/external/tensorflow/tensorflow/core/common_runtime/gpu/ |
D | gpu_event_mgr_test.cc | 100 auto stream_exec = GPUMachineManager()->ExecutorForDevice(0).ValueOrDie(); in TEST() local 101 EventMgr em(stream_exec, GPUOptions()); in TEST() 117 auto stream_exec = GPUMachineManager()->ExecutorForDevice(0).ValueOrDie(); in TEST() local 118 EventMgr em(stream_exec, GPUOptions()); in TEST() 122 std::unique_ptr<se::Stream> stream(new se::Stream(stream_exec)); in TEST() 150 auto stream_exec = GPUMachineManager()->ExecutorForDevice(0).ValueOrDie(); in TEST() local 151 EventMgr em(stream_exec, GPUOptions()); in TEST() 154 std::unique_ptr<se::Stream> stream(new se::Stream(stream_exec)); in TEST() 167 auto stream_exec = GPUMachineManager()->ExecutorForDevice(0).ValueOrDie(); in TEST() local 168 EventMgr em(stream_exec, GPUOptions()); in TEST() [all …]
|
D | gpu_debug_allocator_test.cc | 43 auto stream_exec = in TEST() local 51 ASSERT_TRUE(stream_exec->SynchronousMemcpy(&gpu_array_ptr, &cpu_array[0], in TEST() 71 auto stream_exec = in TEST() 80 ASSERT_TRUE(stream_exec->SynchronousMemcpy( in TEST() 88 stream_exec->SynchronousMemcpy(&gpu_hdr_ptr, &pi, sizeof(float))); in TEST() 107 auto stream_exec = in TEST() 116 ASSERT_TRUE(stream_exec->SynchronousMemcpy( in TEST() 124 stream_exec->SynchronousMemcpy(&gpu_ftr_ptr, &pi, sizeof(float))); in TEST() 140 auto stream_exec = in TEST() local 149 ASSERT_TRUE(stream_exec->SynchronousMemcpy(&cpu_array[0], gpu_array_ptr, in TEST() [all …]
|
D | gpu_host_allocator.h | 29 explicit GpuHostAllocator(se::StreamExecutor* stream_exec, int numa_node, in GpuHostAllocator() argument 33 stream_exec_(stream_exec), in GpuHostAllocator()
|
D | gpu_bfc_allocator.h | 40 explicit GPUMemAllocator(se::StreamExecutor* stream_exec, in GPUMemAllocator() argument 45 stream_exec_(stream_exec), in GPUMemAllocator()
|
/external/tensorflow/tensorflow/compiler/xla/service/gpu/ |
D | buffer_comparator.cc | 84 auto stream_exec = stream->parent(); in Create() local 101 compiler->RunHloPasses(std::move(module), stream_exec, nullptr)); in Create() 102 return compiler->RunBackend(std::move(module), stream_exec, nullptr); in Create() 107 auto device_ordinal = stream_exec->device_ordinal(); in Create() 135 auto stream_exec = stream_->parent(); in CompareEqualImpl() local 137 auto device_ordinal = stream_exec->device_ordinal(); in CompareEqualImpl() 138 ShapedBuffer shaped_test_buffer(shape, shape, stream_exec->platform(), in CompareEqualImpl() 142 run_options.set_device_ordinal(stream_exec->device_ordinal()); in CompareEqualImpl()
|
D | nvptx_compiler.cc | 179 Status OptimizeHloModule(HloModule* hlo_module, se::StreamExecutor* stream_exec, in OptimizeHloModule() argument 273 pipeline.AddPass<CusolverRewriter>(stream_exec, device_allocator); in OptimizeHloModule() 277 if (IsVoltaOrLater(*stream_exec)) { in OptimizeHloModule() 300 LayoutAssignment::InstructionCanChangeLayout, stream_exec); in OptimizeHloModule() 344 pipeline.AddPass<CudnnConvAlgorithmPicker>(stream_exec, device_allocator, in OptimizeHloModule() 619 std::unique_ptr<HloModule> module, se::StreamExecutor* stream_exec, in RunHloPasses() argument 626 OptimizeHloModule(module.get(), stream_exec, device_allocator, this)); in RunHloPasses() 634 std::unique_ptr<HloModule> module, se::StreamExecutor* stream_exec, in RunBackend() argument 638 TF_RET_CHECK(stream_exec != nullptr); in RunBackend() 682 &stream_exec->GetDeviceDescription(), in RunBackend() [all …]
|
D | cudnn_conv_algorithm_picker.cc | 45 se::StreamExecutor* stream_exec) { in GetAlgorithms() argument 51 stream_exec->GetConvolveBackwardFilterAlgorithms(true, &algorithms); in GetAlgorithms() 54 succ = stream_exec->GetConvolveBackwardDataAlgorithms(true, &algorithms); in GetAlgorithms() 58 succ = stream_exec->GetConvolveAlgorithms(true, &algorithms); in GetAlgorithms() 83 tensorflow::mutex_lock LockGpu(const se::StreamExecutor* stream_exec) { in LockGpu() argument 93 std::make_tuple(stream_exec->platform(), in LockGpu() 94 stream_exec->device_ordinal()), in LockGpu()
|
D | cudnn_conv_algorithm_picker.h | 40 CudnnConvAlgorithmPicker(se::StreamExecutor* stream_exec, in CudnnConvAlgorithmPicker() argument 42 : stream_exec_(stream_exec), allocator_(allocator), compiler_(compiler) {} in CudnnConvAlgorithmPicker()
|
D | nvptx_compiler.h | 54 std::unique_ptr<HloModule> module, se::StreamExecutor* stream_exec, 58 std::unique_ptr<HloModule> module, se::StreamExecutor* stream_exec,
|
D | cusolver_rewriter.cc | 202 CusolverRewriter::CusolverRewriter(se::StreamExecutor* stream_exec, in CusolverRewriter() argument 204 : stream_exec_(stream_exec), allocator_(allocator) {} in CusolverRewriter()
|
D | cusolver_rewriter.h | 32 CusolverRewriter(se::StreamExecutor* stream_exec,
|
D | stream_executor_util.h | 31 bool IsVoltaOrLater(const se::StreamExecutor& stream_exec);
|
/external/tensorflow/tensorflow/compiler/xla/service/interpreter/ |
D | compiler.cc | 113 std::unique_ptr<HloModule> hlo_module, se::StreamExecutor* stream_exec, in RunBackend() argument 115 TF_RET_CHECK(stream_exec != nullptr); in RunBackend() 139 std::vector<std::vector<se::StreamExecutor*>> stream_exec, in RunBackendOnModuleGroup() argument 147 std::vector<std::vector<se::StreamExecutor*>> stream_exec, in Compile() argument 156 if (stream_exec.size() != 1 || stream_exec[0].size() != 1) { in Compile() 162 RunHloPasses(std::move(hlo_modules[0]), stream_exec[0][0], in Compile() 166 RunBackend(std::move(module), stream_exec[0][0], device_allocator)); in Compile()
|
D | compiler.h | 47 std::unique_ptr<HloModule> hlo_module, se::StreamExecutor* stream_exec, 55 std::unique_ptr<HloModule> hlo_module, se::StreamExecutor* stream_exec, 59 std::vector<std::vector<se::StreamExecutor*>> stream_exec, 64 std::vector<std::vector<se::StreamExecutor*>> stream_exec,
|
/external/tensorflow/tensorflow/compiler/xla/ |
D | device_util.h | 32 string DeviceIdentifier(se::StreamExecutor* stream_exec) { in DeviceIdentifier() argument 33 return absl::StrCat(stream_exec->platform()->Name(), ":", in DeviceIdentifier() 34 stream_exec->device_ordinal()); in DeviceIdentifier()
|
/external/tensorflow/tensorflow/core/kernels/ |
D | gpu_utils.cc | 60 se::StreamExecutor* stream_exec, in LogConvAutotuneResults() argument 72 *log.mutable_cudnn_version() = GetCudnnVersion(stream_exec); in LogConvAutotuneResults() 73 *log.mutable_compute_capability() = GetComputeCapability(stream_exec); in LogConvAutotuneResults() 83 se::StreamExecutor* stream_exec, in LogFusedConvAutotuneResults() argument 102 *log.mutable_cudnn_version() = GetCudnnVersion(stream_exec); in LogFusedConvAutotuneResults() 103 *log.mutable_compute_capability() = GetComputeCapability(stream_exec); in LogFusedConvAutotuneResults()
|
D | gpu_utils.h | 168 se::StreamExecutor* stream_exec, 175 se::StreamExecutor* stream_exec,
|
D | conv_ops_gpu.h | 144 se::StreamExecutor* stream_exec) const { in ShouldIncludeWinogradNonfusedAlgo() argument 145 auto* dnn_support = stream_exec->AsDnn(); in ShouldIncludeWinogradNonfusedAlgo()
|
/external/tensorflow/tensorflow/stream_executor/gpu/ |
D | gpu_activation.cc | 26 GpuExecutor* ExtractGpuExecutor(StreamExecutor* stream_exec); 34 StreamExecutor* stream_exec) in ScopedActivateExecutorContext() argument 35 : ScopedActivateExecutorContext(ExtractGpuExecutor(stream_exec)) {} in ScopedActivateExecutorContext()
|
D | gpu_activation.h | 45 explicit ScopedActivateExecutorContext(StreamExecutor* stream_exec);
|
/external/tensorflow/tensorflow/stream_executor/ |
D | event.cc | 24 Event::Event(StreamExecutor* stream_exec) in Event() argument 25 : stream_exec_(stream_exec), in Event()
|
D | event.h | 49 explicit Event(StreamExecutor* stream_exec); // NOLINT
|
/external/tensorflow/tensorflow/compiler/xla/service/cpu/ |
D | cpu_compiler.h | 137 std::unique_ptr<HloModule> module, se::StreamExecutor* stream_exec, 141 std::unique_ptr<HloModule> module, se::StreamExecutor* stream_exec,
|
/external/tensorflow/tensorflow/compiler/xla/service/ |
D | compiler.h | 178 std::vector<std::vector<se::StreamExecutor*>> stream_exec, 191 std::vector<std::vector<se::StreamExecutor*>> stream_exec,
|
D | llvm_compiler.h | 79 std::vector<std::vector<se::StreamExecutor*>> stream_exec,
|