Re: Vectorization, SIMD

[Marcus Harnisch]

Hi Gerd

On 23/04/2025 15.50, Gerd Möllmann wrote:
> The "problem" with auto-vectorization is that it fails to optimize loops
> in many cases. (The .org file contains a compiler option with which one
> can see why it fails. By default, clang says nothing.) For example, it
> looks like a break or return in a loop makes it fail. Also a switch,
> and I don't know what else.

Almost 15 years ago I wrote an article about ARM NEON vectorization, 
which may explain some of the issues you are seeing. I would expect 
other SIMD architectures to be affected in similar ways.

> IOW, it's doesn't feel like something that can be relied on, if one
> really wants SIMD, especially if some code is compiled by different
> compilers or different versions of the same compiler.

The main takeaway was that you can't just flip a compiler switch and 
expect things to improve significantly. For that, code has to be written 
with vectorization in mind.

Regards,
Marcus