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 8F6233858403; Tue,  8 Nov 2022 06:11:08 +0000 (GMT)
DKIM-Filter: OpenDKIM Filter v2.11.0 sourceware.org 8F6233858403
DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=gcc.gnu.org;
	s=default; t=1667887868;
	bh=TrNZ3FBetKda6QCOJcbqLO/EonYqIjppJHIt2817LOs=;
	h=From:To:Subject:Date:In-Reply-To:References:From;
	b=qif8DeOo3iZrzrCqz3ehHqWyTOpH6u7+kAAs0h6+ylI/nOk4sb8ZVRwEukfbwoL+0
	 86l8qsEbQBEj8jy1o8vbBNjQw2ubxb+pVd3VdZKSr+u6Ohnx2TRO9Hwh7cQAsSUJfH
	 MvovnaY9ORnb+P70LoKE76a1A7luV835SQYRah78=
From: "unlvsur at live dot com" <gcc-bugzilla@gcc.gnu.org>
To: gcc-bugs@gcc.gnu.org
Subject: [Bug target/107563] __builtin_shufflevector fails to pshufd
 instructions under default x86_64 compilation toggle which is the sse2 one
Date: Tue, 08 Nov 2022 06:11:08 +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: missed-optimization
X-Bugzilla-Severity: normal
X-Bugzilla-Who: unlvsur at live dot com
X-Bugzilla-Status: UNCONFIRMED
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-107563-4-YoahemLfXe@http.gcc.gnu.org/bugzilla/>
In-Reply-To: <bug-107563-4@http.gcc.gnu.org/bugzilla/>
References: <bug-107563-4@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=3D107563

--- Comment #7 from cqwrteur <unlvsur at live dot com> ---
(In reply to Hongtao.liu from comment #6)
> Shufd only handles
>=20
> void foo1(temp_vec_type& v) noexcept
> {
> 	v=3D__builtin_shufflevector(v,v,12,13,14,15,8,9,10,11,4,5,6,7,0,1,2,3);
> }
>=20
> Not the case in #c0.

I am using it for byte swap

actually, clang has a solution

                        using x86_64_v4si [[__gnu__::__vector_size__ (16)]]=
 =3D
int;
                        using x86_64_v16qi [[__gnu__::__vector_size__ (16)]=
] =3D
char;
                        using x86_64_v8hi [[__gnu__::__vector_size__ (16)]]=
 =3D
short;
                        constexpr x86_64_v16qi zero{};
                        if constexpr(sizeof(T)=3D=3D8)
                        {
                                auto
res0{__builtin_ia32_punpcklbw128(temp_vec,zero)};
                                auto
res1{__builtin_ia32_pshufd((x86_64_v4si)res0,78)};
                                auto
res2{__builtin_ia32_pshuflw((x86_64_v8hi)res1,27)};
                                auto res3{__builtin_ia32_pshufhw(res2,27)};
                                auto
res4{__builtin_ia32_punpckhbw128(temp_vec,zero)};
                                auto
res5{__builtin_ia32_pshufd((x86_64_v4si)res4,78)};
                                auto
res6{__builtin_ia32_pshuflw((x86_64_v8hi)res5,27)};
                                auto res7{__builtin_ia32_pshufhw(res6,27)};
                                temp_vec=3D__builtin_ia32_packuswb128(res3,=
res7);
                        }
                        else if constexpr(sizeof(T)=3D=3D4)
                        {
                                auto
res0{__builtin_ia32_punpcklbw128(temp_vec,zero)};
                                auto
res2{__builtin_ia32_pshuflw((x86_64_v8hi)res0,27)};
                                auto res3{__builtin_ia32_pshufhw(res2,27)};
                                auto
res4{__builtin_ia32_punpckhbw128(temp_vec,zero)};
                                auto
res6{__builtin_ia32_pshuflw((x86_64_v8hi)res4,27)};
                                auto res7{__builtin_ia32_pshufhw(res6,27)};
                                temp_vec=3D__builtin_ia32_packuswb128(res3,=
res7);
                        }
                        else if constexpr(sizeof(T)=3D=3D2)
                        {
                                using x86_64_v8hu [[__gnu__::__vector_size__
(16)]] =3D unsigned short;
                                auto res0{(x86_64_v8hu)temp_vec};
                                temp_vec=3D(x86_64_v16qi)((res0>>8)|(res0<<=
8));
                        }=