I've recently started working on a project for a client and through the discussions it was mentioned that they received two IBM Minsky HPC's from IBM as part of a deal for IBM infrastructure, but they haven't used it at all (this was in 2019)
The Minsky's each have 2x Power8 CPUs and 4 Nvidia P100 + NVLink's installed. At a closer look the Nvidia Tesla P100 (which this machine has four of) each ships with3584 CUDA cores. Here is a link to the spec-sheet: https://www.nvidia.com/content/dam/en-zz/Solutions/Data-Center/tesla-p100/pdf/nvidia-tesla-p100-datasheet.pdf
Their server guys need to get back to me to confirm the OS it support. Since it is a HPC it is mostlikely going to be some Linux flavour (when looking at the OpenPower Foundation's website).
The question I have is: Is there anyway I could get Manifold v9 to use these GPU's? and if so how?
I would welcome any suggestions.
I may be stepping out of bounds here, but I believe that Manifold is currently wedded to Microsoft and .NET. It is unclear to me that if these were used as a virtual server that getting the CUDA drivers working in the VM for running Windows, that Manifold, 1. could take advantage of the underlying hardware and 2. that it would/might break the license agreement for Manifold. Perhaps Dimitri could cover the current state of VM usage for running Manifold.
Just last week I tried to run the Manifold Viewer on a wine installed linux laptop. There were some UI issues, but to my surprise it ran. My hope was to provide a remote client to running an example server on my home desktop.
From the GPGPU topic in the User Manual: "Please run modern, 64-bit Manifold in 64-bit Windows to utilize GPGPU within Manifold Release 9 and Viewer."
See also the System Requirements page on the website.