Merge pull request #1934 from RoyXue/GSoC2014_part2

doc/library/config.txt

@@ -320,6 +320,15 @@ import theano and print the config variable, as in:
     For the memory profile, do not print Apply nodes if the size
     of their outputs (in bytes) is lower than this.
 
+.. attribute:: profiling.min_peak_memory
+
+    Bool value: either True or False
+
+    Default False
+
+    Do the memory profile print the min peak memory usage?
+    It only works when profile=True, profile_memory=True
+
 .. attribute:: config.lib.amdlibm
 
     Bool value: either True or False

theano/compile/tests/test_profiling.py

@@ -2,36 +2,57 @@
 Test of memory profiling
 
 """
+import StringIO
+
+import numpy
+
 import theano
 import theano.tensor as T
-import StringIO
 
 
 def test_profiling():
 
-    old1 = theano.config.profile
-    old2 = theano.config.profile_memory
+    config1 = theano.config.profile
+    config2 = theano.config.profile_memory
+    config3 = theano.config.profiling.min_peak_memory
     try:
         theano.config.profile = True
         theano.config.profile_memory = True
+        theano.config.profiling.min_peak_memory = True
+
+        x = [T.dvector("val%i" % i) for i in range(3)]
+
+        z = []
+        z += [T.outer(x[i], x[i+1]).sum(axis=1) for i in range(len(x)-1)]
+        z += [x[i] + x[i+1] for i in range(len(x)-1)]
 
-        x = T.dvector("x")
-        y = T.dvector("y")
-        z = x + y
         p = theano.ProfileStats(False)
+
         if theano.config.mode in ["DebugMode", "DEBUG_MODE"]:
             m = "FAST_RUN"
         else:
             m = None
-        f = theano.function([x, y], z, profile=p, name="test_profiling",
+
+        f = theano.function(x, z, profile=p, name="test_profiling",
                             mode=m)
-        output = f([1, 2, 3, 4], [1, 1, 1, 1])
+
+        inp = [numpy.arange(1024) + 1 for i in range(len(x))]
+        output = f(*inp)
 
         buf = StringIO.StringIO()
         f.profile.summary(buf)
+
+        # regression testing for future algo speed up
+        the_string = buf.getvalue()
+        lines1 = [l for l in the_string.split("\n") if "Max if linker" in l]
+        lines2 = [l for l in the_string.split("\n") if "Minimum peak" in l]
+        assert "Max if linker=cvm(default): 8224KB (16408KB)" in the_string, (lines1, lines2)
+        assert "Minimum peak from all valid apply node order is 8208KB" in the_string, (lines1, lines2)
+
     finally:
-        theano.config.profile = old1
-        theano.config.profile_memory = old2
+        theano.config.profile = config1
+        theano.config.profile_memory = config2
+        theano.config.profiling.min_peak_memory = config3
 
 
 if __name__ == '__main__':

theano/gof/vm.py

@@ -149,6 +149,9 @@ class VM(object):
         if hasattr(self, 'node_cleared_order'):
             profile.node_cleared_order = self.node_cleared_order[:]
 
+        if hasattr(self, 'dependencies'):
+            profile.dependencies = self.dependencies.copy()
+
         # clear the timer info out of the buffers
         for i in xrange(len(self.call_times)):
             self.call_times[i] = 0.0