From 44b7d26526f6465dacaf01a73d16ae13b57390b4 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Maximilian=20M=C3=BCller?= <maximilianm@nvidia.com>
Date: Thu, 5 Oct 2023 00:49:05 +0200
Subject: [PATCH] adding pre pack sync

---
 onnxruntime/core/providers/cuda/nn/conv.cc           | 1 +
 onnxruntime/core/providers/cuda/nn/conv_transpose.cc | 1 +
 2 files changed, 2 insertions(+)

diff --git a/onnxruntime/core/providers/cuda/nn/conv.cc b/onnxruntime/core/providers/cuda/nn/conv.cc
index f1626da36562e..4ccb5a3baeecb 100644
--- a/onnxruntime/core/providers/cuda/nn/conv.cc
+++ b/onnxruntime/core/providers/cuda/nn/conv.cc
@@ -118,6 +118,7 @@ Status Conv<T, NHWC>::PrePack(const Tensor& tensor, int input_idx, AllocatorPtr
       if (!status.IsOK()) {
         return status;
       }
+      CUDA_CALL_THROW(cudaStreamSynchronize(DefaultCudaStream()));
       is_packed = true;
     }
   }
diff --git a/onnxruntime/core/providers/cuda/nn/conv_transpose.cc b/onnxruntime/core/providers/cuda/nn/conv_transpose.cc
index a1c15fe257649..a8831a09605b7 100644
--- a/onnxruntime/core/providers/cuda/nn/conv_transpose.cc
+++ b/onnxruntime/core/providers/cuda/nn/conv_transpose.cc
@@ -77,6 +77,7 @@ Status ConvTranspose<T, NHWC>::PrePack(const Tensor& tensor, int input_idx, Allo
       if (!status.IsOK()) {
         return status;
       }
+      CUDA_CALL_THROW(cudaStreamSynchronize(DefaultCudaStream()));
       is_packed = true;
     }
   }