On 10/7/20 1:56 PM, Martin Liška wrote:
So what do you suggest to fix it :P?
Note that we have a nice reduced test-case here:
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=97295#c6
It's really a mixture of -O1 -fprofile-use and -O0.
Martin
On 10/7/20 1:23 PM, Jan Hubicka wrote:
During WPA we merge 2 functions where one is built with -O3
-fprofile-use while the other one uses -O0. We end up with:
prevailing node:
BitwiseCast (const double aFrom)
{
long unsigned int temp;
long unsigned int D.4528;
:
BitwiseCast (aFrom
> During WPA we merge 2 functions where one is built with -O3
> -fprofile-use while the other one uses -O0. We end up with:
>
> prevailing node:
>
> BitwiseCast (const double aFrom)
> {
> long unsigned int temp;
> long unsigned int D.4528;
>
>:
> BitwiseCast (aFrom_2(D), &temp);
> _4
On 10/7/20 1:18 PM, Martin Liška wrote:
* ipa-utils.c (ipa_merge_profiles):
I forgot to fill up the ChangeLog entry. Fixed.
Martin
>From a5d5d23175d336883972ebef66380cbc4f1b6d54 Mon Sep 17 00:00:00 2001
From: Martin Liska
Date: Wed, 7 Oct 2020 12:30:27 +0200
Subject: [PATCH] IPA: me
On 10/7/20 1:18 PM, Martin Liška wrote:
* ipa-utils.c (ipa_merge_profiles):
I forgot to fill up the ChangeLog entry. Fixed.
Martin
During WPA we merge 2 functions where one is built with -O3
-fprofile-use while the other one uses -O0. We end up with:
prevailing node:
BitwiseCast (const double aFrom)
{
long unsigned int temp;
long unsigned int D.4528;
:
BitwiseCast (aFrom_2(D), &temp);
_4 = temp;
temp ={v} {CLO
