GROMACS 2025 TPR support

[al42and]

Hi!

Not sure what's the best way to give a heads-up (or even whether it's valuable), but in February, GROMACS 2025 will be released (2025-beta is out), which will bump the tpx version to 136.

Related commits:

• 135: https://gitlab.com/gromacs/gromacs/-/commit/b68acdf4#6dea3eb4ce84e382008dfdd11affefe1ac59aa08
• 136: https://gitlab.com/gromacs/gromacs/-/commit/93657d82#6dea3eb4ce84e382008dfdd11affefe1ac59aa08

So, the TPR parser will need to be updated to support the new files. I can make test inputs if that helps, but I don't think I'll be able to submit a full PR in the near future.

[tylerjereddy]

I've been doing this the last few times, it isn't too hard. Is there a compelling/urgent need to add support before the stable release? It may be best for us to wait just in case there are additional mutations to the format before final release.

[al42and]

Is there a compelling/urgent need to add support before the stable release?

No, just wanted to give a heads-up about impeding change and the rough timeframe for that.

[tylerjereddy]

On a related topic, we've also been working on adding support for coordinate and velocity parsing from your .tpr files (see: #4873). I haven't found a better method of developing that than studying the control flow of your C++ .tpr parsing code, but it seems to be coming along slowly as we expand the tpx format history, etc.

[al42and]

The TPR was intended as internal-only format and was, so yeah, we don't have any documentation (either in prose or as some kind of programmatic schema) besides the source itself :( You probably know about gmx dump -s file.tpr to convert the file to plain text, but this is not reversible, so only helpful as debugging/investigation aid. And I don't think there's a shortcut to avoid partially parsing the file to find the beginning of coordinate/velocity data.

There is work in progress to switch trajectories to HDF5/H5md instead of custom formats, but no specific plans for TPR replacement.