Hello Elena,
I tried your suggestions, but with no success.
Because I'm using a virtual machine, I think that maybe my system has some
kind of problem, because I started to have problems also with gnu compilers
(gcc, g++ and gfortran).
I changed from FEDORA 19 to Scientific Linux 6.4 and now I can build both
with gnu and intel compilers.
I'll try again with FEDORA 19 in the near future and if I found the
problem or the solution ( 
) then I'll put a new comment here.
Thanks!
Eduardo Jauch.
Em 01-11-2013 14:29, Elena Pourmal escreveu:
Eduardo,
Could you please try to disable shared libraries? Please use
--disable-shared configure flag when configuring the library. I would also
suggest that you try 1.8.12-pre2
ftp://ftp.hdfgroup.uiuc.edu/pub/outgoing/hdf5/hdf5-1.8.12/pre2
Unfortunately we do not have Intel Parallel Studio here, and I cannot
reproduce the problem.
Thank you!
Elena
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Elena Pourmal The HDF Group http://hdfgroup.org
1800 So. Oak St., Suite 203, Champaign IL 61820
217.531.6112
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
On Nov 1, 2013, at 6:37 AM, Eduardo Jauch <eduardo.jauch@ist.utl.pt> > wrote:
Hello Elena,
I removed the flags and used just ifort, as you suggested, but the result
was the same.
My second attempt was to run without --enable-fortran2003. With it, the
"Mounting test" didin't pass and it repeats without going to the next test.
Without this flag, I solved the problem of "memory" consumed, but I still
get an error:
============================
Fortran API: fortranlib_test Test Log
==========================
FORTRAN tests
==========================
FORTRANLIB_TEST is linked with HDF5 Library version 1.8 release 11
Mounting test
PASSED
Reopen test
PASSED
File open/close test
PASSED
File free space test
PASSED
Dataset test
PASSED
Extendible dataset test
PASSED
Basic dataspace test
PASSED
Reference to object test
PASSED
Reference to dataset region test
PASSED
forrtl: severe (174): SIGSEGV, segmentation fault occurred
Image PC Routine Line
Source
lt-fortranlib_tes 000000000049E349 Unknown Unknown Unknown
lt-fortranlib_tes 000000000049CCC0 Unknown Unknown Unknown
lt-fortranlib_tes 0000000000472A72 Unknown Unknown Unknown
lt-fortranlib_tes 0000000000458983 Unknown Unknown Unknown
lt-fortranlib_tes 000000000042F00B Unknown Unknown Unknown
libpthread.so.0 0000003B1C40EFA0 Unknown Unknown Unknown
libhdf5_fortran.s 00007FEEC7E588FC Unknown Unknown Unknown
libhdf5_fortran.s 00007FEEC7E4CE6D Unknown Unknown Unknown
lt-fortranlib_tes 00000000004203EB Unknown Unknown Unknown
lt-fortranlib_tes 000000000040D5CC Unknown Unknown Unknown
lt-fortranlib_tes 000000000040D1B6 Unknown Unknown Unknown
libc.so.6 0000003B1BC21B75 Unknown Unknown Unknown
lt-fortranlib_tes 000000000040D0A9 Unknown Unknown Unknown
I updated the intel parallel studio to update 1 and did the tests again,
but the results were the same.
So I add some "debug" parameters to try to get more information, and the
result is this:
============================
Fortran API: fortranlib_test Test Log
==========================
FORTRAN tests
==========================
FORTRANLIB_TEST is linked with HDF5 Library version 1.8 release 11
Mounting test
PASSED
Reopen test
PASSED
File open/close test
PASSED
File free space test
PASSED
Dataset test
PASSED
Extendible dataset test
PASSED
Basic dataspace test
PASSED
Reference to object test
PASSED
Reference to dataset region test
PASSED
forrtl: severe (174): SIGSEGV, segmentation fault occurred
Image PC Routine Line
Source
lt-fortranlib_tes 000000000049BC99 Unknown Unknown Unknown
lt-fortranlib_tes 000000000049A610 Unknown Unknown Unknown
lt-fortranlib_tes 00000000004703C2 Unknown Unknown Unknown
lt-fortranlib_tes 00000000004562D3 Unknown Unknown Unknown
lt-fortranlib_tes 000000000042C95B Unknown Unknown Unknown
libpthread.so.0 0000003B1C40EFA0 Unknown Unknown Unknown
libhdf5_fortran.s 00007FB43829BAAC h5sselect_hypersl 993 H5Sf.c
libhdf5_fortran.s 00007FB43828F547 h5s_mp_h5sselect_ 1332
H5Sff.f90
lt-fortranlib_tes 000000000041EC68 test_basic_select 905
tH5Sselect.f90
lt-fortranlib_tes 000000000040D567 MAIN__ 129
fortranlib_test.f90
lt-fortranlib_tes 000000000040D1B6 Unknown Unknown Unknown
libc.so.6 0000003B1BC21B75 Unknown Unknown Unknown
lt-fortranlib_tes 000000000040D0A9 Unknown Unknown Unknown
I tried to see what was happening and put some "write's" and "printf's" in
the source files to understand what is happening (my debug skills in linux
are near absolute zero).
And I found this:
Reference to dataset region test
PASSED
test_basic_select
present(stride).and. present(block)
forrtl: severe (174): SIGSEGV, segmentation fault occurred
Image PC Routine Line
Source
lt-fortranlib_tes 000000000049BD69 Unknown Unknown Unknown
lt-fortranlib_tes 000000000049A6E0 Unknown Unknown Unknown
lt-fortranlib_tes 0000000000470492 Unknown Unknown Unknown
lt-fortranlib_tes 00000000004563A3 Unknown Unknown Unknown
lt-fortranlib_tes 000000000042CA2B Unknown Unknown Unknown
libpthread.so.0 0000003B1C40EFA0 Unknown Unknown Unknown
libhdf5_fortran.s 00007F07945D7D1E h5sselect_hypersl 997 H5Sf.c
libhdf5_fortran.s 00007F07945CB636 h5s_mp_h5sselect_ 1333
H5Sff.f90
lt-fortranlib_tes 000000000041ED38 test_basic_select 905
tH5Sselect.f90
lt-fortranlib_tes 000000000040D5CF MAIN__ 130
fortranlib_test.f90
lt-fortranlib_tes 000000000040D1B6 Unknown Unknown Unknown
libc.so.6 0000003B1BC21B75 Unknown Unknown Unknown
lt-fortranlib_tes 000000000040D0A9 Unknown Unknown Unknown
rank=2, count lenght=1, stride lenght=1
stride address=0xd19400
i=0, t=1
start[t]=3
count[t]=3
stride[t]=1
block[t]=1
i=1, t=0
start[t]=0
count[t]=2
stride[t]=1
block[t]=1
end
rank=2, count lenght=1, stride lenght=1
stride address=(nil)
i=0, t=1
start[t]=0
count[t]=2
So, somehow, nh5sselect_hyperslab_c is being called twice, even if it
shouldn't, at least I think. And in the second (thread?) , the stride array
is not defined.
And this is all, for now.
This should be something related with the intel compilers, because when I
use the gnu compilers (gcc, g++, gfortran) everything works.
But I really need to compile this with intel fortran...
Thanks,
Best regards,
Eduardo Jauch
Em 31-10-2013 21:56, Elena Pourmal escreveu:
Eduardo,
I don't see any reason for Fortran tests to behave such way. Fortran is
a thin wrapper on top of the HDF5 C library. If C tests behaved, so should
Fortran.
I am not sure that if it has something to do with the flags you used
-DDEC$=DEC_ -DMS$=MS_. Those were necessary for the old versions of Linux
and ancient Intel compilers.
Could you please remove those flags and just use
export FC ifort
to configure and build the library?
Thank you!
Elena
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Elena Pourmal The HDF Group http://hdfgroup.org
1800 So. Oak St., Suite 203, Champaign IL 61820
217.531.6112
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
On Oct 31, 2013, at 7:12 AM, Eduardo Jauch <eduardo.jauch@ist.utl.pt> > wrote:
Hello,
I'm trying to compile HDF5 with the new Intel tools, but while the
./config and make works without a problem, in the test phase (make check),
when reaching the Fortran API test, the test uses all my machines memory
(and the swap also), as you can see on this image:
http://img201.imageshack.us/img201/7406/cwk.png
The "scripts" that I used to compile szlib, zlib and the hdf5 are these:
SZLIB:
export CC=icc
export CXX=icpc
export F77=ifort
export CFLAGS='-O3 -xHost -ip'
export CXXFLAGS='-O3 -xHost -ip'
export FFLAGS='-O3 -xHost -ip'
tar -zxvf szip-2.1.tar.gz
cd szip-2.1
./configure --prefix=/home/user/Mohid/szip-2.1
make
make check
make install
ZLIB:
export CC=icc
export CFLAGS='-O3 -xHost -ip'
tar -zxvf zlib-1.2.8.tar.gz
cd zlib-1.2.8
./configure --prefix=/home/user/Mohid/zlib-1.2.8
make
make check
make install
hdf5:
export CC=icc
export FC='ifort -fpp -DDEC$=DEC_ -DMS$=MS_'
export CXX=icpc
tar -zxvf hdf5-1.8.11.tar.gz
cd hdf5-1.8.11
./configure --prefix=/home/user/Mohid/hdf5-1.8.11 --enable-fortran
--enable-fortran2003 --enable-cxx --enable-unsupported \
--with-szlib=/home/user/Mohid/szip-2.1
--with-zlib=/home/user/Mohid/zlib-1.2.8/include,/home/user/Mohid/zlib-1.2.8/lib
--enable-production
make
make check
make install
I used the '-fpp -DDEC$=DEC_ -DMS$=MS_' because it was refered here:
http://www.hdfgroup.org/ftp/HDF5/current/src/unpacked/release_docs/INSTALL
The 'result' of the ./config on the HDF5 is this:
SUMMARY OF THE HDF5 CONFIGURATION
=================================
General Information:
-------------------
HDF5 Version: 1.8.11
Configured on: Thu Oct 31 09:53:29 WET 2013
Configured by: user@localhost.localdomain
Configure mode: production
Host system: x86_64-unknown-linux-gnu
Uname information: Linux localhost.localdomain
3.11.6-200.fc19.x86_64 #1 SMP Fri Oct 18 22:34:18 UTC 2013 x86_64 x86_64
x86_64 GNU/Linux
Byte sex: little-endian
Libraries: static, shared
Installation point: /home/user/Mohid/hdf5-1.8.11
Compiling Options:
------------------
Compilation Mode: production
C Compiler:
/opt/intel/composer_xe_2013_sp1.0.080/bin/intel64/icc ( Intel(R) C Intel(R)
64 Compiler Version 14.0.0.080 Build 20130728)
CFLAGS:
H5_CFLAGS: -std=c99 -O
AM_CFLAGS:
CPPFLAGS:
H5_CPPFLAGS: -D_POSIX_C_SOURCE=199506L -DNDEBUG
-UH5_DEBUG_API
AM_CPPFLAGS: -I/home/user/Mohid/szip-2.1/include
-I/home/user/Mohid/zlib-1.2.8/include -D_LARGEFILE_SOURCE
-D_LARGEFILE64_SOURCE -D_BSD_SOURCE
Shared C Library: yes
Static C Library: yes
Statically Linked Executables: no
LDFLAGS:
H5_LDFLAGS:
AM_LDFLAGS: -L/home/user/Mohid/zlib-1.2.8/lib
-L/home/user/Mohid/szip-2.1/lib
Extra libraries: -lsz -lz -ldl -lm
Archiver: ar
Ranlib: ranlib
Debugged Packages:
API Tracing: no
Languages:
----------
Fortran: yes
Fortran Compiler:
/opt/intel/composer_xe_2013_sp1.0.080/bin/intel64/ifort -fpp -DDEC$=DEC_
-DMS$=MS_ ( Intel(R) Fortran Intel(R) 64 Compiler Version 14.0.0.080 Build
20130728)
Fortran 2003 Compiler: yes
Fortran Flags:
H5 Fortran Flags: -O3
AM Fortran Flags:
Shared Fortran Library: yes
Static Fortran Library: yes
C++: yes
C++ Compiler:
/opt/intel/composer_xe_2013_sp1.0.080/bin/intel64/icpc ( Intel(R) C++
Intel(R) 64 Compiler Version 14.0.0.080 Build 20130728)
C++ Flags:
H5 C++ Flags:
AM C++ Flags:
Shared C++ Library: yes
Static C++ Library: yes
Features:
---------
Parallel HDF5: no
High Level library: yes
Threadsafety: no
Default API Mapping: v18
With Deprecated Public Symbols: yes
I/O filters (external): deflate(zlib),szip(encoder)
I/O filters (internal): shuffle,fletcher32,nbit,scaleoffset
MPE: no
Direct VFD: no
dmalloc: no
Clear file buffers before write: yes
Using memory checker: no
Function Stack Tracing: no
GPFS: no
Strict File Format Checks: no
Optimization Instrumentation: no
Large File Support (LFS): yes
Can you see some reason to this happens?
Any tip?
Thanks,
Best regards,
Eduardo Jauch.
--
TECNICO Lisboa - MARETEC
Secção de Ambiente e Energia
Av. Rovisco Pais
1049-001 Lisboa
Tel.: +351 21 8419424
Fax: +351 21 8417365
URL: www.mohid.com
_______________________________________________
Hdf-forum is for HDF software users discussion.
Hdf-forum@lists.hdfgroup.org
http://mail.lists.hdfgroup.org/mailman/listinfo/hdf-forum_lists.hdfgroup.org
_______________________________________________
Hdf-forum is for HDF software users discussion.Hdf-forum@lists.hdfgroup.orghttp://mail.lists.hdfgroup.org/mailman/listinfo/hdf-forum_lists.hdfgroup.org
--
TECNICO Lisboa - MARETEC
Secção de Ambiente e Energia
Av. Rovisco Pais
1049-001 Lisboa
Tel.: +351 21 8419424
Fax: +351 21 8417365
URL: www.mohid.com
_______________________________________________
Hdf-forum is for HDF software users discussion.
Hdf-forum@lists.hdfgroup.org
http://mail.lists.hdfgroup.org/mailman/listinfo/hdf-forum_lists.hdfgroup.org
_______________________________________________
Hdf-forum is for HDF software users discussion.Hdf-forum@lists.hdfgroup.orghttp://mail.lists.hdfgroup.org/mailman/listinfo/hdf-forum_lists.hdfgroup.org
--
<http://www.ist.utl.pt>
<http://www.mohid.com>
TECNICO Lisboa - MARETEC
Secção de Ambiente e Energia
Av. Rovisco Pais
1049-001 Lisboa
Tel.: +351 21 8419424
Fax: +351 21 8417365
URL: www.mohid.com
_______________________________________________
Hdf-forum is for HDF software users discussion.
Hdf-forum@lists.hdfgroup.org
http://mail.lists.hdfgroup.org/mailman/listinfo/hdf-forum_lists.hdfgroup.org
