Hi,

I am not sure if this is the right category to ask, apologies if not. In the Stan Functions Reference it is written that vectorizations for exp and log are defined elementwise and thus provide only minimal efficiency speedups. I understand from this that such vectorizations are only for the convenience of not having to write a loop but they do not involve SIMD instructions, which is how I usually understand the term vectorization. Is my understanding correct? For example, in Matlab writing ây = exp(x)â is dramatically faster than looping over each element in âxâ. Is there a reason for not vectorizing these functions in Stan and are there any plans to do it?

Thanks,

Louis