I was benchmarking with stream.cu.
I installed:
NVIDIA-Linux-x86_64-180.22-pkg2.run
cuda-sdk-linux-2.10.1215.2015-3233425.run
cudatoolkit_2.1_linux64_ubuntu8.04.run
$ which nvcc
/usr/local/cuda/bin/nvcc
$ nvcc -V
nvcc: NVIDIA ® Cuda compiler driver
Copyright © 2005-2007 NVIDIA Corporation
Built on Wed_Dec__3_16:25:17_PST_2008
Cuda compilation tools, release 2.1, V0.2.1221
So I I compiled and run:
export PATH=/usr/local/cuda/bin:$PATH
export LD_LIBRARY_PATH=/usr/local/cuda/lib
$ nvcc stream.cu
$ ./a.out
STREAM Benchmark implementation in CUDA
Array size (single precision)=2000000
using 128 threads per block, 15625 blocks
Function Rate (MB/s) Avg time Min time Max time
Copy: 48072.2521 0.0003 0.0003 0.0003
Scale: 46250.0786 0.0003 0.0003 0.0003
Add: 52869.3782 0.0005 0.0005 0.0005
Triad: 53572.8026 0.0004 0.0004 0.0005
If I try -O1 it fails, the man page mentions --optimize (-O).
$ nvcc -O1 stream.cu
/usr/include/bits/stdio2.h(35): error: identifier “__builtin_va_arg_pack” is undefined
/usr/include/bits/stdio2.h(66): error: identifier “__builtin_va_arg_pack” is undefined
/usr/include/bits/stdio2.h(99): error: identifier “__builtin_va_arg_pack” is undefined
/usr/include/bits/stdio2.h(105): error: identifier “__builtin_va_arg_pack” is undefined
/usr/include/bits/stdio2.h(159): error: identifier “__builtin_va_arg_pack” is undefined
/usr/include/bits/stdio2.h(167): error: identifier “__builtin_va_arg_pack” is undefined
/usr/include/bits/stdio2.h(174): error: identifier “__builtin_va_arg_pack” is undefined
/usr/include/bits/stdio2.h(182): error: identifier “__builtin_va_arg_pack” is undefined
8 errors detected in the compilation of “/tmp/tmpxft_000055df_00000000-4_stream.cpp1.ii”.
Ideas?