[PATCH] D29092: PR31729: [GVNHoist] Don't hoist unsafe scalars at -Oz

Mehdi AMINI via Phabricator via llvm-commits llvm-commits at lists.llvm.org
Thu Feb 23 16:12:46 PST 2017


mehdi_amini added a comment.

In https://reviews.llvm.org/D29092#685217, @hiraditya wrote:

> In https://reviews.llvm.org/D29092#685017, @majnemer wrote:
>
> > In https://reviews.llvm.org/D29092#682961, @hans wrote:
> >
> > > David, do you have any further concerns? Otherwise I'd like to see this committed so we can resolve the PR for 4.0.
> >
> >
> > I don't understand why we cannot assert isSafeToSpeculativelyExecute if we are not in minsize. If we are given a call instruction, why do we think it is safe to hoist it?
>
>
> A call without any side effects, for example, can be hoisted.


No.
Please read the review I linked that introduce a new attribute to this end, because not having *memory* side effects is not enough.


https://reviews.llvm.org/D29092





More information about the llvm-commits mailing list