https://gcc.gnu.org/bugzilla/show_bug.cgi?id=121662

--- Comment #5 from Richard Biener <rguenth at gcc dot gnu.org> ---
(In reply to Richard Biener from comment #3)
> I remember we run into this issue in SPEC somewhere where an outer loop
> repeatedly triggers fault suppression on a masked store in an inner
> vectorized
> loop.  The code in the vectorizer was contributed by Intel IIRC.
> 
> Note fault suppression also happens for paging, not only for actual access
> violations.

The classical fault is COW of the zero page after allocation btw.

And all this is also a reason for aggressively aligning masked stores so
that partially masked stores do not trigger the same issue (we don't do
such aggressive aligning or prioritizing of aligning of masked stores).

Reply via email to