From c333af2f60033114b459ec78508e116fe8d9677f Mon Sep 17 00:00:00 2001
From: Zhang Jun <ewalker@live.cn>
Date: Mon, 10 Oct 2022 16:40:06 +0800
Subject: [PATCH] [inference] CPU-> GPU async io copy for TensorRT using
 ShareExternalData API (#46636)

---
 paddle/fluid/operators/tensorrt/tensorrt_engine_op.h | 3 +--
 1 file changed, 1 insertion(+), 2 deletions(-)

diff --git a/paddle/fluid/operators/tensorrt/tensorrt_engine_op.h b/paddle/fluid/operators/tensorrt/tensorrt_engine_op.h
index 5ad7980b8af..18ed1d9eaa9 100644
--- a/paddle/fluid/operators/tensorrt/tensorrt_engine_op.h
+++ b/paddle/fluid/operators/tensorrt/tensorrt_engine_op.h
@@ -501,8 +501,7 @@ class TensorRTEngineOp : public framework::OperatorBase {
       // check the input_tensor
       if (!platform::is_gpu_place(t.place())) {
         phi::DenseTensor out;
-        platform::CUDAPlace dst_place;
-        framework::TransDataDevice(t, dst_place, &out);
+        framework::TensorCopy(t, dev_place, dev_ctx, &out);
         t.ShareDataWith(out);
       }
       auto t_shape = phi::vectorize<int64_t>(t.dims());
-- 
GitLab