Re: [AMBER] Question about QM/MM parallelization in sander.MPI

From: Adrian Roitberg via AMBER <amber.ambermd.org>
Date: Wed, 29 Apr 2026 10:54:01 -0400

Hi

There are two levels of parallelism when you do QM/MM in sander.

One is within the QM part itself, where usually different integrals are
computing in different cores, and then they are all put together to get
a single energy and forces.

The second level is the one you described, which is assigning different
atoms to different cores, to compute classical forces and propagate the
dynamics.


Adrian


On 4/29/26 4:43 AM, 张玮豪 via AMBER wrote:
> [External Email]
>
> Dear Amber Developers,
>
> I am currently performing QM/MM molecular dynamics simulations using sander.MPI, and I have some questions regarding the parallelization strategy of the QM part.
>
>
>
> From the output of my simulation, I see messages such as:
>
> "QMMM: Quantum atom + link atom division among threads"
>
> where the QM region (including link atoms) is divided among multiple threads (e.g., 32 threads). Each thread is assigned a subset of QM atoms.
>
>
>
> For classical molecular dynamics simulations, it is straightforward to understand: each thread is responsible for computing the potential energy and forces of all atoms within its assigned spatial region, and the results from all threads are then combined. However, for QM/MM molecular dynamics, due to the special nature of the QM region calculations, it is not possible to treat atoms in different regions independently as in classical MD. In this case, how does sander.MPI actually work?
>
>
>
> Thank you very much for your time and assistance. I truly appreciate your help and any insights you can provide regarding my question.
>
> I look forward to your response.
>
>
>
> Best regards,
> Weihao Zhang
> _______________________________________________
> AMBER mailing list
> AMBER.ambermd.org
> http://lists.ambermd.org/mailman/listinfo/amber

-- 
Dr. Adrian E. Roitberg
Frank E. Harris Professor
Department of Chemistry
University of Florida
roitberg.ufl.edu
352-392-6972
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Wed Apr 29 2026 - 08:00:03 PDT
Custom Search