Merge pull request #6221 from lamblin/advidx_padshape

theano/gpuarray/subtensor.py

@@ -387,7 +387,7 @@ int sub_setarray(GpuArray *dst, GpuArray *src) {
           void *args[2];
           args[0] = &zview->ga;
           args[1] = &%(x)s->ga;
-          if (GpuElemwise_call(iadd, args, GE_BROADCAST) != GA_NO_ERROR) {
+          if (GpuElemwise_call(iadd, args, GE_BROADCAST | GE_PADSHAPE) != GA_NO_ERROR) {
             PyErr_SetString(PyExc_RuntimeError, "Error doing inplace add");
             Py_DECREF(zview);
             %(fail)s
@@ -399,7 +399,7 @@ int sub_setarray(GpuArray *dst, GpuArray *src) {
         parent_version = super(GpuIncSubtensor, self).c_code_cache_version()
         if not parent_version:
             return
-        return parent_version + (9,)
+        return parent_version + (10,)
 
 
 class GpuAdvancedSubtensor1(HideC, tensor.AdvancedSubtensor1):
@@ -1009,7 +1009,7 @@ class GpuAdvancedIncSubtensor1(Op):
               void *args[2];
               args[0] = (void *)&row_x->ga;
               args[1] = (void *)&row_y->ga;
-              ret = GpuElemwise_call(iadd, args, GE_BROADCAST);
+              ret = GpuElemwise_call(iadd, args, GE_BROADCAST | GE_PADSHAPE);
             }
             Py_DECREF(row_x);
             Py_DECREF(row_y);
@@ -1031,7 +1031,7 @@ class GpuAdvancedIncSubtensor1(Op):
                    """ % dict(fail=sub['fail']))
 
     def c_code_cache_version(self):
-        return (4,)
+        return (5,)
 
 
 class GpuAdvancedIncSubtensor1_dev20(GpuKernelBase, HideC,

theano/tensor/tests/test_subtensor.py

@@ -545,8 +545,8 @@ class T_subtensor(unittest.TestCase, utt.TestOptimizationMixin):
 
                 t = op(n[:z, :z], m)
                 gn, gm = theano.tensor.grad(theano.tensor.sum(t), [n, m])
-                utt.verify_grad(lambda m: op(n[:z, :z], m), [mv])
-                utt.verify_grad(lambda nn: op(nn[:z, :z], mv), [data])
+                utt.verify_grad(lambda m: op(n[:z, :z], m), [mv], mode=self.mode)
+                utt.verify_grad(lambda nn: op(nn[:z, :z], mv), [data], mode=self.mode)
 
     def test_grad_0d(self):
         data = np.asarray(rand(2, 3), dtype=self.dtype)