Fix repeat op tests on windows.

Fix recompilation of Ops in other repo then Theano

Add assert about the shape in the graph.

This fix problem of some module in the cache being skipped.

This speed up the loading and the refresh.
The refresh with a local cache with ~650 element went from 0.01s to 0.001s.

fix gh-1738

Small modif to allow the buildbot script to easily submit new results wi...

[MRG]Better error message when a thunk fail: print user code line when node was created

Use gradient_undefined instead of None in MRG uniform

this is more clear and allow to see all of them at once.

Fixed memory leaks in cumsum and cumprod.

Raise the backtrace limit to caught where v1 + v2 is created in user script.
Remove the add_tag_trace from the bt.

Conversion of GpuSoftmax and GpuSoftmaxWithBias to the new backend

Fix unpickling of Elemwise op since openmp changes.

Tag slow tests

Don't try to use the c_code for GpuElemwise on opencl since it's cuda-only.

Replace content of c_init_code by reference to function setup_ext_cuda() in GpuSoftmax and GpuSoftmaxWithBias

Elemwise openmp

Fix Sparse grad crash and implement mixed dtype in sparse Mul/Add

I changed how format is specified. There was no bug as the input
variable are put in the key. Now I make explicit in the op key the
format.

This simplify at a few places the code too.