmpInst::Predicate::FCMP_OGE;
+  case LLVM::FCmpPredicate::olt:
+    return llvm::CmpInst::Predicate::FCMP_OLT;
+  case LLVM::FCmpPredicate::ole:
+    return llvm::CmpInst::Predicate::FCMP_OLE;
+  case LLVM::FCmpPredicate::one:
+    return llvm::CmpInst::Predicate::FCMP_ONE;
+  case LLVM::FCmpPredicate::ord:
+    return llvm::CmpInst::Predicate::FCMP_ORD;
+  case LLVM::FCmpPredicate::ueq:
+    return llvm::CmpInst::Predicate::FCMP_UEQ;
+  case LLVM::FCmpPredicate::ugt:
+    return llvm::CmpInst::Predicate::FCMP_UGT;
+  case LLVM::FCmpPredicate::uge:
+    return llvm::CmpInst::Predicate::FCMP_UGE;
+  case LLVM::FCmpPredicate::ult:
+    return llvm::CmpInst::Predicate::FCMP_ULT;
+  case LLVM::FCmpPredicate::ule:
+    return llvm::CmpInst::Predicate::FCMP_ULE;
+  case LLVM::FCmpPredicate::une:
+    return llvm::CmpInst::Predicate::FCMP_UNE;
+  case LLVM::FCmpPredicate::uno:
+    return llvm::CmpInst::Predicate::FCMP_UNO;
+  case LLVM::FCmpPredicate::_true:
+    return llvm::CmpInst::Predicate::FCMP_TRUE;
+  }
+  llvm_unreachable("incorrect comparison predicate");
+}
+
+static llvm::AtomicRMWInst::BinOp getLLVMAtomicBinOp(AtomicBinOp op) {
+  switch (op) {
+  case LLVM::AtomicBinOp::xchg:
+    return llvm::AtomicRMWInst::BinOp::Xchg;
+  case LLVM::AtomicBinOp::add:
+    return llvm::AtomicRMWInst::BinOp::Add;
+  case LLVM::AtomicBinOp::sub:
+    return llvm::AtomicRMWInst::BinOp::Sub;
+  case LLVM::AtomicBinOp::_and:
+    return llvm::AtomicRMWInst::BinOp::And;
+  case LLVM::AtomicBinOp::nand:
+    return llvm::AtomicRMWInst::BinOp::Nand;
+  case LLVM::AtomicBinOp::_or:
+    return llvm::AtomicRMWInst::BinOp::Or;
+  case LLVM::AtomicBinOp::_xor:
+    return llvm::AtomicRMWInst::BinOp::Xor;
+  case LLVM::AtomicBinOp::max:
+    return llvm::AtomicRMWInst::BinOp::Max;
+  case LLVM::AtomicBinOp::min:
+    return llvm::AtomicRMWInst::BinOp::Min;
+  case LLVM::AtomicBinOp::umax:
+    return llvm::AtomicRMWInst::BinOp::UMax;
+  case LLVM::AtomicBinOp::umin:
+    return llvm::AtomicRMWInst::BinOp::UMin;
+  case LLVM::AtomicBinOp::fadd:
+    return llvm::AtomicRMWInst::BinOp::FAdd;
+  case LLVM::AtomicBinOp::fsub:
+    return llvm::AtomicRMWInst::BinOp::FSub;
+  }
+  llvm_unreachable("incorrect atomic binary operator");
+}
+
+static llvm::AtomicOrdering getLLVMAtomicOrdering(AtomicOrdering ordering) {
+  switch (ordering) {
+  case LLVM::AtomicOrdering::not_atomic:
+    return llvm::AtomicOrdering::NotAtomic;
+  case LLVM::AtomicOrdering::unordered:
+    return llvm::AtomicOrdering::Unordered;
+  case LLVM::AtomicOrdering::monotonic:
+    return llvm::AtomicOrdering::Monotonic;
+  case LLVM::AtomicOrdering::acquire:
+    return llvm::AtomicOrdering::Acquire;
+  case LLVM::AtomicOrdering::release:
+    return llvm::AtomicOrdering::Release;
+  case LLVM::AtomicOrdering::acq_rel:
+    return llvm::AtomicOrdering::AcquireRelease;
+  case LLVM::AtomicOrdering::seq_cst:
+    return llvm::AtomicOrdering::SequentiallyConsistent;
+  }
+  llvm_unreachable("incorrect atomic ordering");
+}
+
+static llvm::FastMathFlags getFastmathFlags(FastmathFlagsInterface &op) {
+  using llvmFMF = llvm::FastMathFlags;
+  using FuncT = void (llvmFMF::*)(bool);
+  const std::pair<FastmathFlags, FuncT> handlers[] = {
+      // clang-format off
+      {FastmathFlags::nnan,     &llvmFMF::setNoNaNs},
+      {FastmathFlags::ninf,     &llvmFMF::setNoInfs},
+      {FastmathFlags::nsz,      &llvmFMF::setNoSignedZeros},
+      {FastmathFlags::arcp,     &llvmFMF::setAllowReciprocal},
+      {FastmathFlags::contract, &llvmFMF::setAllowContract},
+      {FastmathFlags::afn,      &llvmFMF::setApproxFunc},
+      {FastmathFlags::reassoc,  &llvmFMF::setAllowReassoc},
+      {FastmathFlags::fast,     &llvmFMF::setFast},
+      // clang-format on
+  };
+  llvm::FastMathFlags ret;
+  auto fmf = op.fastmathFlags();
+  for (auto it : handlers)
+    if (bitEnumContains(fmf, it.first))
+      (ret.*(it.second))(true);
+  return ret;
+}
+
+namespace {
+/// Dispatcher functional object targeting different overloads of
+/// ModuleTranslation::mapValue.
+// TODO: this is only necessary for compatibility with the code emitted from
+// ODS, remove when ODS is updated (after all dialects have migrated to the new
+// translation mechanism).
+struct MapValueDispatcher {
+  explicit MapValueDispatcher(ModuleTranslation &mt) : moduleTranslation(mt) {}
+
+  llvm::Value *&operator()(mlir::Value v) {
+    return moduleTranslation.mapValue(v);
+  }
+
+  void operator()(mlir::Value m, llvm::Value *l) {
+    moduleTranslation.mapValue(m, l);
+  }
+
+  LLVM::ModuleTranslation &moduleTranslation;
+};
+} // end namespace
+
+static LogicalResult
+convertOperationImpl(Operation &opInst, llvm::IRBuilderBase &builder,
+                     LLVM::ModuleTranslation &moduleTranslation) {
+  auto extractPosition = [](ArrayAttr attr) {
+    SmallVector<unsigned, 4> position;
+    position.reserve(attr.size());
+    for (Attribute v : attr)
+      position.push_back(v.cast<IntegerAttr>().getValue().getZExtValue());
+    return position;
+  };
+
+  llvm::IRBuilder<>::FastMathFlagGuard fmfGuard(builder);
+  if (auto fmf = dyn_cast<FastmathFlagsInterface>(opInst))
+    builder.setFastMathFlags(getFastmathFlags(fmf));
+
+  // TODO: these are necessary for compatibility with the code emitted from ODS,
+  // remove them when ODS is updated (after all dialects have migrated to the
+  // new translation mechanism).
+  MapValueDispatcher mapValue(moduleTranslation);
+  auto lookupValue = [&](mlir::Value v) {
+    return moduleTranslation.lookupValue(v);
+  };
+  auto convertType = [&](Type ty) { return moduleTranslation.convertType(ty); };
+  auto lookupValues = [&](ValueRange vs) {
+    return moduleTranslation.lookupValues(vs);
+  };
+  auto getLLVMConstant = [&](llvm::Type *ty, Attribute attr, Location loc) {
+    return moduleTranslation.getLLVMConstant(ty, attr, loc);
+  };
+
+#include "mlir/Dialect/LLVMIR/LLVMConversions.inc"
+
+  // Emit function calls.  If the "callee" attribute is present, this is a
+  // direct function call and we also need to look up the remapped function
+  // itself.  Otherwise, this is an indirect call and the callee is the first
+  // operand, look it up as a normal value.  Return the llvm::Value representing
+  // the function result, which may be of llvm::VoidTy type.
+  auto convertCall = [&](Operation &op) -> llvm::Value * {
+    auto operands = moduleTranslation.lookupValues(op.getOperands());
+    ArrayRef<llvm::Value *> operandsRef(operands);
+    if (auto attr = op.getAttrOfType<FlatSymbolRefAttr>("callee"))
+      return builder.CreateCall(
+          moduleTranslation.lookupFunction(attr.getValue()), operandsRef);
+    auto *calleePtrType =
+        cast<llvm::PointerType>(operandsRef.front()->getType());
+    auto *calleeType =
+        cast<llvm::FunctionType>(calleePtrType->getElementType());
+    return builder.CreateCall(calleeType, operandsRef.front(),
+                              operandsRef.drop_front());
+  };
+
+  // Emit calls.  If the called function has a result, remap the corresponding
+  // value.  Note that LLVM IR dialect CallOp has either 0 or 1 result.
+  if (isa<LLVM::CallOp>(opInst)) {
+    llvm::Value *result = convertCall(opInst);
+    if (opInst.getNumResults() != 0) {
+      mapValue(opInst.getResult(0), result);
+      return success();
+    }
+    // Check that LLVM call returns void for 0-result functions.
+    return success(result->getType()->isVoidTy());
+  }
+
+  if (auto inlineAsmOp = dyn_cast<LLVM::InlineAsmOp>(opInst)) {
+    // TODO: refactor function type creation which usually occurs in std-LLVM
+    // conversion.
+    SmallVector<Type, 8> operandTypes;
+    operandTypes.reserve(inlineAsmOp.operands().size());
+    for (auto t : inlineAsmOp.operands().getTypes())
+      operandTypes.push_back(t);
+
+    Type resultType;
+    if (inlineAsmOp.getNumResults() == 0) {
+      resultType = LLVM::LLVMVoidType::get(&moduleTranslation.getContext());
+    } else {
+      assert(inlineAsmOp.getNumResults() == 1);
+      resultType = inlineAsmOp.getResultTypes()[0];
+    }
+    auto ft = LLVM::LLVMFunctionType::get(resultType, operandTypes);
+    llvm::InlineAsm *inlineAsmInst =
+        inlineAsmOp.asm_dialect().hasValue()
+            ? llvm::InlineAsm::get(
+                  static_cast<llvm::FunctionType *>(convertType(ft)),
+                  inlineAsmOp.asm_string(), inlineAsmOp.constraints(),
+                  inlineAsmOp.has_side_effects(), inlineAsmOp.is_align_stack(),
+                  convertAsmDialectToLLVM(*inlineAsmOp.asm_dialect()))
+            : llvm::InlineAsm::get(
+                  static_cast<llvm::FunctionType *>(convertType(ft)),
+                  inlineAsmOp.asm_string(), inlineAsmOp.constraints(),
+                  inlineAsmOp.has_side_effects(), inlineAsmOp.is_align_stack());
+    llvm::Value *result =
+        builder.CreateCall(inlineAsmInst, lookupValues(inlineAsmOp.operands()));
+    if (opInst.getNumResults() != 0)
+      mapValue(opInst.getResult(0), result);
+    return success();
+  }
+
+  if (auto invOp = dyn_cast<LLVM::InvokeOp>(opInst)) {
+    auto operands = lookupValues(opInst.getOperands());
+    ArrayRef<llvm::Value *> operandsRef(operands);
+    if (auto attr = opInst.getAttrOfType<FlatSymbolRefAttr>("callee")) {
+      builder.CreateInvoke(moduleTranslation.lookupFunction(attr.getValue()),
+                           moduleTranslation.lookupBlock(invOp.getSuccessor(0)),
+                           moduleTranslation.lookupBlock(invOp.getSuccessor(1)),
+                           operandsRef);
+    } else {
+      auto *calleePtrType =
+          cast<llvm::PointerType>(operandsRef.front()->getType());
+      auto *calleeType =
+          cast<llvm::FunctionType>(calleePtrType->getElementType());
+      builder.CreateInvoke(calleeType, operandsRef.front(),
+                           moduleTranslation.lookupBlock(invOp.getSuccessor(0)),
+                           moduleTranslation.lookupBlock(invOp.getSuccessor(1)),
+                           operandsRef.drop_front());
+    }
+    return success();
+  }
+
+  if (auto lpOp = dyn_cast<LLVM::LandingpadOp>(opInst)) {
+    llvm::Type *ty = convertType(lpOp.getType());
+    llvm::LandingPadInst *lpi =
+        builder.CreateLandingPad(ty, lpOp.getNumOperands());
+
+    // Add clauses
+    for (llvm::Value *operand : lookupValues(lpOp.getOperands())) {
+      // All operands should be constant - checked by verifier
+      if (auto *constOperand = dyn_cast<llvm::Constant>(operand))
+        lpi->addClause(constOperand);
+    }
+    mapValue(lpOp.getResult(), lpi);
+    return success();
+  }
+
+  // Emit branches.  We need to look up the remapped blocks and ignore the block
+  // arguments that were transformed into PHI nodes.
+  if (auto brOp = dyn_cast<LLVM::BrOp>(opInst)) {
+    llvm::BranchInst *branch =
+        builder.CreateBr(moduleTranslation.lookupBlock(brOp.getSuccessor()));
+    moduleTranslation.mapBranch(&opInst, branch);
+    return success();
+  }
+  if (auto condbrOp = dyn_cast<LLVM::CondBrOp>(opInst)) {
+    auto weights = condbrOp.branch_weights();
+    llvm::MDNode *branchWeights = nullptr;
+    if (weights) {
+      // Map weight attributes to LLVM metadata.
+      auto trueWeight =
+          weights.getValue().getValue(0).cast<IntegerAttr>().getInt();
+      auto falseWeight =
+          weights.getValue().getValue(1).cast<IntegerAttr>().getInt();
+      branchWeights =
+          llvm::MDBuilder(moduleTranslation.getLLVMContext())
+              .createBranchWeights(static_cast<uint32_t>(trueWeight),
+                                   static_cast<uint32_t>(falseWeight));
+    }
+    llvm::BranchInst *branch = builder.CreateCondBr(
+        moduleTranslation.lookupValue(condbrOp.getOperand(0)),
+        moduleTranslation.lookupBlock(condbrOp.getSuccessor(0)),
+        moduleTranslation.lookupBlock(condbrOp.getSuccessor(1)), branchWeights);
+    moduleTranslation.mapBranch(&opInst, branch);
+    return success();
+  }
+  if (auto switchOp = dyn_cast<LLVM::SwitchOp>(opInst)) {
+    llvm::MDNode *branchWeights = nullptr;
+    if (auto weights = switchOp.branch_weights()) {
+      llvm::SmallVector<uint32_t> weightValues;
+      weightValues.reserve(weights->size());
+      for (llvm::APInt weight : weights->cast<DenseIntElementsAttr>())
+        weightValues.push_back(weight.getLimitedValue());
+      branchWeights = llvm::MDBuilder(moduleTranslation.getLLVMContext())
+                          .createBranchWeights(weightValues);
+    }
+
+    llvm::SwitchInst *switchInst = builder.CreateSwitch(
+        moduleTranslation.lookupValue(switchOp.value()),
+        moduleTranslation.lookupBlock(switchOp.defaultDestination()),
+        switchOp.caseDestinations().size(), branchWeights);
+
+    auto *ty =
+        llvm::cast<llvm::IntegerType>(convertType(switchOp.value().getType()));
+    for (auto i :
+         llvm::zip(switchOp.case_values()->cast<DenseIntElementsAttr>(),
+                   switchOp.caseDestinations()))
+      switchInst->addCase(
+          llvm::ConstantInt::get(ty, std::get<0>(i).getLimitedValue()),
+          moduleTranslation.lookupBlock(std::get<1>(i)));
+
+    moduleTranslation.mapBranch(&opInst, switchInst);
+    return success();
+  }
+
+  // Emit addressof.  We need to look up the global value referenced by the
+  // operation and store it in the MLIR-to-LLVM value mapping.  This does not
+  // emit any LLVM instruction.
+  if (auto addressOfOp = dyn_cast<LLVM::AddressOfOp>(opInst)) {
+    LLVM::GlobalOp global = addressOfOp.getGlobal();
+    LLVM::LLVMFuncOp function = addressOfOp.getFunction();
+
+    // The verifier should not have allowed this.
+    assert((global || function) &&
+           "referencing an undefined global or function");
+
+    mapValue(addressOfOp.getResult(),
+             global ? moduleTranslation.lookupGlobal(global)
+                    : moduleTranslation.lookupFunction(function.getName()));
+    return success();
+  }
+
+  return failure();
+}
+
+LogicalResult mlir::LLVMDialectLLVMIRTranslationInterface::convertOperation(
+    Operation *op, llvm::IRBuilderBase &builder,
+    LLVM::ModuleTranslation &moduleTranslation) const {
+  return convertOperationImpl(*op, builder, moduleTranslation);
+}

mlir/lib/Target/LLVMIR/LLVMAVX512Intr.cpp

 //===- AVX512Intr.cpp - Convert MLIR LLVM dialect to LLVM intrinsics ------===//
 //
 // 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
 //
 //===----------------------------------------------------------------------===//
 //
 // This file implements a translation between the MLIR LLVM and AVX512 dialects
 // and LLVM IR with AVX intrinsics.
 //
 //===----------------------------------------------------------------------===//
 
 #include "mlir/Dialect/LLVMIR/LLVMAVX512Dialect.h"
+#include "mlir/Target/LLVMIR.h"
 #include "mlir/Target/LLVMIR/ModuleTranslation.h"
 #include "mlir/Translation.h"
 #include "llvm/IR/IntrinsicsX86.h"
 
 using namespace mlir;
 
 namespace {
 class LLVMAVX512ModuleTranslation : public LLVM::ModuleTranslation {
TranslateFromMLIRRegistration reg(
             module, llvmContext, "LLVMDialectModule");
         if (!llvmModule)
           return failure();
 
         llvmModule->print(output, nullptr);
         return success();
       },
       [](DialectRegistry &registry) {
-        registry.insert<LLVM::LLVMAVX512Dialect, LLVM::LLVMDialect>();
+        registry.insert<LLVM::LLVMAVX512Dialect>();
+        registerLLVMDialectTranslation(registry);
       });
 }
 } // namespace mlir

mlir/lib/Target/LLVMIR/LLVMArmNeonIntr.cpp

 //===- ArmNeonIntr.cpp - Convert MLIR LLVM dialect to LLVM intrinsics -----===//
 //
 // 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
 //
 //===----------------------------------------------------------------------===//
 //
 // This file implements a translation between the MLIR LLVM and ArmNeon dialects
 // and LLVM IR with ArmNeon intrinsics.
 //
 //===----------------------------------------------------------------------===//
 
 #include "mlir/Dialect/LLVMIR/LLVMArmNeonDialect.h"
+#include "mlir/Target/LLVMIR.h"
 #include "mlir/Target/LLVMIR/ModuleTranslation.h"
 #include "mlir/Translation.h"
 #include "llvm/IR/IntrinsicsAArch64.h"
 
 using namespace mlir;
 
 namespace {
 class LLVMArmNeonModuleTranslation : public LLVM::ModuleTranslation {
TranslateFromMLIRRegistration reg(
             module, llvmContext, "LLVMDialectModule");
         if (!llvmModule)
           return failure();
 
         llvmModule->print(output, nullptr);
         return success();
       },
       [](DialectRegistry &registry) {
-        registry.insert<LLVM::LLVMArmNeonDialect, LLVM::LLVMDialect>();
+        registry.insert<LLVM::LLVMArmNeonDialect>();
+        registerLLVMDialectTranslation(registry);
       });
 }
 } // namespace mlir

mlir/lib/Target/LLVMIR/LLVMArmSVEIntr.cpp

 //===- LLVMArmSVEIntr.cpp - Convert MLIR LLVM dialect to LLVM intrinsics --===//
 //
 // 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
 //
 //===----------------------------------------------------------------------===//
 //
 // This file implements a translation between the MLIR LLVM and ArmSVE dialects
 // and LLVM IR with Arm SVE intrinsics.
 //
 //===----------------------------------------------------------------------===//
 
 #include "mlir/Dialect/LLVMIR/LLVMArmSVEDialect.h"
+#include "mlir/Target/LLVMIR.h"
 #include "mlir/Target/LLVMIR/ModuleTranslation.h"
 #include "mlir/Translation.h"
 #include "llvm/IR/IntrinsicsAArch64.h"
 
 using namespace mlir;
 
 namespace {
 class LLVMArmSVEModuleTranslation : public LLVM::ModuleTranslation {
TranslateFromMLIRRegistration reg(
             module, llvmContext, "LLVMDialectModule");
         if (!llvmModule)
           return failure();
 
         llvmModule->print(output, nullptr);
         return success();
       },
       [](DialectRegistry &registry) {
-        registry.insert<LLVM::LLVMArmSVEDialect, LLVM::LLVMDialect>();
+        registry.insert<LLVM::LLVMArmSVEDialect>();
+        registerLLVMDialectTranslation(registry);
       });
 }
 } // namespace mlir

mlir/lib/Target/LLVMIR/ModuleTranslation.cpp

 #include "DebugTranslation.h"
 #include "mlir/Dialect/LLVMIR/LLVMDialect.h"
 #include "mlir/Dialect/OpenMP/OpenMPDialect.h"
 #include "mlir/IR/Attributes.h"
 #include "mlir/IR/BuiltinOps.h"
 #include "mlir/IR/BuiltinTypes.h"
 #include "mlir/IR/RegionGraphTraits.h"
 #include "mlir/Support/LLVM.h"
+#include "mlir/Target/LLVMIR/LLVMTranslationInterface.h"
 #include "mlir/Target/LLVMIR/TypeTranslation.h"
 #include "llvm/ADT/TypeSwitch.h"
 
 #include "llvm/ADT/PostOrderIterator.h"
 #include "llvm/ADT/SetVector.h"
 #include "llvm/Frontend/OpenMP/OMPIRBuilder.h"
 #include "llvm/IR/BasicBlock.h"
 #include "llvm/IR/CFG.h"
if (auto stringAttr = attr.dyn_cast<StringAttr>()) {
     return llvm::ConstantDataArray::get(
         llvmModule->getContext(), ArrayRef<char>{stringAttr.getValue().data(),
                                                  stringAttr.getValue().size()});
   }
   emitError(loc, "unsupported constant value");
   return nullptr;
 }
 
-/// Convert MLIR integer comparison predicate to LLVM IR comparison predicate.
-static llvm::CmpInst::Predicate getLLVMCmpPredicate(ICmpPredicate p) {
-  switch (p) {
-  case LLVM::ICmpPredicate::eq:
-    return llvm::CmpInst::Predicate::ICMP_EQ;
-  case LLVM::ICmpPredicate::ne:
-    return llvm::CmpInst::Predicate::ICMP_NE;
-  case LLVM::ICmpPredicate::slt:
-    return llvm::CmpInst::Predicate::ICMP_SLT;
-  case LLVM::ICmpPredicate::sle:
-    return llvm::CmpInst::Predicate::ICMP_SLE;
-  case LLVM::ICmpPredicate::sgt:
-    return llvm::CmpInst::Predicate::ICMP_SGT;
-  case LLVM::ICmpPredicate::sge:
-    return llvm::CmpInst::Predicate::ICMP_SGE;
-  case LLVM::ICmpPredicate::ult:
-    return llvm::CmpInst::Predicate::ICMP_ULT;
-  case LLVM::ICmpPredicate::ule:
-    return llvm::CmpInst::Predicate::ICMP_ULE;
-  case LLVM::ICmpPredicate::ugt:
-    return llvm::CmpInst::Predicate::ICMP_UGT;
-  case LLVM::ICmpPredicate::uge:
-    return llvm::CmpInst::Predicate::ICMP_UGE;
-  }
-  llvm_unreachable("incorrect comparison predicate");
-}
-
-static llvm::CmpInst::Predicate getLLVMCmpPredicate(FCmpPredicate p) {
-  switch (p) {
-  case LLVM::FCmpPredicate::_false:
-    return llvm::CmpInst::Predicate::FCMP_FALSE;
-  case LLVM::FCmpPredicate::oeq:
-    return llvm::CmpInst::Predicate::FCMP_OEQ;
-  case LLVM::FCmpPredicate::ogt:
-    return llvm::CmpInst::Predicate::FCMP_OGT;
-  case LLVM::FCmpPredicate::oge:
-    return llvm::CmpInst::Predicate::FCMP_OGE;
-  case LLVM::FCmpPredicate::olt:
-    return llvm::CmpInst::Predicate::FCMP_OLT;
-  case LLVM::FCmpPredicate::ole:
-    return llvm::CmpInst::Predicate::FCMP_OLE;
-  case LLVM::FCmpPredicate::one:
-    return llvm::CmpInst::Predicate::FCMP_ONE;
-  case LLVM::FCmpPredicate::ord:
-    return llvm::CmpInst::Predicate::FCMP_ORD;
-  case LLVM::FCmpPredicate::ueq:
-    return llvm::CmpInst::Predicate::FCMP_UEQ;
-  case LLVM::FCmpPredicate::ugt:
-    return llvm::CmpInst::Predicate::FCMP_UGT;
-  case LLVM::FCmpPredicate::uge:
-    return llvm::CmpInst::Predicate::FCMP_UGE;
-  case LLVM::FCmpPredicate::ult:
-    return llvm::CmpInst::Predicate::FCMP_ULT;
-  case LLVM::FCmpPredicate::ule:
-    return llvm::CmpInst::Predicate::FCMP_ULE;
-  case LLVM::FCmpPredicate::une:
-    return llvm::CmpInst::Predicate::FCMP_UNE;
-  case LLVM::FCmpPredicate::uno:
-    return llvm::CmpInst::Predicate::FCMP_UNO;
-  case LLVM::FCmpPredicate::_true:
-    return llvm::CmpInst::Predicate::FCMP_TRUE;
-  }
-  llvm_unreachable("incorrect comparison predicate");
-}
-
-static llvm::AtomicRMWInst::BinOp getLLVMAtomicBinOp(AtomicBinOp op) {
-  switch (op) {
-  case LLVM::AtomicBinOp::xchg:
-    return llvm::AtomicRMWInst::BinOp::Xchg;
-  case LLVM::AtomicBinOp::add:
-    return llvm::AtomicRMWInst::BinOp::Add;
-  case LLVM::AtomicBinOp::sub:
-    return llvm::AtomicRMWInst::BinOp::Sub;
-  case LLVM::AtomicBinOp::_and:
-    return llvm::AtomicRMWInst::BinOp::And;
-  case LLVM::AtomicBinOp::nand:
-    return llvm::AtomicRMWInst::BinOp::Nand;
-  case LLVM::AtomicBinOp::_or:
-    return llvm::AtomicRMWInst::BinOp::Or;
-  case LLVM::AtomicBinOp::_xor:
-    return llvm::AtomicRMWInst::BinOp::Xor;
-  case LLVM::AtomicBinOp::max:
-    return llvm::AtomicRMWInst::BinOp::Max;
-  case LLVM::AtomicBinOp::min:
-    return llvm::AtomicRMWInst::BinOp::Min;
-  case LLVM::AtomicBinOp::umax:
-    return llvm::AtomicRMWInst::BinOp::UMax;
-  case LLVM::AtomicBinOp::umin:
-    return llvm::AtomicRMWInst::BinOp::UMin;
-  case LLVM::AtomicBinOp::fadd:
-    return llvm::AtomicRMWInst::BinOp::FAdd;
-  case LLVM::AtomicBinOp::fsub:
-    return llvm::AtomicRMWInst::BinOp::FSub;
-  }
-  llvm_unreachable("incorrect atomic binary operator");
-}
-
-static llvm::AtomicOrdering getLLVMAtomicOrdering(AtomicOrdering ordering) {
-  switch (ordering) {
-  case LLVM::AtomicOrdering::not_atomic:
-    return llvm::AtomicOrdering::NotAtomic;
-  case LLVM::AtomicOrdering::unordered:
-    return llvm::AtomicOrdering::Unordered;
-  case LLVM::AtomicOrdering::monotonic:
-    return llvm::AtomicOrdering::Monotonic;
-  case LLVM::AtomicOrdering::acquire:
-    return llvm::AtomicOrdering::Acquire;
-  case LLVM::AtomicOrdering::release:
-    return llvm::AtomicOrdering::Release;
-  case LLVM::AtomicOrdering::acq_rel:
-    return llvm::AtomicOrdering::AcquireRelease;
-  case LLVM::AtomicOrdering::seq_cst:
-    return llvm::AtomicOrdering::SequentiallyConsistent;
-  }
-  llvm_unreachable("incorrect atomic ordering");
-}
-
 ModuleTranslation::ModuleTranslation(Operation *module,
                                      std::unique_ptr<llvm::Module> llvmModule)
     : mlirModule(module), llvmModule(std::move(llvmModule)),
       debugTranslation(
           std::make_unique<DebugTranslation>(module, *this->llvmModule)),
       ompDialect(module->getContext()->getLoadedDialect("omp")),
-      typeTranslator(this->llvmModule->getContext()) {
+      typeTranslator(this->llvmModule->getContext()),
+      iface(module->getContext()) {
   assert(satisfiesLLVMModule(mlirModule) &&
          "mlirModule should honor LLVM's module semantics.");
 }
 ModuleTranslation::~ModuleTranslation() {
   if (ompBuilder)
     ompBuilder->finalize();
 }
 
return llvm::TypeSwitch<Operation *, LogicalResult>(&opInst)
         return success();
       })
       .Default([&](Operation *inst) {
         return inst->emitError("unsupported OpenMP operation: ")
                << inst->getName();
       });
 }
 
-static llvm::FastMathFlags getFastmathFlags(FastmathFlagsInterface &op) {
-  using llvmFMF = llvm::FastMathFlags;
-  using FuncT = void (llvmFMF::*)(bool);
-  const std::pair<FastmathFlags, FuncT> handlers[] = {
-      // clang-format off
-      {FastmathFlags::nnan,     &llvmFMF::setNoNaNs},
-      {FastmathFlags::ninf,     &llvmFMF::setNoInfs},
-      {FastmathFlags::nsz,      &llvmFMF::setNoSignedZeros},
-      {FastmathFlags::arcp,     &llvmFMF::setAllowReciprocal},
-      {FastmathFlags::contract, &llvmFMF::setAllowContract},
-      {FastmathFlags::afn,      &llvmFMF::setApproxFunc},
-      {FastmathFlags::reassoc,  &llvmFMF::setAllowReassoc},
-      {FastmathFlags::fast,     &llvmFMF::setFast},
-      // clang-format on
-  };
-  llvm::FastMathFlags ret;
-  auto fmf = op.fastmathFlags();
-  for (auto it : handlers)
-    if (bitEnumContains(fmf, it.first))
-      (ret.*(it.second))(true);
-  return ret;
-}
-
 /// Given a single MLIR operation, create the corresponding LLVM IR operation
 /// using the `builder`.  LLVM IR Builder does not have a generic interface so
 /// this has to be a long chain of `if`s calling different functions with a
 /// different number of arguments.
 LogicalResult ModuleTranslation::convertOperation(Operation &opInst,
                                                   llvm::IRBuilder<> &builder) {
-  auto extractPosition = [](ArrayAttr attr) {
-    SmallVector<unsigned, 4> position;
-    position.reserve(attr.size());
-    for (Attribute v : attr)
-      position.push_back(v.cast<IntegerAttr>().getValue().getZExtValue());
-    return position;
-  };
-
-  llvm::IRBuilder<>::FastMathFlagGuard fmfGuard(builder);
-  if (auto fmf = dyn_cast<FastmathFlagsInterface>(opInst))
-    builder.setFastMathFlags(getFastmathFlags(fmf));
-
-#include "mlir/Dialect/LLVMIR/LLVMConversions.inc"
-
-  // Emit function calls.  If the "callee" attribute is present, this is a
-  // direct function call and we also need to look up the remapped function
-  // itself.  Otherwise, this is an indirect call and the callee is the first
-  // operand, look it up as a normal value.  Return the llvm::Value representing
-  // the function result, which may be of llvm::VoidTy type.
-  auto convertCall = [this, &builder](Operation &op) -> llvm::Value * {
-    auto operands = lookupValues(op.getOperands());
-    ArrayRef<llvm::Value *> operandsRef(operands);
-    if (auto attr = op.getAttrOfType<FlatSymbolRefAttr>("callee"))
-      return builder.CreateCall(lookupFunction(attr.getValue()), operandsRef);
-    auto *calleePtrType =
-        cast<llvm::PointerType>(operandsRef.front()->getType());
-    auto *calleeType =
-        cast<llvm::FunctionType>(calleePtrType->getElementType());
-    return builder.CreateCall(calleeType, operandsRef.front(),
-                              operandsRef.drop_front());
-  };
-
-  // Emit calls.  If the called function has a result, remap the corresponding
-  // value.  Note that LLVM IR dialect CallOp has either 0 or 1 result.
-  if (isa<LLVM::CallOp>(opInst)) {
-    llvm::Value *result = convertCall(opInst);
-    if (opInst.getNumResults() != 0) {
-      mapValue(opInst.getResult(0), result);
-      return success();
-    }
-    // Check that LLVM call returns void for 0-result functions.
-    return success(result->getType()->isVoidTy());
-  }
-
-  if (auto inlineAsmOp = dyn_cast<LLVM::InlineAsmOp>(opInst)) {
-    // TODO: refactor function type creation which usually occurs in std-LLVM
-    // conversion.
-    SmallVector<Type, 8> operandTypes;
-    operandTypes.reserve(inlineAsmOp.operands().size());
-    for (auto t : inlineAsmOp.operands().getTypes())
-      operandTypes.push_back(t);
-
-    Type resultType;
-    if (inlineAsmOp.getNumResults() == 0) {
-      resultType = LLVM::LLVMVoidType::get(mlirModule->getContext());
-    } else {
-      assert(inlineAsmOp.getNumResults() == 1);
-      resultType = inlineAsmOp.getResultTypes()[0];
-    }
-    auto ft = LLVM::LLVMFunctionType::get(resultType, operandTypes);
-    llvm::InlineAsm *inlineAsmInst =
-        inlineAsmOp.asm_dialect().hasValue()
-            ? llvm::InlineAsm::get(
-                  static_cast<llvm::FunctionType *>(convertType(ft)),
-                  inlineAsmOp.asm_string(), inlineAsmOp.constraints(),
-                  inlineAsmOp.has_side_effects(), inlineAsmOp.is_align_stack(),
-                  convertAsmDialectToLLVM(*inlineAsmOp.asm_dialect()))
-            : llvm::InlineAsm::get(
-                  static_cast<llvm::FunctionType *>(convertType(ft)),
-                  inlineAsmOp.asm_string(), inlineAsmOp.constraints(),
-                  inlineAsmOp.has_side_effects(), inlineAsmOp.is_align_stack());
-    llvm::Value *result =
-        builder.CreateCall(inlineAsmInst, lookupValues(inlineAsmOp.operands()));
-    if (opInst.getNumResults() != 0)
-      mapValue(opInst.getResult(0), result);
-    return success();
-  }
 
-  if (auto invOp = dyn_cast<LLVM::InvokeOp>(opInst)) {
-    auto operands = lookupValues(opInst.getOperands());
-    ArrayRef<llvm::Value *> operandsRef(operands);
-    if (auto attr = opInst.getAttrOfType<FlatSymbolRefAttr>("callee")) {
-      builder.CreateInvoke(lookupFunction(attr.getValue()),
-                           lookupBlock(invOp.getSuccessor(0)),
-                           lookupBlock(invOp.getSuccessor(1)), operandsRef);
-    } else {
-      auto *calleePtrType =
-          cast<llvm::PointerType>(operandsRef.front()->getType());
-      auto *calleeType =
-          cast<llvm::FunctionType>(calleePtrType->getElementType());
-      builder.CreateInvoke(
-          calleeType, operandsRef.front(), lookupBlock(invOp.getSuccessor(0)),
-          lookupBlock(invOp.getSuccessor(1)), operandsRef.drop_front());
-    }
-    return success();
-  }
-
-  if (auto lpOp = dyn_cast<LLVM::LandingpadOp>(opInst)) {
-    llvm::Type *ty = convertType(lpOp.getType());
-    llvm::LandingPadInst *lpi =
-        builder.CreateLandingPad(ty, lpOp.getNumOperands());
-
-    // Add clauses
-    for (llvm::Value *operand : lookupValues(lpOp.getOperands())) {
-      // All operands should be constant - checked by verifier
-      if (auto *constOperand = dyn_cast<llvm::Constant>(operand))
-        lpi->addClause(constOperand);
-    }
-    mapValue(lpOp.getResult(), lpi);
-    return success();
-  }
-
-  // Emit branches.  We need to look up the remapped blocks and ignore the block
-  // arguments that were transformed into PHI nodes.
-  if (auto brOp = dyn_cast<LLVM::BrOp>(opInst)) {
-    llvm::BranchInst *branch =
-        builder.CreateBr(lookupBlock(brOp.getSuccessor()));
-    mapBranch(&opInst, branch);
+  // TODO(zinenko): this should be the "main" conversion here, remove the
+  // dispatch below.
+  if (succeeded(iface.convertOperation(&opInst, builder, *this)))
     return success();
-  }
-  if (auto condbrOp = dyn_cast<LLVM::CondBrOp>(opInst)) {
-    auto weights = condbrOp.branch_weights();
-    llvm::MDNode *branchWeights = nullptr;
-    if (weights) {
-      // Map weight attributes to LLVM metadata.
-      auto trueWeight =
-          weights.getValue().getValue(0).cast<IntegerAttr>().getInt();
-      auto falseWeight =
-          weights.getValue().getValue(1).cast<IntegerAttr>().getInt();
-      branchWeights =
-          llvm::MDBuilder(llvmModule->getContext())
-              .createBranchWeights(static_cast<uint32_t>(trueWeight),
-                                   static_cast<uint32_t>(falseWeight));
-    }
-    llvm::BranchInst *branch = builder.CreateCondBr(
-        lookupValue(condbrOp.getOperand(0)),
-        lookupBlock(condbrOp.getSuccessor(0)),
-        lookupBlock(condbrOp.getSuccessor(1)), branchWeights);
-    mapBranch(&opInst, branch);
-    return success();
-  }
-  if (auto switchOp = dyn_cast<LLVM::SwitchOp>(opInst)) {
-    llvm::MDNode *branchWeights = nullptr;
-    if (auto weights = switchOp.branch_weights()) {
-      llvm::SmallVector<uint32_t> weightValues;
-      weightValues.reserve(weights->size());
-      for (llvm::APInt weight : weights->cast<DenseIntElementsAttr>())
-        weightValues.push_back(weight.getLimitedValue());
-      branchWeights = llvm::MDBuilder(llvmModule->getContext())
-                          .createBranchWeights(weightValues);
-    }
-
-    llvm::SwitchInst *switchInst =
-        builder.CreateSwitch(lookupValue(switchOp.value()),
-                             lookupBlock(switchOp.defaultDestination()),
-                             switchOp.caseDestinations().size(), branchWeights);
-
-    auto *ty =
-        llvm::cast<llvm::IntegerType>(convertType(switchOp.value().getType()));
-    for (auto i :
-         llvm::zip(switchOp.case_values()->cast<DenseIntElementsAttr>(),
-                   switchOp.caseDestinations()))
-      switchInst->addCase(
-          llvm::ConstantInt::get(ty, std::get<0>(i).getLimitedValue()),
-          lookupBlock(std::get<1>(i)));
-
-    mapBranch(&opInst, switchInst);
-    return success();
-  }
-
-  // Emit addressof.  We need to look up the global value referenced by the
-  // operation and store it in the MLIR-to-LLVM value mapping.  This does not
-  // emit any LLVM instruction.
-  if (auto addressOfOp = dyn_cast<LLVM::AddressOfOp>(opInst)) {
-    LLVM::GlobalOp global = addressOfOp.getGlobal();
-    LLVM::LLVMFuncOp function = addressOfOp.getFunction();
-
-    // The verifier should not have allowed this.
-    assert((global || function) &&
-           "referencing an undefined global or function");
-
-    mapValue(addressOfOp.getResult(), global
-                                          ? globalsMapping.lookup(global)
-                                          : lookupFunction(function.getName()));
-    return success();
-  }
 
   if (ompDialect && opInst.getDialect() == ompDialect)
     return convertOmpOperation(opInst, builder);
 
   return opInst.emitError("unsupported or non-LLVM operation: ")
          << opInst.getName();
 }
 

mlir/tools/mlir-cpu-runner/mlir-cpu-runner.cpp

 //
 //===----------------------------------------------------------------------===//
 
 #include "mlir/Dialect/LLVMIR/LLVMDialect.h"
 #include "mlir/Dialect/OpenMP/OpenMPDialect.h"
 #include "mlir/ExecutionEngine/JitRunner.h"
 #include "mlir/ExecutionEngine/OptUtils.h"
 #include "mlir/IR/Dialect.h"
+#include "mlir/Target/LLVMIR.h"
+
 #include "llvm/Support/InitLLVM.h"
 #include "llvm/Support/TargetSelect.h"
 
 int main(int argc, char **argv) {
   llvm::InitLLVM y(argc, argv);
   llvm::InitializeNativeTarget();
   llvm::InitializeNativeTargetAsmPrinter();
   llvm::InitializeNativeTargetAsmParser();
   mlir::initializeLLVMPasses();
 
   mlir::DialectRegistry registry;
   registry.insert<mlir::LLVM::LLVMDialect, mlir::omp::OpenMPDialect>();
+  mlir::registerLLVMDialectTranslation(registry);
+
   return mlir::JitRunnerMain(argc, argv, registry);
 }

mlir/tools/mlir-cuda-runner/mlir-cuda-runner.cpp

 #include "mlir/Dialect/LLVMIR/LLVMDialect.h"
 #include "mlir/Dialect/LLVMIR/NVVMDialect.h"
 #include "mlir/Dialect/StandardOps/IR/Ops.h"
 #include "mlir/ExecutionEngine/JitRunner.h"
 #include "mlir/ExecutionEngine/OptUtils.h"
 #include "mlir/IR/BuiltinOps.h"
 #include "mlir/Pass/Pass.h"
 #include "mlir/Pass/PassManager.h"
+#include "mlir/Target/LLVMIR.h"
 #include "mlir/Target/NVVMIR.h"
 #include "mlir/Transforms/DialectConversion.h"
 #include "mlir/Transforms/Passes.h"
+
 #include "llvm/Support/InitLLVM.h"
 #include "llvm/Support/TargetSelect.h"
 
 #include "cuda.h"
 
 using namespace mlir;
 
 inline void emit_cuda_error(const llvm::Twine &message, const char *buffer,
int main(int argc, char **argv) {
 
   mlir::JitRunnerConfig jitRunnerConfig;
   jitRunnerConfig.mlirTransformer = runMLIRPasses;
 
   mlir::DialectRegistry registry;
   registry.insert<mlir::LLVM::LLVMDialect, mlir::NVVM::NVVMDialect,
                   mlir::async::AsyncDialect, mlir::gpu::GPUDialect,
                   mlir::StandardOpsDialect>();
+  mlir::registerLLVMDialectTranslation(registry);
 
   return mlir::JitRunnerMain(argc, argv, registry, jitRunnerConfig);
 }

mlir/tools/mlir-rocm-runner/mlir-rocm-runner.cpp

 #include "mlir/Dialect/LLVMIR/ROCDLDialect.h"
 #include "mlir/Dialect/StandardOps/IR/Ops.h"
 #include "mlir/ExecutionEngine/JitRunner.h"
 #include "mlir/ExecutionEngine/OptUtils.h"
 #include "mlir/IR/BuiltinOps.h"
 #include "mlir/Pass/Pass.h"
 #include "mlir/Pass/PassManager.h"
 #include "mlir/Support/FileUtilities.h"
+#include "mlir/Target/LLVMIR.h"
 #include "mlir/Target/ROCDLIR.h"
 #include "mlir/Transforms/DialectConversion.h"
 #include "mlir/Transforms/Passes.h"
 #include "llvm/Support/ErrorOr.h"
 #include "llvm/Support/FileUtilities.h"
 #include "llvm/Support/InitLLVM.h"
 #include "llvm/Support/LineIterator.h"
 #include "llvm/Support/Program.h"
int main(int argc, char **argv) {
   mlir::initializeLLVMPasses();
 
   mlir::JitRunnerConfig jitRunnerConfig;
   jitRunnerConfig.mlirTransformer = runMLIRPasses;
 
   mlir::DialectRegistry registry;
   registry.insert<mlir::LLVM::LLVMDialect, mlir::gpu::GPUDialect,
                   mlir::ROCDL::ROCDLDialect, mlir::StandardOpsDialect>();
+  mlir::registerLLVMDialectTranslation(registry);
 
   return mlir::JitRunnerMain(argc, argv, registry, jitRunnerConfig);
 }

mlir/tools/mlir-spirv-cpu-runner/mlir-spirv-cpu-runner.cpp

int main(int argc, char **argv) {
 
   mlir::JitRunnerConfig jitRunnerConfig;
   jitRunnerConfig.mlirTransformer = runMLIRPasses;
   jitRunnerConfig.llvmModuleBuilder = convertMLIRModule;
 
   mlir::DialectRegistry registry;
   registry.insert<mlir::LLVM::LLVMDialect, mlir::gpu::GPUDialect,
                   mlir::spirv::SPIRVDialect, mlir::StandardOpsDialect>();
+  mlir::registerLLVMDialectTranslation(registry);
 
   return mlir::JitRunnerMain(argc, argv, registry, jitRunnerConfig);
 }

mlir/tools/mlir-vulkan-runner/mlir-vulkan-runner.cpp

 #include "mlir/Dialect/SPIRV/IR/SPIRVDialect.h"
 #include "mlir/Dialect/SPIRV/IR/SPIRVOps.h"
 #include "mlir/Dialect/SPIRV/Transforms/Passes.h"
 #include "mlir/Dialect/StandardOps/IR/Ops.h"
 #include "mlir/ExecutionEngine/JitRunner.h"
 #include "mlir/ExecutionEngine/OptUtils.h"
 #include "mlir/Pass/Pass.h"
 #include "mlir/Pass/PassManager.h"
+#include "mlir/Target/LLVMIR.h"
 #include "llvm/Support/InitLLVM.h"
 #include "llvm/Support/TargetSelect.h"
 
 using namespace mlir;
 
 static LogicalResult runMLIRPasses(ModuleOp module) {
   PassManager passManager(module.getContext());
   applyPassManagerCLOptions(passManager);
int main(int argc, char **argv) {
   mlir::initializeLLVMPasses();
 
   mlir::JitRunnerConfig jitRunnerConfig;
   jitRunnerConfig.mlirTransformer = runMLIRPasses;
 
   mlir::DialectRegistry registry;
   registry.insert<mlir::LLVM::LLVMDialect, mlir::gpu::GPUDialect,
                   mlir::spirv::SPIRVDialect, mlir::StandardOpsDialect>();
+  mlir::registerLLVMDialectTranslation(registry);
 
   return mlir::JitRunnerMain(argc, argv, registry, jitRunnerConfig);
 }

mlir/unittests/ExecutionEngine/Invoke.cpp

 #include "mlir/ExecutionEngine/CRunnerUtils.h"
 #include "mlir/ExecutionEngine/ExecutionEngine.h"
 #include "mlir/ExecutionEngine/MemRefUtils.h"
 #include "mlir/ExecutionEngine/RunnerUtils.h"
 #include "mlir/IR/MLIRContext.h"
 #include "mlir/InitAllDialects.h"
 #include "mlir/Parser.h"
 #include "mlir/Pass/PassManager.h"
+#include "mlir/Target/LLVMIR.h"
 #include "llvm/Support/TargetSelect.h"
 #include "llvm/Support/raw_ostream.h"
 
 #include "gmock/gmock.h"
 
 using namespace mlir;
 
 static struct LLVMInitializer {

 
 TEST(MLIRExecutionEngine, AddInteger) {
   std::string moduleStr = R"mlir(
   func @foo(%arg0 : i32) -> i32 attributes { llvm.emit_c_interface } {
     %res = std.addi %arg0, %arg0 : i32
     return %res : i32
   }
   )mlir";
-  MLIRContext context;
-  registerAllDialects(context);
+  DialectRegistry registry;
+  registerAllDialects(registry);
+  registerLLVMDialectTranslation(registry);
+  MLIRContext context(registry);
   OwningModuleRef module = parseSourceString(moduleStr, &context);
   ASSERT_TRUE(!!module);
   ASSERT_TRUE(succeeded(lowerToLLVMDialect(*module)));
   auto jitOrError = ExecutionEngine::create(*module);
   ASSERT_TRUE(!!jitOrError);
   std::unique_ptr<ExecutionEngine> jit = std::move(jitOrError.get());
   // The result of the function must be passed as output argument.
   int result = 0;
   llvm::Error error =
       jit->invoke("foo", 42, ExecutionEngine::Result<int>(result));
   ASSERT_TRUE(!error);
   ASSERT_EQ(result, 42 + 42);
 }
 
 TEST(MLIRExecutionEngine, SubtractFloat) {
   std::string moduleStr = R"mlir(
   func @foo(%arg0 : f32, %arg1 : f32) -> f32 attributes { llvm.emit_c_interface } {
     %res = std.subf %arg0, %arg1 : f32
     return %res : f32
   }
   )mlir";
-  MLIRContext context;
-  registerAllDialects(context);
+  DialectRegistry registry;
+  registerAllDialects(registry);
+  registerLLVMDialectTranslation(registry);
+  MLIRContext context(registry);
   OwningModuleRef module = parseSourceString(moduleStr, &context);
   ASSERT_TRUE(!!module);
   ASSERT_TRUE(succeeded(lowerToLLVMDialect(*module)));
   auto jitOrError = ExecutionEngine::create(*module);
   ASSERT_TRUE(!!jitOrError);
   std::unique_ptr<ExecutionEngine> jit = std::move(jitOrError.get());
   // The result of the function must be passed as output argument.
   float result = -1;
TEST(NativeMemRefJit, ZeroRankMemref) {
   A[{}] = 0;
   std::string moduleStr = R"mlir(
   func @zero_ranked(%arg0 : memref<f32>) attributes { llvm.emit_c_interface } {
     %cst42 = constant 42.0 : f32
     store %cst42, %arg0[] : memref<f32>
     return
   }
   )mlir";
-  MLIRContext context;
-  registerAllDialects(context);
+  DialectRegistry registry;
+  registerAllDialects(registry);
+  registerLLVMDialectTranslation(registry);
+  MLIRContext context(registry);
   auto module = parseSourceString(moduleStr, &context);
   ASSERT_TRUE(!!module);
   ASSERT_TRUE(succeeded(lowerToLLVMDialect(*module)));
   auto jitOrError = ExecutionEngine::create(*module);
   ASSERT_TRUE(!!jitOrError);
   auto jit = std::move(jitOrError.get());
 
   llvm::Error error = jit->invoke("zero_ranked", &*A);
TEST(NativeMemRefJit, RankOneMemref) {
   std::string moduleStr = R"mlir(
   func @one_ranked(%arg0 : memref<?xf32>) attributes { llvm.emit_c_interface } {
     %cst42 = constant 42.0 : f32
     %cst5 = constant 5 : index
     store %cst42, %arg0[%cst5] : memref<?xf32>
     return
   }
   )mlir";
-  MLIRContext context;
-  registerAllDialects(context);
+  DialectRegistry registry;
+  registerAllDialects(registry);
+  registerLLVMDialectTranslation(registry);
+  MLIRContext context(registry);
   auto module = parseSourceString(moduleStr, &context);
   ASSERT_TRUE(!!module);
   ASSERT_TRUE(succeeded(lowerToLLVMDialect(*module)));
   auto jitOrError = ExecutionEngine::create(*module);
   ASSERT_TRUE(!!jitOrError);
   auto jit = std::move(jitOrError.get());
 
   llvm::Error error = jit->invoke("one_ranked", &*A);
func @rank2_memref(%arg0 : memref<?x?xf32>, %arg1 : memref<?x?xf32>) attributes { llvm.emit_c_interface } {
     %x = constant 2 : index
     %y = constant 1 : index
     %cst42 = constant 42.0 : f32
     store %cst42, %arg0[%y, %x] : memref<?x?xf32>
     store %cst42, %arg1[%x, %y] : memref<?x?xf32>
     return
   }
   )mlir";
-  MLIRContext context;
-  registerAllDialects(context);
+  DialectRegistry registry;
+  registerAllDialects(registry);
+  registerLLVMDialectTranslation(registry);
+  MLIRContext context(registry);
   OwningModuleRef module = parseSourceString(moduleStr, &context);
   ASSERT_TRUE(!!module);
   ASSERT_TRUE(succeeded(lowerToLLVMDialect(*module)));
   auto jitOrError = ExecutionEngine::create(*module);
   ASSERT_TRUE(!!jitOrError);
   std::unique_ptr<ExecutionEngine> jit = std::move(jitOrError.get());
 
   llvm::Error error = jit->invoke("rank2_memref", &*A, &*A);
TEST(NativeMemRefJit, JITCallback) {
   std::string moduleStr = R"mlir(
   func private @callback(%arg0: memref<?x?xf32>, %coefficient: i32)  attributes { llvm.emit_c_interface } 
   func @caller_for_callback(%arg0: memref<?x?xf32>, %coefficient: i32) attributes { llvm.emit_c_interface } {
     %unranked = memref_cast %arg0: memref<?x?xf32> to memref<*xf32>
     call @callback(%arg0, %coefficient) : (memref<?x?xf32>, i32) -> ()
     return
   }
   )mlir";
-  MLIRContext context;
-  registerAllDialects(context);
+  DialectRegistry registry;
+  registerAllDialects(registry);
+  registerLLVMDialectTranslation(registry);
+  MLIRContext context(registry);
   auto module = parseSourceString(moduleStr, &context);
   ASSERT_TRUE(!!module);
   ASSERT_TRUE(succeeded(lowerToLLVMDialect(*module)));
   auto jitOrError = ExecutionEngine::create(*module);
   ASSERT_TRUE(!!jitOrError);
   auto jit = std::move(jitOrError.get());
   // Define any extra symbols so they're available at runtime.
   jit->registerSymbols([&](llvm::orc::MangleAndInterner interner) {