How can I parse the LLVM MCJIT compilation result?

Question

I have a program that I wrote that uses LLVM 3.5 as a JIT compiler, which I am trying to upgrade to use MCJIT in LLVM 3.7. This basically works for me, but I'm struggling to reproduce one debugging function implemented with LLVM 3.5.

I would like to see the host machine code (for example, x86, x64 or ARM, not the LLVM IR) generated by the JIT process; in debug builds I log this when my program is running. With LLVM 3.5, I was able to do this by calling ExecutionEngine :: runJITOnFunction () to populate the llvm :: MachineCodeInfo object, which gave me the starting address and size of the generated code. Then I could parse this code.

I can not find the equivalent in MCJIT. I can get the start address of the function (e.g. via getPointerToFunction ()), but not the size.

I have seen disassemble the memory , but besides the fact that the answers do not have so many details, it looks like it's more about how to parse a sequence of bytes. I know how to do this, my question is: how can I get a sequence of bytes in the first place?

If this helps to make it more specific, please reformulate this question as follows: "How can I extend the Kaleidoscope JIT example to show the machine code (x86, ARM, etc.) that it produces, and not just LLVM IR? "

Thanks.

Answer 1

.

• . MCJIT. :

  class MCJITMemoryManager : public llvm::RTDyldMemoryManager {
  public:
  static std::unique_ptr<MCJITMemoryManager> Create();
  
  MCJITMemoryManager();
  virtual ~MCJITMemoryManager();
  
  // Allocate a memory block of (at least) the given size suitable for
  // executable code. The section_id is a unique identifier assigned by the
  // MCJIT engine, and optionally recorded by the memory manager to access a
  // loaded section.
  byte* allocateCodeSection(uintptr_t size, unsigned alignment,
                            unsigned section_id,
                            llvm::StringRef section_name) override;
  
  // Allocate a memory block of (at least) the given size suitable for data.
  // The SectionID is a unique identifier assigned by the JIT engine, and
  // optionally recorded by the memory manager to access a loaded section.
  byte* allocateDataSection(uintptr_t size, unsigned alignment,
                      unsigned section_id, llvm::StringRef section_name,
                      bool is_readonly) override;
  ...
  }
  

  EngineBuilder:

  std::unique_ptr<MCJITMemoryManager> manager = MCJITMemoryManager::Create();
  llvm::ExecutionEngine* raw = lvm::EngineBuilder(std::move(module))
      .setMCJITMemoryManager(std::move(manager))
      ...
      .create();
  

  , . ( ). , , gdb ( - LLVM).

• llc LLVM IR ( ..). , MCJIT - , , ( , llc).

Answer 2

llvm/Object/SymbolSize.h llvm::object::computeSymbolSizes(ObjectFile&). - ObjectFile.

, :

• , Module ObjectFile, - : class ModuleToObjectFileCompiler { ... // Compile a Module to an ObjectFile. llvm::object::OwningBinary<llvm::object::ObjectFile> operator() (llvm::Module&); };

• operator() of ModuleToObjectFileCompiler, llvm/ExecutionEngine/Orc/CompileUtils.h, SimpleCompiler.

• ModuleToObjectFileCompiler llvm::orc::IRCompileLayer, : new llvm::orc::IRCompileLayer <llvm::orc::ObjectLinkingLayer <llvm::orc::DoNothingOnNotifyLoaded> > (_object_layer, _module_to_object_file);

• operator() of ModuleToObjectFileCompiler ObjectFile, computeSymbolSizes(). std::vector, , Module. . .