From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (qmail 29766 invoked by alias); 14 Jun 2011 15:26:31 -0000 Received: (qmail 29751 invoked by uid 22791); 14 Jun 2011 15:26:29 -0000 X-SWARE-Spam-Status: No, hits=-6.5 required=5.0 tests=AWL,BAYES_00,RCVD_IN_DNSWL_HI,SPF_HELO_PASS,T_RP_MATCHES_RCVD X-Spam-Check-By: sourceware.org Received: from mx1.redhat.com (HELO mx1.redhat.com) (209.132.183.28) by sourceware.org (qpsmtpd/0.43rc1) with ESMTP; Tue, 14 Jun 2011 15:26:14 +0000 Received: from int-mx12.intmail.prod.int.phx2.redhat.com (int-mx12.intmail.prod.int.phx2.redhat.com [10.5.11.25]) by mx1.redhat.com (8.14.4/8.14.4) with ESMTP id p5EFQ4DO026191 (version=TLSv1/SSLv3 cipher=DHE-RSA-AES256-SHA bits=256 verify=OK); Tue, 14 Jun 2011 11:26:05 -0400 Received: from tyan-ft48-01.lab.bos.redhat.com (tyan-ft48-01.lab.bos.redhat.com [10.16.42.4]) by int-mx12.intmail.prod.int.phx2.redhat.com (8.14.4/8.14.4) with ESMTP id p5EFQ4tJ019559 (version=TLSv1/SSLv3 cipher=DHE-RSA-AES256-SHA bits=256 verify=NO); Tue, 14 Jun 2011 11:26:04 -0400 Received: from tyan-ft48-01.lab.bos.redhat.com (localhost.localdomain [127.0.0.1]) by tyan-ft48-01.lab.bos.redhat.com (8.14.4/8.14.4) with ESMTP id p5EFQ35S020230; Tue, 14 Jun 2011 17:26:03 +0200 Received: (from jakub@localhost) by tyan-ft48-01.lab.bos.redhat.com (8.14.4/8.14.4/Submit) id p5EFQ3nh020228; Tue, 14 Jun 2011 17:26:03 +0200 Date: Tue, 14 Jun 2011 15:32:00 -0000 From: Jakub Jelinek To: Richard Guenther Cc: Bernd Schmidt , gcc-patches@gcc.gnu.org Subject: Re: RFC: Fix GCSE exp_equiv_p on MEMs with different MEM_ATTRS (PR rtl-optimization/49390) Message-ID: <20110614152602.GC17079@tyan-ft48-01.lab.bos.redhat.com> Reply-To: Jakub Jelinek References: <20110613184028.GW17079@tyan-ft48-01.lab.bos.redhat.com> <4DF72BC4.5050306@codesourcery.com> MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: User-Agent: Mutt/1.5.21 (2010-09-15) X-IsSubscribed: yes Mailing-List: contact gcc-patches-help@gcc.gnu.org; run by ezmlm Precedence: bulk List-Id: List-Archive: List-Post: List-Help: Sender: gcc-patches-owner@gcc.gnu.org X-SW-Source: 2011-06/txt/msg01064.txt.bz2 On Tue, Jun 14, 2011 at 11:49:08AM +0200, Richard Guenther wrote: > So I'd say we revert your patch for now and if somebody feels like > implementing the above ... Ok, here is what I've bootstrapped/regtested on x86_64-linux and i686-linux and committed to trunk and 4.6 branch: 2011-06-14 Jakub Jelinek PR rtl-optimization/49390 Revert: 2010-06-29 Bernd Schmidt * cse.c (exp_equiv_p): For MEMs, if for_gcse, only compare MEM_ALIAS_SET. * gcc.c-torture/execute/pr49390.c: New test. --- gcc/cse.c.jj (revision 161534) +++ gcc/cse.c (revision 161533) @@ -2669,16 +2669,26 @@ case MEM: if (for_gcse) { - /* Can't merge two expressions in different alias sets, since we - can decide that the expression is transparent in a block when - it isn't, due to it being set with the different alias set. */ - if (MEM_ALIAS_SET (x) != MEM_ALIAS_SET (y)) - return 0; - /* A volatile mem should not be considered equivalent to any other. */ if (MEM_VOLATILE_P (x) || MEM_VOLATILE_P (y)) return 0; + + /* Can't merge two expressions in different alias sets, since we + can decide that the expression is transparent in a block when + it isn't, due to it being set with the different alias set. + + Also, can't merge two expressions with different MEM_ATTRS. + They could e.g. be two different entities allocated into the + same space on the stack (see e.g. PR25130). In that case, the + MEM addresses can be the same, even though the two MEMs are + absolutely not equivalent. + + But because really all MEM attributes should be the same for + equivalent MEMs, we just use the invariant that MEMs that have + the same attributes share the same mem_attrs data structure. */ + if (MEM_ATTRS (x) != MEM_ATTRS (y)) + return 0; } break; --- gcc/testsuite/gcc.c-torture/execute/pr49390.c.jj 2011-06-13 17:28:09.000000000 +0200 +++ gcc/testsuite/gcc.c-torture/execute/pr49390.c 2011-06-13 17:27:49.000000000 +0200 @@ -0,0 +1,88 @@ +/* PR rtl-optimization/49390 */ + +struct S { unsigned int s1; unsigned int s2; }; +struct T { unsigned int t1; struct S t2; }; +struct U { unsigned short u1; unsigned short u2; }; +struct V { struct U v1; struct T v2; }; +struct S a; +char *b; +union { char b[64]; struct V v; } u; +volatile int v; +extern void abort (void); + +__attribute__((noinline, noclone)) void +foo (int x, void *y, unsigned int z, unsigned int w) +{ + if (x != 4 || y != (void *) &u.v.v2) + abort (); + v = z + w; + v = 16384; +} + +__attribute__((noinline, noclone)) void +bar (struct S x) +{ + v = x.s1; + v = x.s2; +} + +__attribute__((noinline, noclone)) int +baz (struct S *x) +{ + v = x->s1; + v = x->s2; + v = 0; + return v + 1; +} + +__attribute__((noinline, noclone)) void +test (struct S *c) +{ + struct T *d; + struct S e = a; + unsigned int f, g; + if (c == 0) + c = &e; + else + { + if (c->s2 % 8192 <= 15 || (8192 - c->s2 % 8192) <= 31) + foo (1, 0, c->s1, c->s2); + } + if (!baz (c)) + return; + g = (((struct U *) b)->u2 & 2) ? 32 : __builtin_offsetof (struct V, v2); + f = c->s2 % 8192; + if (f == 0) + { + e.s2 += g; + f = g; + } + else if (f < g) + { + foo (2, 0, c->s1, c->s2); + return; + } + if ((((struct U *) b)->u2 & 1) && f == g) + { + bar (*c); + foo (3, 0, c->s1, c->s2); + return; + } + d = (struct T *) (b + c->s2 % 8192); + if (d->t2.s1 >= c->s1 && (d->t2.s1 != c->s1 || d->t2.s2 >= c->s2)) + foo (4, d, c->s1, c->s2); + return; +} + +int +main () +{ + struct S *c = 0; + asm ("" : "+r" (c) : "r" (&a)); + u.v.v2.t2.s1 = 8192; + b = u.b; + test (c); + if (v != 16384) + abort (); + return 0; +} Jakub