[MCX Newsletters] MCX Suite v2025 is here, upgrade today!

[Qianqian Fang]

(You received this message because you are a registered user of MCX (https://mcx.space) - an open-source Monte Carlo photon transport simulator - and had indicated that you wanted to be notified for future updates in your registration form. If you are no longer interested in this software, please feel free to click on the unsubscribe link at the bottom of this email. In that case, we apologize for the inconvenience.)

Dear MCX/MMC registered users,

After nearly a year of non-stop development, I am excited to let you know a new stable release, MCX/MMC v2025, has finally arrived.

Please go to our website https://mcx.space, or use the following link to access the latest packages

https://mcx.space/wiki/?Get

MCX/MCXCL/MMC v2025 is a maintenance release, containing fixes to a number of bugs (including a critical bug) to ensure stability and accuracy of our solvers. Key updates are highlighted below:

=== Critical bug fix ===

• A critical bug (#222) was fixed in this release, detailed in https://github.com/fangq/mcx/issues/222. This bug affects all simulations for mcx/mcxcl v2023 and v2024.2. It can produce skewed partial path distributions when a photon's weight drops to numerical 0 after traveling a long path, or in highly absorbing media.

=== Key updates for all components ===

• all OpenCL based simulators, including mcxcl/mcxlabcl/pmcxcl/mmc/mmclab have optimized thread configurations to run about 7x faster on Apple M1/2/3/4 processors.
• mcx/mcxcl/mmc support using -N/--net command to dynamically list and run online simulations hosted on our NeuroJSON.io database, including https://neurojson.org/db/mcx and https://neurojson.org/db/mmc. We expect to further expand these online library of simulations for both mcx and mmc
• for mcx/mcxcl/mmc, --bench flag now has a shortcut, -Q, can be used to list and run built-in benchmarks
• one can now set cfg.flog to 0 or 1 in mcxlab/mmclab to suppress printing the MCX banner
• we've made some deliberations on the desired "ground-truth" CSF optical properties in a newly published paper https://doi.org/10.1117/1.NPh.12.1.015009, particularly, we updated all CSF mua values for our built-in brain simulations
  
• mcx/mcxcl/mmc can use -f followed by nothing or a "-" to read JSON input from pipe, allowing one to use advanced text processors, such as sed/grep/curl/jq to dynamically modify input, such as
  

  curl -s -X GET https://neurojson.io:7777/mcx/aircube60 | jq '.Forward.Dt = 1e-9' | mcx -f

 === MCX/MCX-CL ===

• a bug was fixed to make handling of low-absorption media more accurate (#164)
  
• multi-source simulation can now return the source ID (#217)
• fix a pmcx/pmcxcl bug that fails to use angleinvcdf and invcdf (#233)
• fix a bug that prevents photon-sharing being used in pattern3d source (#221), correctly output detector ID/launch pixel ID
  
• a negative radius of a detector captures but does not save a photon, allowing one to create "ring-shaped" detectors
• for mcxcl/mcxlabcl, multi-GPU simulation was not working since v2024.2, now it is fixed
• for mcxcl, the missing --maxjumdebug flag is now added
  

=== MMC ===

• photon-sharing (#86, #72, #102) is finally supported in GPU/OpenCL based mmc/mmclab, making the GPU backend support nearly all features that CPU/SSE backend  offers, except iMMC
• mmc/mmclab now automatically computes initial-element ID, facenb, elem/nodal volume mesh data if not provided
• add --dumpjson to export an mmc simulation to a single JSON encoded data file for easy sharing
• various bug fixes (#101, #103, #98) for the CUDA-backend in the "trinity" version of mmc/mmclab, matching the outputs for OpenCL backend (thanks to Andrea Farina)
  
• provide mmc-trinity downloads packages that simultaneously supports OpenCL, CUDA and CPU/SSE4/OpenMP
  
• add the missing -H/--maxdetphoton flag
  

I want to particularly highlight that mcx/mmc v2025 is the first release that we started deep integration with our open data portal at https://neurojson.io. With NeuroJSON.io, we intent to build a library of simulation input datasets, anatomical models, sample inputs/outputs that can be freely accessible via the -N/--net flag or REST APIs. In the next release, we will also implement automatic update checking/downloading via this online portal, as well as community submission of new simulations to further grow this library.

One more thing that I want to add here is that we have uploaded 1500x 3D mcx simulations of random domains, each with 5x photon levels from 1e5 to 1e9 photons, created in our 2022 JBO deep-learning based MC denoising paper (https://doi.org/10.1117/1.JBO.27.8.083019) to NeuroJSON.io, at

https://neurojson.org/db/cotilab/MCX_DL_Denoiser_2022

it took many GPU hours to compute this dataset. We hope other colleagues can reuse it if you see a fit - for example, serving as machine learning based training dataset. The entire dataset is about 1.5GB in size, and can be downloaded using a curl script or using our REST-API listed at the bottom of the page.

enjoy MCX v2025! happy modeling.

Qianqian