Missing example for compiling to multiple implementations

[kunaltyagi]

if desired, it can compile the same source code to multiple different implementations

The above line from README suggests this but I couldn't find examples/tutorials/other references regarding the process to do so. Would it be possible to add more information regarding this?

I would also like to generate multiple versions (scalar, avx512, avx2, sse, cuda) to benchmark them (since sometimes avx drops the clock speed and can actually hurt performance in multi-threaded applications). Would it be possible to do this in the same binary?

[mp3guy]

You could split the core code you want to benchmark into a library and then change the compiler options for each version you want to build in CMake?

[kunaltyagi]

then change the compiler options for each version you want to build in CMake

That would not be ideal (for my case, but a sound advice). Ideally, I would want a single binary to preserve existing pipeline (such as comparing results).

Is it possible to pass a "tag" to tell enoki which version to prefer using overloading? Something like how C++17's execution policy would be ideal. (Then I can use testing frameworks to run everything with type parametrization)

[mp3guy]

You can probably do it in a single binary. You can conditionally name each of the method symbols based on the compilation options and statically link against all of them.