Quantum ESPRESSO: One Further Step toward the Exascale.

Area of Science:

• Computational physics and chemistry
• Materials science
• Quantum mechanics

Background:

• Quantum ESPRESSO is a popular open-source suite for electronic-structure calculations.
• Traditional calculations are computationally intensive, limiting system sizes and simulation times.
• The increasing availability of Graphics Processing Units (GPUs) offers potential for significant acceleration.

Purpose of the Study:

• To review the current status of GPU porting for Quantum ESPRESSO.
• To highlight recent advancements in accelerating the software suite.
• To present performance benchmarks of GPU-accelerated codes.

Main Methods:

• Utilizing OpenACC and CUDA Fortran for code offloading to GPUs.
• Focusing on the porting of main Quantum ESPRESSO codes, including linear-response functionalities.
• Conducting extensive performance benchmarks across diverse GPU-accelerated architectures.

Main Results:

• Successful porting of key Quantum ESPRESSO codes to GPUs has been achieved.
• Demonstrated significant performance improvements, particularly for linear-response calculations.
• Extensive benchmarks provide insights into performance gains on various GPU hardware.

Conclusions:

• GPU acceleration via OpenACC and CUDA Fortran is effectively implemented in Quantum ESPRESSO.
• The enhanced suite offers substantial speedups for electronic-structure calculations.
• This advancement is expected to benefit researchers in computational materials science and related fields.