[Bug libc/789] New: HP_TIMING_ACCUM broken on ia32 ?

[Bug libc/789] New: HP_TIMING_ACCUM broken on ia32 ?

[castet dot matthieu at free dot fr]

Hi,

I am trying to use the hp-timing macro for some benchmark, but the
HP_TIMING_ACCUM seem broken on ia32.


When compiling, gcc give me that error :

error: impossible register constraint in `asm'

Somebody have an idea how to correct it ?

Thanks

Matthieu

PS : I am using gcc 3.3.5 (Debian 1:3.3.5-8) and 2.3.4.

-- 
           Summary: HP_TIMING_ACCUM broken on ia32 ?
           Product: glibc
           Version: 2.3.4
            Status: NEW
          Severity: normal
          Priority: P2
         Component: libc
        AssignedTo: gotom at debian dot or dot jp
        ReportedBy: castet dot matthieu at free dot fr
                CC: glibc-bugs at sources dot redhat dot com


http://sources.redhat.com/bugzilla/show_bug.cgi?id=789

------- You are receiving this mail because: -------
You are on the CC list for the bug, or are watching someone who is.

[Bug libc/789] HP_TIMING_ACCUM broken on ia32 ?

[gotom at debian dot or dot jp]

------- Additional Comments From gotom at debian dot or dot jp  2005-03-15 00:30 -------
Do you have a short sample code to show us your problem?


-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |WAITING


http://sources.redhat.com/bugzilla/show_bug.cgi?id=789

------- You are receiving this mail because: -------
You are on the CC list for the bug, or are watching someone who is.

[Bug libc/789] HP_TIMING_ACCUM broken on ia32 ?

[jakub at redhat dot com]

------- Additional Comments From jakub at redhat dot com  2005-03-15 07:19 -------
Does the http://sources.redhat.com/ml/libc-alpha/2005-03/msg00053.html
patch help?
FYI, HP_TIMING_ACCUM is never used by glibc, so it might as well be simply nuked.
You can use HP_TIMING_ACCUM_NT...

-- 


http://sources.redhat.com/bugzilla/show_bug.cgi?id=789

------- You are receiving this mail because: -------
You are on the CC list for the bug, or are watching someone who is.

[Bug libc/789] HP_TIMING_ACCUM broken on ia32 ?

[castet dot matthieu at free dot fr]

------- Additional Comments From castet dot matthieu at free dot fr  2005-03-15 08:28 -------
http://sources.redhat.com/ml/libc-alpha/2005-03/msg00053.html doesn't help.
I can't use HP_TIMING_ACCUM_NT, because I use it in thread environement.
Actually I use atomic operation on 32 bits variable because my diff is enough
small :

old_time = atomic_exchange_and_add(&buf->stat_time_low, (unsigned int)(diff));
if ((old_time + (unsigned int)(diff)) < old_time)
   atomic_increment(&buf->stat_time_high);

-- 


http://sources.redhat.com/bugzilla/show_bug.cgi?id=789

------- You are receiving this mail because: -------
You are on the CC list for the bug, or are watching someone who is.

[Bug libc/789] HP_TIMING_ACCUM broken on ia32 ?

[castet dot matthieu at free dot fr]

[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #1: Type: text/plain, Size: 1967 bytes --]


------- Additional Comments From castet dot matthieu at free dot fr  2005-03-15 08:35 -------
Created an attachment (id=434)
 --> (http://sources.redhat.com/bugzilla/attachment.cgi?id=434&action=view)
testcase

in src/nptl :

$ gcc /tmp/test.c -std=gnu99 -O2 -Wall -Winline -Wstrict-prototypes
-Wwrite-strings -g -mpreferred-stack-boundary=2     -I../include -I.
-I/home/castetm/trace/contrib/libc_compil/build/nptl -I.. -I../libio 
-I/home/castetm/trace/contrib/libc_compil/build -I../sysdeps/i386/elf
-I../nptl/sysdeps/unix/sysv/linux/i386/i686
-I../nptl/sysdeps/unix/sysv/linux/i386 -I../nptl/sysdeps/unix/sysv/linux
-I../nptl/sysdeps/pthread -I../sysdeps/pthread -I../nptl/sysdeps/unix/sysv
-I../nptl/sysdeps/unix -I../nptl/sysdeps/i386/i686 -I../nptl/sysdeps/i386
-I../sysdeps/unix/sysv/linux/i386 -I../sysdeps/unix/sysv/linux -I../sysdeps/gnu
-I../sysdeps/unix/common -I../sysdeps/unix/mman -I../sysdeps/unix/inet
-I../sysdeps/unix/sysv/i386 -I../sysdeps/unix/sysv -I../sysdeps/unix/i386
-I../sysdeps/unix -I../sysdeps/posix -I../sysdeps/i386/i686/fpu
-I../sysdeps/i386/i686 -I../sysdeps/i386/i486 -I../nptl/sysdeps/i386/i486
-I../sysdeps/i386/fpu -I../sysdeps/i386 -I../sysdeps/wordsize-32
-I../sysdeps/ieee754/ldbl-96 -I../sysdeps/ieee754/dbl-64
-I../sysdeps/ieee754/flt-32 -I../sysdeps/ieee754 -I../sysdeps/generic/elf
-I../sysdeps/generic  -D_LIBC_REENTRANT -D_LIBC_REENTRANT -include
../include/libc-symbols.h   -DNOT_IN_libc=1 -DIS_IN_libpthread=1 -o /tmp/test
/tmp/test.c:4: attention : function declaration isn't a prototype
/tmp/test.c: Dans la fonction « main »:
/tmp/test.c:20: attention : implicit declaration of function `printf'
/tmp/test.c:18: error: impossible register constraint in `asm'
/tmp/test.c:21: embrouillé par les erreurs précédentes, abandon

-- 


http://sources.redhat.com/bugzilla/show_bug.cgi?id=789

------- You are receiving this mail because: -------
You are on the CC list for the bug, or are watching someone who is.

[Bug libc/789] HP_TIMING_ACCUM broken on ia32 ?

[jakub at redhat dot com]

------- Additional Comments From jakub at redhat dot com  2005-03-15 10:15 -------
http://sources.redhat.com/ml/libc-alpha/2005-03/msg00056.html

BTW, your testcase is broken, 1) you have not initialized stat_time
2) you only do HP_TIMING_NOW (end); in the cycle, so you count first cycle
10 times, second one 9 times and so on.

-- 


http://sources.redhat.com/bugzilla/show_bug.cgi?id=789

------- You are receiving this mail because: -------
You are on the CC list for the bug, or are watching someone who is.

[Bug libc/789] HP_TIMING_ACCUM broken on ia32 ?

[castet dot matthieu at free dot fr]

------- Additional Comments From castet dot matthieu at free dot fr  2005-03-15 14:27 -------
Now the testcase compile (my file still don't compile because of missing 
register in class  Q_REGS).

But if you run the testcase, it loop forever :(.

-- 


http://sources.redhat.com/bugzilla/show_bug.cgi?id=789

------- You are receiving this mail because: -------
You are on the CC list for the bug, or are watching someone who is.

[Bug libc/789] HP_TIMING_ACCUM broken on ia32 ?

[castet dot matthieu at free dot fr]

------- Additional Comments From castet dot matthieu at free dot fr  2005-03-15 14:34 -------
Created an attachment (id=435)
 --> (http://sources.redhat.com/bugzilla/attachment.cgi?id=435&action=view)
corrected testcase

There something funny using HP_TIMING_ACCUM_NT : depending if stat_time is
local or global, the number of clock change : it takes 298 for local vs 711 for
global...

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
 Attachment #434 is|0                           |1
           obsolete|                            |


http://sources.redhat.com/bugzilla/show_bug.cgi?id=789

------- You are receiving this mail because: -------
You are on the CC list for the bug, or are watching someone who is.

[Bug libc/789] HP_TIMING_ACCUM broken on ia32 ?

[cvs-commit at gcc dot gnu dot org]

------- Additional Comments From cvs-commit at gcc dot gnu dot org  2005-03-15 22:31 -------
Subject: Bug 789

CVSROOT:	/cvs/glibc
Module name:	libc
Changes by:	roland@sources.redhat.com	2005-03-15 22:31:22

Modified files:
	sysdeps/i386/i686: hp-timing.h 

Log message:
	2005-03-15  Jakub Jelinek  <jakub@redhat.com>
	
	[BZ #789]
	* sysdeps/i386/i686/hp-timing.h (HP_TIMING_ACCUM): Fix asm constraints.
	Remove memory clobber.

Patches:
http://sources.redhat.com/cgi-bin/cvsweb.cgi/libc/sysdeps/i386/i686/hp-timing.h.diff?cvsroot=glibc&r1=1.6&r2=1.7



-- 


http://sources.redhat.com/bugzilla/show_bug.cgi?id=789

------- You are receiving this mail because: -------
You are on the CC list for the bug, or are watching someone who is.

[Bug libc/789] HP_TIMING_ACCUM broken on ia32 ?

[castet dot matthieu at free dot fr]

------- Additional Comments From castet dot matthieu at free dot fr  2005-03-16 09:14 -------
thanks, it works.

Now there something wrong with dl_hp_timing_overhead : it is used in
HP_TIMING_ACCUM, but not in HP_TIMING_ACCUM_NT. Also it is initialised in
HP_TIMING_DIFF_INIT, so we could think it is used in HP_TIMING_DIFF...

PS : __temp1 is a useless var

-- 


http://sources.redhat.com/bugzilla/show_bug.cgi?id=789

------- You are receiving this mail because: -------
You are on the CC list for the bug, or are watching someone who is.

[Bug libc/789] HP_TIMING_ACCUM broken on ia32 ?

[castet dot matthieu at free dot fr]

------- Additional Comments From castet dot matthieu at free dot fr  2005-03-22 12:09 -------
it is still broken because it make no difference between __PIC__ and no __PIC__.

I think __arch_compare_and_exchange_val_64_acq should be fixed and we should use it.

-- 


http://sources.redhat.com/bugzilla/show_bug.cgi?id=789

------- You are receiving this mail because: -------
You are on the CC list for the bug, or are watching someone who is.

[Bug libc/789] HP_TIMING_ACCUM broken on ia32 ?

[drepper at redhat dot com]

------- Additional Comments From drepper at redhat dot com  2005-09-23 19:10 -------
This is all a non-issue.  The code isn't used.

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|WAITING                     |RESOLVED
         Resolution|                            |WORKSFORME


http://sourceware.org/bugzilla/show_bug.cgi?id=789

------- You are receiving this mail because: -------
You are on the CC list for the bug, or are watching someone who is.

[Bug libc/789] HP_TIMING_ACCUM broken on ia32 ?

[castet dot matthieu at free dot fr]

------- Additional Comments From castet dot matthieu at free dot fr  2005-09-23 20:16 -------
So why are not removed it ?

-- 


http://sourceware.org/bugzilla/show_bug.cgi?id=789

------- You are receiving this mail because: -------
You are on the CC list for the bug, or are watching someone who is.