Remote 3D-accelerated OpenGL apps from Windows server

Question

I want to start using "optimized connection" feature of xlwings (v 0.11.4), but I would like to be able to kill Python process when I close the workbook (even if Excel instance is still running). I couldn't find any information on this in topic in xlwings documentation. Anyone had a similar problem and found a solution?

Solution 1

Thinanywhere has an RDP plugin that they claim can use the GPU to accelerate OpenGL on the server.

In addition, Citrix recently released an add-on to XenApp 6.5 that does allow GPU sharing for OpenGL. They have supported Direct3D GPU sharing for a while (with the proper regkey set) on XenApp as well.

Solution 2

Citrix XenDesktop supports GPU-accelerated OpenGL via the Citrix Receiver (universal thin client application) for Linux, OS X, Android and Windows and would support your use-case on the operating systems you mentioned.

The Microsoft equivalent, RemoteFX (on RDP), while it does support GPUs and OpenGL 1.1, is more limited on the operating systems supported. Unsure if you can use any standard Linux RDP client with it.

Comments

[Felix Zumstein]

make sure you clean up your global excel object references: stackoverflow.com/a/37571308/918626

[matetam]

My problem is with Python process (not Excel zombie) running until I close Excel instance. It looks like this, I open Excel (Python process does not yet exist), then I open a workbook, which within its AutoOpen function calls Python to start the environment). I see a new Python process starting and being up entire time this Excel instance is running. This Python process will only end when I close Excel instance. I would like however to have the same result (end Python process) when I close the workbook (with Excel instance still running).

[Felix Zumstein]

You can try to call the function KillPy which is part of the addin.

[Rex]

You could do a couple things with XenDesktop. You could assign a physical desktop with the GPU to a user and they would have direct access to the GPU that way and, with the HDX license, would be able to do full DirectX/OpenGL. You could also add a hypervisor with GPU passthrough to assign the GPU to a virtual desktop. Again, have access to the GPU. Both of these solutions are limited to having only one user having access to the GPU at a time (unless you had multiple GPU's in the remote system).

[Lorin Hochstein]

I tried RemoteFX on RDP, but I could not confirm that it was actually accelerating the OpenGL 1.1. Subjectively, I couldn't tell a difference in performance, and querying OpenGL to get rendering info gave the same response as when not using RemoteFX.

[Pascal]

Welcome to SO, please don't just post links as an anwser.

[Admin]

It also relies on hardware acceleration on the client, which is explicitly what is not wanted here.