(if-conversion could directly generate masked load/stores of course and not use a scratch-pad at all in that case).
[Abe wrote:]
IMO that`s a great idea, but I don`t know how to do it. Hints would be welcome. In particular, how does one
>> "generate masked load/stores" at the GIMPLE level? [Richard Biener wrote:]
It already does that, see predicate_mem_writes.
> You should definitely preserve that path Thanks. Yes, we have not intentionally disabled that.
On what hardware did you test?
AMD64 arch., Intel implementation. Nothing fancy AFAIK in the flags to make it super-specific, e.g. "-march=nocona" or "-march=native". Occasionally using AVX-2 flags as specified by some test cases. Regards, Abe
