PulseExploreJournal ClubDebatesTrendingResearchersJournals
Instagram
HomeExploreJournal ClubTrending
Synapse
⌘+K
Synapse
May 17, 2026The Journal of Chemical Physics0 citations

Designing quantum chemistry algorithms with just-in-time compilation

View Full Paper
XWXiaojie WuQSQiming SunYWYuanheng Wang

Key Points

  • This research aims to enhance the efficiency of electron repulsion integral computations using just-in-time compilation.
  • Introduced JIT compilation to integral kernels for Gaussian-type orbitals.
  • Evaluated algorithms for Coulomb and exchange matrices on an NVIDIA A100-80G GPU.
  • Implemented a compact CUDA core with only ∼1000 lines of code.
  • Achieved a 2× speedup for the small 6-31G* basis set on GPU4PySCF v1.4 over the previous state.
  • Improved JK evaluations by up to 4× with the large def2-TZVPP basis set.
  • Single-precision implementation yielded a 3× speedup over the prior state-of-the-art.

Abstract

We introduce just-in-time (JIT) compilation to the integral kernels for Gaussian-type orbitals to enhance the efficiency of electron repulsion integral computations. For Coulomb and exchange (JK) matrices, JIT-based algorithms yield a 2× speedup for the small 6-31G* basis set over GPU4PySCF v1.4 on an NVIDIA A100-80G GPU. By incorporating a novel algorithm designed for orbitals with high angular momentum, the efficiency of JK evaluations with the large def2-TZVPP basis set is improved by up to 4×. The core CUDA implementation is compact, comprising only ∼1000 lines of code, including support for single-precision arithmetic. Furthermore, the single-precision implementation achieves a 3× speedup over the previous state-of-the-art.

Ask AI
Helpful
Bookmark
Share
View Full Paper

Cite This Study

Wu et al. (2026) studied this question.

synapsesocial.com/papers/6a095c2c7880e6d24efe23b1https://doi.org/10.1063/5.0314659
Ask AI
Helpful
Bookmark
Share
View Full Paper