From mboxrd@z Thu Jan  1 00:00:00 1970
Return-Path: <gcc-patches-return-289572-listarch-gcc-patches=gcc.gnu.org@gcc.gnu.org>
Received: (qmail 7219 invoked by alias); 15 Apr 2011 12:41:34 -0000
Received: (qmail 7205 invoked by uid 22791); 15 Apr 2011 12:41:31 -0000
X-SWARE-Spam-Status: No, hits=-2.3 required=5.0	tests=AWL,BAYES_00,DKIM_SIGNED,DKIM_VALID,DKIM_VALID_AU,FREEMAIL_FROM,RCVD_IN_DNSWL_LOW,RFC_ABUSE_POST,TW_TM,T_TO_NO_BRKTS_FREEMAIL
X-Spam-Check-By: sourceware.org
Received: from mail-ww0-f51.google.com (HELO mail-ww0-f51.google.com) (74.125.82.51)    by sourceware.org (qpsmtpd/0.43rc1) with ESMTP; Fri, 15 Apr 2011 12:41:23 +0000
Received: by wwf26 with SMTP id 26so3040529wwf.8        for <gcc-patches@gcc.gnu.org>; Fri, 15 Apr 2011 05:41:22 -0700 (PDT)
MIME-Version: 1.0
Received: by 10.227.169.140 with SMTP id z12mr2056371wby.89.1302871281907; Fri, 15 Apr 2011 05:41:21 -0700 (PDT)
Received: by 10.227.0.140 with HTTP; Fri, 15 Apr 2011 05:41:21 -0700 (PDT)
In-Reply-To: <g4r597eemb.fsf@linaro.org>
References: <g4pqorfvwp.fsf@linaro.org>	<g4r597eemb.fsf@linaro.org>
Date: Fri, 15 Apr 2011 12:44:00 -0000
Message-ID: <BANLkTi=Y-+uOexkL1rqVz=TseCm2=spsAg@mail.gmail.com>
Subject: Re: [8/9] Testsuite: split tests for strided accesses
From: Richard Guenther <richard.guenther@gmail.com>
To: gcc-patches@gcc.gnu.org, patches@lianro.org, richard.sandiford@linaro.org
Content-Type: text/plain; charset=ISO-8859-1
Content-Transfer-Encoding: quoted-printable
X-IsSubscribed: yes
Mailing-List: contact gcc-patches-help@gcc.gnu.org; run by ezmlm
Precedence: bulk
List-Id: <gcc-patches.gcc.gnu.org>
List-Archive: <http://gcc.gnu.org/ml/gcc-patches/>
List-Post: <mailto:gcc-patches@gcc.gnu.org>
List-Help: <mailto:gcc-patches-help@gcc.gnu.org>
Sender: gcc-patches-owner@gcc.gnu.org
X-SW-Source: 2011-04/txt/msg01176.txt.bz2

On Tue, Apr 12, 2011 at 4:19 PM, Richard Sandiford
<richard.sandiford@linaro.org> wrote:
> The next patch introduces separate vect_stridedN target selectors
> for each tested stride factor N. =A0At the moment, some tests contain
> several independent loops that have different stride factors.
> It's easier to make the next change if we put these loops into
> separate tests.
>
> Tested on x86_64-linux-gnu and arm-linux-gnueabi. =A0OK to install?

Ok.

Thanks,
Richard.

> Richard
>
>
> gcc/testsuite/
> =A0 =A0 =A0 =A0* gcc.dg/vect/slp-11.c: Split into...
> =A0 =A0 =A0 =A0* gcc.dg/vect/slp-11a.c, gcc.dg/vect/slp-11b.c,
> =A0 =A0 =A0 =A0gcc.dg/vect/slp-11c.c: ...these tests.
> =A0 =A0 =A0 =A0* gcc.dg/vect/slp-12a.c: Split 4-stride loop into...
> =A0 =A0 =A0 =A0* gcc.dg/vect/slp-12c.c: ...this new test.
> =A0 =A0 =A0 =A0* gcc.dg/vect/slp-19.c: Split into...
> =A0 =A0 =A0 =A0* gcc.dg/vect/slp-19a.c, gcc.dg/vect/slp-19b.c,
> =A0 =A0 =A0 =A0gcc.dg/vect/slp-19c.c: ...these new tests.
>
> Index: gcc/testsuite/gcc.dg/vect/slp-11.c
> =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=
=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=
=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D
> --- gcc/testsuite/gcc.dg/vect/slp-11.c =A02011-04-12 15:18:24.000000000 +=
0100
> +++ /dev/null =A0 2011-03-23 08:42:11.268792848 +0000
> @@ -1,113 +0,0 @@
> -/* { dg-require-effective-target vect_int } */
> -
> -#include <stdarg.h>
> -#include "tree-vect.h"
> -
> -#define N 8
> -
> -int
> -main1 ()
> -{
> - =A0int i;
> - =A0unsigned int out[N*8], a0, a1, a2, a3, a4, a5, a6, a7, b1, b0, b2, b=
3, b4, b5, b6, b7;
> - =A0unsigned int in[N*8] =3D {0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,1=
7,18,19,20,21,22,23,24,25,26,27,28,29,30,31,32,33,34,35,36,37,38,39,40,41,4=
2,43,44,45,46,47,48,49,50,51,52,53,54,55,56,57,58,59,60,61,62,63};
> - =A0float out2[N*8];
> -
> - =A0/* Different operations - not SLPable. =A0*/
> - =A0for (i =3D 0; i < N; i++)
> - =A0 =A0{
> - =A0 =A0 =A0a0 =3D in[i*8] + 5;
> - =A0 =A0 =A0a1 =3D in[i*8 + 1] * 6;
> - =A0 =A0 =A0a2 =3D in[i*8 + 2] + 7;
> - =A0 =A0 =A0a3 =3D in[i*8 + 3] + 8;
> - =A0 =A0 =A0a4 =3D in[i*8 + 4] + 9;
> - =A0 =A0 =A0a5 =3D in[i*8 + 5] + 10;
> - =A0 =A0 =A0a6 =3D in[i*8 + 6] + 11;
> - =A0 =A0 =A0a7 =3D in[i*8 + 7] + 12;
> -
> - =A0 =A0 =A0b0 =3D a0 * 3;
> - =A0 =A0 =A0b1 =3D a1 * 2;
> - =A0 =A0 =A0b2 =3D a2 * 12;
> - =A0 =A0 =A0b3 =3D a3 * 5;
> - =A0 =A0 =A0b4 =3D a4 * 8;
> - =A0 =A0 =A0b5 =3D a5 * 4;
> - =A0 =A0 =A0b6 =3D a6 * 3;
> - =A0 =A0 =A0b7 =3D a7 * 2;
> -
> - =A0 =A0 =A0out[i*8] =3D b0 - 2;
> - =A0 =A0 =A0out[i*8 + 1] =3D b1 - 3;
> - =A0 =A0 =A0out[i*8 + 2] =3D b2 - 2;
> - =A0 =A0 =A0out[i*8 + 3] =3D b3 - 1;
> - =A0 =A0 =A0out[i*8 + 4] =3D b4 - 8;
> - =A0 =A0 =A0out[i*8 + 5] =3D b5 - 7;
> - =A0 =A0 =A0out[i*8 + 6] =3D b6 - 3;
> - =A0 =A0 =A0out[i*8 + 7] =3D b7 - 7;
> - =A0 =A0}
> -
> - =A0/* check results: =A0*/
> - =A0for (i =3D 0; i < N; i++)
> - =A0 =A0{
> - =A0 =A0 =A0if (out[i*8] !=3D =A0(in[i*8] + 5) * 3 - 2
> - =A0 =A0 =A0 =A0 || out[i*8 + 1] !=3D (in[i*8 + 1] * 6) * 2 - 3
> - =A0 =A0 =A0 =A0 || out[i*8 + 2] !=3D (in[i*8 + 2] + 7) * 12 - 2
> - =A0 =A0 =A0 =A0 || out[i*8 + 3] !=3D (in[i*8 + 3] + 8) * 5 - 1
> - =A0 =A0 =A0 =A0 || out[i*8 + 4] !=3D (in[i*8 + 4] + 9) * 8 - 8
> - =A0 =A0 =A0 =A0 || out[i*8 + 5] !=3D (in[i*8 + 5] + 10) * 4 - 7
> - =A0 =A0 =A0 =A0 || out[i*8 + 6] !=3D (in[i*8 + 6] + 11) * 3 - 3
> - =A0 =A0 =A0 =A0 || out[i*8 + 7] !=3D (in[i*8 + 7] + 12) * 2 - 7)
> - =A0 =A0 =A0 abort ();
> - =A0 =A0}
> -
> - =A0/* Requires permutation - not SLPable. =A0*/
> - =A0for (i =3D 0; i < N*2; i++)
> - =A0 =A0{
> - =A0 =A0 =A0out[i*4] =3D (in[i*4] + 2) * 3;
> - =A0 =A0 =A0out[i*4 + 1] =3D (in[i*4 + 2] + 2) * 7;
> - =A0 =A0 =A0out[i*4 + 2] =3D (in[i*4 + 1] + 7) * 3;
> - =A0 =A0 =A0out[i*4 + 3] =3D (in[i*4 + 3] + 3) * 4;
> - =A0 =A0}
> -
> - =A0/* check results: =A0*/
> - =A0for (i =3D 0; i < N*2; i++)
> - =A0 =A0{
> - =A0 =A0 =A0if (out[i*4] !=3D =A0(in[i*4] + 2) * 3
> - =A0 =A0 =A0 =A0 || out[i*4 + 1] !=3D (in[i*4 + 2] + 2) * 7
> - =A0 =A0 =A0 =A0 || out[i*4 + 2] !=3D (in[i*4 + 1] + 7) * 3
> - =A0 =A0 =A0 =A0 || out[i*4 + 3] !=3D (in[i*4 + 3] + 3) * 4)
> - =A0 =A0 =A0 =A0abort ();
> - =A0 =A0}
> -
> - =A0/* Different operations - not SLPable. =A0*/
> - =A0for (i =3D 0; i < N*4; i++)
> - =A0 =A0{
> - =A0 =A0 =A0out2[i*2] =3D ((float) in[i*2] * 2 + 6) ;
> - =A0 =A0 =A0out2[i*2 + 1] =3D (float) (in[i*2 + 1] * 3 + 7);
> - =A0 =A0}
> -
> - =A0/* check results: =A0*/
> - =A0for (i =3D 0; i < N*4; i++)
> - =A0 =A0{
> - =A0 =A0 =A0if (out2[i*2] !=3D =A0((float) in[i*2] * 2 + 6)
> - =A0 =A0 =A0 =A0 || out2[i*2 + 1] !=3D (float) (in[i*2 + 1] * 3 + 7))
> - =A0 =A0 =A0 =A0abort ();
> - =A0 =A0}
> -
> -
> - =A0return 0;
> -}
> -
> -int main (void)
> -{
> - =A0check_vect ();
> -
> - =A0main1 ();
> -
> - =A0return 0;
> -}
> -
> -/* { dg-final { scan-tree-dump-times "vectorized 3 loops" 1 "vect" =A0{ =
target { { vect_uintfloat_cvt && vect_strided } && =A0vect_int_mult } } } }=
 */
> -/* { dg-final { scan-tree-dump-times "vectorized 2 loops" 1 "vect" =A0{ =
target { { { ! vect_uintfloat_cvt } && vect_strided } && =A0vect_int_mult }=
 } } } */
> -/* { dg-final { scan-tree-dump-times "vectorized 0 loops" 1 "vect" =A0{t=
arget =A0{ ! { vect_int_mult && vect_strided } } } } } =A0*/
> -/* { dg-final { scan-tree-dump-times "vectorizing stmts using SLP" 0 =A0=
"vect" =A0} } */
> -/* { dg-final { cleanup-tree-dump "vect" } } */
> -
> Index: gcc/testsuite/gcc.dg/vect/slp-11a.c
> =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=
=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=
=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D
> --- /dev/null =A0 2011-03-23 08:42:11.268792848 +0000
> +++ gcc/testsuite/gcc.dg/vect/slp-11a.c 2011-04-12 15:18:25.000000000 +01=
00
> @@ -0,0 +1,75 @@
> +/* { dg-require-effective-target vect_int } */
> +
> +#include <stdarg.h>
> +#include "tree-vect.h"
> +
> +#define N 8
> +
> +int
> +main1 ()
> +{
> + =A0int i;
> + =A0unsigned int out[N*8], a0, a1, a2, a3, a4, a5, a6, a7, b1, b0, b2, b=
3, b4, b5, b6, b7;
> + =A0unsigned int in[N*8] =3D {0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,1=
7,18,19,20,21,22,23,24,25,26,27,28,29,30,31,32,33,34,35,36,37,38,39,40,41,4=
2,43,44,45,46,47,48,49,50,51,52,53,54,55,56,57,58,59,60,61,62,63};
> +
> + =A0/* Different operations - not SLPable. =A0*/
> + =A0for (i =3D 0; i < N; i++)
> + =A0 =A0{
> + =A0 =A0 =A0a0 =3D in[i*8] + 5;
> + =A0 =A0 =A0a1 =3D in[i*8 + 1] * 6;
> + =A0 =A0 =A0a2 =3D in[i*8 + 2] + 7;
> + =A0 =A0 =A0a3 =3D in[i*8 + 3] + 8;
> + =A0 =A0 =A0a4 =3D in[i*8 + 4] + 9;
> + =A0 =A0 =A0a5 =3D in[i*8 + 5] + 10;
> + =A0 =A0 =A0a6 =3D in[i*8 + 6] + 11;
> + =A0 =A0 =A0a7 =3D in[i*8 + 7] + 12;
> +
> + =A0 =A0 =A0b0 =3D a0 * 3;
> + =A0 =A0 =A0b1 =3D a1 * 2;
> + =A0 =A0 =A0b2 =3D a2 * 12;
> + =A0 =A0 =A0b3 =3D a3 * 5;
> + =A0 =A0 =A0b4 =3D a4 * 8;
> + =A0 =A0 =A0b5 =3D a5 * 4;
> + =A0 =A0 =A0b6 =3D a6 * 3;
> + =A0 =A0 =A0b7 =3D a7 * 2;
> +
> + =A0 =A0 =A0out[i*8] =3D b0 - 2;
> + =A0 =A0 =A0out[i*8 + 1] =3D b1 - 3;
> + =A0 =A0 =A0out[i*8 + 2] =3D b2 - 2;
> + =A0 =A0 =A0out[i*8 + 3] =3D b3 - 1;
> + =A0 =A0 =A0out[i*8 + 4] =3D b4 - 8;
> + =A0 =A0 =A0out[i*8 + 5] =3D b5 - 7;
> + =A0 =A0 =A0out[i*8 + 6] =3D b6 - 3;
> + =A0 =A0 =A0out[i*8 + 7] =3D b7 - 7;
> + =A0 =A0}
> +
> + =A0/* check results: =A0*/
> + =A0for (i =3D 0; i < N; i++)
> + =A0 =A0{
> + =A0 =A0 =A0if (out[i*8] !=3D =A0(in[i*8] + 5) * 3 - 2
> + =A0 =A0 =A0 =A0 || out[i*8 + 1] !=3D (in[i*8 + 1] * 6) * 2 - 3
> + =A0 =A0 =A0 =A0 || out[i*8 + 2] !=3D (in[i*8 + 2] + 7) * 12 - 2
> + =A0 =A0 =A0 =A0 || out[i*8 + 3] !=3D (in[i*8 + 3] + 8) * 5 - 1
> + =A0 =A0 =A0 =A0 || out[i*8 + 4] !=3D (in[i*8 + 4] + 9) * 8 - 8
> + =A0 =A0 =A0 =A0 || out[i*8 + 5] !=3D (in[i*8 + 5] + 10) * 4 - 7
> + =A0 =A0 =A0 =A0 || out[i*8 + 6] !=3D (in[i*8 + 6] + 11) * 3 - 3
> + =A0 =A0 =A0 =A0 || out[i*8 + 7] !=3D (in[i*8 + 7] + 12) * 2 - 7)
> + =A0 =A0 =A0 abort ();
> + =A0 =A0}
> +
> + =A0return 0;
> +}
> +
> +int main (void)
> +{
> + =A0check_vect ();
> +
> + =A0main1 ();
> +
> + =A0return 0;
> +}
> +
> +/* { dg-final { scan-tree-dump-times "vectorized 1 loops" 1 "vect" { tar=
get { vect_strided && vect_int_mult } } } } */
> +/* { dg-final { scan-tree-dump-times "vectorized 0 loops" 1 "vect" { tar=
get { ! { vect_strided && vect_int_mult } } } } } */
> +/* { dg-final { scan-tree-dump-times "vectorizing stmts using SLP" 0 "ve=
ct" } } */
> +/* { dg-final { cleanup-tree-dump "vect" } } */
> Index: gcc/testsuite/gcc.dg/vect/slp-11b.c
> =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=
=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=
=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D
> --- /dev/null =A0 2011-03-23 08:42:11.268792848 +0000
> +++ gcc/testsuite/gcc.dg/vect/slp-11b.c 2011-04-12 15:18:25.000000000 +01=
00
> @@ -0,0 +1,49 @@
> +/* { dg-require-effective-target vect_int } */
> +
> +#include <stdarg.h>
> +#include "tree-vect.h"
> +
> +#define N 8
> +
> +int
> +main1 ()
> +{
> + =A0int i;
> + =A0unsigned int out[N*8], a0, a1, a2, a3, a4, a5, a6, a7, b1, b0, b2, b=
3, b4, b5, b6, b7;
> + =A0unsigned int in[N*8] =3D {0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,1=
7,18,19,20,21,22,23,24,25,26,27,28,29,30,31,32,33,34,35,36,37,38,39,40,41,4=
2,43,44,45,46,47,48,49,50,51,52,53,54,55,56,57,58,59,60,61,62,63};
> +
> + =A0/* Requires permutation - not SLPable. =A0*/
> + =A0for (i =3D 0; i < N*2; i++)
> + =A0 =A0{
> + =A0 =A0 =A0out[i*4] =3D (in[i*4] + 2) * 3;
> + =A0 =A0 =A0out[i*4 + 1] =3D (in[i*4 + 2] + 2) * 7;
> + =A0 =A0 =A0out[i*4 + 2] =3D (in[i*4 + 1] + 7) * 3;
> + =A0 =A0 =A0out[i*4 + 3] =3D (in[i*4 + 3] + 3) * 4;
> + =A0 =A0}
> +
> + =A0/* check results: =A0*/
> + =A0for (i =3D 0; i < N*2; i++)
> + =A0 =A0{
> + =A0 =A0 =A0if (out[i*4] !=3D =A0(in[i*4] + 2) * 3
> + =A0 =A0 =A0 =A0 || out[i*4 + 1] !=3D (in[i*4 + 2] + 2) * 7
> + =A0 =A0 =A0 =A0 || out[i*4 + 2] !=3D (in[i*4 + 1] + 7) * 3
> + =A0 =A0 =A0 =A0 || out[i*4 + 3] !=3D (in[i*4 + 3] + 3) * 4)
> + =A0 =A0 =A0 =A0abort ();
> + =A0 =A0}
> +
> + =A0return 0;
> +}
> +
> +int main (void)
> +{
> + =A0check_vect ();
> +
> + =A0main1 ();
> +
> + =A0return 0;
> +}
> +
> +/* { dg-final { scan-tree-dump-times "vectorized 1 loops" 1 "vect" { tar=
get { vect_strided && vect_int_mult } } } } */
> +/* { dg-final { scan-tree-dump-times "vectorized 0 loops" 1 "vect" { tar=
get { ! { vect_strided && vect_int_mult } } } } } */
> +/* { dg-final { scan-tree-dump-times "vectorizing stmts using SLP" 0 "ve=
ct" } } */
> +/* { dg-final { cleanup-tree-dump "vect" } } */
> Index: gcc/testsuite/gcc.dg/vect/slp-11c.c
> =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=
=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=
=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D
> --- /dev/null =A0 2011-03-23 08:42:11.268792848 +0000
> +++ gcc/testsuite/gcc.dg/vect/slp-11c.c 2011-04-12 15:18:25.000000000 +01=
00
> @@ -0,0 +1,46 @@
> +/* { dg-require-effective-target vect_int } */
> +
> +#include <stdarg.h>
> +#include "tree-vect.h"
> +
> +#define N 8
> +
> +int
> +main1 ()
> +{
> + =A0int i;
> + =A0unsigned int in[N*8] =3D {0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,1=
7,18,19,20,21,22,23,24,25,26,27,28,29,30,31,32,33,34,35,36,37,38,39,40,41,4=
2,43,44,45,46,47,48,49,50,51,52,53,54,55,56,57,58,59,60,61,62,63};
> + =A0float out[N*8];
> +
> + =A0/* Different operations - not SLPable. =A0*/
> + =A0for (i =3D 0; i < N*4; i++)
> + =A0 =A0{
> + =A0 =A0 =A0out[i*2] =3D ((float) in[i*2] * 2 + 6) ;
> + =A0 =A0 =A0out[i*2 + 1] =3D (float) (in[i*2 + 1] * 3 + 7);
> + =A0 =A0}
> +
> + =A0/* check results: =A0*/
> + =A0for (i =3D 0; i < N*4; i++)
> + =A0 =A0{
> + =A0 =A0 =A0if (out[i*2] !=3D =A0((float) in[i*2] * 2 + 6)
> + =A0 =A0 =A0 =A0 || out[i*2 + 1] !=3D (float) (in[i*2 + 1] * 3 + 7))
> + =A0 =A0 =A0 =A0abort ();
> + =A0 =A0}
> +
> +
> + =A0return 0;
> +}
> +
> +int main (void)
> +{
> + =A0check_vect ();
> +
> + =A0main1 ();
> +
> + =A0return 0;
> +}
> +
> +/* { dg-final { scan-tree-dump-times "vectorized 1 loops" 1 "vect" { tar=
get { { vect_uintfloat_cvt && vect_strided } && vect_int_mult } } } } */
> +/* { dg-final { scan-tree-dump-times "vectorized 0 loops" 1 "vect" { tar=
get { ! { { vect_uintfloat_cvt && vect_strided } && vect_int_mult } } } } }=
 */
> +/* { dg-final { scan-tree-dump-times "vectorizing stmts using SLP" 0 =A0=
"vect" =A0} } */
> +/* { dg-final { cleanup-tree-dump "vect" } } */
> Index: gcc/testsuite/gcc.dg/vect/slp-12a.c
> =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=
=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=
=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D
> --- gcc/testsuite/gcc.dg/vect/slp-12a.c 2011-04-12 15:18:24.000000000 +01=
00
> +++ gcc/testsuite/gcc.dg/vect/slp-12a.c 2011-04-12 15:18:25.000000000 +01=
00
> @@ -11,7 +11,7 @@ main1 ()
> =A0 int i;
> =A0 unsigned int out[N*8], a0, a1, a2, a3, a4, a5, a6, a7, b1, b0, b2, b3=
, b4, b5, b6, b7;
> =A0 unsigned int in[N*8] =3D {0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17=
,18,19,20,21,22,23,24,25,26,27,28,29,30,31,32,33,34,35,36,37,38,39,40,41,42=
,43,44,45,46,47,48,49,50,51,52,53,54,55,56,57,58,59,60,61,62,63};
> - =A0unsigned int ia[N], ib[N*2];
> + =A0unsigned int ia[N];
>
> =A0 for (i =3D 0; i < N; i++)
> =A0 =A0 {
> @@ -61,27 +61,6 @@ main1 ()
> =A0 =A0 =A0 =A0abort ();
> =A0 =A0 }
>
> - =A0for (i =3D 0; i < N*2; i++)
> - =A0 =A0{
> - =A0 =A0 =A0out[i*4] =3D (in[i*4] + 2) * 3;
> - =A0 =A0 =A0out[i*4 + 1] =3D (in[i*4 + 1] + 2) * 7;
> - =A0 =A0 =A0out[i*4 + 2] =3D (in[i*4 + 2] + 7) * 3;
> - =A0 =A0 =A0out[i*4 + 3] =3D (in[i*4 + 3] + 7) * 7;
> -
> - =A0 =A0 =A0ib[i] =3D 7;
> - =A0 =A0}
> -
> - =A0/* check results: =A0*/
> - =A0for (i =3D 0; i < N*2; i++)
> - =A0 =A0{
> - =A0 =A0 =A0if (out[i*4] !=3D =A0(in[i*4] + 2) * 3
> - =A0 =A0 =A0 =A0 || out[i*4 + 1] !=3D (in[i*4 + 1] + 2) * 7
> - =A0 =A0 =A0 =A0 || out[i*4 + 2] !=3D (in[i*4 + 2] + 7) * 3
> - =A0 =A0 =A0 =A0 || out[i*4 + 3] !=3D (in[i*4 + 3] + 7) * 7
> - =A0 =A0 =A0 =A0 || ib[i] !=3D 7)
> - =A0 =A0 =A0 =A0abort ();
> - =A0 =A0}
> -
> =A0 return 0;
> =A0}
>
> @@ -94,11 +73,8 @@ int main (void)
> =A0 return 0;
> =A0}
>
> -/* { dg-final { scan-tree-dump-times "vectorized 2 loops" 1 "vect" =A0{t=
arget { vect_strided && vect_int_mult} } } } */
> -/* { dg-final { scan-tree-dump-times "vectorized 1 loops" 1 "vect" =A0{t=
arget { {! {vect_strided}} && vect_int_mult } } } } */
> -/* { dg-final { scan-tree-dump-times "vectorized 0 loops" 1 "vect" =A0{t=
arget =A0{ ! vect_int_mult } } } } */
> -/* { dg-final { scan-tree-dump-times "vectorizing stmts using SLP" 2 "ve=
ct" {target { vect_strided && vect_int_mult } } } } */
> -/* { dg-final { scan-tree-dump-times "vectorizing stmts using SLP" 1 "ve=
ct" {target { {! {vect_strided}} && vect_int_mult } } } } */
> -/* { dg-final { scan-tree-dump-times "vectorizing stmts using SLP" 0 "ve=
ct" {target =A0{ ! vect_int_mult } } } } */
> +/* { dg-final { scan-tree-dump-times "vectorized 1 loops" 1 "vect" { tar=
get { vect_strided && vect_int_mult } } } } */
> +/* { dg-final { scan-tree-dump-times "vectorized 0 loops" 1 "vect" { tar=
get { ! { vect_strided && vect_int_mult } } } } } */
> +/* { dg-final { scan-tree-dump-times "vectorizing stmts using SLP" 1 "ve=
ct" { target { vect_strided && vect_int_mult } } } } */
> +/* { dg-final { scan-tree-dump-times "vectorizing stmts using SLP" 0 "ve=
ct" { target { ! { vect_strided && vect_int_mult } } } } } */
> =A0/* { dg-final { cleanup-tree-dump "vect" } } */
> -
> Index: gcc/testsuite/gcc.dg/vect/slp-12c.c
> =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=
=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=
=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D
> --- /dev/null =A0 2011-03-23 08:42:11.268792848 +0000
> +++ gcc/testsuite/gcc.dg/vect/slp-12c.c 2011-04-12 15:18:25.000000000 +01=
00
> @@ -0,0 +1,53 @@
> +/* { dg-require-effective-target vect_int } */
> +
> +#include <stdarg.h>
> +#include "tree-vect.h"
> +
> +#define N 8
> +
> +int
> +main1 ()
> +{
> + =A0int i;
> + =A0unsigned int out[N*8], a0, a1, a2, a3, a4, a5, a6, a7, b1, b0, b2, b=
3, b4, b5, b6, b7;
> + =A0unsigned int in[N*8] =3D {0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,1=
7,18,19,20,21,22,23,24,25,26,27,28,29,30,31,32,33,34,35,36,37,38,39,40,41,4=
2,43,44,45,46,47,48,49,50,51,52,53,54,55,56,57,58,59,60,61,62,63};
> + =A0unsigned int ia[N*2];
> +
> + =A0for (i =3D 0; i < N*2; i++)
> + =A0 =A0{
> + =A0 =A0 =A0out[i*4] =3D (in[i*4] + 2) * 3;
> + =A0 =A0 =A0out[i*4 + 1] =3D (in[i*4 + 1] + 2) * 7;
> + =A0 =A0 =A0out[i*4 + 2] =3D (in[i*4 + 2] + 7) * 3;
> + =A0 =A0 =A0out[i*4 + 3] =3D (in[i*4 + 3] + 7) * 7;
> +
> + =A0 =A0 =A0ia[i] =3D 7;
> + =A0 =A0}
> +
> + =A0/* check results: =A0*/
> + =A0for (i =3D 0; i < N*2; i++)
> + =A0 =A0{
> + =A0 =A0 =A0if (out[i*4] !=3D =A0(in[i*4] + 2) * 3
> + =A0 =A0 =A0 =A0 || out[i*4 + 1] !=3D (in[i*4 + 1] + 2) * 7
> + =A0 =A0 =A0 =A0 || out[i*4 + 2] !=3D (in[i*4 + 2] + 7) * 3
> + =A0 =A0 =A0 =A0 || out[i*4 + 3] !=3D (in[i*4 + 3] + 7) * 7
> + =A0 =A0 =A0 =A0 || ia[i] !=3D 7)
> + =A0 =A0 =A0 =A0abort ();
> + =A0 =A0}
> +
> + =A0return 0;
> +}
> +
> +int main (void)
> +{
> + =A0check_vect ();
> +
> + =A0main1 ();
> +
> + =A0return 0;
> +}
> +
> +/* { dg-final { scan-tree-dump-times "vectorized 1 loops" 1 "vect" =A0{ =
target { vect_int_mult } } } } */
> +/* { dg-final { scan-tree-dump-times "vectorized 0 loops" 1 "vect" =A0{ =
target { ! vect_int_mult } } } } */
> +/* { dg-final { scan-tree-dump-times "vectorizing stmts using SLP" 1 "ve=
ct" { target vect_int_mult } } } */
> +/* { dg-final { scan-tree-dump-times "vectorizing stmts using SLP" 0 "ve=
ct" { target { ! vect_int_mult } } } } */
> +/* { dg-final { cleanup-tree-dump "vect" } } */
> Index: gcc/testsuite/gcc.dg/vect/slp-19.c
> =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=
=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=
=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D
> --- gcc/testsuite/gcc.dg/vect/slp-19.c =A02011-04-12 15:18:24.000000000 +=
0100
> +++ /dev/null =A0 2011-03-23 08:42:11.268792848 +0000
> @@ -1,154 +0,0 @@
> -/* { dg-require-effective-target vect_int } */
> -
> -#include <stdarg.h>
> -#include "tree-vect.h"
> -
> -#define N 16
> -
> -int
> -main1 ()
> -{
> - =A0unsigned int i;
> - =A0unsigned int out[N*8];
> - =A0unsigned int in[N*8] =3D {0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,1=
7,18,19,20,21,22,23,24,25,26,27,28,29,30,31,32,33,34,35,36,37,38,39,40,41,4=
2,43,44,45,46,47,48,49,50,51,52,53,54,55,56,57,58,59,60,61,62,63,0,1,2,3,4,=
5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20,21,22,23,24,25,26,27,28,29,30,31=
,32,33,34,35,36,37,38,39,40,41,42,43,44,45,46,47,48,49,50,51,52,53,54,55,56=
,57,58,59,60,61,62,63};
> - =A0unsigned int ia[N*2], a0, a1, a2, a3;
> -
> - =A0for (i =3D 0; i < N; i++)
> - =A0 =A0{
> - =A0 =A0 =A0out[i*8] =3D in[i*8];
> - =A0 =A0 =A0out[i*8 + 1] =3D in[i*8 + 1];
> - =A0 =A0 =A0out[i*8 + 2] =3D in[i*8 + 2];
> - =A0 =A0 =A0out[i*8 + 3] =3D in[i*8 + 3];
> - =A0 =A0 =A0out[i*8 + 4] =3D in[i*8 + 4];
> - =A0 =A0 =A0out[i*8 + 5] =3D in[i*8 + 5];
> - =A0 =A0 =A0out[i*8 + 6] =3D in[i*8 + 6];
> - =A0 =A0 =A0out[i*8 + 7] =3D in[i*8 + 7];
> -
> - =A0 =A0 =A0ia[i] =3D in[i*8 + 2];
> - =A0 =A0}
> -
> - =A0/* check results: =A0*/
> - =A0for (i =3D 0; i < N; i++)
> - =A0 =A0{
> - =A0 =A0 =A0if (out[i*8] !=3D =A0in[i*8]
> - =A0 =A0 =A0 =A0 || out[i*8 + 1] !=3D in[i*8 + 1]
> - =A0 =A0 =A0 =A0 || out[i*8 + 2] !=3D in[i*8 + 2]
> - =A0 =A0 =A0 =A0 || out[i*8 + 3] !=3D in[i*8 + 3]
> - =A0 =A0 =A0 =A0 || out[i*8 + 4] !=3D in[i*8 + 4]
> - =A0 =A0 =A0 =A0 || out[i*8 + 5] !=3D in[i*8 + 5]
> - =A0 =A0 =A0 =A0 || out[i*8 + 6] !=3D in[i*8 + 6]
> - =A0 =A0 =A0 =A0 || out[i*8 + 7] !=3D in[i*8 + 7]
> - =A0 =A0 =A0 =A0 || ia[i] !=3D in[i*8 + 2])
> - =A0 =A0 =A0 abort ();
> - =A0 =A0}
> -
> - =A0for (i =3D 0; i < N*2; i++)
> - =A0 =A0{
> - =A0 =A0 =A0a0 =3D in[i*4] + 1;
> - =A0 =A0 =A0a1 =3D in[i*4 + 1] + 2;
> - =A0 =A0 =A0a2 =3D in[i*4 + 2] + 3;
> - =A0 =A0 =A0a3 =3D in[i*4 + 3] + 4;
> -
> - =A0 =A0 =A0out[i*4] =3D a0;
> - =A0 =A0 =A0out[i*4 + 1] =3D a1;
> - =A0 =A0 =A0out[i*4 + 2] =3D a2;
> - =A0 =A0 =A0out[i*4 + 3] =3D a3;
> -
> - =A0 =A0 =A0ia[i] =3D a2;
> - =A0 =A0}
> -
> - =A0/* check results: =A0*/
> - =A0for (i =3D 0; i < N*2; i++)
> - =A0 =A0{
> - =A0 =A0 =A0if (out[i*4] !=3D =A0in[i*4] + 1
> - =A0 =A0 =A0 =A0 || out[i*4 + 1] !=3D in[i*4 + 1] + 2
> - =A0 =A0 =A0 =A0 || out[i*4 + 2] !=3D in[i*4 + 2] + 3
> - =A0 =A0 =A0 =A0 || out[i*4 + 3] !=3D in[i*4 + 3] + 4
> - =A0 =A0 =A0 =A0 || ia[i] !=3D in[i*4 + 2] + 3)
> - =A0 =A0 =A0 =A0abort ();
> - =A0 =A0}
> -
> - =A0/* The last stmt requires interleaving of not power of 2 size - not
> - =A0 =A0 vectorizable. =A0*/
> - =A0for (i =3D 0; i < N/2; i++)
> - =A0 =A0{
> - =A0 =A0 =A0out[i*12] =3D in[i*12];
> - =A0 =A0 =A0out[i*12 + 1] =3D in[i*12 + 1];
> - =A0 =A0 =A0out[i*12 + 2] =3D in[i*12 + 2];
> - =A0 =A0 =A0out[i*12 + 3] =3D in[i*12 + 3];
> - =A0 =A0 =A0out[i*12 + 4] =3D in[i*12 + 4];
> - =A0 =A0 =A0out[i*12 + 5] =3D in[i*12 + 5];
> - =A0 =A0 =A0out[i*12 + 6] =3D in[i*12 + 6];
> - =A0 =A0 =A0out[i*12 + 7] =3D in[i*12 + 7];
> - =A0 =A0 =A0out[i*12 + 8] =3D in[i*12 + 8];
> - =A0 =A0 =A0out[i*12 + 9] =3D in[i*12 + 9];
> - =A0 =A0 =A0out[i*12 + 10] =3D in[i*12 + 10];
> - =A0 =A0 =A0out[i*12 + 11] =3D in[i*12 + 11];
> -
> - =A0 =A0 =A0ia[i] =3D in[i*12 + 7];
> - =A0 =A0}
> -
> - =A0/* check results: =A0*/
> - =A0for (i =3D 0; i < N/2; i++)
> - =A0 =A0{
> - =A0 =A0 =A0if (out[i*12] !=3D =A0in[i*12]
> - =A0 =A0 =A0 =A0 || out[i*12 + 1] !=3D in[i*12 + 1]
> - =A0 =A0 =A0 =A0 || out[i*12 + 2] !=3D in[i*12 + 2]
> - =A0 =A0 =A0 =A0 || out[i*12 + 3] !=3D in[i*12 + 3]
> - =A0 =A0 =A0 =A0 || out[i*12 + 4] !=3D in[i*12 + 4]
> - =A0 =A0 =A0 =A0 || out[i*12 + 5] !=3D in[i*12 + 5]
> - =A0 =A0 =A0 =A0 || out[i*12 + 6] !=3D in[i*12 + 6]
> - =A0 =A0 =A0 =A0 || out[i*12 + 7] !=3D in[i*12 + 7]
> - =A0 =A0 =A0 =A0 || out[i*12 + 8] !=3D in[i*12 + 8]
> - =A0 =A0 =A0 =A0 || out[i*12 + 9] !=3D in[i*12 + 9]
> - =A0 =A0 =A0 =A0 || out[i*12 + 10] !=3D in[i*12 + 10]
> - =A0 =A0 =A0 =A0 || out[i*12 + 11] !=3D in[i*12 + 11]
> - =A0 =A0 =A0 =A0 || ia[i] !=3D in[i*12 + 7])
> - =A0 =A0 =A0 =A0abort ();
> - =A0 =A0}
> -
> - =A0/* Hybrid SLP with unrolling by 2. =A0*/
> - =A0for (i =3D 0; i < N; i++)
> - =A0 =A0{
> - =A0 =A0 =A0out[i*6] =3D in[i*6];
> - =A0 =A0 =A0out[i*6 + 1] =3D in[i*6 + 1];
> - =A0 =A0 =A0out[i*6 + 2] =3D in[i*6 + 2];
> - =A0 =A0 =A0out[i*6 + 3] =3D in[i*6 + 3];
> - =A0 =A0 =A0out[i*6 + 4] =3D in[i*6 + 4];
> - =A0 =A0 =A0out[i*6 + 5] =3D in[i*6 + 5];
> -
> - =A0 =A0 =A0ia[i] =3D i;
> - =A0 =A0}
> -
> - =A0/* check results: =A0*/
> - =A0for (i =3D 0; i < N/2; i++)
> - =A0 =A0{
> - =A0 =A0 =A0if (out[i*6] !=3D =A0in[i*6]
> - =A0 =A0 =A0 =A0 || out[i*6 + 1] !=3D in[i*6 + 1]
> - =A0 =A0 =A0 =A0 || out[i*6 + 2] !=3D in[i*6 + 2]
> - =A0 =A0 =A0 =A0 || out[i*6 + 3] !=3D in[i*6 + 3]
> - =A0 =A0 =A0 =A0 || out[i*6 + 4] !=3D in[i*6 + 4]
> - =A0 =A0 =A0 =A0 || out[i*6 + 5] !=3D in[i*6 + 5]
> - =A0 =A0 =A0 =A0 || ia[i] !=3D i)
> - =A0 =A0 =A0 =A0abort ();
> - =A0 =A0}
> -
> -
> - =A0return 0;
> -}
> -
> -int main (void)
> -{
> - =A0check_vect ();
> -
> - =A0main1 ();
> -
> - =A0return 0;
> -}
> -
> -/* { dg-final { scan-tree-dump-times "vectorized 3 loops" 1 "vect" { tar=
get =A0vect_strided =A0} } } */
> -/* { dg-final { scan-tree-dump-times "vectorized 1 loops" 1 "vect" { tar=
get =A0{ ! { vect_strided } } } } } */
> -/* { dg-final { scan-tree-dump-times "vectorizing stmts using SLP" 3 "ve=
ct" =A0{ target =A0vect_strided =A0} } } */
> -/* { dg-final { scan-tree-dump-times "vectorizing stmts using SLP" 1 "ve=
ct" =A0{ target { ! { vect_strided } } } } } */
> -/* { dg-final { cleanup-tree-dump "vect" } } */
> -
> Index: gcc/testsuite/gcc.dg/vect/slp-19a.c
> =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=
=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=
=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D
> --- /dev/null =A0 2011-03-23 08:42:11.268792848 +0000
> +++ gcc/testsuite/gcc.dg/vect/slp-19a.c 2011-04-12 15:18:25.000000000 +01=
00
> @@ -0,0 +1,61 @@
> +/* { dg-require-effective-target vect_int } */
> +
> +#include <stdarg.h>
> +#include "tree-vect.h"
> +
> +#define N 16
> +
> +int
> +main1 ()
> +{
> + =A0unsigned int i;
> + =A0unsigned int out[N*8];
> + =A0unsigned int in[N*8] =3D {0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,1=
7,18,19,20,21,22,23,24,25,26,27,28,29,30,31,32,33,34,35,36,37,38,39,40,41,4=
2,43,44,45,46,47,48,49,50,51,52,53,54,55,56,57,58,59,60,61,62,63,0,1,2,3,4,=
5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20,21,22,23,24,25,26,27,28,29,30,31=
,32,33,34,35,36,37,38,39,40,41,42,43,44,45,46,47,48,49,50,51,52,53,54,55,56=
,57,58,59,60,61,62,63};
> + =A0unsigned int ia[N*2];
> +
> + =A0for (i =3D 0; i < N; i++)
> + =A0 =A0{
> + =A0 =A0 =A0out[i*8] =3D in[i*8];
> + =A0 =A0 =A0out[i*8 + 1] =3D in[i*8 + 1];
> + =A0 =A0 =A0out[i*8 + 2] =3D in[i*8 + 2];
> + =A0 =A0 =A0out[i*8 + 3] =3D in[i*8 + 3];
> + =A0 =A0 =A0out[i*8 + 4] =3D in[i*8 + 4];
> + =A0 =A0 =A0out[i*8 + 5] =3D in[i*8 + 5];
> + =A0 =A0 =A0out[i*8 + 6] =3D in[i*8 + 6];
> + =A0 =A0 =A0out[i*8 + 7] =3D in[i*8 + 7];
> +
> + =A0 =A0 =A0ia[i] =3D in[i*8 + 2];
> + =A0 =A0}
> +
> + =A0/* check results: =A0*/
> + =A0for (i =3D 0; i < N; i++)
> + =A0 =A0{
> + =A0 =A0 =A0if (out[i*8] !=3D =A0in[i*8]
> + =A0 =A0 =A0 =A0 || out[i*8 + 1] !=3D in[i*8 + 1]
> + =A0 =A0 =A0 =A0 || out[i*8 + 2] !=3D in[i*8 + 2]
> + =A0 =A0 =A0 =A0 || out[i*8 + 3] !=3D in[i*8 + 3]
> + =A0 =A0 =A0 =A0 || out[i*8 + 4] !=3D in[i*8 + 4]
> + =A0 =A0 =A0 =A0 || out[i*8 + 5] !=3D in[i*8 + 5]
> + =A0 =A0 =A0 =A0 || out[i*8 + 6] !=3D in[i*8 + 6]
> + =A0 =A0 =A0 =A0 || out[i*8 + 7] !=3D in[i*8 + 7]
> + =A0 =A0 =A0 =A0 || ia[i] !=3D in[i*8 + 2])
> + =A0 =A0 =A0 abort ();
> + =A0 =A0}
> +
> + =A0return 0;
> +}
> +
> +int main (void)
> +{
> + =A0check_vect ();
> +
> + =A0main1 ();
> +
> + =A0return 0;
> +}
> +
> +/* { dg-final { scan-tree-dump-times "vectorized 1 loops" 1 "vect" { tar=
get vect_strided } } } */
> +/* { dg-final { scan-tree-dump-times "vectorized 0 loops" 1 "vect" { tar=
get { ! vect_strided } } } } */
> +/* { dg-final { scan-tree-dump-times "vectorizing stmts using SLP" 1 "ve=
ct" { target vect_strided } } } */
> +/* { dg-final { scan-tree-dump-times "vectorizing stmts using SLP" 0 "ve=
ct" { target { ! vect_strided } } } } */
> +/* { dg-final { cleanup-tree-dump "vect" } } */
> Index: gcc/testsuite/gcc.dg/vect/slp-19b.c
> =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=
=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=
=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D
> --- /dev/null =A0 2011-03-23 08:42:11.268792848 +0000
> +++ gcc/testsuite/gcc.dg/vect/slp-19b.c 2011-04-12 15:18:25.000000000 +01=
00
> @@ -0,0 +1,58 @@
> +/* { dg-require-effective-target vect_int } */
> +
> +#include <stdarg.h>
> +#include "tree-vect.h"
> +
> +#define N 16
> +
> +int
> +main1 ()
> +{
> + =A0unsigned int i;
> + =A0unsigned int out[N*8];
> + =A0unsigned int in[N*8] =3D {0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,1=
7,18,19,20,21,22,23,24,25,26,27,28,29,30,31,32,33,34,35,36,37,38,39,40,41,4=
2,43,44,45,46,47,48,49,50,51,52,53,54,55,56,57,58,59,60,61,62,63,0,1,2,3,4,=
5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20,21,22,23,24,25,26,27,28,29,30,31=
,32,33,34,35,36,37,38,39,40,41,42,43,44,45,46,47,48,49,50,51,52,53,54,55,56=
,57,58,59,60,61,62,63};
> + =A0unsigned int ia[N*2], a0, a1, a2, a3;
> +
> + =A0for (i =3D 0; i < N*2; i++)
> + =A0 =A0{
> + =A0 =A0 =A0a0 =3D in[i*4] + 1;
> + =A0 =A0 =A0a1 =3D in[i*4 + 1] + 2;
> + =A0 =A0 =A0a2 =3D in[i*4 + 2] + 3;
> + =A0 =A0 =A0a3 =3D in[i*4 + 3] + 4;
> +
> + =A0 =A0 =A0out[i*4] =3D a0;
> + =A0 =A0 =A0out[i*4 + 1] =3D a1;
> + =A0 =A0 =A0out[i*4 + 2] =3D a2;
> + =A0 =A0 =A0out[i*4 + 3] =3D a3;
> +
> + =A0 =A0 =A0ia[i] =3D a2;
> + =A0 =A0}
> +
> + =A0/* check results: =A0*/
> + =A0for (i =3D 0; i < N*2; i++)
> + =A0 =A0{
> + =A0 =A0 =A0if (out[i*4] !=3D =A0in[i*4] + 1
> + =A0 =A0 =A0 =A0 || out[i*4 + 1] !=3D in[i*4 + 1] + 2
> + =A0 =A0 =A0 =A0 || out[i*4 + 2] !=3D in[i*4 + 2] + 3
> + =A0 =A0 =A0 =A0 || out[i*4 + 3] !=3D in[i*4 + 3] + 4
> + =A0 =A0 =A0 =A0 || ia[i] !=3D in[i*4 + 2] + 3)
> + =A0 =A0 =A0 =A0abort ();
> + =A0 =A0}
> +
> + =A0return 0;
> +}
> +
> +int main (void)
> +{
> + =A0check_vect ();
> +
> + =A0main1 ();
> +
> + =A0return 0;
> +}
> +
> +/* { dg-final { scan-tree-dump-times "vectorized 1 loops" 1 "vect" { tar=
get vect_strided } } } */
> +/* { dg-final { scan-tree-dump-times "vectorized 0 loops" 1 "vect" { tar=
get { ! vect_strided } } } } */
> +/* { dg-final { scan-tree-dump-times "vectorizing stmts using SLP" 1 "ve=
ct" { target vect_strided } } } */
> +/* { dg-final { scan-tree-dump-times "vectorizing stmts using SLP" 0 "ve=
ct" { target { ! vect_strided } } } } */
> +/* { dg-final { cleanup-tree-dump "vect" } } */
> Index: gcc/testsuite/gcc.dg/vect/slp-19c.c
> =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=
=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=
=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D
> --- /dev/null =A0 2011-03-23 08:42:11.268792848 +0000
> +++ gcc/testsuite/gcc.dg/vect/slp-19c.c 2011-04-12 15:18:25.000000000 +01=
00
> @@ -0,0 +1,95 @@
> +/* { dg-require-effective-target vect_int } */
> +
> +#include <stdarg.h>
> +#include "tree-vect.h"
> +
> +#define N 16
> +
> +int
> +main1 ()
> +{
> + =A0unsigned int i;
> + =A0unsigned int out[N*8];
> + =A0unsigned int in[N*8] =3D {0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,1=
7,18,19,20,21,22,23,24,25,26,27,28,29,30,31,32,33,34,35,36,37,38,39,40,41,4=
2,43,44,45,46,47,48,49,50,51,52,53,54,55,56,57,58,59,60,61,62,63,0,1,2,3,4,=
5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20,21,22,23,24,25,26,27,28,29,30,31=
,32,33,34,35,36,37,38,39,40,41,42,43,44,45,46,47,48,49,50,51,52,53,54,55,56=
,57,58,59,60,61,62,63};
> + =A0unsigned int ia[N*2], a0, a1, a2, a3;
> +
> + =A0/* The last stmt requires interleaving of not power of 2 size - not
> + =A0 =A0 vectorizable. =A0*/
> + =A0for (i =3D 0; i < N/2; i++)
> + =A0 =A0{
> + =A0 =A0 =A0out[i*12] =3D in[i*12];
> + =A0 =A0 =A0out[i*12 + 1] =3D in[i*12 + 1];
> + =A0 =A0 =A0out[i*12 + 2] =3D in[i*12 + 2];
> + =A0 =A0 =A0out[i*12 + 3] =3D in[i*12 + 3];
> + =A0 =A0 =A0out[i*12 + 4] =3D in[i*12 + 4];
> + =A0 =A0 =A0out[i*12 + 5] =3D in[i*12 + 5];
> + =A0 =A0 =A0out[i*12 + 6] =3D in[i*12 + 6];
> + =A0 =A0 =A0out[i*12 + 7] =3D in[i*12 + 7];
> + =A0 =A0 =A0out[i*12 + 8] =3D in[i*12 + 8];
> + =A0 =A0 =A0out[i*12 + 9] =3D in[i*12 + 9];
> + =A0 =A0 =A0out[i*12 + 10] =3D in[i*12 + 10];
> + =A0 =A0 =A0out[i*12 + 11] =3D in[i*12 + 11];
> +
> + =A0 =A0 =A0ia[i] =3D in[i*12 + 7];
> + =A0 =A0}
> +
> + =A0/* check results: =A0*/
> + =A0for (i =3D 0; i < N/2; i++)
> + =A0 =A0{
> + =A0 =A0 =A0if (out[i*12] !=3D =A0in[i*12]
> + =A0 =A0 =A0 =A0 || out[i*12 + 1] !=3D in[i*12 + 1]
> + =A0 =A0 =A0 =A0 || out[i*12 + 2] !=3D in[i*12 + 2]
> + =A0 =A0 =A0 =A0 || out[i*12 + 3] !=3D in[i*12 + 3]
> + =A0 =A0 =A0 =A0 || out[i*12 + 4] !=3D in[i*12 + 4]
> + =A0 =A0 =A0 =A0 || out[i*12 + 5] !=3D in[i*12 + 5]
> + =A0 =A0 =A0 =A0 || out[i*12 + 6] !=3D in[i*12 + 6]
> + =A0 =A0 =A0 =A0 || out[i*12 + 7] !=3D in[i*12 + 7]
> + =A0 =A0 =A0 =A0 || out[i*12 + 8] !=3D in[i*12 + 8]
> + =A0 =A0 =A0 =A0 || out[i*12 + 9] !=3D in[i*12 + 9]
> + =A0 =A0 =A0 =A0 || out[i*12 + 10] !=3D in[i*12 + 10]
> + =A0 =A0 =A0 =A0 || out[i*12 + 11] !=3D in[i*12 + 11]
> + =A0 =A0 =A0 =A0 || ia[i] !=3D in[i*12 + 7])
> + =A0 =A0 =A0 =A0abort ();
> + =A0 =A0}
> +
> + =A0/* Hybrid SLP with unrolling by 2. =A0*/
> + =A0for (i =3D 0; i < N; i++)
> + =A0 =A0{
> + =A0 =A0 =A0out[i*6] =3D in[i*6];
> + =A0 =A0 =A0out[i*6 + 1] =3D in[i*6 + 1];
> + =A0 =A0 =A0out[i*6 + 2] =3D in[i*6 + 2];
> + =A0 =A0 =A0out[i*6 + 3] =3D in[i*6 + 3];
> + =A0 =A0 =A0out[i*6 + 4] =3D in[i*6 + 4];
> + =A0 =A0 =A0out[i*6 + 5] =3D in[i*6 + 5];
> +
> + =A0 =A0 =A0ia[i] =3D i;
> + =A0 =A0}
> +
> + =A0/* check results: =A0*/
> + =A0for (i =3D 0; i < N/2; i++)
> + =A0 =A0{
> + =A0 =A0 =A0if (out[i*6] !=3D =A0in[i*6]
> + =A0 =A0 =A0 =A0 || out[i*6 + 1] !=3D in[i*6 + 1]
> + =A0 =A0 =A0 =A0 || out[i*6 + 2] !=3D in[i*6 + 2]
> + =A0 =A0 =A0 =A0 || out[i*6 + 3] !=3D in[i*6 + 3]
> + =A0 =A0 =A0 =A0 || out[i*6 + 4] !=3D in[i*6 + 4]
> + =A0 =A0 =A0 =A0 || out[i*6 + 5] !=3D in[i*6 + 5]
> + =A0 =A0 =A0 =A0 || ia[i] !=3D i)
> + =A0 =A0 =A0 =A0abort ();
> + =A0 =A0}
> +
> + =A0return 0;
> +}
> +
> +int main (void)
> +{
> + =A0check_vect ();
> +
> + =A0main1 ();
> +
> + =A0return 0;
> +}
> +
> +/* { dg-final { scan-tree-dump-times "vectorized 1 loops" 1 "vect" } } */
> +/* { dg-final { scan-tree-dump-times "vectorizing stmts using SLP" 1 "ve=
ct" } } */
> +/* { dg-final { cleanup-tree-dump "vect" } } */
>