Add (py)CUDA test/info script

[pgleeson]

A simple python script to check the local installation of (py)CUDA would be very helpful, giving info on current configuration, hardware specs etc.

[pgleeson]

Getting an error when running python generic_demo.py

Process LPU-2:
Traceback (most recent call last):
  File "/usr/lib/python2.7/multiprocessing/process.py", line 258, in _bootstrap
    self.run()
  File "/home/padraig/NK/src/neurokernel/neurokernel/core.py", line 772, in run
    self.pre_run()
  File "/home/padraig/NK/src/neurokernel/neurokernel/LPU/LPU.py", line 262, in pre_run
    self._initialize_gpu_ds()
  File "/home/padraig/NK/src/neurokernel/neurokernel/LPU/LPU.py", line 606, in _initialize_gpu_ds
    len(self.input_neuron_list), np.float64)
  File "/home/padraig/NK/local/lib/python2.7/site-packages/pycuda/gpuarray.py", line 932, in zeros
    result.fill(zero)
  File "/home/padraig/NK/local/lib/python2.7/site-packages/pycuda/gpuarray.py", line 525, in fill
    func = elementwise.get_fill_kernel(self.dtype)
  File "<string>", line 2, in get_fill_kernel
  File "/home/padraig/NK/local/lib/python2.7/site-packages/pycuda/tools.py", line 423, in context_dependent_memoize
    result = func(*args)
  File "/home/padraig/NK/local/lib/python2.7/site-packages/pycuda/elementwise.py", line 486, in get_fill_kernel
    "fill")
  File "/home/padraig/NK/local/lib/python2.7/site-packages/pycuda/elementwise.py", line 157, in get_elwise_kernel
    arguments, operation, name, keep, options, **kwargs)
  File "/home/padraig/NK/local/lib/python2.7/site-packages/pycuda/elementwise.py", line 143, in get_elwise_kernel_and_types
    keep, options, **kwargs)
  File "/home/padraig/NK/local/lib/python2.7/site-packages/pycuda/elementwise.py", line 71, in get_elwise_module
    options=options, keep=keep)
  File "/home/padraig/NK/local/lib/python2.7/site-packages/pycuda/compiler.py", line 251, in __init__
    arch, code, cache_dir, include_dirs)
  File "/home/padraig/NK/local/lib/python2.7/site-packages/pycuda/compiler.py", line 241, in compile
    return compile_plain(source, options, keep, nvcc, cache_dir)
  File "/home/padraig/NK/local/lib/python2.7/site-packages/pycuda/compiler.py", line 73, in compile_plain
    checksum.update(preprocess_source(source, options, nvcc).encode("utf-8"))
  File "/home/padraig/NK/local/lib/python2.7/site-packages/pycuda/compiler.py", line 52, in preprocess_source
    cmdline, stderr=stderr)
CompileError: nvcc preprocessing of /tmp/tmpUYPlDl.cu failed
[command: nvcc --preprocess -arch sm_12 -I/home/padraig/NK/lib/python2.7/site-packages/pycuda/cuda /tmp/tmpUYPlDl.cu --compiler-options -P]
[stderr:
In file included from /usr/local/cuda/bin/../include/cuda_runtime.h:59:0,
                 from <command-line>:0:
/usr/local/cuda/bin/../include/host_config.h:82:2: error: #error -- unsupported GNU version! gcc 4.5 and up are not supported!
 #error -- unsupported GNU version! gcc 4.5 and up are not supported!
  ^
]

[pgleeson]

I managed to get some of the nk examples running successfully by updating my version of the CUDA libraries & pycuda (so ignore the error above), but a helper script for checking the local installation & compatibility would be very useful.

[lebedov]

I added more information on how to examine the available GPUs to the documentation and how to get PyCUDA to install properly. I also added a script (check_cuda_version.py) to the master repo for checking the installed CUDA version; we can combine it with the other installation steps into a single script to streamline the installation procedure.

[pgleeson]

Great, thanks. Worked for me.

Would be a good idea too as you say to integrate it into the set-up scripts to tell users whether the minimal requirements have been satisfied.