Weird error when compiling

Accelerated Computing CUDA CUDA Programming and Performance
1

Hi,

I am trying to compile an application that did it well under CUDA 3.2. Now I have updated to 4.0 and things seem broken. This is what happens when trying to compile something that depends on Intel TBB libraries (NOTE: The amd stuff is a non sense here that comes from the PARSEC stuff I am using… it has nothing to do here):

nvcc -L/home/jrbcast/parsec-2.1_cuda/pkgs/libs/gsl/inst/amd64-linux.icc/lib -L/usr/local/cuda/lib64 -lcuda -L/home/jrbcast/parsec-2.1_cuda/pkgs/libs/imagick/inst/amd64-linux.icc/lib -L/home/jrbcast/intel/composerxe-2011.3.174/lib64 -L/home/jrbcast/intel/composerxe-2011.3.174/lib -L/home/jrbcast/parsec-2.1_cuda/pkgs/apps/ferret/obj/amd64-linux.icc/parsec/lib -lrt -lm -lgsl -lgslcblas -lstdc++ -lcuda -lcudart -lcublas -lcass -lgsl -lgslcblas -lpthread -ltbb -I/home/jrbcast/parsec-2.1_cuda/pkgs/apps/ferret/src/include/ -c /home/jrbcast/parsec-2.1_cuda/pkgs/apps/ferret/obj/amd64-linux.icc/benchmark/ferret-tbb6.cu -o /home/jrbcast/parsec-2.1_cuda/pkgs/apps/ferret/obj/amd64-linux.icc/parsec/obj/ferret-tbb6.o
/home/jrbcast/intel/composerxe-2011.3.174/tbb/include/tbb/atomic.h: In member function ‘I tbb::internal::atomic_impl_with_arithmetic<I, D, StepType>::fetch_and_add(D)’:
/home/jrbcast/intel/composerxe-2011.3.174/tbb/include/tbb/atomic.h:223: error: expected primary-expression before ‘)’ token
/home/jrbcast/intel/composerxe-2011.3.174/tbb/include/tbb/atomic.h:223: error: expected ‘;’
/home/jrbcast/intel/composerxe-2011.3.174/tbb/include/tbb/atomic.h:223: error: ‘__T3’ has not been declared
make: *** [/home/jrbcast/parsec-2.1_cuda/pkgs/apps/ferret/obj/amd64-linux.icc/parsec/obj/ferret-tbb6.o] Error 1

Any hints?

Regards,

2

I am facing exactly the same problem. I tried to find a minimal program the reproduces the error and ended up with

#include "tbb/atomic.h"

int main() {

  return 0;

}

Compiling it on my Ubuntu 10.4 box with

nvcc -v -o tbb -I /usr/local/tbb30_20110427oss/include/ tbb.cu

I get

#$ _SPACE_= 

#$ _CUDART_=cudart

#$ _HERE_=/usr/local/cuda/bin

#$ _THERE_=/usr/local/cuda/bin

#$ _TARGET_SIZE_=64

#$ TOP=/usr/local/cuda/bin/..

#$ LD_LIBRARY_PATH=/usr/local/cuda/bin/../lib:/usr/local/cuda/bin/../extools/lib:/home/bauke/local/lib:/usr/local/cuda/lib64:/usr/local/intel/composerxe-2011.1.107/composerxe-2011.1.107/compiler/lib/intel64:/usr/local/intel/composerxe-2011.1.107/composerxe-2011.1.107/mkl/lib/intel64:/usr/local/intel/composerxe-2011.1.107/composerxe-2011.1.107/compiler/lib/intel64:/usr/local/intel/composerxe-2011.1.107/composerxe-2011.1.107/mpirt/lib/intel64:/usr/local/intel/composerxe-2011.1.107/composerxe-2011.1.107/ipp/../compiler/lib/intel64:/usr/local/intel/composerxe-2011.1.107/composerxe-2011.1.107/ipp/lib/intel64:/usr/local/intel/composerxe-2011.1.107/composerxe-2011.1.107/compiler/lib/intel64:/usr/local/intel/composerxe-2011.1.107/composerxe-2011.1.107/mkl/lib/intel64:/usr/local/intel/composerxe-2011.1.107/composerxe-2011.1.107/tbb/lib/intel64//cc4.1.0_libc2.4_kernel2.6.16.21

#$ PATH=/usr/local/cuda/bin/../open64/bin:/usr/local/cuda/bin:/usr/local/n1ge6/bin/lx24-amd64:/home/bauke/local/bin:/usr/local/cuda/bin:/usr/local/intel/composerxe-2011.1.107/composerxe-2011.1.107/bin/intel64:/usr/local/intel/composerxe-2011.1.107/composerxe-2011.1.107/mpirt/bin/intel64:/usr/local/texlive/2010/bin/x86_64-linux:/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin:/usr/games:/usr/local/texlive/2010/texmf/doc/info

#$ INCLUDES="-I/usr/local/cuda/bin/../include" "-I/usr/local/cuda/bin/../include/cudart"  

#$ LIBRARIES=  "-L/usr/local/cuda/bin/../lib64" -lcudart

#$ CUDAFE_FLAGS=

#$ OPENCC_FLAGS=

#$ PTXAS_FLAGS=

#$ gcc -D__CUDA_ARCH__=100 -E -x c++ -DCUDA_NO_SM_13_DOUBLE_INTRINSICS  -DCUDA_FLOAT_MATH_FUNCTIONS -DCUDA_NO_SM_11_ATOMIC_INTRINSICS -DCUDA_NO_SM_12_ATOMIC_INTRINSICS  -D__CUDACC__ -C  -I"/usr/local/tbb30_20110427oss/include/" "-I/usr/local/cuda/bin/../include" "-I/usr/local/cuda/bin/../include/cudart"   -include "cuda_runtime.h" -m64 -o "/tmp/tmpxft_00002eb4_00000000-4_tbb.cpp1.ii" "tbb.cu" 

#$ cudafe --m64 --gnu_version=40403 -tused --no_remove_unneeded_entities  --gen_c_file_name "/tmp/tmpxft_00002eb4_00000000-1_tbb.cudafe1.c" --stub_file_name "/tmp/tmpxft_00002eb4_00000000-1_tbb.cudafe1.stub.c" --gen_device_file_name "/tmp/tmpxft_00002eb4_00000000-1_tbb.cudafe1.gpu" --include_file_name "/tmp/tmpxft_00002eb4_00000000-3_tbb.fatbin.c" "/tmp/tmpxft_00002eb4_00000000-4_tbb.cpp1.ii" 

#$ gcc -D__CUDA_ARCH__=100 -E -x c -DCUDA_NO_SM_13_DOUBLE_INTRINSICS  -DCUDA_FLOAT_MATH_FUNCTIONS -DCUDA_NO_SM_11_ATOMIC_INTRINSICS -DCUDA_NO_SM_12_ATOMIC_INTRINSICS  -D__CUDACC__ -C  -D__CUDA_FTZ -I"/usr/local/tbb30_20110427oss/include/" "-I/usr/local/cuda/bin/../include" "-I/usr/local/cuda/bin/../include/cudart"   -m64 -o "/tmp/tmpxft_00002eb4_00000000-5_tbb.cpp2.i" "/tmp/tmpxft_00002eb4_00000000-1_tbb.cudafe1.gpu" 

#$ cudafe --m64 --gnu_version=40403 --c  --gen_c_file_name "/tmp/tmpxft_00002eb4_00000000-6_tbb.cudafe2.c" --stub_file_name "/tmp/tmpxft_00002eb4_00000000-6_tbb.cudafe2.stub.c" --gen_device_file_name "/tmp/tmpxft_00002eb4_00000000-6_tbb.cudafe2.gpu" --include_file_name "/tmp/tmpxft_00002eb4_00000000-3_tbb.fatbin.c" "/tmp/tmpxft_00002eb4_00000000-5_tbb.cpp2.i" 

#$ gcc -D__CUDA_ARCH__=100 -E -x c -DCUDA_NO_SM_13_DOUBLE_INTRINSICS  -DCUDA_FLOAT_MATH_FUNCTIONS -DCUDA_NO_SM_11_ATOMIC_INTRINSICS -DCUDA_NO_SM_12_ATOMIC_INTRINSICS  -D__CUDABE__  -D__CUDA_FTZ -I"/usr/local/tbb30_20110427oss/include/" "-I/usr/local/cuda/bin/../include" "-I/usr/local/cuda/bin/../include/cudart"   -m64 -o "/tmp/tmpxft_00002eb4_00000000-7_tbb.cpp3.i" "/tmp/tmpxft_00002eb4_00000000-6_tbb.cudafe2.gpu" 

#$ filehash -s " " "/tmp/tmpxft_00002eb4_00000000-7_tbb.cpp3.i" > "/tmp/tmpxft_00002eb4_00000000-8_tbb.hash"

#$ gcc -E -x c++ -D__CUDACC__ -C  -I"/usr/local/tbb30_20110427oss/include/" "-I/usr/local/cuda/bin/../include" "-I/usr/local/cuda/bin/../include/cudart"   -include "cuda_runtime.h" -m64 -o "/tmp/tmpxft_00002eb4_00000000-9_tbb.cpp4.ii" "tbb.cu" 

#$ cudafe++ --m64 --gnu_version=40403 --parse_templates  --gen_c_file_name "/tmp/tmpxft_00002eb4_00000000-1_tbb.cudafe1.cpp" --stub_file_name "/tmp/tmpxft_00002eb4_00000000-1_tbb.cudafe1.stub.c" "/tmp/tmpxft_00002eb4_00000000-9_tbb.cpp4.ii" 

#$ nvopencc  -TARG:compute_10 -m64 -OPT:ftz=1 -CG:ftz=1 -CG:prec_div=0 -CG:prec_sqrt=0  "/tmp/tmpxft_00002eb4_00000000-10_tbb" "/tmp/tmpxft_00002eb4_00000000-7_tbb.cpp3.i"  -o "/tmp/tmpxft_00002eb4_00000000-2_tbb.ptx"

#$ ptxas  -arch=sm_10 -m64  "/tmp/tmpxft_00002eb4_00000000-2_tbb.ptx"  -o "/tmp/tmpxft_00002eb4_00000000-11_tbb.sm_10.cubin" 

#$ fatbinary --create="/tmp/tmpxft_00002eb4_00000000-3_tbb.fatbin" --key="1abd7ece1f0930d3" --ident="tbb.cu" -cuda "--image=profile=compute_10,file=/tmp/tmpxft_00002eb4_00000000-2_tbb.ptx" "--image=profile=sm_10,file=/tmp/tmpxft_00002eb4_00000000-11_tbb.sm_10.cubin" --embedded-fatbin="/tmp/tmpxft_00002eb4_00000000-3_tbb.fatbin.c" 

#$ rm /tmp/tmpxft_00002eb4_00000000-3_tbb.fatbin

#$ gcc -D__CUDA_ARCH__=100 -E -x c++ -DCUDA_NO_SM_13_DOUBLE_INTRINSICS  -DCUDA_FLOAT_MATH_FUNCTIONS -DCUDA_NO_SM_11_ATOMIC_INTRINSICS -DCUDA_NO_SM_12_ATOMIC_INTRINSICS   -D__CUDA_FTZ -I"/usr/local/tbb30_20110427oss/include/" "-I/usr/local/cuda/bin/../include" "-I/usr/local/cuda/bin/../include/cudart"   -m64 -o "/tmp/tmpxft_00002eb4_00000000-12_tbb.ii" "/tmp/tmpxft_00002eb4_00000000-1_tbb.cudafe1.cpp" 

#$ gcc -c -x c++ -I"/usr/local/tbb30_20110427oss/include/" "-I/usr/local/cuda/bin/../include" "-I/usr/local/cuda/bin/../include/cudart"   -fpreprocessed -m64 -o "/tmp/tmpxft_00002eb4_00000000-13_tbb.o" "/tmp/tmpxft_00002eb4_00000000-12_tbb.ii" 

/usr/local/tbb30_20110427oss/include/tbb/atomic.h: In member function ‘I tbb::internal::atomic_impl_with_arithmetic<I, D, StepType>::fetch_and_add(D)’:

/usr/local/tbb30_20110427oss/include/tbb/atomic.h:231: error: ‘__T3’ has not been declared

# --error 0x1 --

I would like to switch from CUDA 3.2 to 4.0. This error, however, is a real show-stopper for me because my applications use both CUDA and Intel Threading Building Blocks. Any hints how to fix this?

3

Hi,

I finally found the solution. I don’t know the exact reason, but it seems that nvcc does not automatically detect now if you are dealing with C or C++ code. Therefore, you must add the correct flag to overcome the problem:

Try this: nvcc --x c++ …

Good luck,

Jose.

4

This solved my problem. Great!

Heiko