From 36de178aaee72e3a7e3baf9d0220bc773383f1f0 Mon Sep 17 00:00:00 2001
From: Wilber <jiweibo@baidu.com>
Date: Mon, 11 Jan 2021 15:05:31 +0800
Subject: [PATCH] Cherry-pick  30194 30164 30201(#30202)

---
 paddle/fluid/inference/api/paddle_pass_builder.cc | 4 ++++
 paddle/fluid/inference/tensorrt/CMakeLists.txt    | 5 +++++
 paddle/fluid/operators/shape_op.cc                | 1 +
 paddle/fluid/operators/shape_op.cu                | 3 ++-
 4 files changed, 12 insertions(+), 1 deletion(-)

diff --git a/paddle/fluid/inference/api/paddle_pass_builder.cc b/paddle/fluid/inference/api/paddle_pass_builder.cc
index 95303e7e85..4b0bfbe2af 100644
--- a/paddle/fluid/inference/api/paddle_pass_builder.cc
+++ b/paddle/fluid/inference/api/paddle_pass_builder.cc
@@ -220,8 +220,12 @@ void CpuPassStrategy::EnableMKLDNN() {
              // "fc_mkldnn_pass",
              // "fc_act_mkldnn_fuse_pass",
              "batch_norm_act_fuse_pass",
+#ifndef _WIN32
+             // TODO(intel): Please fix the bug on windows.
+             // https://github.com/PaddlePaddle/Paddle/issues/29710
              "mkldnn_inplace_pass",  // This pass should be activated after
                                      // fuses
+#endif
          })) {
       passes_.push_back(pass);
     }
diff --git a/paddle/fluid/inference/tensorrt/CMakeLists.txt b/paddle/fluid/inference/tensorrt/CMakeLists.txt
index 4f3da10f6e..d1d146b2ce 100644
--- a/paddle/fluid/inference/tensorrt/CMakeLists.txt
+++ b/paddle/fluid/inference/tensorrt/CMakeLists.txt
@@ -1,4 +1,9 @@
+# Compiling with WITH_PYTHON=ON and WITH_TENSORRT=ON failed on windows. Temporarily add paddle_inference_api dependency to solve the problem
+if(WIN32)
+nv_library(tensorrt_engine SRCS engine.cc trt_int8_calibrator.cc DEPS ${GLOB_OPERATOR_DEPS} framework_proto device_context boost paddle_inference_api)
+else()
 nv_library(tensorrt_engine SRCS engine.cc trt_int8_calibrator.cc DEPS ${GLOB_OPERATOR_DEPS} framework_proto device_context boost)
+endif()
 nv_library(tensorrt_op_teller SRCS op_teller.cc DEPS framework_proto device_context boost)
 nv_test(test_tensorrt SRCS test_tensorrt.cc DEPS dynload_cuda device_context dynamic_loader)
 nv_test(test_tensorrt_engine SRCS test_engine.cc DEPS dynload_cuda tensorrt_engine)
diff --git a/paddle/fluid/operators/shape_op.cc b/paddle/fluid/operators/shape_op.cc
index 0ecf9bfb5d..d8ec12659f 100644
--- a/paddle/fluid/operators/shape_op.cc
+++ b/paddle/fluid/operators/shape_op.cc
@@ -69,5 +69,6 @@ REGISTER_OPERATOR(
     paddle::framework::EmptyGradOpMaker<paddle::framework::OpDesc>,
     paddle::framework::EmptyGradOpMaker<paddle::imperative::OpBase>);
 REGISTER_OP_CPU_KERNEL(shape, ops::ShapeKernel<bool>, ops::ShapeKernel<int>,
+                       ops::ShapeKernel<int8_t>, ops::ShapeKernel<uint8_t>,
                        ops::ShapeKernel<int64_t>, ops::ShapeKernel<float>,
                        ops::ShapeKernel<double>);
diff --git a/paddle/fluid/operators/shape_op.cu b/paddle/fluid/operators/shape_op.cu
index 5d50b17818..fce723c784 100644
--- a/paddle/fluid/operators/shape_op.cu
+++ b/paddle/fluid/operators/shape_op.cu
@@ -16,7 +16,8 @@ limitations under the License. */
 
 REGISTER_OP_CUDA_KERNEL(
     shape, paddle::operators::ShapeKernel<bool>,
-    paddle::operators::ShapeKernel<int>,
+    paddle::operators::ShapeKernel<int>, paddle::operators::ShapeKernel<int8_t>,
+    paddle::operators::ShapeKernel<uint8_t>,
     paddle::operators::ShapeKernel<int64_t>,
     paddle::operators::ShapeKernel<float>,
     paddle::operators::ShapeKernel<double>,
-- 
GitLab