Dear Axel,

I have been following your suggestion.
Every time, I use a script to build which has

rm -fR ~/build
mkdir ~/build
cd ~/build
$PICSRC/configure ~/paramSets/case001

So there is no question of any outdated build residing inside the build directory.
As before every build attempt, the build directory is re-created.
It could be some other issue, which I am not able to find out.

Also, could pthread_create error be due to cuda-7.5 and gcc-4.8 and glibc?
Today, I will try with cuda-7.0 and same version of gcc and glibc and let you know.

Regards,
Ajit



-----Original Message-----
From: picongpu-users@hzdr.de [mailto:picongpu-users@hzdr.de] On Behalf Of Huebl, Axel
Sent: Wednesday, April 27, 2016 8:43 PM
To: picongpu-users@hzdr.de
Subject: Re: [PIConGPU-Users] [PIConGPU-Users] PIConGPU compilation -lpthread error

Dear Ajit,

looks like an older, outdated build using CMake resides inside build/
Just remove it like this:

cd $HOME/build
rm -rf ../build/*

$PICSRC/configure ~/paramSets/case001
make -j install


Best,
Axel

On 27.04.2016 12:35, Dr. Ajit Upadhyay wrote:

Dear René,

Thanks for such a prompt reply. The log files, I sent were from the
release-0.1.2 branch. However, I did try "master"  as well, but the error
remains same.

$PICSRC/configure ~/paramSets/case001

in release-0.1.2 branch as well as master branch gives following error:-


CMake Error at CMakeLists.txt:38 (add_subdirectory):
   The binary directory
     /home/ajitup/build/build_picongpu
   is already used to build a source directory.  It cannot be used to build
source directory
     /home/ajitup/code.PIConGPU/picongpu-master/src/picongpu
Specify a unique binary directory name.

CMake Error at CMakeLists.txt:38 (add_subdirectory):
   The binary directory
     /home/ajitup/build/build_picongpu
   is already used to build a source directory.  It cannot be used to build
source directory
     /home/ajitup/code.PIConGPU/picongpu-master/src/picongpu
   Specify a unique binary directory name.


Whereas, there is no quota set by HPC system and I am able to write/create
a file/folder in ~/build directory. my PICSRC is ~/src/picongpu and I
build the code in ~/build directory. Further error is in CMakeError.log :


Determining if the pthread_create exist failed with the following output:
Change Dir: /home/ajitup/build/CMakeFiles/CMakeTmp

Run Build Command:"/usr/bin/gmake" "cmTC_d035f/fast"
/usr/bin/gmake -f CMakeFiles/cmTC_d035f.dir/build.make
CMakeFiles/cmTC_d035f.dir/build
gmake[1]: Entering directory '/home/ajitup/build/CMakeFiles/CMakeTmp'
Building C object CMakeFiles/cmTC_d035f.dir/CheckSymbolExists.c.o
/usr/bin/cc    -fPIC    -o CMakeFiles/cmTC_d035f.dir/CheckSymbolExists.c.o
   -c /home/ajitup/build/CMakeFiles/CMakeTmp/CheckSymbolExists.c
Linking C executable cmTC_d035f
/usr/local/bin/cmake -E cmake_link_script
CMakeFiles/cmTC_d035f.dir/link.txt --verbose=1
/usr/bin/cc  -fPIC     CMakeFiles/cmTC_d035f.dir/CheckSymbolExists.c.o  -o
cmTC_d035f -rdynamic
CMakeFiles/cmTC_d035f.dir/CheckSymbolExists.c.o: In function `main':
CheckSymbolExists.c:(.text+0x1b): undefined reference to `pthread_create'
collect2: error: ld returned 1 exit status
CMakeFiles/cmTC_d035f.dir/build.make:97: recipe for target 'cmTC_d035f'
failed
gmake[1]: *** [cmTC_d035f] Error 1
gmake[1]: Leaving directory '/home/ajitup/build/CMakeFiles/CMakeTmp'
Makefile:126: recipe for target 'cmTC_d035f/fast' failed
gmake: *** [cmTC_d035f/fast] Error 2

File /home/ajitup/build/CMakeFiles/CMakeTmp/CheckSymbolExists.c:
/* */
#include <pthread.h>

int main(int argc, char** argv)
{
   (void)argv;
#ifndef pthread_create
   return ((int*)(&pthread_create))[argc];
#else
   (void)argc;
   return 0;
#endif
}


I have attached some test files which I compiled to check for
pthread_create function using    $cc -lpthread ***.c -o ***.out
and they compiled. A search for libpthread shows it in /lib as well as
/lib64 in system-root. I have gcc 4.8.5, and cuda 7.5. I did try dev
branch, but as you suggested it will require enabling c++11 support. The
error in dev branch is:

Please enable C++11 with `-std=c++11` or `-DCMAKE_CXX_STANDARD=11`  when
compiling with nvcc 7.5

I will try adding c++11 support and let you know, but the CMakeError.log
shows lpthread issue even in dev branch.

regards,
ajit

Attachements:  Cmake log files for release-branch
                test codes for lpthread
                CMakeLists.txt from release-0.1.2

---------------------------- Original Message ----------------------------
Subject: Re: Help on PIConGPU?
From:    René Widera <r.widera@hzdr.de>
Date:    Wed, April 27, 2016 2:14 pm
To:      "Dr. Ajit Upadhyay" <ajitup@rrcat.gov.in>
------------------------------------------------------------------
Dear Dr. Ajit Upadhyay,

thank you for participate our project PIConGPU.

Please subscribe at our mailing list [1] for future issues with
picongpu. The mailing list helps the community if other run in equal
issues.

Currently I have only looked very fast to your problem, I will take a
deeper analysis tomorrow.
Which version of PIConGPU you used? master or dev
For cuda 7.5 please switch to `dev` because as I remember we need to
enable C++11 for cuda 7.5 else PIConGPU will not compile and the master
is not C++11 compatible.
Please keep in mind that the branch `dev` is our working branch and
interfaces can be changed.
There will be a new release in one or two weeks available.

I looked first at your file `build_output_shell` and found the error
```
CMake Error at CMakeLists.txt:38 (add_subdirectory):
    The binary directory
      /home/ajitup/build/build_picongpu
```
This looks like you are not allowed to create directories in `~/build`.
Maybe your HPC System set a quota for your home directory which results
in the case that no new files can be created. Could you please check
this behavior.

pthread issue:

Could you please provide me with the cmake script and all needed files.
It looks like the linker flag `lpthread` is missing in your example.

If some of the issues are solved till tomorrow please give me a feedback.

btw: please skip the `Dr.` before my name but I would be glad to get a
honorary degree ;-)

best regards,
René

[1]
https://github.com/ComputationalRadiationPhysics/picongpu/tree/dev#users

On 27.04.2016 09:43, Dr. Ajit Upadhyay wrote:

Dear Dr. René Widera,

I downloaded the release-0.1.2 branch of code and tried to compile it

on a HP Z840 (Intel Xeon E5-2630v3 2.4 1866 8C x 2, 128GB memory)
workstation with two Nividia Quadro 4200 GPUs (compatible with CUDA).
The OS on my workstation is openSuSe Leap 42.1 with cuda-7.5. I have
tested some examples from the nvidia-cuda library and they are working
fine.

I have compiled all the libraries required by PIConGPU successfully and

placed them as per the instructions in /picongpu/src/install.md file.
However, when I try to compile the code itself, it gives a strange
error> of pthread_create function (please look at the attached
CMakeError.log).

I did check that the relevant library is in its place
(/lib/libpthread-2.19.so, glibc-2.19-19.1_x86_64). I did google for the

error but could not resolve this issue as the problem seems to be with
"configure" script of the picongpu code.

I made a test program from CMakeError.log file and compiled it

separately and it compiles (please have a look at pthread_check_log, a
few lines of code are listed and it compiles all these which uses this
function). I have attached  the output dumped on the linux-terminal in
file build_output_shell. Other files created by cmake are also
attached.

Could you please point out where is the fault in my setup of this code
or what should I do to run this code.

Regards,


#############################################################
This message is sent to you because you are subscribed to
   the mailing list <picongpu-users@hzdr.de>.
To unsubscribe, E-mail to: <picongpu-users-off@hzdr.de>
To switch to the DIGEST mode, E-mail to <picongpu-users-digest@hzdr.de>
To switch to the INDEX mode, E-mail to <picongpu-users-index@hzdr.de>
Send administrative queries to  <picongpu-users-request@hzdr.de>