Transformer-based pre-trained models of code (PTMC) have been widely used and have achieved state-of-the-art performance in many mission-critical applications. However, they can be vulnerable to adversarial attacks through identifier substitution or coding style transformation, which can significantly reduce accuracy and may incur security concerns. Although several approaches have been proposed to generate adversarial examples for PTMC, the effectiveness and efficiency of these approaches, especially for different code intelligence tasks, is not well understood. To bridge this gap, this study systematically analyzes five state-of-the-art adversarial attack approaches from three perspectives: effectiveness, efficiency, and the quality of generated examples. The results showed that none of the five approaches balanced all these perspectives. Particularly, approaches with a high attack success rate tend to be time-consuming, and the adversarial code they generate often lack naturalness, and vice versa. To address this limitation, we explored the impact of perturbing identifiers under different contexts and found that identifier substitution within FOR and IF statements was the most effective. Based on these findings, we proposed a new approach that prioritizes different types of statements for various tasks and utilize beam search to generate adversarial examples. Evaluation results show that it outperforms the state-of-the-art ALERT in terms of effectiveness and efficiency while preserving the naturalness of the generated adversarial examples.