Make local_dot22_to_dot22scalar opt work in one more case.

theano/tensor/blas.py

@@ -1906,7 +1906,12 @@ def local_dot22_to_dot22scalar(node):
     d = node.inputs[dot22_idx]
     i_scalar = [_as_scalar(x, dtype=d.dtype) for x in node.inputs]
     if not any(i_scalar):
-        i_mul = [x.owner and x.owner.op == T.mul for x in node.inputs]
+        # Check witch input of node is a mul with scalar inputs.
+        # We could reuse this scalar value.
+        i_mul = [x.owner and x.owner.op == T.mul and
+                 any([_as_scalar(x_i, dtype=d.dtype)
+                   for x_i in x.owner.inputs])
+                 for x in node.inputs]
         if not any(i_mul):
             #no scalar in input and no multiplication
             #if their was a multiplication we couls reorder the graph
@@ -1916,11 +1921,11 @@ def local_dot22_to_dot22scalar(node):
         #maybe we can reorder the graph as this mul have a mul in input.
         #The canonizer should have merged those mul together.
         #We support only 1 additional level of mul.
-        mul_idx = i_mul.index(True)  # we take the first mul!
+        mul_idx = i_mul.index(True)  # The first one should always work
         m = node.inputs[mul_idx]
 
         if any([_as_scalar(x, dtype=d.dtype)
-                for x in m.owner.inputs]):
+                for x in m.owner.inputs]):  # This should be always True
             scalar_idx = -1
             for i, x in enumerate(m.owner.inputs):
                 if _as_scalar(x, dtype=d.dtype) and (theano.scalar.upcast(

theano/tensor/tests/test_blas.py

@@ -1067,8 +1067,8 @@ def test_local_dot22_to_dot22scalar():
         T.mul(_dot22(A, A), T.mul(m, y, z), m),
         T.mul(_dot22(A, A), m, T.mul(m, y, z)),
 
-        #Case that isn't opt for now.
-        #T.mul(_dot22(A, A), (r * m), (m * x)),
+        #Case that opt later in gh-1515
+        T.mul(_dot22(A, A), (r * m), (m * x)),
     ]):
         node2 = theano.tensor.blas.local_dot22_to_dot22scalar.transform(
             node.owner)