Cross compilers, project assembly for x86-64 and ARM, and actually development for multiple platforms?

Question

GCC

The idea of ​​the project arose, and everything seems to be going well, the research ended more or less successfully and it seems like it’s worth starting to implement it, but it wasn’t there.
I have no experience in developing cross-platform software at all (some simple qt programs don't count), so I don't even know where to start.
The program itself is the core for other programs. In fact, this is a separate daemon process that performs some kind of work. No gui. Several modules must perform quite a lot of calculations, so there's nothing to be done, you have to write separately platform-specific code for the same arm (opencl is not supported on androids). We also need support for automatic vectorization of calculations. Since the project is created for the sake of interest, somehow I don’t really want to buy a license for intel parallel studio. And of the free ones, it's either gcc or clang (which has everything bad with windows and no neon support).
cmake - convenient, but as far as I remember, there are difficulties with setting up a custom compiler for windows (although visual studio may work, since version 2011+ it seems like the vs compiler supports automatic vectorization, but I don’t know if this is the case in the express edition), nevertheless would like to use gcc. And honestly, I still have little idea how to set it all up, how to work with incremental builds later, how to build additional libraries, divide them into modules, etc.
Ideally, it would be nice to find a small codebase project that supports the arm and x86-64 architectures (the operating systems currently required are windows, windows phone, linux, android). Again, if anyone has experience, tell me what kind of workdflow should be on such projects, which moments are the most difficult, and which ones can be solved.

Answer 1

Express is an IDE. It has nothing to do with assembly utilities.
Ideally - you yourself named - Qt. In order not to sweat too much, I can suggest an interesting trick for parallel computing - read everything in the shader, stuffing the information into textures. And then extract the result from the same textures. This is the most cross-platform way.
Well, or you have to write a little platform-dependent code. Al so