From mboxrd@z Thu Jan  1 00:00:00 1970
Return-Path: <libc-hacker-return-9532-listarch-libc-hacker=sources.redhat.com@sourceware.org>
Received: (qmail 12650 invoked by alias); 1 Sep 2009 15:54:10 -0000
Received: (qmail 12631 invoked by uid 22791); 1 Sep 2009 15:54:09 -0000
X-SWARE-Spam-Status: No, hits=-2.3 required=5.0
	tests=AWL,BAYES_00
X-Spam-Check-By: sourceware.org
Received: from sunsite.ms.mff.cuni.cz (HELO sunsite.mff.cuni.cz) (195.113.15.26)
    by sourceware.org (qpsmtpd/0.43rc1) with ESMTP; Tue, 01 Sep 2009 15:54:02 +0000
Received: from sunsite.mff.cuni.cz (localhost [127.0.0.1])
	by sunsite.mff.cuni.cz (8.14.3/8.14.3) with ESMTP id n81FTYqi027346;
	Tue, 1 Sep 2009 17:29:34 +0200
Received: (from jakub@localhost)
	by sunsite.mff.cuni.cz (8.14.3/8.14.3/Submit) id n81FTY0x027345;
	Tue, 1 Sep 2009 17:29:34 +0200
Date: Tue, 01 Sep 2009 15:54:00 -0000
From: Jakub Jelinek <jakub@redhat.com>
To: Ulrich Drepper <drepper@redhat.com>
Cc: Glibc hackers <libc-hacker@sources.redhat.com>
Subject: [PATCH] Fix x86_64 bits/mathinline.h for -m32 compilation
Message-ID: <20090901152934.GB3611@sunsite.ms.mff.cuni.cz>
Reply-To: Jakub Jelinek <jakub@redhat.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
User-Agent: Mutt/1.5.19 (2009-01-05)
Mailing-List: contact libc-hacker-help@sourceware.org; run by ezmlm
Precedence: bulk
List-Id: <libc-hacker.sourceware.org>
List-Subscribe: <mailto:libc-hacker-subscribe@sourceware.org>
List-Archive: <http://sourceware.org/ml/libc-hacker/>
List-Post: <mailto:libc-hacker@sourceware.org>
List-Help: <mailto:libc-hacker-help@sourceware.org>, <http://sourceware.org/ml/#faqs>
Sender: libc-hacker-owner@sourceware.org
X-SW-Source: 2009-09/txt/msg00000.txt.bz2

Hi!

With the latest bits/mathinline.h,
echo '#include <math.h>' | gcc -S -xc - -o - -std=c99 -O2 -m32
doesn't compile:
/usr/include/bits/mathinline.h:39: error: impossible constraint in 'asm'
/usr/include/bits/mathinline.h:46: error: impossible constraint in 'asm'
The problem is that SSE{,2} assembly and "x" constraint is used even in
32-bit code.

Here is an untested fix:

2009-09-01  Jakub Jelinek  <jakub@redhat.com>

	* sysdeps/x86_64/fpu/bits/mathinline.h: Include bits/wordsize.h.
	(__signbitf, __signbit): Only use SSE inline asm for 64-bit.

diff --git a/sysdeps/x86_64/fpu/bits/mathinline.h b/sysdeps/x86_64/fpu/bits/mathinline.h
index ece0f02..dc58f67 100644
--- a/sysdeps/x86_64/fpu/bits/mathinline.h
+++ b/sysdeps/x86_64/fpu/bits/mathinline.h
@@ -22,6 +22,8 @@
 # error "Never use <bits/mathinline.h> directly; include <math.h> instead."
 #endif
 
+#include <bits/wordsize.h>
+
 #ifndef __extern_inline
 # define __MATH_INLINE __inline
 #else
@@ -35,16 +37,26 @@
 __MATH_INLINE int
 __NTH (__signbitf (float __x))
 {
+#if __WORDSIZE == 32
+  __extension__ union { float __f; int __i; } __u = { __f: __x };
+  return __u.__i < 0;
+#else
   int __m;
   __asm ("pmovmskb %1, %0" : "=r" (__m) : "x" (__x));
   return __m & 0x8;
+#endif
 }
 __MATH_INLINE int
 __NTH (__signbit (double __x))
 {
+#if __WORDSIZE == 32
+  __extension__ union { double __d; int __i[2]; } __u = { __d: __x };
+  return __u.__i[1] < 0;
+#else
   int __m;
   __asm ("pmovmskb %1, %0" : "=r" (__m) : "x" (__x));
   return __m & 0x80;
+#endif
 }
 __MATH_INLINE int
 __NTH (__signbitl (long double __x))

	Jakub