add the local_cut_gpu_host_gpu optimizer into the canonicalize phase.

This allow all other optimization to work correctly. This allow the insertion of gemm into gpu code in the deep learning tutorial of the mlp to have a speed up of 3x.

add the local_cut_gpu_host_gpu optimizer into the canonicalize phase.
f96de3b3 · Frederic Bastien · bb96ebe8 · f96de3b3
--- a/theano/sandbox/cuda/opt.py
+++ b/theano/sandbox/cuda/opt.py
@@ -70,6 +70,9 @@ gpu_cut_copies.register('cut_gpu_host_transfers', local_cut_gpu_host_gpu,
        'fast_run', 'inplace', 'gpu')
 gpu_cut_copies.register('cut_gpu_constant_transfers', tensor.opt.constant_folding, 
        'fast_run', 'gpu')
+#register it into canonicalize to allow other optimization to work without
+#botering with this useless pattern.
+compile.optdb['canonicalize'].register('local_cut_gpu_host_gpu', local_cut_gpu_host_gpu, 'fast_run')
 @register_opt()
 @local_optimizer([])