2 molecules simultaneously computation

[Bandersnatch]

potentially offtopic
Most of the time client (cpu) computes 2 molecules simultaneously (or even 3), and just sometimes only 1 molecule. Is it software optimisation only/most effective cpu usage, or 2 molecules are related to each other and must be computed as conjuncted pair?

[JimboPalmer]

As always, every single time, our answers would be better if we could see the 'system' portion of your log.

The current default setup for FAH is to have one CPU client running on as many threads as is wise, and one GPU client per whitelisted video card.

You CAN configure multiple CPU clients, but in general you will do less scientific work.

Each client works on one molecule. So the number of molecules worked on is the number of clients. (FAH frequently downloads the next Work Unit when 99% done with the current one, but does not start that one until the current one finishes)

[7im]

The picture looks like 2 or 3, but it is always only one protein.

[bruce]

The picture looks like 2 or 3, but it is always only one protein.

This is a known bug in the viewer.

Mathematically, the analysis takes place in a box, or rather an infinite number of boxes stacked up in 3 dimensions. Mathematically, the protein is in each of the boxes, but the viewer only shows one. If the protein happens to cross one or more boundaries, it also crosses the opposite boundary and any atoms that disappear (in the viewer) reappear entering the box from the opposite side.

The viewer needs to be smart enough to glue those pieces back together before it displays it.

Have you tried Folding Atomata - a 3rd-party protein viewer for F@h? I recommend it.

[Jesse_V]

Thanks Bruce, but my viewer doesn't piece those pieces back together because I don't have access to enough information to solve the problem. I do think my viewer looks better than the official F@h one, but then I'm biased.

[bruce]

I knew you were working on it and that you needed more information, but I had hoped that you'd figured out a way to guess how the pieces needed to fit together. Unfortunately it looks like it's not a high enough priority for PG.

The proteins I fold seem to stay inside the box when I look at them.

[davidcoton]

Is this related to wrapping Christmas presents? When I fold paper round a box, the box doesn't always stay inside....

D.

[bruce]

No, it's more like trying to wrap a new puppy/rabbit which keeps pushing its way out of the box long before you have time to get the wrapping paper.

Mathematically speaking, there are prisms which make the part that has escaped the box appear like it's still inside the box.

[Bandersnatch]

As always, every single time, our answers would be better if we could see the 'system' portion of your log.

The current default setup for FAH is to have one CPU client running on as many threads as is wise, and one GPU client per whitelisted video card.

You CAN configure multiple CPU clients, but in general you will do less scientific work.

Each client works on one molecule. So the number of molecules worked on is the number of clients. (FAH frequently downloads the next Work Unit when 99% done with the current one, but does not start that one until the current one finishes)

For example, project 6099 screenshot

[Bandersnatch]

The picture looks like 2 or 3, but it is always only one protein.

This is a known bug in the viewer.

Mathematically, the analysis takes place in a box, or rather an infinite number of boxes stacked up in 3 dimensions. Mathematically, the protein is in each of the boxes, but the viewer only shows one. If the protein happens to cross one or more boundaries, it also crosses the opposite boundary and any atoms that disappear (in the viewer) reappear entering the box from the opposite side.

The viewer needs to be smart enough to glue those pieces back together before it displays it.

Have you tried Folding Atomata - a 3rd-party protein viewer for F@h? I recommend it.

Thank you. I read about it, but there is no win porting yet. Jesse_V wrote might be it will be ported to javascript, so I'll be able to try