00-theano-advancedinctensor.patch

--- theano/sandbox/cuda/opt.py	2017-03-21 18:45:53.532335945 +0100
+++ theano/sandbox/cuda/opt.py	2017-03-21 18:45:53.532335945 +0100
@@ -1111,7 +1111,7 @@
 
                 gpu_op = GpuAdvancedIncSubtensor1(**props_dict)
             else:
-                gpu_op = GpuAdvancedIncSubtensor1_dev20(**props_dict)
+                gpu_op = GpuAdvancedIncSubtensor1(**props_dict)
             return [gpu_op(as_cuda_ndarray_variable(x),
                            as_cuda_ndarray_variable(y), *coords)]
 
@@ -1149,7 +1149,7 @@
             if (compute_capability < 2 or y.ndim != 2 or x.ndim != 2):
                 gpu_op = GpuAdvancedIncSubtensor1(**node.op._props_dict())
             else:
-                gpu_op = GpuAdvancedIncSubtensor1_dev20(**node.op._props_dict())
+                gpu_op = GpuAdvancedIncSubtensor1(**node.op._props_dict())
             return [host_from_gpu(gpu_op(gpu_x, gpu_y, *coords))]
     return False
 
--- theano/gpuarray/opt.py	2017-03-21 18:42:35.589317691 +0100
+++ theano/gpuarray/opt.py	2017-03-21 18:42:35.589317691 +0100
@@ -68,8 +68,7 @@
 from .subtensor import (GpuIncSubtensor, GpuSubtensor,
                         GpuAdvancedSubtensor,
                         GpuAdvancedSubtensor1,
-                        GpuAdvancedIncSubtensor1,
-                        GpuAdvancedIncSubtensor1_dev20)
+                        GpuAdvancedIncSubtensor1)
 from .opt_util import alpha_merge, output_merge, pad_dims, unpad_dims
 from .reduction import GpuMaxAndArgmax
 from .linalg import (GpuCusolverSolve, cusolver_available)
@@ -1056,7 +1055,7 @@
     if compute_capability >= 2 and x.ndim == 1 and y.ndim == 0:
         x = x.dimshuffle(0, 'x')
         y = y.dimshuffle('x', 'x')
-        ret = GpuAdvancedIncSubtensor1_dev20(
+        ret = GpuAdvancedIncSubtensor1(
             set_instead_of_inc=set_instead_of_inc)(x, y, ilist)
         ret = GpuDimShuffle(ret.type.broadcastable, [0])(ret)
         return ret
@@ -1064,15 +1063,14 @@
         return GpuAdvancedIncSubtensor1(
             set_instead_of_inc=set_instead_of_inc)
     else:
-        return GpuAdvancedIncSubtensor1_dev20(
+        return GpuAdvancedIncSubtensor1(
             set_instead_of_inc=set_instead_of_inc)
 
 
 @register_inplace()
-@local_optimizer([GpuAdvancedIncSubtensor1, GpuAdvancedIncSubtensor1_dev20])
+@local_optimizer([GpuAdvancedIncSubtensor1])
 def local_advincsub1_gpua_inplace(node):
-    if isinstance(node.op, (GpuAdvancedIncSubtensor1,
-                            GpuAdvancedIncSubtensor1_dev20)):
+    if isinstance(node.op, (GpuAdvancedIncSubtensor1,)):
         if not node.op.inplace:
             return [node.op.clone_inplace()(*node.inputs)]