Accelerate inference period in op Cache method (#43857)

paddle/fluid/framework/operator.cc

@@ -1380,6 +1380,11 @@ void OperatorWithKernel::RunImpl(const Scope& scope,
     RuntimeContext ctx(Inputs(), Outputs(), scope);
     RunImpl(scope, place, &ctx);
     pre_scope_ = cur_scope;
+  } else if (run_phi_kernel_ && impl_ != nullptr && !need_prepare_data_ &&
+             !need_prepare_phi_data_) {
+    if (!all_kernels_must_compute_runtime_shape_)
+      this->Info().infer_shape_(impl_->getRuntimeInferShapeContext());
+    (*pt_kernel_)(impl_->getKernelContext());
   } else {
     if (runtime_ctx_.get() == nullptr || pre_scope_ != cur_scope) {
       std::lock_guard<std::mutex> lock(cache_update_mutex_);