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 30A173858D32; Mon, 17 Oct 2022 11:09:57 +0000 (GMT)
DKIM-Filter: OpenDKIM Filter v2.11.0 sourceware.org 30A173858D32
DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=gcc.gnu.org;
	s=default; t=1666004997;
	bh=A8dKsCaB8w3OfT38cteCJ26HEZ54ez3CI+ghuX5q7YM=;
	h=From:To:Subject:Date:In-Reply-To:References:From;
	b=r2lWnsiPm9yk300hc7R+b2nc1HgNNIjc0IsX14IenbQK2+UOCkEN4uk/4byFtoVoi
	 NkQfRwaMxBRj/DnVgXl/zHDKPAwZlsrj0ExnyNTn/pxnfIDS78dcM9s8RllRKW6Rnt
	 a8XZWAxnlY1I5YpkaTvpicVcqaUKt7jx38u7X1WM=
From: "rguenth at gcc dot gnu.org" <gcc-bugzilla@gcc.gnu.org>
To: gcc-bugs@gcc.gnu.org
Subject: [Bug target/107257] [13 Regression] Wrong code at -O2 on
 x86_64-linux-gnu since r13-857-gf1652e3343b1ec47
Date: Mon, 17 Oct 2022 11:09:55 +0000
X-Bugzilla-Reason: CC
X-Bugzilla-Type: changed
X-Bugzilla-Watch-Reason: None
X-Bugzilla-Product: gcc
X-Bugzilla-Component: target
X-Bugzilla-Version: 13.0
X-Bugzilla-Keywords: 
X-Bugzilla-Severity: normal
X-Bugzilla-Who: rguenth at gcc dot gnu.org
X-Bugzilla-Status: NEW
X-Bugzilla-Resolution: 
X-Bugzilla-Priority: P1
X-Bugzilla-Assigned-To: unassigned at gcc dot gnu.org
X-Bugzilla-Target-Milestone: 13.0
X-Bugzilla-Flags: 
X-Bugzilla-Changed-Fields: 
Message-ID: <bug-107257-4-eqEiQqEBMe@http.gcc.gnu.org/bugzilla/>
In-Reply-To: <bug-107257-4@http.gcc.gnu.org/bugzilla/>
References: <bug-107257-4@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=3D107257

--- Comment #6 from Richard Biener <rguenth at gcc dot gnu.org> ---
With -fno-tree-slp-vectorize the failure goes away.  diff in optimized:

 __attribute__((noipa, noinline, noclone, no_icf))
 void q ()
 {
-  <bb 2> [local count: 1073742492]:
-  MEM <unsigned long> [(int *)&d] =3D 18380021091030725580;
+  int b_lsm.15;
+  int f;
+  _Bool _5;
+
+  <bb 2> [local count: 1073741824]:
+  MEM <vector(2) int> [(int *)&d] =3D { -15535156, -15535156 };
   d[2] =3D -15535156;
+
+  <bb 3> [local count: 8687551919]:
+  # f_30 =3D PHI <f_25(5), 0(2)>
+  if (f_30 =3D=3D 0)
+    goto <bb 5>; [1.43%]
+  else
+    goto <bb 4>; [98.57%]
+
+  <bb 4> [local count: 4698759900]:
+  if (f_30 !=3D 2)
+    goto <bb 5>; [79.66%]
+  else
+    goto <bb 6>; [20.34%]
+
+  <bb 5> [local count: 7731921808]:
+  # f_25 =3D PHI <2(4), 1(3)>
+  goto <bb 3>; [100.00%]
+
+  <bb 6> [local count: 1073742492]:
   a =3D 2;
-  b =3D -1;
+  _5 =3D f_30 =3D=3D 0;
+  b_lsm.15_33 =3D _5 ? 2 : -1;
+  b =3D b_lsm.15_33;
   return;

 }

both variants look OK so instead RTL expansion looks fishy(?)

;; b =3D b_lsm.15_33;

(insn 28 26 29 (parallel [
            (set (reg:CCC 17 flags)
                (ne:CCC (reg/v:SI 83 [ f ])
                    (const_int 0 [0])))
            (set (reg:SI 89)
                (neg:SI (reg/v:SI 83 [ f ])))
        ]) -1
     (nil))

(insn 29 28 30 (parallel [
            (set (reg:SI 90)
                (neg:SI (ltu:SI (reg:CCC 17 flags)
                        (const_int 0 [0]))))
            (clobber (reg:CC 17 flags))
        ]) -1
     (nil))

(insn 30 29 31 (parallel [
            (set (reg:SI 90)
                (ior:SI (reg:SI 90)
                    (const_int 2 [0x2])))
            (clobber (reg:CC 17 flags))
        ]) -1
     (nil))

(insn 31 30 32 (set (reg:SI 88 [ b_lsm.15 ])
        (reg:SI 90)) -1
     (nil))

(insn 32 31 0 (set (mem/c:SI (symbol_ref:DI ("b") [flags 0x2]  <var_decl
0x7ffff6526d80 b>) [1 b+0 S4 A32])
        (reg:SI 88 [ b_lsm.15 ])) -1
     (nil))

and we eventually optimize this to

q:
.LFB2:
        .cfi_startproc
        movq    .LC0(%rip), %rax
        movl    $-15535156, d+8(%rip)
        movl    $2, a(%rip)
        movq    %rax, d(%rip)
        movl    $2, b(%rip)
        ret=