[andy_]

Is native Windows worth supporting as a platform for CFD for:

1) solvers on remote clusters
2) solvers and pre/post processing on desktop/workstations
3) pre/post processing on desktops/workstations with remote solvers

I appreciate it is possible to get engineering software to run on native Windows after a fashion but is the effort worthwhile? Issues like not fully supporting complex numbers in the runtime just baffles me when it comes to simulation software.

PS by supporting native Windows I am excluding running linux programs via cygwin, WSL or equivalents. I am including msys though which uses linux tools to build native programs.

Does anyone have a reasonable idea of the proportion of Windows vs Linux licenses sold by commercial CFD providers? Do some commercial CFD providers fail to support Windows in the way a fair few open source simulation programs do?

[LuckyTran]

It is worthwhile and I daresay mandatory for engineering. Commercial CFD providers do not discriminate against user system when they sell licenses because they know up front that both platforms MUST be supported with the SAME license.

All the big bois fully support windows "natively," via the installation of appropriate Microsoft Visual C++ #### libraries. High performance superclusters are dominated by linux installations, but desktops, workshops, and day-to-day CFD is still dominated by Windows in the workplace simply because Windows dominates in general. E.g. microsoft teams

So to answer the question, far-and-large, commercial CFD companies do not neglect Windows and do not enforce linux on her customers. Also, it is 2024. We have cross-platform video games. It is archaic and silly to think that something as simple as a license would still be platform specific.

[andy_]

Originally Posted by LuckyTran

It is worthwhile and I daresay mandatory for engineering. Commercial CFD providers do not discriminate against user system when they sell licenses because they know up front that both platforms MUST be supported with the SAME license.

All the big bois fully support windows "natively," via the installation of appropriate Microsoft Visual C++ #### libraries. High performance superclusters are dominated by linux installations, but desktops, workshops, and day-to-day CFD is still dominated by Windows in the workplace simply because Windows dominates in general. E.g. microsoft teams

So to answer the question, far-and-large, commercial CFD companies do not neglect Windows and do not enforce linux on her customers. Also, it is 2024. We have cross-platform video games. It is archaic and silly to think that something as simple as a license would still be platform specific.

I may have muddied my question somewhat by trying to avoid prompting. There is no question of running linux on windows. The question is how to support running CFD on windows going forward given MS seem to have stopped developing HPC support 6 years ago, Intel and AMD HPC support is developing without ms tools, there are issues with ms runtime for HPC which one can get round to some extent with effort but is the effort worthwhile given where things seem to be going, etc...

A frontend on windows can be addressed in number of ways and isn't particularly problematic. What is more problematic is how to support user supplied functionality in the backend/solver when the solver is running on windows. There are options and windows is a secondary platform so although the tail isn't going to wag the dog I would like the tail to wag if it doesn't involve too much effort.

There is little requirement to support windows HPC clusters though it seems to be relatively straightforward using non-MS software and tools. There is a requirement to support running smaller CFD jobs on windows workstations/desktops hence the interest in getting a feel for how many currently use windows rather than unix/linux/apple and of these how many have a significant dependency on MS tools and runtime.