[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [Intel Compiler] Optimisation flag does not like __va_arg_pack / un
From: |
Jaroslav Hajek |
Subject: |
Re: [Intel Compiler] Optimisation flag does not like __va_arg_pack / undefined reference to __builtin_is(finite|nan|inf) |
Date: |
Wed, 21 Apr 2010 06:51:43 +0200 |
On Tue, Apr 20, 2010 at 3:07 PM, Vincent Letocart
<address@hidden> wrote:
>
>
> I'm struggling to get octave compiled with the Intel Compiler
> Suite (11.1-069) on a Linux SLES 11.0.
>
> Here is the context:
>
> > uname -a
> Linux tethys 2.6.27.45-0.1-default #1 SMP 2010-02-22 16:49:47 +0100
> x86_64 x86_64 x86_64 GNU/Linux
>
> > icc --version
> icc (ICC) 11.1 20100203
> Copyright (C) 1985-2010 Intel Corporation. All rights reserved.
>
> > ifort --version
> ifort (IFORT) 11.1 20100203
> Copyright (C) 1985-2010 Intel Corporation. All rights reserved.
>
> > icpc --version
> icpc (ICC) 11.1 20100203
> Copyright (C) 1985-2010 Intel Corporation. All rights reserved.
>
> > gcc --version
> gcc (SUSE Linux) 4.3.2 [gcc-4_3-branch revision 141291]
> Copyright (C) 2008 Free Software Foundation, Inc.
> This is free software; see the source for copying conditions. There
> is NO
> warranty; not even for MERCHANTABILITY or FITNESS FOR A PARTICULAR
> PURPOSE.
>
>
> I have actually two sort of errors. I also chosed for reproducing the errors
> with a very minimal
> set of flags activated:
>
> ./configure CPP=cpp CC=icc F77=ifort CXX=icpc CPPFLAGS='-w' CFLAGS='-w'
> FFLAGS='-w' CXXFLAGS='-w'
> then
> make [or make -j 8, whatever]
>
>
>
>
> An error is related to something called __va_arg_pack. Executing the command
> by hand reproduces the
> error [here for instance, with octave-3.2.4]:
>
> src> icpc -c -w -I/usr/include/freetype2 -fPIC -I. -I..
> -I../liboctave -I../src -I../libcruft/misc -DHAVE_CONFIG_H -Wall -W -Wshadow
> -Wold-style-cast -Wformat -w -I/usr/include/freetype2 -pthread
> -fmessage-length=0 -O2 -Wall -D_FORTIFY_SOURCE=2 -fstack-protector
> -funwind-tables -fasynchronous-unwind-tables -g -fno-strict-aliasing
> -D_THREAD_SAFE -D_REENTRANT ./DLD-FUNCTIONS/fltk_backend.cc -o
> pic/fltk_backend.o
> /usr/include/bits/wchar2.h(290): error: identifier
> "__builtin_va_arg_pack" is undefined
> __fmt, __va_arg_pack ());
> ^
> /usr/include/bits/wchar2.h(291): error: identifier
> "__builtin_va_arg_pack" is undefined
> return __swprintf_alias (__s, __n, __fmt, __va_arg_pack ());
> ^
> /usr/include/bits/wchar2.h(340): error: identifier
> "__builtin_va_arg_pack" is undefined
> return __wprintf_chk (__USE_FORTIFY_LEVEL - 1, __fmt,
> __va_arg_pack ());
> ^
> /usr/include/bits/wchar2.h(347): error: identifier
> "__builtin_va_arg_pack" is undefined
> __va_arg_pack ());
> ^
> /usr/include/bits/stdio2.h(35): error: identifier
> "__builtin_va_arg_pack" is undefined
> __bos (__s), __fmt, __va_arg_pack
> ());
> ^
> /usr/include/bits/stdio2.h(66): error: identifier
> "__builtin_va_arg_pack" is undefined
> __bos (__s), __fmt, __va_arg_pack
> ());
> ^
> /usr/include/bits/stdio2.h(99): error: identifier
> "__builtin_va_arg_pack" is undefined
> __va_arg_pack ());
> ^
> /usr/include/bits/stdio2.h(105): error: identifier
> "__builtin_va_arg_pack" is undefined
> return __printf_chk (__USE_FORTIFY_LEVEL - 1, __fmt, __va_arg_pack
> ());
> ^
> /usr/include/bits/stdio2.h(159): error: identifier
> "__builtin_va_arg_pack" is undefined
> __va_arg_pack ());
> ^
> /usr/include/bits/stdio2.h(167): error: identifier
> "__builtin_va_arg_pack" is undefined
> __va_arg_pack ());
> ^
> /usr/include/bits/stdio2.h(174): error: identifier
> "__builtin_va_arg_pack" is undefined
> __va_arg_pack ());
> ^
> /usr/include/bits/stdio2.h(182): error: identifier
> "__builtin_va_arg_pack" is undefined
> __va_arg_pack ());
> ^
> compilation aborted for ./DLD-FUNCTIONS/fltk_backend.cc (code 2)
>
>
> Strangely, replacing '-O2' by '-O0' in the command line above causes the
> compilation of that
> file to succeed.
>
> Another error involves some of the "builtin functions"" coming with the last
> gcc versions.
> Again the octave version does not matter: 3.2.2, 3.2.4 or 3.3.51, I see no
> way to
> bypass the problem.
>
>
> icpc -w -I/usr/include/freetype2 -I. -I.. -I../liboctave -I../src
> -I../libcruft/misc -DHAVE_CONFIG_H -Wall -W -Wshadow -Wold-style-cast
> -Wformat -w -I/usr/include/freetype2 -pthread -rdynamic \
> -L.. -fPIC -o octave \
> main.o \
> -L../liboctave -L../libcruft -L../src -Wl,-rpath
> -Wl,/usr/local/lib/octave-3.2.4 \
> -loctinterp -loctave -lcruft \
> \
> \
> -lfftw3 -lfftw3f -L/usr/X11R6/lib -lGL -lGLU \
> -lX11 -lreadline -lncurses -ldl -lz -lm -lfreetype -lz
> -L/usr/X11R6/lib -lGL -lGLU -lz -L/opt/intel/Compiler/11.1/069/lib/intel64
> -limf -lm -lfreetype -L/usr/X11R6/lib -lGL -lGLU
> -L/opt/intel/Compiler/11.1/069/mkl/lib/em64t -L/usr/local/lib64
> -L/usr/local/lib64/../lib64 -L/usr/lib64/gcc/x86_64-suse-linux/4.3
> -L/usr/lib64/gcc/x86_64-suse-linux/4.3/../../../../lib64 -L/lib/../lib64
> -L/usr/lib/../lib64
> -L/usr/lib64/gcc/x86_64-suse-linux/4.3/../../../../x86_64-suse-linux/lib
> -L/usr/lib64/gcc/x86_64-suse-linux/4.3/../../.. -L/lib64 -L/lib -L/usr/lib64
> -L/usr/lib -lifport -lifcore -lsvml -lipgo -lirc -lpthread -lirc_s -ldl
> ../src/liboctinterp.so: undefined reference to `__builtin_isfinite'
> ../src/liboctinterp.so: undefined reference to `__builtin_isnan'
> ../src/liboctinterp.so: undefined reference to `__builtin_isinf'
> make[2]: *** [octave] Error 1
> make[2]: Leaving directory `/usr/local/src/octave-3.2.4/src'
>
>
> In the list of alternatives, I made an attempt to set optimization flag to
> -O0, an attempt
> to redefine those __builtin* symbols at link time, but I got no optimistic
> result (other
> symbols not found, or an usable octave binary).
>
>
>
> Does anyone know a way to avoid those problems ? Magic flag, or even some
> substitution
> to apply to the configure/Makefile ? What other information should be useful ?
>
>
> Thanks for your time
>
> VincneT.
>
> --
Intel C++ is designed to work with gcc's headers, so it's supposed to
either recognize gcc's builtins, or use a replacement header. So I
think it's a problem in Intel C++. A quick search turned this up:
http://software.intel.com/en-us/forums/showthread.php?t=64188
it seems to confirm it's an Intel C++ bug. However, there's no
definite hint of a working version.
The only advice I can give you is to compile gcc 4.1.x, that works for
me with Intel C++ 11.0.
hth
--
RNDr. Jaroslav Hajek, PhD
computing expert & GNU Octave developer
Aeronautical Research and Test Institute (VZLU)
Prague, Czech Republic
url: www.highegg.matfyz.cz