diff options
| author |  Peter Hawkins <phawkins@google.com> | 2018-08-09 16:12:33 -0700 |
|---|---|---|
| committer |  TensorFlower Gardener <gardener@tensorflow.org> | 2018-08-09 16:17:24 -0700 |
| commit | b306f5f9458feddbdb89b7db557cb74dc9408d07 (patch) | |
| tree | c816564723de457711e9bbabe98617d3f2e121ad /tensorflow/stream_executor | |
| parent | 243f6e636c93d63884c574ae9b61d397726189ed (diff) | |
[TF:XLA] Add a real implementation of XlaDevice::Sync() so Session::Run() will correctly wait for all computations to complete on an XLA device before termination.
[TF:XLA] Change the XlaTensor definition event to be a shared pointer to an stream_executor::Event. This allows many tensors to share the same definition event.
PiperOrigin-RevId: 208128264
Diffstat (limited to 'tensorflow/stream_executor')
| -rw-r--r-- | tensorflow/stream_executor/host/host_gpu_executor.h | 2 |
1 files changed, 1 insertions, 1 deletions
diff --git a/tensorflow/stream_executor/host/host_gpu_executor.h b/tensorflow/stream_executor/host/host_gpu_executor.h index 858396ef96..7ba1f18101 100644 --- a/tensorflow/stream_executor/host/host_gpu_executor.h +++ b/tensorflow/stream_executor/host/host_gpu_executor.h @@ -88,7 +88,7 @@ class HostExecutor : public internal::StreamExecutorInterface { uint64 size) override; // No "synchronize all activity" implemented for this platform at the moment. - bool SynchronizeAllActivity() override { return false; } + bool SynchronizeAllActivity() override { return true; } bool SynchronousMemZero(DeviceMemoryBase *location, uint64 size) override; bool SynchronousMemSet(DeviceMemoryBase *location, int value, |