port GpuConv to the new gpu back-end

Make the gpuarray opt for conv use it.

OpFromGraph documentation and cleanup

GpuIncSubtensor

Fix test in the daily buildbot by doing the comparison on ndarray, not C...

Try to detect when gcc incorrectly identifies processors as -avx

Fix display of test outputs and known failures in theano.test()

Fix compilation crash instroduced in gh-1673 on 32 bit python.

Do not use python code for elemwise with >= 32 inputs

NumPy is not able to deal with that many inputs to a ufunc,
and will segfault or raise an Exception.

[MRG] Add c_code for MaxAndArgmax