From patchwork Sat Aug 7 15:52:23 2010 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 8bit X-Patchwork-Submitter: "H.J. Lu" X-Patchwork-Id: 61182 Return-Path: X-Original-To: incoming@patchwork.ozlabs.org Delivered-To: patchwork-incoming@bilbo.ozlabs.org Received: from sourceware.org (server1.sourceware.org [209.132.180.131]) by ozlabs.org (Postfix) with SMTP id 49D27B6F0D for ; Sun, 8 Aug 2010 01:52:34 +1000 (EST) Received: (qmail 2763 invoked by alias); 7 Aug 2010 15:52:32 -0000 Received: (qmail 2755 invoked by uid 22791); 7 Aug 2010 15:52:31 -0000 X-SWARE-Spam-Status: No, hits=-1.8 required=5.0 tests=AWL, BAYES_00, DKIM_SIGNED, DKIM_VALID, DKIM_VALID_AU, FREEMAIL_FROM, RCVD_IN_DNSWL_NONE, TW_ZJ X-Spam-Check-By: sourceware.org Received: from mail-qw0-f47.google.com (HELO mail-qw0-f47.google.com) (209.85.216.47) by sourceware.org (qpsmtpd/0.43rc1) with ESMTP; Sat, 07 Aug 2010 15:52:26 +0000 Received: by qwg8 with SMTP id 8so6198772qwg.20 for ; Sat, 07 Aug 2010 08:52:23 -0700 (PDT) MIME-Version: 1.0 Received: by 10.220.60.204 with SMTP id q12mr9050284vch.41.1281196343104; Sat, 07 Aug 2010 08:52:23 -0700 (PDT) Received: by 10.220.195.133 with HTTP; Sat, 7 Aug 2010 08:52:23 -0700 (PDT) In-Reply-To: References: Date: Sat, 7 Aug 2010 08:52:23 -0700 Message-ID: Subject: Re: PATCH: PR target/45213: "suffix or operands invalid for `push'" triggered by optimisations on x86_64 From: "H.J. Lu" To: Uros Bizjak Cc: gcc-patches@gcc.gnu.org X-IsSubscribed: yes Mailing-List: contact gcc-patches-help@gcc.gnu.org; run by ezmlm Precedence: bulk List-Id: List-Unsubscribe: List-Archive: List-Post: List-Help: Sender: gcc-patches-owner@gcc.gnu.org Delivered-To: mailing list gcc-patches@gcc.gnu.org On Sat, Aug 7, 2010 at 3:45 AM, Uros Bizjak wrote: > On Sat, Aug 7, 2010 at 12:11 PM, Uros Bizjak wrote: > >>> "pushq $imm32S" only takes 32bit signed extended immediate. You can't push >>> 0xbf800000. Instead, you push -1082130432 or 0xffffffffbf800000.  This >>> patch makes it signed.  OK for trunk/4.5/4.4? >> >> No, see the comment in real.h: >> >> /* IN is a REAL_VALUE_TYPE.  OUT is a long.  */ >> #define REAL_VALUE_TO_TARGET_SINGLE(IN, OUT) \ >>  ((OUT) = real_to_target (NULL, &(IN), mode_for_size (32, MODE_FLOAT, 0))) > > IMO, this is correct patch, to also generate correct extension on ILP32 hosts. > > Index: i386.c > =================================================================== > --- i386.c      (revision 162975) > +++ i386.c      (working copy) > @@ -12921,7 +12921,7 @@ > >       if (ASSEMBLER_DIALECT == ASM_ATT) >        putc ('$', file); > -      fprintf (file, "0x%08lx", (long unsigned int) l); > +      fprintf (file, "0x%08llx", (unsigned long long) (int) l); >     } > >   /* These float cases don't actually occur as immediate operands.  */ > SF is 4 bytes, the same as SI. pushq takes 32bit signed extended immediate. It is safe for pushq since the upper 4 bytes, which are all 1s, are unused. For everything else, we use 32bit unsigned int. We don't want movl $0xffffffffbf800000, (%rsp) How does this patch look? diff --git a/gcc/config/i386/i386.c b/gcc/config/i386/i386.c index 204211a..4914299 100644 --- a/gcc/config/i386/i386.c +++ b/gcc/config/i386/i386.c @@ -12921,7 +12921,12 @@ ix86_print_operand (FILE *file, rtx x, int code) if (ASSEMBLER_DIALECT == ASM_ATT) putc ('$', file); - fprintf (file, "0x%08lx", (long unsigned int) l); + /* 'q' is only used on pushq which takes 32bit signed extended + immediate. */ + if (code == 'q') + fprintf (file, "%d", (int) l); + else + fprintf (file, "0x%08x", (unsigned int) l); } /* These float cases don't actually occur as immediate operands. */