tests pass with new casting

theano/tensor/nnet/corr.py

@@ -75,6 +75,34 @@ class BaseCorrMM(gof.Op):
             str(self.subsample),
             str(self.filter_dilation))
 
+    def cast(self, in1, in2):
+         """
+         Upcast input variables if neccesary.
+         """
+         float_types = ['float32', 'float64']
+         dtype_1 = in1.type.dtype
+         dtype_2 = in2.type.dtype
+ 
+         assert dtype_1 in float_types
+         assert dtype_2 in float_types
+
+         if dtype_1 == 'float64' or dtype_2 == 'float64':
+             dtype_o = 'float64'
+         else:
+             dtype_o = 'float32'
+ 
+         if dtype_1 != dtype_o:
+             out1 = in1.astype(dtype_o)
+         else:
+             out1 = in1
+ 
+         if dtype_2 != dtype_o:
+             out2 = in2.astype(dtype_o)
+         else:
+             out2 = in2
+ 
+         return out1, out2
+
     def c_support_code(self):
         return blas_header_text()
 
@@ -383,6 +411,7 @@ class CorrMM(BaseCorrMM):
     def make_node(self, img, kern):
         img = as_tensor_variable(img)
         kern = as_tensor_variable(kern)
+        img, kern = self.cast(img, kern)
         if img.type.ndim != 4:
             raise TypeError('img must be 4D tensor')
         if kern.type.ndim != 4:
@@ -445,6 +474,7 @@ class CorrMM_gradWeights(BaseCorrMM):
     def make_node(self, img, topgrad, shape=None):
         img = as_tensor_variable(img)
         topgrad = as_tensor_variable(topgrad)
+        img, topgrad = self.cast(img, topgrad)
         if img.type.ndim != 4:
             raise TypeError('img must be 4D tensor')
         if topgrad.type.ndim != 4:
@@ -546,6 +576,7 @@ class CorrMM_gradInputs(BaseCorrMM):
     def make_node(self, kern, topgrad, shape=None):
         kern = as_tensor_variable(kern)
         topgrad = as_tensor_variable(topgrad)
+        kern, topgrad = self.cast(kern, topgrad)
         if kern.type.ndim != 4:
             raise TypeError('kern must be 4D tensor')
         if topgrad.type.ndim != 4: