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 D5ED03857C7A; Wed, 16 Jun 2021 05:53:01 +0000 (GMT)
DKIM-Filter: OpenDKIM Filter v2.11.0 sourceware.org D5ED03857C7A
From: "luoxhu at gcc dot gnu.org" <gcc-bugzilla@gcc.gnu.org>
To: gcc-bugs@gcc.gnu.org
Subject: [Bug target/100866] PPC: Inefficient code for vec_revb(vector
 unsigned short) < P9
Date: Wed, 16 Jun 2021 05:53:01 +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: 8.3.1
X-Bugzilla-Keywords: missed-optimization
X-Bugzilla-Severity: enhancement
X-Bugzilla-Who: luoxhu at gcc dot gnu.org
X-Bugzilla-Status: NEW
X-Bugzilla-Resolution: 
X-Bugzilla-Priority: P3
X-Bugzilla-Assigned-To: unassigned at gcc dot gnu.org
X-Bugzilla-Target-Milestone: ---
X-Bugzilla-Flags: 
X-Bugzilla-Changed-Fields: 
Message-ID: <bug-100866-4-FIS6G0PuB3@http.gcc.gnu.org/bugzilla/>
In-Reply-To: <bug-100866-4@http.gcc.gnu.org/bugzilla/>
References: <bug-100866-4@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
X-BeenThere: gcc-bugs@gcc.gnu.org
X-Mailman-Version: 2.1.29
Precedence: list
List-Id: Gcc-bugs mailing list <gcc-bugs.gcc.gnu.org>
List-Unsubscribe: <https://gcc.gnu.org/mailman/options/gcc-bugs>,
 <mailto:gcc-bugs-request@gcc.gnu.org?subject=unsubscribe>
List-Archive: <https://gcc.gnu.org/pipermail/gcc-bugs/>
List-Post: <mailto:gcc-bugs@gcc.gnu.org>
List-Help: <mailto:gcc-bugs-request@gcc.gnu.org?subject=help>
List-Subscribe: <https://gcc.gnu.org/mailman/listinfo/gcc-bugs>,
 <mailto:gcc-bugs-request@gcc.gnu.org?subject=subscribe>
X-List-Received-Date: Wed, 16 Jun 2021 05:53:01 -0000

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=3D100866

--- Comment #5 from luoxhu at gcc dot gnu.org ---
(In reply to Segher Boessenkool from comment #4)
> This PR is specifically about the vec_revb builtin.  But yes, we should
> look at what is generated for all other code (having only the builtin
> generate good code is suboptimal for a generic thing like this), and for
> other sizes as well.

Sorry I don't quite understand what you mean. IMO vec_revb is expanded by
CODE_FOR_revb_v8hi through revb_<mode> pattern. So this is where we should
change to make better code generation...=20
For V8HI, it is natural to use vspltish 8+vrlh to turn
{0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15} to
{1,0,3,2,5,4,7,6,9,8,11,10,13,12,15,14}.

But for V4SI, we need use vspltish+vrlh to turn it to
{1,0,3,2,5,4,7,6,9,8,11,10,13,12,15,14} first, and a "vrlw 16" to turn it t=
o=20
{3,2,1,0,7,6,5,4,11,10,9,8,15,14,13,12}. I am not sure whether this is bett=
er
than lvx+xxlnor+vperm especially for V2DI&V1TI with additional "vrld 32" or
"vrld 32"+"vrlq 64"? (Those are all operations on register without load from
memory like lvx.)


bt 5
#0  gen_revb_v8hi (operand0=3D0x7ffff4d4ce40, operand1=3D0x7ffff4d4cf60) at
../../gcc/gcc/config/rs6000/vsx.md:5858
#1  0x0000000010b05360 in insn_gen_fn::operator()<rtx_def*, rtx_def*>
(this=3D0x130ab188 <insn_data+163016>) at../../gcc/gcc/recog.h:407
#2  0x0000000011aa1e30 in rs6000_expand_unop_builtin (icode=3DCODE_FOR_revb=
_v8hi,
exp=3D<call_expr 0x7ffff4f509a0>
, target=3D0x7ffff4d4ce40) at ../../gcc/gcc/config/rs6000/rs6000-call.c:9451
#3  0x0000000011ab27a4 in rs6000_expand_builtin (exp=3D<call_expr
0x7ffff4f509a0>, target=3D0x7ffff4d4ce40, subtarget=3D0x0, mode=3DE_V8HImod=
e,
ignore=3D0) at ../../gcc/gcc/config/rs6000/rs6000-call.c:13157
#4  0x0000000010815268 in expand_builtin (exp=3D<call_expr 0x7ffff4f509a0>,
target=3D0x7ffff4d4ce40, subtarget=3D0x0, mode=3DE_V8HImode, ignore=3D0) at
../../gcc/gcc/builtins.c:9559=