This is the mail archive of the gcc@gcc.gnu.org mailing list for the GCC project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: PR 11319 analyzed (Re: Patch fixing 3.3 bug PR 9745 and PR 10021)


On Tuesday, July 15, 2003, at 03:10 PM, Jim Wilson wrote:
Why only the checks in true_dependence and not also output and anti
dependence? We probably should have a function overlapping_memrefs_p
similar to nonoverlapping_memrefs_p except checking for overlaps instead
of nonoverlaps, and call it everyplace where we call
nonoverlapping_memregs_p.

That's one approach. The problem as I see it is that there are 3 possible
relationships between two memrefs - overlap, don't overlap, and can't tell -
and the interface is designed with all boolean functions. IMO it would
be better to make most of the functions tri-valued.



Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]