Remove commented calls to register_tracks since that mechanism isn't needed.

theano/sandbox/cuda/opt.py

@@ -770,9 +770,6 @@ def local_gpu_advanced_subtensor1(node):
     return False
 
 
-#tensor.opt.local_inplace_incsubtensor1.add_track(GpuAdvancedIncSubtensor1)
-#tensor.opt.local_inplace_incsubtensor1.add_track(GpuAdvancedIncSubtensor1_dev20)
-
 @register_opt()
 @local_optimizer([gpu_from_host, tensor.AdvancedIncSubtensor1])
 def local_gpu_advanced_incsubtensor1(node):
@@ -847,8 +844,6 @@ def local_gpu_advanced_incsubtensor1(node):
     return False
 
 
-#tensor.opt.local_inplace_setsubtensor.add_track(GpuIncSubtensor)
-
 @register_opt()
 @local_optimizer([gpu_from_host, tensor.IncSubtensor])
 def local_gpu_incsubtensor(node):