path: root/tensorflow/stream_executor/cuda/cuda_gpu_executor.h

// The CUDA implementation of the StreamExecutorInterface functionality.
// CUDA inclusions are ideally confined to this implementation file.
//
// The notions from the StreamExecutor basically correspond to the CUDA streams
// programming model provided by the libcuda.so driver APIs, so we don't have
// to do much more than wrap the calls to the libraries appropriately.
#ifndef TENSORFLOW_STREAM_EXECUTOR_CUDA_CUDA_GPU_EXECUTOR_H_
#define TENSORFLOW_STREAM_EXECUTOR_CUDA_CUDA_GPU_EXECUTOR_H_

#include <map>
#include <set>

#include "tensorflow/stream_executor/cuda/cuda_kernel.h"
#include "tensorflow/stream_executor/event.h"
#include "tensorflow/stream_executor/lib/status.h"
#include "tensorflow/stream_executor/lib/statusor.h"
#include "tensorflow/stream_executor/platform.h"
#include "tensorflow/stream_executor/platform/mutex.h"
#include "tensorflow/stream_executor/platform/port.h"
#include "tensorflow/stream_executor/platform/thread_annotations.h"
#include "tensorflow/stream_executor/stream_executor_internal.h"

namespace perftools {
namespace gputools {
namespace blas {
class BlasSupport;
}
namespace internal {
class RngSupport;
}  // namespace internal
}  // namespace gputools
}  // namespace perftools

namespace perftools {
namespace gputools {
namespace cuda {

// CUDA-platform implementation of the platform-agnostic
// StreamExecutorInferface.
class CUDAExecutor : public internal::StreamExecutorInterface {
 public:
  // sub_platform indicates the subplatform used in this executor; it must
  // be a CUDA type.
  explicit CUDAExecutor(const PluginConfig &plugin_config)
      : device_(0),
        context_(nullptr),
        device_ordinal_(0),
        cc_major_(0),
        cc_minor_(0),
        plugin_config_(plugin_config) {}

  // See the corresponding StreamExecutor methods for method comments on the
  // following overrides.

  ~CUDAExecutor() override;

  port::Status Init(int device_ordinal, DeviceOptions device_options) override;

  bool GetKernel(const MultiKernelLoaderSpec &spec,
                 KernelBase *kernel) override;

  bool Launch(Stream *stream, const ThreadDim &thread_dims,
              const BlockDim &block_dims, const KernelBase &k,
              const std::vector<KernelArg> &args) override;

  void *Allocate(uint64 size) override;

  void *AllocateSubBuffer(DeviceMemoryBase *mem, uint64 offset_bytes,
                          uint64 size_bytes) override;

  void Deallocate(DeviceMemoryBase *mem) override;

  // CUDA allocation/registration functions are necessary because the driver
  // internally sets up buffers for DMA operations (and page locks them).
  // There's no external interface for us to otherwise control these DMA
  // settings.
  void *HostMemoryAllocate(uint64 size) override {
    return CUDADriver::HostAllocate(context_, size);
  }

  void HostMemoryDeallocate(void *location) override {
    return CUDADriver::HostDeallocate(context_, location);
  }

  bool HostMemoryRegister(void *location, uint64 size) override;

  bool HostMemoryUnregister(void *location) override;

  bool SynchronizeAllActivity() override;

  bool SynchronousMemZero(DeviceMemoryBase *location, uint64 size) override;

  bool SynchronousMemSet(DeviceMemoryBase *location, int value,
                         uint64 size) override;

  bool SynchronousMemcpy(DeviceMemoryBase *gpu_dst, const void *host_src,
                         uint64 size) override;

  bool SynchronousMemcpy(void *host_dst, const DeviceMemoryBase &gpu_src,
                         uint64 size) override;

  bool SynchronousMemcpyDeviceToDevice(DeviceMemoryBase *gpu_dst,
                                       const DeviceMemoryBase &gpu_src,
                                       uint64 size) override;

  bool MemZero(Stream *stream, DeviceMemoryBase *location,
               uint64 size) override;
  bool Memset32(Stream *stream, DeviceMemoryBase *location, uint32 pattern,
                uint64 size) override;

  bool Memcpy(Stream *stream, void *host_dst, const DeviceMemoryBase &gpu_src,
              uint64 size) override;

  bool Memcpy(Stream *stream, DeviceMemoryBase *gpu_dst, const void *host_src,
              uint64 size) override;

  bool MemcpyDeviceToDevice(Stream *stream, DeviceMemoryBase *gpu_dst,
                            const DeviceMemoryBase &gpu_src,
                            uint64 size) override;

  bool HostCallback(Stream *stream, std::function<void()> callback) override;

  bool AllocateStream(Stream *stream) override;

  void DeallocateStream(Stream *stream) override;

  bool CreateStreamDependency(Stream *dependent, Stream *other) override;

  bool AllocateTimer(Timer *timer) override;

  void DeallocateTimer(Timer *timer) override;

  bool StartTimer(Stream *stream, Timer *timer) override;

  bool StopTimer(Stream *stream, Timer *timer) override;

  port::Status AllocateEvent(Event *event) override;

  port::Status DeallocateEvent(Event *event) override;

  port::Status RecordEvent(Stream *stream, Event *event) override;

  port::Status WaitForEvent(Stream *stream, Event *event) override;

  Event::Status PollForEventStatus(Event *event) override;

  bool BlockHostUntilDone(Stream *stream) override;

  int PlatformDeviceCount() override { return CUDADriver::GetDeviceCount(); }

  port::Status EnablePeerAccessTo(StreamExecutorInterface *other) override;

  bool CanEnablePeerAccessTo(StreamExecutorInterface *other) override;

  SharedMemoryConfig GetDeviceSharedMemoryConfig() override;

  port::Status SetDeviceSharedMemoryConfig(SharedMemoryConfig config) override;

  bool DeviceMemoryUsage(int64 *free, int64 *total) const override;

  // Search for the symbol and returns a device pointer and size.
  // Returns false if symbol does not exist.
  bool GetSymbol(const string& symbol_name, void **mem, size_t *bytes) override;

  DeviceDescription *PopulateDeviceDescription() const override;

  // Populates the block_dim_limit by querying the device driver API. If an
  // error occurs at any point while asking the driver for block dim limits, it
  // will be only partially populated as a result, and an error will be logged.
  bool FillBlockDimLimit(BlockDim *block_dim_limit) const;

  KernelArg DeviceMemoryToKernelArg(
      const DeviceMemoryBase &gpu_mem) const override;

  bool SupportsBlas() const override;

  blas::BlasSupport *CreateBlas() override;

  bool SupportsFft() const override;

  fft::FftSupport *CreateFft() override;

  bool SupportsRng() const override;

  rng::RngSupport *CreateRng() override;

  bool SupportsDnn() const override;

  dnn::DnnSupport *CreateDnn() override;

  void *CudaContextHack() override;

  CUcontext cuda_context();

 private:
  // Attempts to find a more specific version of the file indicated by
  // filename by looking for compute-capability-specific suffixed versions; i.e.
  // looking for "foo.ptx" will check to see if "foo.ptx.cc30.ptx" is present if
  // we're on a compute capability 3.0 machine.
  bool FindOnDiskForComputeCapability(port::StringPiece filename,
                                      port::StringPiece canonical_suffix,
                                      string *found_filename) const;

  // Host callback landing routine invoked by CUDA.
  // data: User-provided callback provided to HostCallback() above, captured
  //       as a std::function<void()>. Allocated/initialized inside
  //       HostCallback() and owned and deleted by this call.
  static void InternalHostCallback(CUstream stream, CUresult status,
                                   void *data);

  // Collects metadata for the specified kernel.
  bool GetKernelMetadata(CUDAKernel *cuda_kernel,
                         KernelMetadata *kernel_metadata);

  // Determines if the given kernel's occupancy could be improved by only
  // slightly reducing its register usage. If so, a message is emitted to the
  // INFO log. The warning threshold is controlled by the flag
  // register_occupancy_warning_threshold.
  void OccupancyCheck(const KernelBase &kernel, const ThreadDim &thread_dims,
                      const BlockDim &block_dims);

  // Guards the on-disk-module mapping.
  mutex disk_modules_mu_;

  // Mapping from filename to CUmodule, if it was already retrieved.
  // Multiple CUfunctions are usually obtained from a single CUmodule so we
  // attempt to hit in this mapping first, before retrieving it.
  std::map<string, CUmodule> disk_modules_ GUARDED_BY(disk_modules_mu_);

  // Guards the in-memory-module mapping.
  mutex in_memory_modules_mu_;

  std::map<const char *, CUmodule> in_memory_modules_
      GUARDED_BY(in_memory_modules_mu_);

  // Guards the launched kernel set.
  mutex launched_kernels_mu_;

  // Keeps track of the set of launched kernels. Currently used to suppress the
  // occupancy check on subsequent launches.
  std::set<CUfunction> launched_kernels_ GUARDED_BY(launched_kernels_mu_);

  // Handle for the CUDA device being operated on. Immutable
  // post-initialization.
  CUdevice device_;

  // Handle for session with the library/driver. Immutable post-initialization.
  CUcontext context_;

  // The device ordinal value that this executor was initialized with; recorded
  // for use in getting device metadata. Immutable post-initialization.
  int device_ordinal_;

  // The major verion of the compute capability for device_.
  int cc_major_;

  // The minor verion of the compute capability for device_.
  int cc_minor_;

  // The plugin configuration associated with this instance.
  PluginConfig plugin_config_;

  SE_DISALLOW_COPY_AND_ASSIGN(CUDAExecutor);
};

}  // namespace cuda
}  // namespace gputools
}  // namespace perftools

#endif  // TENSORFLOW_STREAM_EXECUTOR_CUDA_CUDA_GPU_EXECUTOR_H_