/external/tensorflow/tensorflow/compiler/xla/service/interpreter/ |
D | executor.h | 50 using Args = absl::Span<const DeviceMemoryBase>; 72 void *AllocateSubBuffer(DeviceMemoryBase *mem, uint64 offset_bytes, 74 void Deallocate(DeviceMemoryBase *mem) override; 83 bool Memcpy(Stream *stream, void *host_dst, const DeviceMemoryBase &pop_src, 85 bool Memcpy(Stream *stream, DeviceMemoryBase *pop_dst, const void *host_src, 87 bool MemcpyDeviceToDevice(Stream *stream, DeviceMemoryBase *pop_dst, in MemcpyDeviceToDevice() 88 const DeviceMemoryBase &host_src, in MemcpyDeviceToDevice() 93 bool MemZero(Stream *stream, DeviceMemoryBase *location, in MemZero() 97 bool Memset(Stream *stream, DeviceMemoryBase *location, uint8 pattern, in Memset() 101 bool Memset32(Stream *stream, DeviceMemoryBase *location, uint32 pattern, in Memset32() [all …]
|
D | executor.cc | 38 void *XlaInterpreterExecutor::AllocateSubBuffer(DeviceMemoryBase *parent, in AllocateSubBuffer() 44 void XlaInterpreterExecutor::Deallocate(DeviceMemoryBase *mem) { in Deallocate() 51 const DeviceMemoryBase &dev_src, in Memcpy() 60 bool XlaInterpreterExecutor::Memcpy(Stream *stream, DeviceMemoryBase *dev_dst, in Memcpy() 70 DeviceMemoryBase *dev_dst, const void *host_src, uint64 size) { in SynchronousMemcpy() 76 void *host_dst, const DeviceMemoryBase &dev_src, uint64 size) { in SynchronousMemcpy()
|
/external/tensorflow/tensorflow/stream_executor/ |
D | device_memory.h | 56 class DeviceMemoryBase { 61 explicit DeviceMemoryBase(void *opaque = nullptr, uint64 size = 0, 75 bool operator<(const DeviceMemoryBase &other) const { 92 bool IsSameAs(const DeviceMemoryBase &other) const { in IsSameAs() 119 class DeviceMemory final : public DeviceMemoryBase { 122 DeviceMemory() : DeviceMemoryBase(nullptr, 0) {} in DeviceMemory() 127 explicit DeviceMemory(const DeviceMemoryBase &other) in DeviceMemory() 128 : DeviceMemoryBase(const_cast<DeviceMemoryBase &>(other).opaque(), in DeviceMemory() 151 DeviceMemoryBase::Reset(opaque, bytes); in ResetFromByteSize() 163 DeviceMemory(void *opaque, uint64 size) : DeviceMemoryBase(opaque, size) {} in DeviceMemory() [all …]
|
D | stream_executor_internal.h | 199 virtual void *AllocateSubBuffer(DeviceMemoryBase *parent, uint64 offset, 201 virtual void Deallocate(DeviceMemoryBase *mem) = 0; 216 virtual bool SynchronousMemZero(DeviceMemoryBase *location, uint64 size) = 0; 217 virtual bool SynchronousMemSet(DeviceMemoryBase *location, int value, 219 virtual port::Status SynchronousMemcpy(DeviceMemoryBase *gpu_dst, 222 const DeviceMemoryBase &gpu_src, 225 DeviceMemoryBase *gpu_dst, const DeviceMemoryBase &gpu_src, 227 virtual bool MemZero(Stream *stream, DeviceMemoryBase *location, 229 virtual bool Memset(Stream *stream, DeviceMemoryBase *location, uint8 pattern, in Memset() 233 virtual bool Memset32(Stream *stream, DeviceMemoryBase *location, [all …]
|
D | stream_executor_pimpl.h | 191 port::StatusOr<DeviceMemoryBase> GetUntypedSymbol( 202 void Deallocate(DeviceMemoryBase *mem); 249 bool SynchronousMemZero(DeviceMemoryBase *location, 254 bool SynchronousMemSet(DeviceMemoryBase *location, int value, 261 bool SynchronousMemcpy(DeviceMemoryBase *device_dst, const void *host_src, 268 bool SynchronousMemcpy(void *host_dst, const DeviceMemoryBase &device_src, 273 DeviceMemoryBase *device_dst); 280 DeviceMemoryBase *device_dst) { in SynchronousMemcpyH2D() 287 port::Status SynchronousMemcpyD2H(const DeviceMemoryBase &device_src, 303 bool SynchronousMemcpy(DeviceMemoryBase *device_dst, [all …]
|
D | temporary_memory_manager.cc | 32 DeviceMemoryBase device_memory = it->first; in ForceDeallocateAll() 38 const DeviceMemoryBase& device_memory, uint64 generation, bool must_exist) { in MarkFinalized() 56 DeviceMemoryBase device_memory = it->first; in DeallocateFinalizedTemporaries() 67 bool TemporaryMemoryManager::IsFinalized(const DeviceMemoryBase& device_memory, in IsFinalized() 83 bool TemporaryMemoryManager::HasAllocated(const DeviceMemoryBase& device_memory, in HasAllocated() 97 DeviceMemoryBase device_memory = in AllocateArrayBase()
|
D | temporary_device_memory.h | 62 DeviceMemoryBase* mutable_device_memory(); 65 const DeviceMemoryBase& device_memory() const; 93 TemporaryDeviceMemoryBase(Stream* parent, DeviceMemoryBase device_memory, 97 DeviceMemoryBase device_memory_;
|
D | temporary_memory_manager.h | 77 void MarkFinalized(const DeviceMemoryBase& device_memory, uint64 generation, 91 bool IsFinalized(const DeviceMemoryBase& device_memory, 99 bool HasAllocated(const DeviceMemoryBase& device_memory, 117 std::map<DeviceMemoryBase, TemporaryMemoryRecord> records_ GUARDED_BY(mutex_);
|
D | stream_executor_pimpl.cc | 252 void StreamExecutor::Deallocate(DeviceMemoryBase *mem) { in Deallocate() 518 port::StatusOr<DeviceMemoryBase> StreamExecutor::GetUntypedSymbol( in GetUntypedSymbol() 525 return DeviceMemoryBase(opaque, bytes); in GetUntypedSymbol() 604 bool StreamExecutor::SynchronousMemZero(DeviceMemoryBase *location, in SynchronousMemZero() 612 bool StreamExecutor::SynchronousMemSet(DeviceMemoryBase *location, int value, in SynchronousMemSet() 621 bool StreamExecutor::SynchronousMemcpy(DeviceMemoryBase *device_dst, in SynchronousMemcpy() 639 const DeviceMemoryBase &device_src, in SynchronousMemcpy() 653 bool StreamExecutor::SynchronousMemcpy(DeviceMemoryBase *device_dst, in SynchronousMemcpy() 654 const DeviceMemoryBase &device_src, in SynchronousMemcpy() 669 const DeviceMemoryBase &device_src, int64 size, void *host_dst) { in SynchronousMemcpyD2H() [all …]
|
/external/tensorflow/tensorflow/compiler/xla/service/gpu/ |
D | buffer_allocations.h | 45 se::DeviceMemoryBase address); 56 absl::flat_hash_map<BufferAllocation::Index, se::DeviceMemoryBase> 71 se::DeviceMemoryBase GetDeviceAddress( 76 se::DeviceMemoryBase GetDeviceAddress( 79 se::DeviceMemoryBase GetTempBufferBase() const { return temp_buffer_base_; } in GetTempBufferBase() 83 Status TearDown(const std::set<se::DeviceMemoryBase>& live_addresses); 96 se::DeviceMemoryBase buffer); 101 std::vector<se::DeviceMemoryBase> buffers_; 104 se::DeviceMemoryBase temp_buffer_base_;
|
D | buffer_allocations.cc | 36 se::DeviceMemoryBase address) { in RegisterBuffer() 61 if (se::DeviceMemoryBase* address = in Build() 78 se::DeviceMemoryBase buffer_address; in Build() 129 const std::set<se::DeviceMemoryBase>& live_addresses) { in TearDown() 136 se::DeviceMemoryBase buffer_address = GetDeviceAddress(allocation.index()); in TearDown() 153 se::DeviceMemoryBase BufferAllocations::GetDeviceAddress( in GetDeviceAddress() 160 se::DeviceMemoryBase BufferAllocations::GetDeviceAddress( in GetDeviceAddress() 162 se::DeviceMemoryBase base = GetDeviceAddress(buffer_slice.index()); in GetDeviceAddress() 165 return se::DeviceMemoryBase( in GetDeviceAddress() 171 se::DeviceMemoryBase buffer) { in SetBuffer()
|
D | cudnn_conv_runner.h | 54 absl::Span<se::DeviceMemoryBase> operand_buffers, 55 se::DeviceMemoryBase result_buffer, 56 se::DeviceMemoryBase scratch_buf, se::Stream* stream, 60 absl::Span<se::DeviceMemoryBase> operand_buffers, 61 se::DeviceMemoryBase result_buffer,
|
D | cudnn_conv_runner.cc | 31 using se::DeviceMemoryBase; 58 se::DeviceMemoryBase bias_buf; 59 se::DeviceMemoryBase side_input_buf; // nullable 66 se::DeviceMemoryBase input_buf; 67 se::DeviceMemoryBase filter_buf; 68 se::DeviceMemoryBase output_buf; 82 explicit ScratchBufAllocator(se::DeviceMemoryBase scratch) in ScratchBufAllocator() 108 se::DeviceMemoryBase scratch_; 322 absl::Span<se::DeviceMemoryBase> operand_buffers, in GetCudnnConvParams() 323 se::DeviceMemoryBase result_buffer) { in GetCudnnConvParams() [all …]
|
/external/tensorflow/tensorflow/stream_executor/host/ |
D | host_gpu_executor.h | 64 void *AllocateSubBuffer(DeviceMemoryBase *mem, uint64 offset_bytes, 66 void Deallocate(DeviceMemoryBase *mem) override; 75 bool Memcpy(Stream *stream, void *host_dst, const DeviceMemoryBase &gpu_src, 77 bool Memcpy(Stream *stream, DeviceMemoryBase *gpu_dst, const void *host_src, 79 bool MemcpyDeviceToDevice(Stream *stream, DeviceMemoryBase *gpu_dst, 80 const DeviceMemoryBase &host_src, 83 bool MemZero(Stream *stream, DeviceMemoryBase *location, 85 bool Memset(Stream *stream, DeviceMemoryBase *location, uint8 pattern, 87 bool Memset32(Stream *stream, DeviceMemoryBase *location, uint32 pattern, 92 bool SynchronousMemZero(DeviceMemoryBase *location, uint64 size) override; [all …]
|
D | host_gpu_executor.cc | 44 void *HostExecutor::AllocateSubBuffer(DeviceMemoryBase *parent, in AllocateSubBuffer() 49 void HostExecutor::Deallocate(DeviceMemoryBase *mem) { in Deallocate() 55 bool HostExecutor::SynchronousMemZero(DeviceMemoryBase *location, uint64 size) { in SynchronousMemZero() 60 bool HostExecutor::SynchronousMemSet(DeviceMemoryBase *location, int value, in SynchronousMemSet() 67 const DeviceMemoryBase &gpu_src, uint64 size) { in Memcpy() 76 bool HostExecutor::Memcpy(Stream *stream, DeviceMemoryBase *gpu_dst, in Memcpy() 87 DeviceMemoryBase *gpu_dst, in MemcpyDeviceToDevice() 88 const DeviceMemoryBase &gpu_src, in MemcpyDeviceToDevice() 100 bool HostExecutor::MemZero(Stream *stream, DeviceMemoryBase *location, in MemZero() 110 bool HostExecutor::Memset(Stream *stream, DeviceMemoryBase *location, in Memset() [all …]
|
/external/tensorflow/tensorflow/compiler/xla/service/ |
D | owning_device_memory.h | 46 explicit OwningDeviceMemory(se::DeviceMemoryBase mem, int device_ordinal, in OwningDeviceMemory() 56 other.mem_ = se::DeviceMemoryBase(); in OwningDeviceMemory() 68 other.mem_ = se::DeviceMemoryBase(); 103 se::DeviceMemoryBase AsDeviceMemoryBase() { in AsDeviceMemoryBase() 104 return se::DeviceMemoryBase(opaque(), size(), /*is_sub_buffer=*/false); in AsDeviceMemoryBase() 109 TF_MUST_USE_RESULT se::DeviceMemoryBase Forget() { in Forget() 114 se::DeviceMemoryBase mem(mem_); in Forget() 115 mem_ = se::DeviceMemoryBase(); in Forget() 124 se::DeviceMemoryBase mem_;
|
D | shaped_buffer.h | 72 const se::DeviceMemoryBase& root_buffer() const { in root_buffer() 78 const se::DeviceMemoryBase& buffer(const ShapeIndex& index) const { in buffer() 83 void set_buffer(const se::DeviceMemoryBase& buffer, const ShapeIndex& index) { in set_buffer() 90 void set_buffers(ShapeTree<se::DeviceMemoryBase> buffers) { in set_buffers() 97 const ShapeTree<se::DeviceMemoryBase>& buffers() const { return buffers_; } in buffers() 98 ShapeTree<se::DeviceMemoryBase>& buffers() { return buffers_; } in buffers() 119 ShapeTree<se::DeviceMemoryBase> buffers_; 169 *buffers_.mutable_element(index) = se::DeviceMemoryBase(); in set_buffer()
|
D | transfer_manager.h | 168 const se::DeviceMemoryBase& dest, 172 const se::DeviceMemoryBase& source, 178 const se::DeviceMemoryBase& dest, 182 const se::DeviceMemoryBase& source, 270 const se::DeviceMemoryBase& source, 279 se::DeviceMemoryBase* destination); 285 se::Stream* stream, absl::Span<const se::DeviceMemoryBase> elements, 286 const Shape& shape, se::DeviceMemoryBase* region) = 0;
|
D | maybe_owning_device_memory.h | 34 explicit MaybeOwningDeviceMemory(se::DeviceMemoryBase unowned) in MaybeOwningDeviceMemory() 39 MaybeOwningDeviceMemory& operator=(se::DeviceMemoryBase unowned) { 53 se::DeviceMemoryBase AsDeviceMemoryBase(); 65 absl::variant<OwningDeviceMemory, se::DeviceMemoryBase> mem_;
|
D | transfer_manager.cc | 112 se::Stream* stream, const Shape& shape, const se::DeviceMemoryBase& source, in TransferArrayFromDevice() 141 const se::DeviceMemoryBase& dest, in TransferArrayToDevice() 156 const se::DeviceMemoryBase& dest, in TransferArrayToDeviceAsync() 178 se::Stream* stream, const Shape& shape, const se::DeviceMemoryBase& source, in TransferArrayFromDevice() 247 se::DeviceMemoryBase device_memory = device_buffer.buffer(index); in WriteTupleIndexTablesAsync() 251 std::vector<se::DeviceMemoryBase> elements; in WriteTupleIndexTablesAsync() 270 se::DeviceMemoryBase device_memory = device_buffer.buffer({}); in WriteRootTupleIndexTable() 274 std::vector<se::DeviceMemoryBase> elements; in WriteRootTupleIndexTable() 284 se::Stream* stream, const se::DeviceMemoryBase& source, int64 size, in TransferBufferFromDevice() 298 se::DeviceMemoryBase* destination) { in TransferBufferToDevice() [all …]
|
D | shaped_buffer_test.cc | 69 return OwningDeviceMemory(se::DeviceMemoryBase(buf, size), device_ordinal, in Allocate() 73 Status Deallocate(int device_ordinal, se::DeviceMemoryBase mem) override { in Deallocate() 121 [&](const xla::ShapeIndex& index, se::DeviceMemoryBase* buffer) { in TEST() 127 ShapeTree<se::DeviceMemoryBase> buffers = sb.buffers(); in TEST() 134 const se::DeviceMemoryBase& buffer) { in TEST() 142 [&](const xla::ShapeIndex& index, const se::DeviceMemoryBase& buffer) { in TEST()
|
/external/tensorflow/tensorflow/stream_executor/gpu/ |
D | gpu_executor.h | 91 void* AllocateSubBuffer(DeviceMemoryBase* mem, uint64 offset_bytes, 94 void Deallocate(DeviceMemoryBase* mem) override; 122 bool SynchronousMemZero(DeviceMemoryBase* location, uint64 size) override; 124 bool SynchronousMemSet(DeviceMemoryBase* location, int value, 127 port::Status SynchronousMemcpy(DeviceMemoryBase* gpu_dst, 131 const DeviceMemoryBase& gpu_src, 134 port::Status SynchronousMemcpyDeviceToDevice(DeviceMemoryBase* gpu_dst, 135 const DeviceMemoryBase& gpu_src, 138 bool MemZero(Stream* stream, DeviceMemoryBase* location, 140 bool Memset(Stream* stream, DeviceMemoryBase* location, uint8 pattern, [all …]
|
/external/tensorflow/tensorflow/core/common_runtime/gpu/ |
D | gpu_debug_allocator_test.cc | 50 se::DeviceMemory<int64> gpu_array_ptr{se::DeviceMemoryBase{gpu_array}}; in TEST() 79 se::DeviceMemoryBase{gpu_array}}; in TEST() 84 se::DeviceMemoryBase{gpu_array - 1}}; in TEST() 115 se::DeviceMemoryBase{gpu_array}}; in TEST() 121 se::DeviceMemoryBase{gpu_array + s}}; in TEST() 148 se::DeviceMemory<float> gpu_array_ptr{se::DeviceMemoryBase{gpu_array}}; in TEST() 195 se::DeviceMemory<float> gpu_array_ptr{se::DeviceMemoryBase{gpu_array}}; in TEST()
|
D | gpu_util.cc | 60 using se::DeviceMemoryBase; 161 DeviceMemoryBase gpu_src_ptr(src_ptr, total_bytes); in SetProtoFromGPU() 215 DeviceMemoryBase gpu_src_ptr(src_ptr, total_bytes); in DeviceToDeviceCopy() 217 DeviceMemoryBase gpu_dst_ptr(dst_ptr, total_bytes); in DeviceToDeviceCopy() 282 DeviceMemoryBase gpu_src_ptr(src_ptr, total_bytes); in CopyGPUTensorToCPU() 329 DeviceMemoryBase gpu_dst_ptr(dst_ptr, total_bytes); in CopyCPUTensorToGPU() 382 DeviceMemoryBase gpu_ptr(ptr, num_bytes); in MemoryDebugString() 440 DeviceMemoryBase gpu_src_ptr(src_ptr, total_bytes); in CopyGPUTensorToSameGPU() 442 DeviceMemoryBase gpu_dst_ptr(dst_ptr, total_bytes); in CopyGPUTensorToSameGPU()
|
/external/tensorflow/tensorflow/stream_executor/rocm/ |
D | rocm_gpu_executor.cc | 81 static hipDeviceptr_t AsROCmDevicePtr(const DeviceMemoryBase& gpu_mem) { in AsROCmDevicePtr() 86 static hipDeviceptr_t AsROCmDevicePtr(DeviceMemoryBase* gpu_mem) { in AsROCmDevicePtr() 429 void* GpuExecutor::AllocateSubBuffer(DeviceMemoryBase* mem, uint64 offset_bytes, in AllocateSubBuffer() 435 void GpuExecutor::Deallocate(DeviceMemoryBase* mem) { in Deallocate() 460 bool GpuExecutor::SynchronousMemZero(DeviceMemoryBase* location, uint64 size) { in SynchronousMemZero() 470 bool GpuExecutor::SynchronousMemSet(DeviceMemoryBase* location, int value, in SynchronousMemSet() 485 port::Status GpuExecutor::SynchronousMemcpy(DeviceMemoryBase* gpu_dst, in SynchronousMemcpy() 492 const DeviceMemoryBase& gpu_src, in SynchronousMemcpy() 499 DeviceMemoryBase* gpu_dst, const DeviceMemoryBase& gpu_src, uint64 size) { in SynchronousMemcpyDeviceToDevice() 504 bool GpuExecutor::MemZero(Stream* stream, DeviceMemoryBase* location, in MemZero() [all …]
|