Merge pull request #5597 from notoraptor/windows-fixes

theano/gpuarray/reduction.py

@@ -61,6 +61,8 @@ class GpuMaxAndArgmax(Op):
             #endif
         #endif
 
+        int err = 0;
+
         unsigned  %(name)s_redux_len = PyTuple_GET_SIZE(%(axes)s);
         unsigned* %(name)s_axes_to_reduce = (unsigned*)malloc(%(name)s_redux_len * sizeof(unsigned));
         for (unsigned i = 0; i < %(name)s_redux_len; ++i) {
@@ -114,10 +116,12 @@ class GpuMaxAndArgmax(Op):
                 PyErr_SetString(PyExc_RuntimeError, "GpuMaxAndArgmax: unable to set argmax to 0 when input is a scalar.");
                 %(fail)s
             }
-        } else if (GA_NO_ERROR !=
+        } else if (GA_NO_ERROR != (err =
             GpuArray_maxandargmax(&%(max)s->ga, &%(argmax)s->ga, &%(X)s->ga, %(name)s_redux_len, %(name)s_axes_to_reduce)
-        ) {
-            PyErr_SetString(PyExc_RuntimeError, "GpuMaxAndArgmax: unable to compute gpuarray maxandargmax.");
+        )) {
+            PyErr_Format(PyExc_RuntimeError,
+                "GpuMaxAndArgmax: unable to compute gpuarray maxandargmax: error %%d: %%s (%%s).",
+                err, gpuarray_error_str(err), GpuArray_error(&%(X)s->ga, err));
             %(fail)s
         }
         """
@@ -137,4 +141,4 @@ class GpuMaxAndArgmax(Op):
         """ % {'name': name, 'X': inputs[0]}
 
     def c_code_cache_version(self):
-        return (1,)
+        return (1, 1)

theano/gpuarray/tests/test_reduction.py

@@ -14,7 +14,7 @@ from .. import GpuArrayType
 import math
 
 # Number of values to be used in test tensors (except with 0-D tensors!).
-test_size = 10000000
+test_size = 10000
 
 # NB: This order of "unsorted axes" is arbitrary and is here
 # just to have the same informations on profile output

theano/tensor/tests/test_subtensor.py

@@ -533,7 +533,7 @@ class T_subtensor(unittest.TestCase, utt.TestOptimizationMixin):
         data = rand(4, 2, 3)
         idx = [2, 2, 0, 0, 1, 1]
         n = self.shared(data)
-        t = n[self.shared(numpy.asarray(idx))[::2]]
+        t = n[self.shared(numpy.asarray(idx).astype('int64'))[::2]]
         self.assertTrue(isinstance(t.owner.op, tensor.AdvancedSubtensor1))
         val = self.eval_output_and_check(t, op_type=self.adv_sub1, length=2)
         utt.assert_allclose(data[idx[::2]], val)