Investigating the superiority of Intel oneAPI IFX compiler on Intel CPUs using different optimization levels: A case study on a CFD system

Feisal Alaswad, Poovammal Eswaran · 2023

Computational fluid dynamics (CFD) simulations are crucial in scientific and engineering domains. The compiler choice impacts CFD application performance, especially on specific processor architectures. This research compares the performance of a Fortran-based CFD system compiled with Intel OneAPI IFX against other commonly used compilers on Intel CPUs. Intel’s OneAPI IFX compiler is designed for Intel CPUs and offers advanced optimizations. A Typical Fortran CFD system was used for the study, compiling it with IFX, Gfortran, NAG, and PGI compilers. Performance indicators, including execution time, memory usage, and computational precision, were collected. Our research findings revealed that the IFX compiler regularly outperforms the other compilers on Intel CPUs. This research underscores the significant performance benefits of employing the IFX compiler to compile Fortran-based CFD systems. IFX compiler achieves high decimal precision, reaching up to 15 decimal digits, and it also excels in time execution. Compared to other compilers like NAG, PGI, and GFortran, the OneAPI IFX compiler demonstrates notable improvements in execution time, with a 7.7% enhancement without code optimization and up to 3.7% with maximum code optimization on the Ubuntu operating system and with a 38.4% enhancement without code optimization and up to 30.8% with maximum code optimization on the Windows operating system. These findings highlight the superior performance and efficiency of the IFX compiler, making it a compelling choice for optimizing CFD simulations. The IFX compiler is recommended for CFD researchers and professionals to optimize Fortran-based CFD codes for better simulation performance and quicker time-to-solution, specifically on Intel hardware.

Read the paper · More papers on PaperTik