Graphics output not responding respectively throwing error about fluidx3d

[ProjectPhysX]

With the WINDOWS_GAPHICS option you'll have to compile in Windows directly, for example with the MSVC compiler in Visual Studio Community. Compiling that in WSL will not work, as that's a linux environment.

The CONSOLE_GRAPHICS option works in both Windows and Linux/WSL, provided your GPU appears as OpenCL device in WSL. Once the ASCII graphics window and pixel counter appears, the simulation is paused and by default grid flags are visualized. In the 3D Taylor-Green setup there is no solid geometry, so no flags are visualized and the window initially looks black. Press key 4 to enable Q-criterion isosurface visualization, or the other visualization modes with keys 2 or 3, then you should see something. Press P to start/pause the simulation. You can rotate the camera with W/A/S/D.

To disable interactive graphics and instead write rendered images to the hard drive, use GRTAPHICS, and in src/setup.cpp in the main_setup() function you'll need to make a loop to alternatingly run a specified number of LBM time steps and write images. For example:

key_4 = true; // enable Q-criterion isosurface rendering by default
lbm.graphics.set_camera_centered(-60.0f, 20.0f, 0.0f, 2.5f); // set camera
lbm.run(0u); // initialize simulation
while(lbm.get_t()<18000u) {
	lbm.graphics.write_frame_png(); // render frame and write it to hard drive in bin/export/
	lbm.run(30u); // run 30 LBM time steps
}

[gittigittibangbang]

Thank you, I got it working with the GRAPHICS option, and I'll try getting the WINDOWS_GRAPHICS option to work, too. Pictures of the cylinder model look great, but the ones from the F1 car, Concorde or 747 model tend to fill up with isosurfaces completely, even when using varying values of the Q-criterion. Do you have a recommendation for that? By the way, is it possible to create cutting planes and legends?

And again, thanks for creating and publishing your solver, it could really be the kick in the ass for some bigger solver developers to modernize their codes! It's really great work :)

[ProjectPhysX]

What you see is the simulation blowing up at large Reynolds numbers. For these you need to enable SUBGRID in src/defines.hpp!

[gittigittibangbang]

Ah, that makes sense, thank you.

Another quick question: I compiled the benchmark files on my home PC to run them on a workstation, which works fine. However, using the same workflow to compile the exe (using MinGW bash environment to use ./make.sh) on the workstation, the solver does not run stating that no OpenCL devices are available. Since the files created on my personal PC worked without a problem on the workstation using either the GPU or CPU, this seems odd. Do you happen to know what the issue might be? The GPU drivers are installed, of course.