add profile of theano.grad

theano/compile/profiling.py

@@ -654,6 +654,9 @@ class ProfileStats(object):
         # The validation time is a subset of optimizer_time
         assert self.validate_time < self.optimizer_time
 
+    def summary_globals(self, file):
+        print >> file, 'Time in all call to theano.grad() %es' % theano.gradient.grad_time
+
     def summary_memory(self, file, N=None):
         fct_memory = {}  # fgraph->dict(node->[outputs size])
         fct_shapes = {}  # fgraph->dict(node->[outputs shapes]))
@@ -1209,6 +1212,7 @@ class ProfileStats(object):
     def summary(self, file=sys.stderr, n_ops_to_print=20,
                 n_apply_to_print=20):
         self.summary_function(file)
+        self.summary_globals(file)
         local_time = sum(self.apply_time.values())
         if local_time > 0:
             self.summary_class(file, n_ops_to_print)

theano/gradient.py

@@ -10,6 +10,7 @@ __docformat__ = "restructuredtext en"
 import __builtin__
 from itertools import izip
 import logging
+import time
 import warnings
 _logger = logging.getLogger('theano.gradient')
 
@@ -36,6 +37,8 @@ tensor = None
 
 _msg_retType = 'op.grad(...) returned a non-list'
 
+grad_time = 0
+
 
 def format_as(use_list, use_tuple, outputs):
     """
@@ -412,6 +415,7 @@ def grad(cost, wrt, consider_constant=None,
              or Variable in all cases.
 
     """
+    t0 = time.time()
     global tensor
     if tensor is None:
         from theano import tensor
@@ -556,8 +560,12 @@ def grad(cost, wrt, consider_constant=None,
         rval = tuple(rval)
     elif not using_list:
         rval, = rval
+    t1 = time.time()
+    global grad_time
+    grad_time = t1 - t0
     return rval
 
+
 def subgraph_grad(wrt, end, start=None, cost=None, details=False):
     '''
     With respect to `wrt`, computes gradients of cost and/or from