[llvm-dev] Fwd: [PATCH] D79314: PR45733 : global weak symbol are excluded from the optimization .

[Umesh Kalappa via llvm-dev]

Hi Eli  and All,

Thank you for the suggestions, the subjected review is blocking our
internal release.

We would like to know community suggestions over the problem and fix , that
we suggested.

Appreciate your time and thank you
~Umesh

---------- Forwarded message ---------
From: Umesh Kalappa via Phabricator <reviews at reviews.llvm.org>
Date: Wed, May 13, 2020 at 10:02 AM
Subject: [PATCH] D79314: PR45733 : global weak symbol are excluded from the
optimization .
To: <umesh.kalappa0 at gmail.com>, <renatoriolino at gmail.com>, <
spatel at rotateright.com>, <tpr.ll at botech.co.uk>, <efriedma at quicinc.com>
Cc: <llvm-commits at lists.llvm.org>, <88888yl at gmail.com>, <
Samuel.j.knapp at btinternet.com>, <florian_hahn at apple.com>


umesh.kalappa0 added a comment.

In D79314#2032099 <https://reviews.llvm.org/D79314#2032099>, @efriedma
wrote:

> Okay, so assuming that optimization is invalid, there's a bug in clang
*somewhere*.  You haven't explained why this is the right fix.
>
> LLVM IR is a distinct language with its own specification.  The
definition of weak_odr is very explicitly written out.  If you're going to
change LLVM IR optimizations, you need to justify the change in terms of
that specification, not C code.




In D79314#2032099 <https://reviews.llvm.org/D79314#2032099>, @efriedma
wrote:

> Okay, so assuming that optimization is invalid, there's a bug in clang
*somewhere*.  You haven't explained why this is the right fix.


We are bailing out the optimization, if the global has the weak_odr
linkage  ,which is explicit in the condition check .

> LLVM IR is a distinct language with its own specification.  The
definition of weak_odr is very explicitly written out.  If you're going to
change LLVM IR optimizations, you need to justify the change in terms of
that specification, not C code.

Yes, don't propagate the constant for the weak_odr / weak globals (in SCCP
).


Repository:
  rL LLVM

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D79314/new/

https://reviews.llvm.org/D79314




<https://www.avast.com/sig-email?utm_medium=email&utm_source=link&utm_campaign=sig-email&utm_content=webmail>
Virus-free.
www.avast.com
<https://www.avast.com/sig-email?utm_medium=email&utm_source=link&utm_campaign=sig-email&utm_content=webmail>
<#DAB4FAD8-2DD7-40BB-A1B8-4E2AA1F9FDF2>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-dev/attachments/20200515/4be7fcfd/attachment.html>