Can't compile on CUDA 3.2

[examachine]

On Mac OS X I get the following errors with the new RC:


sirius:cupp malfunct$ make
[ 9%] Built target cupp
[ 18%] Built target kernel_memory1d
[ 27%] Building CXX object examples/memory1d/CMakeFiles/
memory1d_example.dir/memory1d.o
In file included from /Users/malfunct/Source/cupp/include/cupp/
device.h:14,
from /Users/malfunct/Source/cupp/examples/memory1d/
memory1d.cpp:9:
/usr/local/cuda/include/cuda_runtime.h:148: warning: unused parameter
‘flags’
/Users/malfunct/Source/cupp/include/cupp/kernel_impl/
kernel_launcher_impl.h: In member function ‘void
cupp::kernel_impl::kernel_launcher_impl<T>::configure_call() [with F_
= void (*)(cupp::deviceT::memory1d<int, cupp::memory1d<int> >*)]’:
/Users/malfunct/Source/cupp/examples/memory1d/memory1d.cpp:57:
instantiated from here
/Users/malfunct/Source/cupp/include/cupp/kernel_impl/
kernel_launcher_impl.h:193: error: invalid conversion from ‘int’ to
‘CUstream_st*’
/Users/malfunct/Source/cupp/include/cupp/kernel_impl/
kernel_launcher_impl.h:193: error: initializing argument 4 of
‘cudaError_t cudaConfigureCall(dim3, dim3, size_t, CUstream_st*)’
make[2]: *** [examples/memory1d/CMakeFiles/memory1d_example.dir/
memory1d.o] Error 1
make[1]: *** [examples/memory1d/CMakeFiles/memory1d_example.dir/all]
Error 2
make: *** [all] Error 2

[Jens Breitbart]

Hi,

sorry, there was a bug in the kernel.h and kernel_launcher_impl.h. I have attached you the two files to fix this issue. They compile without a problem, however I currently have problems with my CUDA system so I can't test the binaries.

Best,

Jens

--
You received this message because you are subscribed to the Google Groups "CuPP" group.
To post to this group, send email to cu...@googlegroups.com.
To unsubscribe from this group, send email to cupp+uns...@googlegroups.com.
For more options, visit this group at http://groups.google.com/group/cupp?hl=en.

--

Wanted: Men/women for hazardous architectures. Low wages. Bitter cold. Long hours of software development. Safe return doubtful. Honor and recognition in the event of success.

[Eray Ozkural]

Hi Jens,

Thanks for your interest. I will try to test them. Much appreciated!!

I really like the direction you're going, though I think with CUDA 3.2 much more could be done, especially since there is a working kernel malloc function. I'm right now hacking thrust so that I can construct vectors inside kernels, if that works all right it will make my life easier, I think a similar enhancement would be possible also for cupp.

Cheers,

Eray Ozkural, PhD candidate.  Comp. Sci. Dept., Bilkent University, Ankara
http://groups.yahoo.com/group/ai-philosophy
http://myspace.com/arizanesil http://myspace.com/malfunct

[Jens Breitbart]

Hi Eray,

yes that is true, allocating memory is a good addition and should make things easier in some cases. I had no chance of playing with this feature yet, so I am not sure how expensive it is and if one should really use it. Have you made any experiments and measured its performance?

Best,

Jens

[Eray Ozkural]

No, not really, I didn't use it inside a container yet. But working on it.