Can I use MCNP on ARM devices?

[Nikel]

TL;DR

I’m a first-year Physics student planning to learn MCNP this summer and looking for a new laptop. Since MCNP is made for x86 processors, I'm trying to find out if it has native ARM support, or if there are reliable workarounds to run it smoothly on an ARM-based laptop. I'm asking here assuming this is the best place to find MCNP users who might have faced the same issue.

Hello everyone,

As a freshman Physics student, I plan to learn MCNP and improve my skills in this field over the summer break. This might not be the ideal place to ask, but I am currently looking for a laptop that can smoothly run MCNP for this purpose.

My main question is regarding the architecture compatibility: since the program is natively compiled for x86 processors, what are my options for running it on an ARM-based laptop? Alternatively, does MCNP offer native support for ARM-based systems?

As I mentioned at the beginning, this might not be the exact forum for such a query, but I wanted to ask here because I assume this is where MCNP users are most likely to be concentrated. I am hoping someone who has encountered a similar dilemma or found a workaround could share their insights.

Thank you in advance for your help!

 

[Astronuc]

One has to look at the operating system (OS) for a given computer.

MCNP should run on any ARM-based laptop, whether an Apple system using (Apple's OS or MacOS) or on any Windows based system as long as MCNP is compiled on a system using the same or equivalent OS.

Normally, one obtains an MCNP executable, so the executable has to be compiled using a compiler program compatible with the OS.

 

[Alex A]

My information may be out of date, but I've checked RSICC and the situation does not seem to have changed. The only option that would let you run it locally on an ARM machine is to request the source code and compile it. The person to ask if this is feasible is the person teaching the MCNP course. If it needs to be done formally through RSICC for version 6, source code requests have more checks and things are a lot slower. If you are outside the US it may be tougher than if you are inside, and requests are hugely expensive for non US universities (free for those in the US, over a thousand dollars otherwise). A very powerful laptop is not needed and remote access to a computing cluster (which either has an existing install, or you install) may be normal anyway. Talk to the teacher as soon as possible, if you are outside the US prepare to be disappointed, if you are inside the US...maybe? Either way good luck!

I've read @Astronuc's reply and I suspect he has misunderstood what an executable is (they are fixed for an architecture and OS). Compiling requires the source code. For my part I've assumed that the macos binary is x86 only - I may be wrong about that. Again, ask your teacher.

 

[Nikel]

My information may be out of date, but I've checked RSICC and the situation does not seem to have changed. The only option that would let you run it locally on an ARM machine is to request the source code and compile it. The person to ask if this is feasible is the person teaching the MCNP course. If it needs to be done formally through RSICC for version 6, source code requests have more checks and things are a lot slower. If you are outside the US it may be tougher than if you are inside, and requests are hugely expensive for non US universities (free for those in the US, over a thousand dollars otherwise). A very powerful laptop is not needed and remote access to a computing cluster (which either has an existing install, or you install) may be normal anyway. Talk to the teacher as soon as possible, if you are outside the US prepare to be disappointed, if you are inside the US...maybe? Either way good luck!

I've read @Astronuc's reply and I suspect he has misunderstood what an executable is (they are fixed for an architecture and OS). Compiling requires the source code. For my part I've assumed that the macos binary is x86 only - I may be wrong about that. Again, ask your teacher.

Thank you for the detailed response. First of all, I am based outside the US, and my intention is actually to learn MCNP on my own through online resources rather than taking a formal university course.

Based on both of your replies, it seems clear to me now that there is no native ARM support available out of the box. If running it on an ARM-based system requires requesting the source code and compiling/tweaking it myself, that is definitely beyond my current technical reach :D

Given the situation, it looks like my best bet is to avoid the hassle entirely and go with a modern laptop powered by a standard x86 AMD or Intel processor.

 

[Astronuc]

If you are outside the US it may be tougher than if you are inside,

Yes, the code is 'Export Controlled'; it is restricted outside the US. Generally the source code is not provided.

so the executable has to be compiled using a compiler program compatible with the OS.

i should have been more clear. I was not refering to the user compiling the source, but rather someone at RSICC or LANL would have to compile the source code on a compatible (or same) architecture and OS.

I've read @Astronuc's reply and I suspect he has misunderstood what an executable is (they are fixed for an architecture and OS).

I understood. One would have to request an executable compiled using a compatible OS and architecture. It can be expensive even for those inside the US.

From LANL

Because the MCNP code is export controlled, the distribution through RSICC is governed by US laws and Department of Energy regulations. US citizens may request either the executable-only distribution package or the distribution package that includes source coding. Non-US citizens should generally request the executable-only distribution package to avoid possible delays in approval.

The MCNP6 distribution package from RSICC includes executables for Linux, macOS, and Windows. The package also includes documentation (release notes & user manual), numerous verification and validation problems, a variety of utility programs and scripts, and numerous ACE-formatted nuclear data libraries.

https://mcnp.lanl.gov/how_to_get_the_mcnp_code.html

An alternatives is OpenMC.

OpenMC is an open source Monte Carlo neutron and photon simulation transport code. Initially developed by the Computational Reactor Physics Group at MIT in 2011 as part of a project to develop scalable parallel algorithms for future exascale supercomputers, it has been contributed to by various universities, laboratories, and other institutions since its release to the public in December 2012. Unlike other major Monte Carlo transport codes such as MCNP or Serpent, it is not export controlled. It has been used in labs including the Consortium for Advanced Simulation of LWRs and the ANL Center for Exascale Simulation of Advanced Reactors.

Features of OpenMC include the ability to perform fixed source, k-eigenvalue, and subcritical multiplication calculations on models built with Constructive solid geometry or CAD. It also features large Python and C/C++ APIs that expand its features.

https://en.wikipedia.org/wiki/OpenMC
https://openmc.org/

OpenMC is a community-developed Monte Carlo neutron and photon transport code. It is capable of performing fixed source, k-eigenvalue, and subcritical multiplication calculations on models built using either a constructive solid geometry or CAD representation. A flexible and efficient tally system enables a wide variety of physical quantities to be tallied and analyzed. OpenMC can run in parallel using a hybrid MPI and OpenMP programming model and has been extensively tested on leadership class supercomputers.