You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Per Arm's blog post, there is now a feature -moutline-atomics in GCC10 (and, I think, Clang12) that implements out-of-line atomics - eg, certain atomic actions in armv8 (not armv8.1) compile to runtime calls that check whether the processor is new enough to use LSE atomics (or needs LL/SC). Ideally, c4t should be able to pass -moutline-atomics or -mno-outline-atomics when on armv8 and a sufficiently recent compiler version.
This might be a turning point to consider a more principled way of teaching c4t about perturbable compiler optimisations. I'll sketch a new issue for this.
The text was updated successfully, but these errors were encountered:
Per Arm's blog post, there is now a feature
-moutline-atomics
in GCC10 (and, I think, Clang12) that implements out-of-line atomics - eg, certain atomic actions inarmv8
(notarmv8.1
) compile to runtime calls that check whether the processor is new enough to use LSE atomics (or needs LL/SC). Ideally, c4t should be able to pass-moutline-atomics
or-mno-outline-atomics
when onarmv8
and a sufficiently recent compiler version.This might be a turning point to consider a more principled way of teaching c4t about perturbable compiler optimisations. I'll sketch a new issue for this.
The text was updated successfully, but these errors were encountered: