[XPU] update xdnn to 0907. (#45777)

* [XPU] update xdnn to 0906. test=kunlun

* [XPU] update xdnn to 0907. test=kunlun

cmake/external/xpu.cmake

@@ -10,7 +10,7 @@ set(XPU_RT_LIB_NAME "libxpurt.so")
 if(NOT DEFINED XPU_BASE_URL)
   set(XPU_BASE_URL_WITHOUT_DATE
       "https://baidu-kunlun-product.cdn.bcebos.com/KL-SDK/klsdk-dev")
-  set(XPU_BASE_URL "${XPU_BASE_URL_WITHOUT_DATE}/20220831")
+  set(XPU_BASE_URL "${XPU_BASE_URL_WITHOUT_DATE}/20220907")
 else()
   set(XPU_BASE_URL "${XPU_BASE_URL}")
 endif()
@@ -19,7 +19,7 @@ endif()
 if(NOT DEFINED XPU_XDNN_BASE_URL)
   set(XPU_XDNN_BASE_URL_WITHOUT_DATE
       "https://klx-sdk-release-public.su.bcebos.com/xdnn/dev")
-  set(XPU_XDNN_BASE_URL "${XPU_XDNN_BASE_URL_WITHOUT_DATE}/20220831")
+  set(XPU_XDNN_BASE_URL "${XPU_XDNN_BASE_URL_WITHOUT_DATE}/20220907")
 else()
   set(XPU_XDNN_BASE_URL "${XPU_XDNN_BASE_URL}")
 endif()

paddle/phi/kernels/xpu/rnn_grad_kernel.cc

@@ -92,14 +92,14 @@ void RnnGradKernel(const Context& dev_ctx,
 
   std::vector<std::vector<const T*>> parameter_lists;
   parameter_lists.resize(num_layers);
-  reset_parameter_vector(weight_list, num_layers, is_bidirec, &parameter_lists);
+  ResetParameterVector(weight_list, num_layers, is_bidirec, &parameter_lists);
 
   for (unsigned int i = 0; i < weight_grad_list.size(); ++i) {
     dev_ctx.template Alloc<T>(weight_grad_list[i]);
   }
   std::vector<std::vector<T*>> parameter_lists_grad;
   parameter_lists_grad.resize(num_layers);
-  reset_parameter_vector(
+  ResetParameterVector(
       weight_grad_list, num_layers, is_bidirec, &parameter_lists_grad);
 
   // allocate the memory and initization the x_grad

paddle/phi/kernels/xpu/rnn_kernel.cc

@@ -89,7 +89,7 @@ void RnnKernel(const Context& dev_ctx,
   // weightlist
   std::vector<std::vector<const T*>> parameter_lists;
   parameter_lists.resize(num_layers);
-  reset_parameter_vector(weight_list, num_layers, is_bidirec, &parameter_lists);
+  ResetParameterVector(weight_list, num_layers, is_bidirec, &parameter_lists);
 
   // init the output and allocate the memory
   dev_ctx.template Alloc<T>(out);

paddle/phi/kernels/xpu/rnn_util.h

@@ -19,10 +19,10 @@
 namespace phi {
 
 template <typename TensorType, typename T>
-void reset_parameter_vector(const std::vector<TensorType>& raw_params_vec,
-                            const int& num_layers,
-                            const bool& is_bidirec,
-                            std::vector<std::vector<T*>>* params_vec) {
+void ResetParameterVector(const std::vector<TensorType>& raw_params_vec,
+                          const int& num_layers,
+                          const bool& is_bidirec,
+                          std::vector<std::vector<T*>>* params_vec) {
   // the parameter raw seuquence is [FWhi, FWhh, BWhi, BWhh] * num_layers
   // + [FBhi, FBhh, BBhi, BBhh] * num_layers, we will reset the parameter to
   // ([FWhi, FWhh, FBhi, FBhh] + [BWhi, BWhh, BBhi, BBhh]) * num_layers

python/paddle/fluid/tests/unittests/xpu/get_test_cover_info.py

@@ -91,7 +91,10 @@ xpu_test_op_type_white_list = [
     "lamb_float16",
     "lars_momentum_float32",
     "resnet_unit",
-    "resnet_unit_grad"
+    "resnet_unit_grad",
+    "c_embedding_float32",  # unittests of collective ops do not using xpu testing framework
+    "c_sync_comm_stream_float32",
+    "c_sync_calc_stream_float32",
 ]
 xpu_test_device_op_white_list = []
 xpu_test_device_op_type_white_list = []