On Thu, 2015-05-07 at 21:44 +0530, Bala Subramani wrote:
> Hi Friends,
> I have NVIDIA K20C tesla card in my GPU workstation. i have installed
> serial and parallel version of amber, but when i configured my cuda version
> of amber11, it configured correctly, when i execute *make cuda. *It throws
> an error:
>
>
> make[3]: Entering directory
> `/home/abhinav/Tools/AMBER11/amber11/src/pmemd/src/cuda'
> cpp -traditional -P -DBINTRAJ -DDIRFRC_EFS -DDIRFRC_COMTRANS
> -DDIRFRC_NOVEC -DFFTLOADBAL_2PROC -DPUBFFT -DCUDA cuda_info.fpp
> cuda_info.f90
> gfortran -O3 -mtune=generic -DCUDA -I/usr/local/cuda-7.0//include -IB40C
> -IB40C/KernelCommon -c cuda_info.f90
> Warning: Nonexistent include directory "B40C"
> Warning: Nonexistent include directory "B40C/KernelCommon"
> gcc -O3 -mtune=generic -DSYSV -D_FILE_OFFSET_BITS=64 -D_LARGEFILE_SOURCE
> -DBINTRAJ -DCUDA -I/usr/local/cuda-7.0//include -IB40C -IB40C/KernelCommon
> -c gpu.cpp
> gcc -O3 -mtune=generic -DSYSV -D_FILE_OFFSET_BITS=64 -D_LARGEFILE_SOURCE
> -DBINTRAJ -DCUDA -I/usr/local/cuda-7.0//include -IB40C -IB40C/KernelCommon
> -c gputypes.cpp
> /usr/local/bin/gcc4.4 -use_fast_math -O3 -gencode
> arch=compute_13,code=sm_13 -gencode arch=compute_20,code=sm_20 -DCUDA
> -I/usr/local/cuda-7.0//include -IB40C -IB40C/KernelCommon -c
> kForcesUpdate.cu
> make[3]: execvp: /usr/local/bin/gcc4.4: Permission denied
> make[3]: *** [kForcesUpdate.o] Error 127
> make[3]: Leaving directory
> `/home/abhinav/Tools/AMBER11/amber11/src/pmemd/src/cuda'
> make[2]: *** [-L/usr/local/cuda-7.0//lib64] Error 2
> make[2]: Leaving directory
> `/home/abhinav/Tools/AMBER11/amber11/src/pmemd/src'
> make[1]: *** [cuda] Error 2
> make[1]: Leaving directory `/home/abhinav/Tools/AMBER11/amber11/src/pmemd'
> make: *** [cuda] Error 2
Is this really Amber 11? I think both your video card and NVIDIA
toolkits are too new to compile pmemd.cuda from Amber 11. You will need
to upgrade to Amber 14 in order to use pmemd.cuda on your system, I
think.
I can't say I've seen the error you're reporting, but if I had to guess
it would be that your CUDA toolkit installation is somehow broken
(perhaps it's trying to use a host compiler that doesn't exist?)
HTH,
Jason
--
Jason M. Swails
BioMaPS,
Rutgers University
Postdoctoral Researcher
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Thu May 07 2015 - 12:30:02 PDT
