Target.hpp

// Copyright (c) 2017 Graphcore Ltd. All rights reserved.
 
#ifndef poplar_Target_hpp
#define poplar_Target_hpp
 
#include <gccs/CompilerFeatures.hpp>
#include <poplar/GlobalExchangeConstraints.hpp>
#include <poplar/IpuLinkConfiguration.hpp>
#include <poplar/IpuLinkTopology.hpp>
#include <poplar/OptionFlags.hpp>
#include <poplar/StringRef.hpp>
#include <poplar/TargetType.hpp>
#include <poplar/Type.hpp>
 
#include <memory>
#include <vector>
 
namespace poplar {
namespace core {
class Target;
struct TargetOptions;
} // namespace core
 
class Target {
public:
  Target();
  ~Target();
  Target(const Target &);
  Target(Target &&) noexcept;
 
  Target(std::istream &in);
 
  Target &operator=(const Target &);
  Target &operator=(Target &&) noexcept;
  bool operator==(const Target &) const;
  bool operator!=(const Target &) const;
  bool operator<(const Target &) const;
 
  std::size_t hash() const;
 
  void serialize(std::ostream &out) const;
 
  TargetType getTargetType() const;
  std::string getTargetSystemString() const;
  StringRef getTargetArchString() const;
  unsigned getNumIPUs() const;
  unsigned getTilesPerIPU() const;
  unsigned getNumWorkerContexts() const;
  unsigned getBytesPerTile() const;
  unsigned getExchangeBytesPerCycle() const;
  unsigned getMemcpyBytesPerCycle() const;
  unsigned getMinIPUSyncDelay() const;
  unsigned getGlobalSyncCycles() const;
  const std::vector<unsigned> &getMemoryElementOffsets() const;
  unsigned getInterleavedMemoryElementIndex() const;
  const std::vector<GlobalExchangeConstraint> &
  getGlobalExchangeConstraints() const;
  // The number of stride bits.
  unsigned getNumStrideBits() const;
  unsigned getDataPathWidth() const;
  unsigned getFp8ConvUnitMaxPipelineDepth() const;
  unsigned getFp16ConvUnitMaxPipelineDepth() const;
  unsigned getFp32ConvUnitMaxPipelineDepth() const;
  unsigned getFp8ConvUnitInputLoadElemsPerCycle() const;
  unsigned getFp16ConvUnitInputLoadElemsPerCycle() const;
  unsigned getFp32ConvUnitInputLoadElemsPerCycle() const;
  unsigned getFp16InFp16OutConvUnitsPerTile() const;
  unsigned getFp16InFp32OutConvUnitsPerTile() const;
  unsigned getFp32InFp32OutConvUnitsPerTile() const;
  unsigned getFp8InFp16OutConvUnitsPerTile() const;
  unsigned getConvUnitCoeffLoadBytesPerCycle() const;
  // rpt counters max value.
  unsigned getRptCountMax() const;
 
  bool supportsExchangeBusSharing() const;
 
  unsigned getTilesPerSharedExchangeBus() const;
 
  unsigned getNumTiles() const;
 
  std::uint64_t getMemoryBytes() const;
 
  unsigned getFloatVectorWidth() const;
 
  unsigned getHalfVectorWidth() const;
 
  unsigned getQuarterVectorWidth() const;
 
  unsigned getQuarterMetadataVectorWidth() const;
 
  unsigned getVectorWidth(const poplar::Type &type) const;
 
  unsigned getWeightsPerConvUnit(const Type &type) const;
 
  unsigned getConvUnitInputLoadElemsPerCycle(const Type &type) const;
 
  unsigned getConvUnitMaxPipelineDepth(const Type &partialsType) const;
  unsigned getNumConvUnits(const Type &activationsType,
                           const Type &partialsType) const;
 
  unsigned getMaxIPUSyncDelay() const;
 
  double getTileClockFrequency() const;
 
  unsigned getNumTilesPerXBContext() const;
 
  unsigned getNumContextsPerXB() const;
 
  unsigned getTileHostExchangeXB(unsigned tile) const;
 
  unsigned getTileHostExchangeContext(unsigned tile) const;
 
  unsigned getTileHostExchangeContextPosition(unsigned tile) const;
 
  std::size_t getTypeSize(const Type &) const;
 
  std::size_t getAtomicStoreGranularity() const;
 
  uint32_t makeFpIctlValue(bool inv, bool div0, bool oflo, bool esr,
                           bool nanoo) const;
 
  unsigned getFpIctlRegIndex() const;
 
  unsigned getDbgDataRegIndex() const;
 
  IpuLinkConfiguration getIpuLinkConfiguration() const;
 
  IpuLinkTopology getIpuLinkTopology() const;
 
  unsigned getIpuLinkDomainSize() const;
 
  // Return the number of IPUs in an instance. Internal use only.
  unsigned getInstanceSize() const;
 
  // Return the state of gateway mode in the PCI Complex. Internal use only.
  bool getGatewayMode() const;
 
  Target createVirtualTarget(unsigned numIPUs, unsigned tilesPerIPU) const;
 
  static Target createCPUTarget(bool accurateHalf = false,
                                unsigned numIPUs = 1);
 
  static Target createIPUTarget(StringRef system, const OptionFlags &opts = {});
 
  static Target createIPUTarget(unsigned numIPUs, StringRef system,
                                const OptionFlags &opts = {});
 
  static Target createIPUTarget(unsigned numIPUs, unsigned tilesPerIPU,
                                StringRef system, const OptionFlags &opts = {});
 
  GC_DEPRECATED static Target createIPUTarget(unsigned numIPUs,
                                              StringRef system,
                                              const core::TargetOptions &opts);
 
  GC_DEPRECATED static Target createIPUTarget(unsigned numIPUs,
                                              unsigned tilesPerIPU,
                                              StringRef system,
                                              const core::TargetOptions &opts);
  // Implementation
  Target(std::unique_ptr<core::Target> &&) noexcept;
  core::Target &getImpl() const { return *impl; }
 
  // Internal use only
  const core::TargetOptions &getTargetOptions() const;
 
private:
  std::unique_ptr<core::Target> impl;
};
 
void copyDeviceHalfToFloat(const Target &target, const void *src, float *dst,
                           std::size_t numElements);
 
void copyFloatToDeviceHalf(const Target &target, const float *src, void *dst,
                           std::size_t numElements);
 
void copyDeviceHalfToDouble(const Target &target, const void *src, double *dst,
                            std::size_t numElements);
 
void copyDoubleToDeviceHalf(const Target &target, const double *src, void *dst,
                            std::size_t numElements);
 
} // namespace poplar
 
namespace std {
 
template <> struct hash<poplar::Target> {
  size_t operator()(const poplar::Target &t) const { return t.hash(); };
};
 
} // namespace std
 
#endif // poplar_Target_hpp