llvm.org GIT mirror llvm / release_37 lib / Target / AMDGPU / AMDGPUTargetTransformInfo.h

Tree @release_37 (Download .tar.gz)

AMDGPUTargetTransformInfo.h @release_37raw · history · blame

//===-- AMDGPUTargetTransformInfo.h - AMDGPU specific TTI -------*- C++ -*-===//
//                     The LLVM Compiler Infrastructure
// This file is distributed under the University of Illinois Open Source
// License. See LICENSE.TXT for details.
/// \file
/// This file a TargetTransformInfo::Concept conforming object specific to the
/// AMDGPU target machine. It uses the target's detailed information to
/// provide more precise answers to certain TTI queries, while letting the
/// target independent and default TTI implementations handle the rest.


#include "AMDGPU.h"
#include "AMDGPUTargetMachine.h"
#include "llvm/Analysis/TargetTransformInfo.h"
#include "llvm/CodeGen/BasicTTIImpl.h"
#include "llvm/Target/TargetLowering.h"

namespace llvm {

class AMDGPUTTIImpl : public BasicTTIImplBase<AMDGPUTTIImpl> {
  typedef BasicTTIImplBase<AMDGPUTTIImpl> BaseT;
  typedef TargetTransformInfo TTI;
  friend BaseT;

  const AMDGPUSubtarget *ST;
  const AMDGPUTargetLowering *TLI;

  const AMDGPUSubtarget *getST() const { return ST; }
  const AMDGPUTargetLowering *getTLI() const { return TLI; }

  explicit AMDGPUTTIImpl(const AMDGPUTargetMachine *TM, const DataLayout &DL)
      : BaseT(TM, DL), ST(TM->getSubtargetImpl()),
        TLI(ST->getTargetLowering()) {}

  // Provide value semantics. MSVC requires that we spell all of these out.
      : BaseT(static_cast<const BaseT &>(Arg)), ST(Arg.ST), TLI(Arg.TLI) {}
      : BaseT(std::move(static_cast<BaseT &>(Arg))), ST(std::move(Arg.ST)),
        TLI(std::move(Arg.TLI)) {}

  bool hasBranchDivergence() { return true; }

  void getUnrollingPreferences(Loop *L, TTI::UnrollingPreferences &UP);

  TTI::PopcntSupportKind getPopcntSupport(unsigned TyWidth) {
    assert(isPowerOf2_32(TyWidth) && "Ty width must be power of 2");
    return ST->hasBCNT(TyWidth) ? TTI::PSK_FastHardware : TTI::PSK_Software;

  unsigned getNumberOfRegisters(bool Vector);
  unsigned getRegisterBitWidth(bool Vector);
  unsigned getMaxInterleaveFactor(unsigned VF);

} // end namespace llvm