Protein Viewer not working with GPU only slots

[informatorius]

Protein Viewer not working with GPU only slots
https://foldingforum.org/viewtopic.php?f=104&t=29138

[jcoffland]

After closing the fah-client ticket I realized this may be a client issue. Regardless we should have only one issue for it.

Can you try the new v7.4.16 beta. I made some changes there that may fix this issue.

[informatorius]

Also with v7.4.16 beta the viewer only shows a demo protein but not a real one. I'm only using GPU slots.

[jcoffland]

It's either a problem with the core not outputting the correct visualization data, the client not reading it correctly or the viewer having problems communicating with the client. Is this with core 0x21?

[bb30994]

Question: How often does Core_21 write visualization data?

One possibility is that it only writes the data when it writes a checkpoint. Checkpoints are often every 12.5% or 20% which (A) would take a long time to register as a viewable trajectory and (B) may be insufficient to create a smooth visualization.

I think there's an enhancement request for Core_21 to separate Write_data_for_Upload and Wirte_data_For_Restart (aka.checkpoint) so maybe the Write_Visualization needs to be considered, at the same time.

(If that's not the problem, it might be something else missing in V7.4.16.)

[informatorius]

ChristianVirtual from foldingforum did a workaround to get the Protein Viewer working
https://github.com/ChristianVirtual/FAH_WrapperGPUTrajectory
Maybe this is helpful.

[jcoffland]

The problem has been fixed in core 0x21 but the fixed version of the core has not yet been released.

[bb30994]

What about the GPU projects that use Core 0x18?

[jcoffland]

I thought 0x18 visualization worked. I worked with Yutong to get 0x17 and 0x18 cores to output JSON visualization format. Report back if it's not working.

[bb30994]

There's a related discussion in
https://github.com/FoldingAtHome/openmm-core/issues/156
and I reported some testing in Windows with Core 0x21 and FAHClient 7.4.16

The Viewer SHOULD be able to visualize proteins being folded by all active FAHCores (Currently 0xa4, 0xa7, and 0x21) in either Windows or Linux or OS-X but it appears there are several inconsistencies to the location and format of the viewer frames ... and whether this is handled by FAHClient v7.4.x is not known.

[informatorius]

Does not work with FahClient v7.5.0 beta

[hitboxr]

Has any progress been made on this? I'd love to help out but without source for the client I'm not sure where to start

[informatorius]

Currently FAHViewer 7.4.4 works with FAHclient 7.5.1 and FahCore_22 GPU work units.

[hitboxr]

I'm not certain how to tell what version of the viewer I have because it doesn't seem to be listed anywhere in the app, but I have the 7.5.1 client and the viewer does not display anything other than demo for me regardless of which core type. Do I need to add something to the viewer command in preferences?

[informatorius]

See solution at FoldingAtHome/fah-issues#1283

[hitboxr]

The build of fah-viewer master posted there works perfectly for me with both 0xA7 and 0x22 cores, cheers for the link. This issue should probably be closed, should it not?

[informatorius]

Only after a new FahClient update was released and the solution can be confirmed there.