https://gcc.gnu.org/bugzilla/show_bug.cgi?id=90838
Richard Biener <rguenth at gcc dot gnu.org> changed:
What |Removed |Added
----------------------------------------------------------------------------
CC| |rguenth at gcc dot gnu.org
--- Comment #3 from Richard Biener <rguenth at gcc dot gnu.org> ---
I think if we start to pattern-match this kind of stuff we have to be careful
to do it once only and for all table-based stuff so we can classify
each used table and not need to re-scan it for each and every use.
Alternatively
we need to de-couple the table classification and stick the classification
result somewhere (the varpool node for example) so we can "easily" re-use it.
The first approach means it would need to be a IPA pass.
What about code that has the table in an automatic variable
[const] int table[64] = { ... };
? That might end up inline expanded from gimplification, copied
from a .LC0 constant or whatnot.
