Skip to content

Conversation

schwalbe10
Copy link

Summary

This paper presents a performance analysis of modern DFT implementations for quantum chemistry calculations, comparing SIESTA, PySCF, and GPU4PySCF on 38-atom chlorinated phosphoric acid systems.

Key Results

  • PySCF: 3.7× speedup compared to SIESTA
  • GPU4PySCF: 390× speedup with 98% cost reduction
  • Focus on environmentally-friendly alternatives to fluorinated materials

Code Availability

https://github.com/schwalbe10/quantum-chemistry-acceleration

This is a draft submission for review.

Copy link

github-actions bot commented Jun 12, 2025

Curvenote Preview

Directory Preview Checks Updated (UTC)
papers/kyohei_sahara 🔍 Inspect 20 checks passed (1 optional) Aug 7, 2025, 8:48 AM

@scipy-conference scipy-conference deleted a comment from github-actions bot Jun 12, 2025
@rowanc1 rowanc1 added paper This indicates that the PR in question is a paper draft This triggers Curvenote Preview actions labels Jun 12, 2025
@mihaimaruseac
Copy link
Collaborator

Hello,

I was assigned to review this. I will do a pass by the weekend, overall on a quick glance this looks good.

@ameyxd
Copy link
Contributor

ameyxd commented Jun 18, 2025

Inviting reviewers: @nyetsche and @mihaimaruseac

# Ensure that this title is the same as the one in `myst.yml`
title: "Quantum Chemistry Acceleration: Comparative Performance Analysis of Modern DFT Implementations"
abstract: |
This proceeding examines the acceleration of quantum chemistry calculations through
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Suggested change
This proceeding examines the acceleration of quantum chemistry calculations through
This article examines the acceleration of quantum chemistry calculations through

- **Enhanced statistical reliability**: Larger sample sizes enabling more robust conclusions and confident predictions in materials design studies through increased sampling of molecular configurations and property distributions
- **Integration with modern workflows**: Seamless compatibility with machine learning approaches and automated high-throughput computational screening pipelines, facilitating development of predictive models for materials discovery

The methodology established here provides a foundation for accelerating quantum chemistry calculations across various molecular systems, particularly those involving compounds with heteroatoms and solvation effects.

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This is not my scientific domain so perhaps heteroatoms and solvation effects can be defined in parentheses? If the definition is obvious a domain expert, then no need.

Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This is addressed in the latest edits: "The methodology established here provides a foundation for accelerating quantum chemistry calculations across various molecular systems, particularly those involving compounds with heteroatoms (atoms other than carbon and hydrogen) and solvation effects (interactions between solutes and surrounding solvent molecules)."

SIESTA have been widely adopted due to their robust implementations and
extensive feature sets. However, these frameworks often present computational
bottlenecks when applied to complex systems requiring extensive parameter space
exploration.

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

these frameworks - is it just SIESTA, or are there others?

As a general comment to the paper, I come away with the understanding that PySCF and GPU4PySCF are much faster and cheaper than SIESTA, but I'm not entirely sure why they're faster. Is it improved parallelization (MPI and particularly on GPUs)? Algorithmic breakthroughs? Is there any case where SIESTA would still be a better choice due to the problem/scientific aspects?

There's probably not enough space to go into details on the PySCF/GPU4PySCF architecture or algorithms, but maybe a sentence or two would help, example: PySCF uses the parallel Foo-Bar Algorithm [2017, Foo et al) to calculate the Frobius coefficient effectively.

Copy link
Contributor

@cdlindsey cdlindsey Aug 29, 2025

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This has been addressed in the latest edits. "PySCF achieves performance improvements through efficient integral computation algorithms and optimized parallel processing using OpenMP and MPI implementations, while GPU4PySCF further accelerates calculations by leveraging GPU parallelization for computationally intensive operations like two-electron integrals and matrix operations."

@nyetsche
Copy link

Not my field, so I can't speak of the scientific merits, but overall a solid paper.

@ameyxd
Copy link
Contributor

ameyxd commented Jul 2, 2025

@cdlindsey will serve as editor for this paper.

Copy link
Collaborator

@mihaimaruseac mihaimaruseac left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Thank you!

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
assigned-editor assigned-reviewer draft This triggers Curvenote Preview actions paper This indicates that the PR in question is a paper
Projects
None yet
Development

Successfully merging this pull request may close these issues.

6 participants