Fix compilation crash reported in gh-1518 for gpujoin.

theano/sandbox/cuda/basic_ops.py

@@ -5,7 +5,7 @@ import sys
 import numpy
 
 import theano
-from theano import Type, Apply
+from theano import gof, Type, Apply
 from theano import tensor, scalar, config
 from theano.compat.six import StringIO
 from theano.scalar import Scalar
@@ -2960,7 +2960,6 @@ class GpuJoin(tensor.Join, GpuOp):
         str = """
         const int axis = PyInt_AsLong((PyObject*)%(axis)s);
         const int nd = %(nd)s;
-        int shape_%(input_1)s[nd];
         int shape_out[nd];
         int width_sum = 0;
         int errorcode;
@@ -2973,11 +2972,6 @@ class GpuJoin(tensor.Join, GpuOp):
         start = NULL;
         stop = NULL;
 
-        for(int i = 0; i<nd; i+=1)
-        {
-            shape_%(input_1)s[i] = CudaNdarray_HOST_DIMS(%(input_1)s)[i];
-            shape_out[i] = shape_%(input_1)s[i];
-        }
         """ % locals()
 
         # getting the shapes of all the involved tensors (input[1:])
@@ -2986,7 +2980,7 @@ class GpuJoin(tensor.Join, GpuOp):
         # shape_%(cdna)s[nd] is initialized before, to prevent following
         # error: jump to label __label_9 crosses initialization of
         # shape_%(cdna)s[nd]
-        for i, cdna in enumerate(inputs[2:]):
+        for i, cdna in enumerate(gof.utils.uniq(inputs[1:])):
             str += """
             int shape_%(cdna)s[nd];
             """ % locals()
@@ -2998,8 +2992,14 @@ class GpuJoin(tensor.Join, GpuOp):
         if(full_slice == NULL){
             %(fail)s;
         }
+
+        for(int i = 0; i<nd; i+=1)
+        {
+            shape_%(input_1)s[i] = CudaNdarray_HOST_DIMS(%(input_1)s)[i];
+            shape_out[i] = shape_%(input_1)s[i];
+        }
         """ % locals()
-        for i, cdna in enumerate(inputs[2:]):
+        for i, cdna in enumerate(gof.utils.uniq(inputs[2:])):
             str += """
             for(int i = 0; i<nd; i+=1)
             {
@@ -3090,7 +3090,7 @@ class GpuJoin(tensor.Join, GpuOp):
         return str
 
     def c_code_cache_version(self):
-        return (4,)
+        return (5,)
 
 gpu_join = GpuJoin()
 

theano/tensor/tests/test_basic.py

@@ -3294,11 +3294,29 @@ class T_Join_and_Split(unittest.TestCase):
         """
         Regression test for a crash that used to happen when rebroadcasting.
         """
-        x = tensor.TensorType(floatX, [False, False, True])()
-        u = tensor.TensorType(floatX, [False, False, True])()
+        x = tensor.TensorType(self.floatX, [False, False, True])()
+        u = tensor.TensorType(self.floatX, [False, False, True])()
         # This line used to crash.
         z = tensor.concatenate([x, -u], axis=2)
 
+    def test_concatenate_same(self):
+        """
+        Test that we can concatenate the same tensor multiple time.
+
+        In the past it was broken on the GPU.
+        """
+        rng = numpy.random.RandomState(seed=utt.fetch_seed())
+        T_shared = self.shared(rng.rand(3, 4).astype(self.floatX))
+        Tout = tensor.concatenate([T_shared, T_shared])
+        f = function([], Tout, mode=self.mode)
+        out = f()
+        if theano.config.mode != 'FAST_COMPILE':
+            assert [True for node in f.maker.fgraph.toposort() if isinstance(
+                node.op, self.join_op)]
+        assert numpy.allclose(out,
+                              numpy.concatenate([T_shared.get_value(),
+                                                 T_shared.get_value()]))
+
 
 class test_comparison(unittest.TestCase):
     """Test <, >, <=, >=, == and !=