[llvm-dev] 回复: ScalarEvolution invariants around wrapping flags

[Sanjoy Das via llvm-dev]

On Mon, Sep 30, 2019 at 8:06 AM 陈 正 via llvm-dev
<llvm-dev at lists.llvm.org> wrote:
> Thank for reverting it.
>
> For the original issue in https://reviews.llvm.org/D64868, if we want to make sure nsw can be correctly added to increment SCEV, we need to also check isLoopEntryGuardedByCond? Is this a right way to go?

I don't see how that would help, but I can be more specific if you are
more specific. :)  How exactly are you suggesting we use
isLoopEntryGuardedByCond?

-- Sanjoy

>
> BRS//
> Chen Zheng
> ________________________________
> 发件人: llvm-dev <llvm-dev-bounces at lists.llvm.org> 代表 Tim Northover via llvm-dev <llvm-dev at lists.llvm.org>
> 发送时间: 2019年9月30日 15:45
> 收件人: Sanjoy Das <sanjoy at playingwithpointers.com>
> 抄送: LLVM Developers Mailing List <llvm-dev at lists.llvm.org>
> 主题: Re: [llvm-dev] ScalarEvolution invariants around wrapping flags
>
> On Sun, 29 Sep 2019 at 19:23, Sanjoy Das <sanjoy at playingwithpointers.com> wrote:
> > Your reasoning sounds correct to me.  Let's revert for now?
>
> Sounds good, I've reverted it with r373184 (and r373185 to polly,
> which fixed up some tests after the original commit).
>
> > I don't think there is an easy fix, we'll have to do a global "must be
> > executed" analysis to reapply the patch soundly.  And that's difficult
> > since any external functional call can call "exit(0)".
>
> Oh yes, that didn't even occur to me.
>
> Cheers.
>
> Tim.
> _______________________________________________
> LLVM Developers mailing list
> llvm-dev at lists.llvm.org
> https://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-dev
> _______________________________________________
> LLVM Developers mailing list
> llvm-dev at lists.llvm.org
> https://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-dev