diff options
Diffstat (limited to 'tensorflow/compiler/xla/service/gpu/copy_thunk.h')
-rw-r--r-- | tensorflow/compiler/xla/service/gpu/copy_thunk.h | 7 |
1 files changed, 5 insertions, 2 deletions
diff --git a/tensorflow/compiler/xla/service/gpu/copy_thunk.h b/tensorflow/compiler/xla/service/gpu/copy_thunk.h index 8b128386f6..91564b520a 100644 --- a/tensorflow/compiler/xla/service/gpu/copy_thunk.h +++ b/tensorflow/compiler/xla/service/gpu/copy_thunk.h @@ -18,6 +18,7 @@ limitations under the License. #include "tensorflow/compiler/xla/service/buffer_assignment.h" #include "tensorflow/compiler/xla/service/gpu/buffer_allocations.h" +#include "tensorflow/compiler/xla/service/gpu/hlo_execution_profiler.h" #include "tensorflow/compiler/xla/service/gpu/thunk.h" #include "tensorflow/compiler/xla/service/hlo_instruction.h" #include "tensorflow/core/platform/stream_executor_no_cuda.h" @@ -40,7 +41,8 @@ class HostToDeviceCopyThunk : public Thunk { HostToDeviceCopyThunk& operator=(const HostToDeviceCopyThunk&) = delete; Status ExecuteOnStream(const BufferAllocations& buffer_allocations, - se::Stream* stream) override; + se::Stream* stream, + HloExecutionProfiler* profiler) override; private: const void* source_address_; @@ -63,7 +65,8 @@ class DeviceToDeviceCopyThunk : public Thunk { DeviceToDeviceCopyThunk& operator=(const DeviceToDeviceCopyThunk&) = delete; Status ExecuteOnStream(const BufferAllocations& buffer_allocations, - se::Stream* stream) override; + se::Stream* stream, + HloExecutionProfiler* profiler) override; private: const BufferAllocation::Slice source_buffer_; |