Key points are not available for this paper at this time.
In the realm of computer science research, compiler optimization is regarded as one of the core techniques for program optimization, greatly propelling the advancement of high-performance computing. During the process of optimizing a program through compilation, it is acknowledged that while the selected compiler from a candidate set may exhibit optimal effects on overall program optimization, it may not necessarily yield the most optimized results for certain code segments within the program. To address this issue, this paper proposes and practically implements a multi-compiler fusion optimization framework (MCFOF) based on functions using ROSE compiler infrastructure. MCFOF selects the optimal compiler for each function within the program, thus integrating the strengths of multiple compilers. Experimental results demonstrate that MCFOF significantly enhances code performance compared to three selected general-purpose compilers and a polyhedral com-piler on multiple platforms. Specifically, it achieves an average performance improvement of 33.2% on the x86 64 platform and 13.4% on the ARMv8 platform. Furthermore, these results confirm the framework's platform independence.
Wang et al. (Fri,) studied this question.