added optimization in tensor/blas to use dot22 for broadcasted tensors

theano/tensor/blas.py

@@ -571,6 +571,10 @@ def _is_real_matrix(res):
             and res.type.ndim == 2 \
             and res.type.broadcastable[0] == False \
             and res.type.broadcastable[1] == False #cope with tuple vs. list
+def _is_real_vector(res):
+    return res.type.dtype in ('float32', 'float64') \
+            and res.type.ndim == 1 \
+            and res.type.broadcastable[0] == False 
 
 def _beta_L_plus_alpha_M(beta, L, alpha, M, recurse_flip = True):
     #print 'BETA L + ALPHA M', beta, L, alpha, M, recurse_flip
@@ -805,13 +809,13 @@ class Dot22(GemmRelated):
     This is a specialization of the more general Dot()
     """
     def make_node(self, x, y):
-        if not _is_real_matrix(x):
+        if x.type.ndim != 2 or x.type.dtype not in ('float32', 'float64'):
             raise TypeError(x)
-        if not _is_real_matrix(x):
+        if y.type.ndim != 2 or y.type.dtype not in ('float32', 'float64'):
             raise TypeError(y)
         if y.type.dtype != x.type.dtype:
             raise TypeError('dtype mismatch to Dot22')
-        bz = [False, False]
+        bz = (x.type.broadcastable[0], y.type.broadcastable[1])
         outputs = [T.tensor(x.type.dtype, bz)]
         return Apply(self, [x,y], outputs)
 
@@ -870,14 +874,26 @@ _dot22 = Dot22()
 
 @local_optimizer([T.dot])
 def local_dot_to_dot22(node):
-    if node.op == T.dot:
+    if node.op != T.dot:
+        return
+
     x,y = node.inputs
-        if _is_real_matrix(x) and _is_real_matrix(y) and y.type.dtype == x.type.dtype:
+    if y.type.dtype != x.type.dtype:
+        # TODO: upcast one so the types match
+        info('Not optimizing dot with inputs', x, y, x.type, y.type)
+        return
+    print 'asdfasdf'
+    if y.type.dtype.startswith('float'):
+        if _is_real_matrix(x) and _is_real_matrix(y):
             return [_dot22(*node.inputs)]
-        else:
+        if _is_real_matrix(x) and _is_real_vector(y):
+            return [_dot22(x, y.dimshuffle(0,'x')).dimshuffle(0)]
+        if _is_real_vector(x) and _is_real_matrix(y):
+            return [_dot22(x.dimshuffle('x',0), y).dimshuffle(1)]
+        if _is_real_vector(x) and _is_real_vector(x):
+            return [_dot22(x.dimshuffle('x',0), y.dimshuffle(0,'x')).dimshuffle()]
+
     info('Not optimizing dot with inputs', x, y, x.type, y.type)
-    else:
-        return False
 
 @local_optimizer([gemm_no_inplace])
 def local_inplace_gemm(node):