The basic idea is to apply autocasting to constants but not to shared scalars.

Joint work with Fred.

call the Alloc instance don't return useless Alloc. This will be discussed in an email on the mailing list.

1) scipy is as fast as our implementation now (+- 10%)
2) the optimized graph hides shape mismatches (and currently can cause a
segfault)

refactored convolution test on the gpu to don't execution version of code that will execute the refenrece version because the optimization can't be applyed.

Made the type of exception raised in that case an InconsistencyError (was
ValueError) and made the GemmOptimizer catch that exception when trying to
insert gemms into a graph.