Re: [gmx-users] OpenSuse 12.1 + CUDA Installation Error

2013-07-24 Thread Carlos Bueno
To install the CUDA I first downloaded and installed the drivers for the
Nvidia cards. The command nvidia-smi doen't give me any problems
> nvidia-smi
Wed Jul 24 17:36:28 2013
+--+

| NVIDIA-SMI 4.310.44   Driver Version: 310.44 |

|---+--+--+
| GPU  Name | Bus-IdDisp.  | Volatile Uncorr.
ECC |
| Fan  Temp  Perf  Pwr:Usage/Cap| Memory-Usage | GPU-Util  Compute
M. |
|===+==+==|
|   0  GeForce GTX 670  | :01:00.0 N/A |
 N/A |
| 16%   57C  N/A N/A /  N/A |   3%   67MB / 2047MB | N/A
 Default |
+---+--+--+
|   1  GeForce GTX 670  | :02:00.0 N/A |
 N/A |
| 71%   82C  N/A N/A /  N/A |   8%  157MB / 2047MB | N/A
 Default |
+---+--+--+


+-+
| Compute processes:   GPU
Memory |
|  GPU   PID  Process name Usage
   |
|=|
|0Not Supported
  |
|1Not Supported
  |
+-+
Then, I downloaded the CUDA5 release for OpenSuse 12.1 from this page
https://developer.nvidia.com/cuda-downloads
but didn't reinstall the driver because that would downgrade my driver.
Now the cuda lib is found in /usr/local/cuda

I added all the repositories and installed all you told me.
And that solved the problem for one of the computers
The others still have an error in make:

[ 55%] Building NVCC (Device) object
src/mdlib/nbnxn_cuda/./nbnxn_cuda_generated_nbnxn_cuda.cu.o
/home/cuda2/Programas/gromacs-4.6.3/include/types/nbnxn_pairlist.h(216):
error: identifier "nbnxn_alloc_t" is undefined

/home/cuda2/Programas/gromacs-4.6.3/include/types/nbnxn_pairlist.h(217):
error: identifier "nbnxn_free_t" is undefined

2 errors detected in the compilation of
"/tmp/tmpxft_76ad_-11_nbnxn_cuda.compute_20.cpp2.i".
CMake Error at CMakeFiles/nbnxn_cuda_generated_nbnxn_cuda.cu.o.cmake:256
(message):
  Error generating file

/home/cuda2/Programas/gromacs-4.6.3/build/src/mdlib/nbnxn_cuda/./nbnxn_cuda_generated_nbnxn_cuda.cu.o


make[2]: *** [src/mdlib/nbnxn_cuda/./nbnxn_cuda_generated_nbnxn_cuda.cu.o]
Error 1
make[1]: *** [src/mdlib/nbnxn_cuda/CMakeFiles/nbnxn_cuda.dir/all] Error 2
make: *** [all] Error 2


Thank you very much
I appreciate your help





2013/7/24 Mirco Wahab 

> On 24.07.2013 04:08, Carlos Bueno wrote:
>
>> *Hi,*
>> *I keep getting errors when I try to install gromacs in OpenSuse 12.1.*
>> *I have installed cuda 5.0 and the nvidia cards. **I have tried with
>> different parameters for cmake:*
>>
>
> How did you install Cuda5? What did you install and how?
>
> I have one OpenSuSE 12.1 box that works as a cluster head
> and serves to gpu-clusters. Every software except the
> nvidia gpu driver is installed on this box for usage by
> the nodes.
>
> Installing a gpu-ready OpenSuSE 12.1 involves (for example, YMMD)
>  * add community- and extra-repositories through yast/repositories,
>these involve (here):
>(a) included in yast/repositories (activate only)
> Packman Repository
> openSUSE BuildService - devel:languages:perl
> openSUSE BuildService - devel:languages:python
> Education
> science
>(b) extra (add manually or ba script)
> devel:/tools
> devel:/libraries:/c_c++
> devel:/gcc
>you can do the latter by script:
>  #
>  #!/bin/sh
>
> Uri[1]="http://download.**opensuse.org/repositories/**
> devel:/tools/openSUSE_12.1/<http://download.opensuse.org/repositories/devel:/tools/openSUSE_12.1/>"
>Name[1]="devel:/tools"
>
> Uri[2]="http://download.**opensuse.org/repositories/**
> devel:/libraries:/c_c++/**openSUSE_12.1/<http://download.opensuse.org/repositories/devel:/libraries:/c_c++/openSUSE_12.1/>"
> Name[2]="devel:/libraries:/c_**c++"
>
> Uri[3]="http://download.**opensuse.org/repositories/**
> devel:/gcc/openSUSE_12.1/<http://download.opensuse.org/repositories/devel:/gcc/openSUSE_12.1/>"
>  Name[3]="devel:/gcc"
>  #
>  for i in 1 2 3; do
> zypper --gpg-auto-import-keys ar ${Uri[i]} ${Name[i]}
> zypper modifyrepo --refresh ${Name[i]}
>  done
>  #
>
>  * install gcc 4.6 / g++ 4.6 through yast
>  * install

[gmx-users] OpenSuse 12.1 + CUDA Installation Error

2013-07-23 Thread Carlos Bueno
*Hi,*
*I keep getting errors when I try to install gromacs in OpenSuse 12.1.*
*I have installed cuda 5.0 and the nvidia cards. **I have tried with
different parameters for cmake:*

>cmake .. -DGMX_BUILD_OWN_FFTW=ON -DGMX_GPU=ON
-DCUDA_TOOLKIT_ROOT_DIR=/usr/local/cuda
*When I make, this error appears:*
 >make
[ 65%] Building C object
src/gmxlib/CMakeFiles/gmx.dir/nonbonded/nb_kernel_avx_256_single/nb_kernel_ElecEwSh_VdwLJSh_GeomW3W3_avx_256_single.c.o
make[2]: *** No hay ninguna regla para construir el objetivo
`//home/cuda/Programas/gromacs-4.6.3/build/src/contrib/fftw/gmxfftw-prefix/lib/libfftw3f.a',
necesario para `src/gmxlib/libgmx.so.8'.  Alto.
 make[1]: *** [src/gmxlib/CMakeFiles/gmx.dir/all] Error 2
make: *** [all] Error 2

*So I decided to compile the fftw library myself with the command:*
./configure --enable-single --enable-shared --enable-sse2 --enable-avx
*
*
*and I think that worked but then this error appears:*

>cmake .. -DGMX_GPU=ON -DCUDA_TOOLKIT_ROOT_DIR=/usr/local/cuda
>make
[ 64%] Building NVCC (Device) object
src/mdlib/nbnxn_cuda/./nbnxn_cuda_generated_nbnxn_cuda.cu.o
/home/cuda3/Programas/gromacs-4.6.3/include/types/nbnxn_pairlist.h(216):
error: identifier "nbnxn_alloc_t" is undefined

/home/cuda3/Programas/gromacs-4.6.3/include/types/nbnxn_pairlist.h(217):
error: identifier "nbnxn_free_t" is undefined

2 errors detected in the compilation of
"/tmp/tmpxft_76f4_-11_nbnxn_cuda.compute_20.cpp2.i".
CMake Error at CMakeFiles/nbnxn_cuda_generated_nbnxn_cuda.cu.o.cmake:256
(message):
  Error generating file

/home/cuda3/Programas/gromacs-4.6.3/build/src/mdlib/nbnxn_cuda/./nbnxn_cuda_generated_nbnxn_cuda.cu.o


make[2]: *** [src/mdlib/nbnxn_cuda/./nbnxn_cuda_generated_nbnxn_cuda.cu.o]
Error 1
make[1]: *** [src/mdlib/nbnxn_cuda/CMakeFiles/nbnxn_cuda.dir/all] Error 2
make: *** [all] Error 2

*I need to install with cuda, nevertheless I tried without to see if that
would work:*

>cmake ..
>make

Linking CXX shared library libgmx.so
[ 65%] Built target gmx
make: *** [all] Error 2

*And to see if I had not installed the library correctly I tried*

>cmake .. -DGMX_BUILD_OWN_FFTW=ON
>make

[ 65%] Building C object
src/gmxlib/CMakeFiles/gmx.dir/nonbonded/nb_kernel_avx_256_single/nb_kernel_ElecEwSh_VdwLJSh_GeomW3W3_avx_256_single.c.o
make[2]: *** No hay ninguna regla para construir el objetivo
`//home/cuda/Programas/gromacs-4.6.3/build/src/contrib/fftw/gmxfftw-prefix/lib/libfftw3f.a',
necesario para `src/gmxlib/libgmx.so.8'.  Alto.
make[1]: *** [src/gmxlib/CMakeFiles/gmx.dir/all] Error 2
make: *** [all] Error 2

*I think the problem is related to the PC's (I have three clones) because I
have installed Gromacs+CUDA in Ubuntu 12.04 without problems.*
*Any suggestions would be very appreciated.*
*Thanks*
-- 
gmx-users mailing listgmx-users@gromacs.org
http://lists.gromacs.org/mailman/listinfo/gmx-users
* Please search the archive at 
http://www.gromacs.org/Support/Mailing_Lists/Search before posting!
* Please don't post (un)subscribe requests to the list. Use the 
www interface or send it to gmx-users-requ...@gromacs.org.
* Can't post? Read http://www.gromacs.org/Support/Mailing_Lists