From mboxrd@z Thu Jan  1 00:00:00 1970
Return-Path: <gcc-bugzilla@gcc.gnu.org>
Received: by sourceware.org (Postfix, from userid 48)
	id 3580E3858C50; Fri, 28 Oct 2022 09:30:48 +0000 (GMT)
DKIM-Filter: OpenDKIM Filter v2.11.0 sourceware.org 3580E3858C50
DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=gcc.gnu.org;
	s=default; t=1666949448;
	bh=wqkMK2HYbLiR9M7jZuXDTQu4SMyhC1zziqC6KAPYZbI=;
	h=From:To:Subject:Date:In-Reply-To:References:From;
	b=CH1s/+RSVMJQktTL9EZqruLOXFrJa7MYaHgn5NIAcJYhxgHk8WlPXXu+ZlWn9+T+6
	 3hw1+WbMUiPP367pzuFo5Az6unMkckQlMb54fm25A908TALGe1JHM1H79i4zGibxXX
	 2/SIyzo0zDXHp8br7DyWIZrxvtz/UIy5afy4/aAI=
From: "rguenth at gcc dot gnu.org" <gcc-bugzilla@gcc.gnu.org>
To: gcc-bugs@gcc.gnu.org
Subject: [Bug middle-end/107389] Always propagate __builtin_assume_aligned
Date: Fri, 28 Oct 2022 09:30:46 +0000
X-Bugzilla-Reason: CC
X-Bugzilla-Type: changed
X-Bugzilla-Watch-Reason: None
X-Bugzilla-Product: gcc
X-Bugzilla-Component: middle-end
X-Bugzilla-Version: 12.2.1
X-Bugzilla-Keywords: missed-optimization
X-Bugzilla-Severity: enhancement
X-Bugzilla-Who: rguenth at gcc dot gnu.org
X-Bugzilla-Status: ASSIGNED
X-Bugzilla-Resolution: 
X-Bugzilla-Priority: P3
X-Bugzilla-Assigned-To: rguenth at gcc dot gnu.org
X-Bugzilla-Target-Milestone: ---
X-Bugzilla-Flags: 
X-Bugzilla-Changed-Fields: attachments.created
Message-ID: <bug-107389-4-23NMNbsSZK@http.gcc.gnu.org/bugzilla/>
In-Reply-To: <bug-107389-4@http.gcc.gnu.org/bugzilla/>
References: <bug-107389-4@http.gcc.gnu.org/bugzilla/>
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable
X-Bugzilla-URL: http://gcc.gnu.org/bugzilla/
Auto-Submitted: auto-generated
MIME-Version: 1.0
List-Id: <gcc-bugs.sourceware.org>

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=3D107389

--- Comment #5 from Richard Biener <rguenth at gcc dot gnu.org> ---
Created attachment 53784
  --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=3D53784&action=3Dedit
prototype

This implements an -O0 fold-builtins pass.  I've disabled some but not all
"optimizations" and instead of just throwing away __builtin_assume_aligned
I'm processing it at -O0 (the machinery from CCP relies on a lattice, with
optimization we should at least merge with the alignment info on the LHS).

On s390 I then see

foo:
.LFB0:
        .cfi_startproc
        stmg    %r11,%r15,88(%r15)
        .cfi_offset 11, -72
        .cfi_offset 12, -64
        .cfi_offset 13, -56
        .cfi_offset 14, -48
        .cfi_offset 15, -40
        aghi    %r15,-176
        .cfi_def_cfa_offset 336
        lgr     %r11,%r15
        .cfi_def_cfa_register 11
        stg     %r2,168(%r11)
        stg     %r3,160(%r11)
        lg      %r1,160(%r11)
        lpq     %r2,0(%r1)
        lg      %r1,168(%r11)
        stmg    %r2,%r3,0(%r1)
        lg      %r2,168(%r11)
        lmg     %r11,%r15,264(%r11)
        .cfi_restore 15
        .cfi_restore 14
        .cfi_restore 13
        .cfi_restore 12
        .cfi_restore 11
        .cfi_def_cfa 15, 160
        br      %r14
        .cfi_endproc

specifically I did not disable __atomic_add_fetch_* optimizations to
.ATOMIC_ADD_FETCH_CMP_0 and friends and also kept optimizing
stack_save/restore.=