[mvapich-discuss] v1.8.a2 and cuda

[Steve Heistand]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

I hadnt seen it was actually adding in a -lcuda somewhere.
that would have been useful. seems the actual cuda stuff hadnt
been installed on the machine I was compiling on. Moving over
to something with the drivers and libs helped greatly.
PGI is still complaining at build time but at least it configures fine.
Trying Intel now.

thanks much

steve


On 02/09/2012 12:21 PM, Jonathan Perkins wrote:
> Thanks.
> 
> I took a quick looks and I see that the following command failed...
> 
> configure:9310: icc -o conftest    -g -DNDEBUG -DNVALGRIND
> -I/u/scicon/tools/opt/cuda_4.1/include
> -I/u/scicon/tools/build/mvapich2-1.8a2/src/mpl/include
> -I/u/scicon/tools/build/mvapich2-1.8a2/src/mpl/include
> -I/u/scicon/tools/build/mvapich2-1.8a2/src/openpa/src
> -I/u/scicon/tools/build/mvapich2-1.8a2/src/openpa/src
> -L/u/scicon/tools/opt/cuda_4.1/lib64
> -L/u/scicon/tools/opt/cuda_4.1/lib   conftest.c -lcuda    >&5
> ld: cannot find -lcuda
> 
> Do you have libcuda.a or libcuda.so installed in
> /u/scicon/tools/opt/cuda_4.1/lib64 or .../lib?
> 
> I see that on our system that libcuda.so is installed in /usr/lib64
> whereas libcudart and friends are installed in /usr/local/cuda/lib64.
> 
> It may be that libcuda is not where you're expecting it?
> 
> On Thu, Feb 9, 2012 at 3:09 PM, Steve Heistand <steve.heistand at nasa.gov> wrote:
> attached config.log
> 
> thanks
> 
> s
> 
> 
> On 02/09/2012 12:07 PM, Jonathan Perkins wrote:
>>>> Hi Steve, can you please send me your config.log?
>>>>
>>>> On Thu, Feb 9, 2012 at 2:04 PM, Steve Heistand <steve.heistand at nasa.gov> wrote:
>>>> hi folks,
>>>>
>>>> so Im trying to enable the cuda functionality and running into some trouble
>>>> at configure time.
>>>> Compilerwise Im trying both Intel V11 something and the latest PGI V12
>>>> and both do the same thing.
>>>>
>>>> I have the latest cuda 4.1 installed in the directory shown in the configure line.
>>>> which is:
>>>>
>>>> ./configure --prefix=/u/scicon/tools/opt/mvapich2_1.8_intel_11 --enable-f77 --enable-fc
>>>> --enable-cxx --enable-romio --enable-smpcoll --enable-threads=runtime --enable-rdma-cm
>>>> --enable-hybrid --enable-cuda --with-device=ch3:mrail --with-rdma=gen2 --enable-shared
>>>> --enable-g=dbg --enable-debuginfo --with-cuda=/u/scicon/tools/opt/cuda_4.1
>>>>
>>>> the end of the configure output is:
>>>> checking cuda.h usability... yes
>>>> checking cuda.h presence... yes
>>>> checking for cuda.h... yes
>>>> checking for library containing cuPointerGetAttribute... no
>>>> configure: error: Could not link with cuda
>>>>
>>>> in the config.log for the last error shown its not even trying to link in any
>>>> cuda libraries. has the -L entries but no -l ones.
>>>>
>>>> anyone having good luck with the latest mvapich2 and cuda?
>>>>
>>>> (without the --enable-cuda the above configure works fine and have installed that
>>>> version of mvapich2 which works fine)
>>>>
>>>> thanks
>>>>
>>>> steve
>>>>
> _______________________________________________
> mvapich-discuss mailing list
> mvapich-discuss at cse.ohio-state.edu
> http://mail.cse.ohio-state.edu/mailman/listinfo/mvapich-discuss
>>>>>
> 

- -- 
************************************************************************
 Steve Heistand                          NASA Ames Research Center
 email: steve.heistand at nasa.gov          Steve Heistand/Mail Stop 258-6
 ph: (650) 604-4369                      Bldg. 258, Rm. 232-5
 Scientific & HPC Application            P.O. Box 1
 Development/Optimization                Moffett Field, CA 94035-0001
************************************************************************
 "Any opinions expressed are those of our alien overlords, not my own."
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v2.0.14 (GNU/Linux)

iEYEARECAAYFAk80MuEACgkQoBCTJSAkVrFG6gCePagSccTRRsr764XZoncWf+d+
00sAoJ2lZGGmoRX7joqU0IB2jYkc3gSs
=ow4C
-----END PGP SIGNATURE-----