Merge pull request #1719 from nouiz/gpu_windows

theano/compile/profiling.py

@@ -342,7 +342,7 @@ class ProfileStats(object):
         es += ['   %2s ']
 
         hs += ['<#call>']
-        es += ['  %4d  ']
+        es += ['  %5d  ']
 
         hs += ['<#apply>']
         es += ['  %4d  ']
@@ -362,9 +362,12 @@ class ProfileStats(object):
                 continue
             tot += t
             ftot = tot * 100 / local_time
+            # Remove the useless start and end of the class name:
+            # "<class 'theano.sandbox.cuda.blas.GpuDot22'>" -> "theano.sandbox.cuda.blas.GpuDot22"
+            class_name = str(a)[8:-2][:maxlen]
             print >> file, format_str % (f, ftot, t, t / nb_call,
                                          impl, nb_call,
-                                         nb_apply, str(a)[:maxlen])
+                                         nb_apply, class_name)
             # While this carries over less information, it is arranged such
             # that it way more readeable that the previous output of the
             # profiler

theano/tensor/subtensor.py

@@ -644,8 +644,14 @@ class Subtensor(Op):
         len_is_slice = len(is_slice)
 
         len_subtensor_spec = spec_pos()
+        subensor_spec = "npy_intp subtensor_spec[%(len_subtensor_spec)s];" % locals()
+        if len_subtensor_spec == 0:
+            subensor_spec = "npy_intp * subtensor_spec = NULL;"
 
-        is_slice_init = ",".join([str(s) for s in is_slice])
+        if is_slice:
+            is_slice_init = "int is_slice[] = {" + ",".join([str(s) for s in is_slice]) + "};"
+        else:
+            is_slice_init = "int* is_slice = NULL;"
         subtensor_init = "\n".join(init_cmds)
 
         x, = inputs[:1]
@@ -673,8 +679,8 @@ class Subtensor(Op):
         // The subtensor is created by iterating over the dimensions
         // and updating stride, shape, and data pointers
 
-        int is_slice[] = {%(is_slice_init)s};
-        npy_intp subtensor_spec[%(len_subtensor_spec)s];
+        %(is_slice_init)s
+        %(subensor_spec)s
         %(subtensor_init)s;
         int spec_pos = 0; //position in subtensor_spec
         int inner_ii = 0; // the current dimension of zview