Hello

would it be beneficial to offer also AVX2 app? Or are there no major advantages in performance compared to SSE3/AVX?

I would guess not much of any difference. TN-Grid compiled 3 different apps from the same source code with just the change in SIMD instruction flags.

They offer SSE2, AVX2 and FMA applications. The AVX2 and FMA applications are about equal but the SSE2 application loses out by about 10% in performance compared to the higher level SIMD instructions sets.

Without specific changes in the code optimized for each SIMD class, I doubt there would be much difference between the AVX app and the proposed AVX2 app.

---

A proud member of the OFA (Old Farts Association)

The current app uses hand-written optimized code for AVX & FMA. Unfortunately, there are no major advantages AVX2 can provide for this specific algorithm in comparison to AVX. Simply enabling AVX2 in the compiler won't give any speedup.
AVX512 could have some features to potentially squeeze more performance on processors that support it, but the code has to be written specifically for this instruction set.