Re: [AMBER] cuda-5.0/lib64/libcufft.so: undefined reference to `__isoc99_sscanf@GLIBC_2.7'

From: Ross Walker <ross.rosswalker.co.uk>
Date: Wed, 21 Nov 2012 10:03:22 -0800

Hi Thomas,

This unfortunately is the problem with being on the hemorrhaging side of
the bleeding edge which is why I stay away from things like Fedora. gcc
3.4 is probably too old for AMBER. I tend to use GCC4 or 4.4 on Redhat 5
or 6 and stick with intel 11 since any intel XX.0.XX compiler is always
likely to be buggy in my experience.

Right now it looks like your mpi is configured for ifort which is why you
get all the link errors. Note the actual intel compiler is crashing on
compilation of binrestart.F90 - probably because the module file related
to that was previously compiled with GCC. I would suggest first seeing if
you can get GCC 4.4 or 4.6 installed and use that. If you can't then stick
with 3.4 which might work BUT either way you will need to rebuild your
mpich2 to use the gcc, g++ and gfortran compilers - You set the
environment variables CC=gcc, CXX=g++, FC=gfortran, F90=gfortan and then
run the mpich2 configure. Note, make sure your path and LD_LIBRARY_PATH
point to the correct gnu compiler version you want to use.

Alternatively consider using Redhat 6 (or the free CentOS 6) which is
officially supported by NVIDIA and just works. :-)

All the best
Ross



On 11/20/12 3:01 PM, "Thomas Evangelidis" <tevang3.gmail.com> wrote:

>OK, AMBER compiles on the cluster with MVAPICH2, CUDA-4.2 and GNU
>compilers
>but it fails on my laptop (Fedora 17 64bit, GeForce GT 650M). MPICH2 fails
>to compile AMBER, e.g. for pemd.MPI I get:
>
>/tmp/ifort5nUvG0.i90(742): catastrophic error: Too many errors, exiting
>compilation aborted for binrestart.F90 (code 1)
>
>In contrast, it is compiles with OpenMPI.
>
>On the other hand, pmemd.CUDA version yields an error like inlining failed
>in call to 'bool b40c::IsKeysOnly()' which is produced from file
>amber12/src/pmemd/src/cuda/B40C/kernel/radixsort_kernel_common.cu. Below
>is
>the full stack trace. Is there any change to make it work??
>
>sudo ln -s /usr/bin/gcc34 /usr/local/cuda/bin/gcc # I did this because
>gcc-4.7.2 is not supported by CUDA-4.2 (gcc34 = 3.4.6)
>configure -cuda gnu
>make install
>
>make[5]: Entering directory
>`/home/thomas/Programs/amber12/src/pmemd/src/cuda'
>/usr/local/cuda/bin/nvcc -use_fast_math -gencode
>arch=compute_13,code=sm_13
>-gencode arch=compute_20,code=sm_20 -gencode arch=compute_30,code=sm_30
>-DCUDA -Duse_SPFP -I/usr/local/cuda/include -IB40C -IB40C/KernelCommon -c
>kForcesUpdate.cu
>./kForcesUpdate.cu(87): Advisory: Loop was not unrolled, cannot deduce
>loop
>trip count
>./kForcesUpdate.cu(127): Advisory: Loop was not unrolled, cannot deduce
>loop trip count
>/usr/local/cuda/bin/nvcc -use_fast_math -gencode
>arch=compute_13,code=sm_13
>-gencode arch=compute_20,code=sm_20 -gencode arch=compute_30,code=sm_30
>-DCUDA -Duse_SPFP -I/usr/local/cuda/include -IB40C -IB40C/KernelCommon -c
>kCalculateLocalForces.cu
>/usr/local/cuda/bin/nvcc -use_fast_math -gencode
>arch=compute_13,code=sm_13
>-gencode arch=compute_20,code=sm_20 -gencode arch=compute_30,code=sm_30
>-DCUDA -Duse_SPFP -I/usr/local/cuda/include -IB40C -IB40C/KernelCommon -c
>kCalculateGBBornRadii.cu
>/usr/local/cuda/bin/nvcc -use_fast_math -gencode
>arch=compute_13,code=sm_13
>-gencode arch=compute_20,code=sm_20 -gencode arch=compute_30,code=sm_30
>-DCUDA -Duse_SPFP -I/usr/local/cuda/include -IB40C -IB40C/KernelCommon -c
>kCalculatePMENonbondEnergy.cu
>/usr/local/cuda/bin/nvcc -use_fast_math -gencode
>arch=compute_13,code=sm_13
>-gencode arch=compute_20,code=sm_20 -gencode arch=compute_30,code=sm_30
>-DCUDA -Duse_SPFP -I/usr/local/cuda/include -IB40C -IB40C/KernelCommon -c
>kCalculateGBNonbondEnergy1.cu
>/usr/local/cuda/bin/nvcc -use_fast_math -gencode
>arch=compute_13,code=sm_13
>-gencode arch=compute_20,code=sm_20 -gencode arch=compute_30,code=sm_30
>-DCUDA -Duse_SPFP -I/usr/local/cuda/include -IB40C -IB40C/KernelCommon -c
>kNLRadixSort.cu
>B40C/kernel/radixsort_kernel_common.cu: In static member function `static
>bool b40c::BaseRadixSortingEnactor<K, V, Storage>::KeysOnly() [with K =
>unsigned int, V = unsigned int, Storage =
>b40c::MultiCtaRadixSortStorage<unsigned int, unsigned int>]':
>B40C/kernel/radixsort_kernel_common.cu:134: sorry, unimplemented: inlining
>failed in call to 'bool b40c::IsKeysOnly() [with V = unsigned int]':
>function not considered for inlining
>tmpxft_00004c7f_00000000-6_kNLRadixSort.compute_13.cudafe1.stub.c:106:
>sorry, unimplemented: called from here
>make[5]: *** [kNLRadixSort.o] Error 1
>make[5]: Leaving directory
>`/home/thomas/Programs/amber12/src/pmemd/src/cuda'
>make[4]: *** [cuda/cuda.a] Error 2
>make[4]: Leaving directory `/home/thomas/Programs/amber12/src/pmemd/src'
>make[3]: *** [cuda] Error 2
>make[3]: Leaving directory `/home/thomas/Programs/amber12/src/pmemd'
>make[2]: *** [cuda] Error 2
>make[2]: Leaving directory `/home/thomas/Programs/amber12/src'
>make[1]: [cuda] Error 2 (ignored)
>make[1]: Leaving directory `/home/thomas/Programs/amber12/AmberTools/src'
>make[1]: Entering directory `/home/thomas/Programs/amber12/src'
>Starting installation of Amber12 (cuda) at Wed Nov 21 00:52:15 EET 2012.
>cd pmemd && make cuda
>make[2]: Entering directory `/home/thomas/Programs/amber12/src/pmemd'
>make -C src/ cuda
>make[3]: Entering directory `/home/thomas/Programs/amber12/src/pmemd/src'
>make -C ./cuda
>make[4]: Entering directory
>`/home/thomas/Programs/amber12/src/pmemd/src/cuda'
>/usr/local/cuda/bin/nvcc -use_fast_math -gencode
>arch=compute_13,code=sm_13
>-gencode arch=compute_20,code=sm_20 -gencode arch=compute_30,code=sm_30
>-DCUDA -Duse_SPFP -I/usr/local/cuda/include -IB40C -IB40C/KernelCommon -c
>kNLRadixSort.cu
>B40C/kernel/radixsort_kernel_common.cu: In static member function `static
>bool b40c::BaseRadixSortingEnactor<K, V, Storage>::KeysOnly() [with K =
>unsigned int, V = unsigned int, Storage =
>b40c::MultiCtaRadixSortStorage<unsigned int, unsigned int>]':
>B40C/kernel/radixsort_kernel_common.cu:134: sorry, unimplemented: inlining
>failed in call to 'bool b40c::IsKeysOnly() [with V = unsigned int]':
>function not considered for inlining
>tmpxft_00004f77_00000000-6_kNLRadixSort.compute_13.cudafe1.stub.c:106:
>sorry, unimplemented: called from here
>make[4]: *** [kNLRadixSort.o] Error 1
>make[4]: Leaving directory
>`/home/thomas/Programs/amber12/src/pmemd/src/cuda'
>make[3]: *** [cuda/cuda.a] Error 2
>make[3]: Leaving directory `/home/thomas/Programs/amber12/src/pmemd/src'
>make[2]: *** [cuda] Error 2
>make[2]: Leaving directory `/home/thomas/Programs/amber12/src/pmemd'
>make[1]: *** [cuda] Error 2
>make[1]: Leaving directory `/home/thomas/Programs/amber12/src'
>make: *** [install] Error 2
>_______________________________________________
>AMBER mailing list
>AMBER.ambermd.org
>http://lists.ambermd.org/mailman/listinfo/amber



_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Wed Nov 21 2012 - 10:30:03 PST
Custom Search