Hi,
I just wanted to test NPP for the first time. Therefore I modified my library (currently only IPP (library) and OpenCV (library and test application) included as external dependencies). To add NPP I added all CUDA dependencies (Cuda 2.3) and npp. When i link the library against libnpp64.so i get a segfault right away when i call the test application. I do not include any headers, nor do i call npp functions. When calling npp functions the behaviour is of course the same.
System:
OS: Ubuntu 9.10
gcc: 4.3.4
Cuda: 2.3
(GeForce GTX 280)
The readme file of NPP says that the library to link agains is common/lib/libnpp-mt.lib. Well this is true for Windows but for Linux this should be libnpp64.so?
Debugging with gdb does not really help:
#0 0x00007ffff4c9218e in ?? () from /usr/local/cuda/lib64/libcudart.so.2
#1 0x00007ffff4c67e79 in __cudaRegisterFatBinary () from /usr/local/cuda/lib64/libcudart.so.2
#2 0x00007ffff3770ea0 in __sti____cudaRegisterAll_48_tmpxft_0000241d_00000000_4_WarpT
ransform_cpp1_ii_980a611b () from /usr/local/cuda/NVIDIA_NPP/common/lib/libnpp64.so.1
#3 0x00007ffff37a360b in global constructors keyed to _Z19getInverseTransformILi0EEvPKdPd () from /usr/local/cuda/NVIDIA_NPP/common/lib/libnpp64.so.1
#4 0x00007ffff3817ed6 in __do_global_ctors_aux () from /usr/local/cuda/NVIDIA_NPP/common/lib/libnpp64.so.1
#5 0x00007ffff371c04b in _init () from /usr/local/cuda/NVIDIA_NPP/common/lib/libnpp64.so.1
#6 0x00007fffffffd8b8 in ?? ()
#7 0x00007ffff7ded6e9 in call_init (l=0x7ffff7e934e0, argc=-195902584, argv=0x7fffffffd8a8, env=0x7fffffffd8b8) at dl-init.c:70
#8 0x00007ffff7ded86f in _dl_init (main_map=0x7ffff7ffb000, argc=1, argv=0x7fffffffd8a8, env=0x7fffffffd8b8) at dl-init.c:134
#9 0x00007ffff7ddfb2a in _dl_start_user () from /lib64/ld-linux-x86-64.so.2
#10 0x0000000000000001 in ?? ()
#11 0x00007fffffffdbc3 in ?? ()
#12 0x0000000000000000 in ?? ()
Thanks for any hints so that i can play around with NPP.
Best,
Manuel