Skip to content

Conversation

@Qubitol
Copy link
Collaborator

@Qubitol Qubitol commented Dec 16, 2025

This fixes #1068.

With these changes, when selecting cppauto we check from the most vectorised (avx512vl) option going through all lower vectorisation levels progressively (avx2, sse4_2) until we find one that is supported.
We fallback to cppnone if nothing compatible is found.

This allows deducing cppauto in a very robust way, covering cases like Ivy Bridge microarchitecture (supports avx, but not avx2, and in this case the deduction would select cppsse4, and not cppavx2 like it was happening before - eventually crashing).

@roiser
Copy link
Member

roiser commented Dec 16, 2025

Perfect. Thanks a lot @Qubitol after the ci has run through please merge

@Qubitol Qubitol merged commit 28c39c8 into madgraph5:master Dec 16, 2025
167 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

cppauto and AVX/AVX2 support + question

2 participants