bug-coreutils
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

bug#13118: [PATCH] factor: disable longlong.h logic for x32 builds


From: Pádraig Brady
Subject: bug#13118: [PATCH] factor: disable longlong.h logic for x32 builds
Date: Thu, 20 Dec 2012 02:29:50 +0000
User-agent: Mozilla/5.0 (X11; Linux x86_64; rv:13.0) Gecko/20120615 Thunderbird/13.0.1

On 12/07/2012 08:55 PM, Pádraig Brady wrote:
On 12/07/2012 08:44 PM, Mike Frysinger wrote:
The current x86_64 asm code does not work for x32 ABIs, so disable it
until someone can fix it.  Simply deleting the q suffix is not enough.

* src/longlong.h: Check for __ILP32__ for x86_64 targets.
---
  src/longlong.h | 2 +-
  1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/src/longlong.h b/src/longlong.h
index f2b2c49..baf001c 100644
--- a/src/longlong.h
+++ b/src/longlong.h
@@ -1006,7 +1006,7 @@ extern UWtype __MPN(udiv_qrnnd) (UWtype *, UWtype, 
UWtype, UWtype);
  #endif
  #endif /* 80x86 */

-#if defined (__amd64__) && W_TYPE_SIZE == 64
+#if defined (__amd64__) && W_TYPE_SIZE == 64 && !defined (__ILP32__)
  #define add_ssaaaa(sh, sl, ah, al, bh, bl) \
    __asm__ ("addq %5,%q1\n\tadcq %3,%q0"                    \
         : "=r" (sh), "=&r" (sl)                    \


I think we may go for something more general here,
as we're having all sorts of problems in this area.
http://git.sv.gnu.org/gitweb/?p=coreutils.git;a=history;f=src/longlong.h

It seems that we need _LP64 defined,
to enable W_TYPE_SIZE == 64 code safely.

I've not come up with anything more generic for the moment,
so will go with this slightly adjusted patch for now.
Was this a runtime or build time issue BTW?
What was the failure mode if runtime?

thanks,
Pádraig.

commit 2178e2ad8dce68862ea9545328aae999700c54cc
Author: Mike Frysinger <address@hidden>
Date:   Fri Dec 7 15:44:18 2012 -0500

    factor: disable x86_64 assembly code for x32 builds

    The current x86_64 asm code does not work for x32 (__ILP32__) ABIs,
    so disable it.  Note simply deleting the q suffix is not enough.

    * src/longlong.h: Restrict x86_64 assembly to _LP64 targets,
    which is consistent with other checks in longlong.h and
    avoids this code on x32.

diff --git a/src/longlong.h b/src/longlong.h
index f2b2c49..1792045 100644
--- a/src/longlong.h
+++ b/src/longlong.h
@@ -683,7 +683,7 @@ extern UWtype __MPN(udiv_qrnnd) (UWtype *, UWtype, UWtype, 
UWtype);
   && defined (_PA_RISC2_0) && defined (_LP64)
 /* Note the _PA_RISC2_0 above is to exclude this code from GCC with
    default -march options which doesn't support these instructions.
-   Also the width check for 'long' is to avoid ilp32 runtimes where
+   Also the width check for 'long' is to avoid ILP32 runtimes where
    GNU/Linux and narrow HP-UX kernels are known to have issues with
    clobbering of context between the add and add,dc instructions.  */
 #define add_ssaaaa(sh, sl, ah, al, bh, bl) \
@@ -1006,7 +1006,9 @@ extern UWtype __MPN(udiv_qrnnd) (UWtype *, UWtype, 
UWtype, UWtype);
 #endif
 #endif /* 80x86 */

-#if defined (__amd64__) && W_TYPE_SIZE == 64
+#if defined (__amd64__) && W_TYPE_SIZE == 64 && defined (_LP64)
+/* Note the width check for 'long' is to avoid ILP32 runtimes (x32)
+   where the ABI is known to be incompatible with the following.  */
 #define add_ssaaaa(sh, sl, ah, al, bh, bl) \
   __asm__ ("addq %5,%q1\n\tadcq %3,%q0"                                        
\
           : "=r" (sh), "=&r" (sl)                                      \






reply via email to

[Prev in Thread] Current Thread [Next in Thread]