Fix crash as we used too much block in some cases.

theano/sandbox/cuda/basic_ops.py

@@ -1388,7 +1388,7 @@ class GpuCAReduce(GpuOp):
             dim3 n_threads(
                     std::min(CudaNdarray_HOST_DIMS(%(x)s)[0],
                             NUM_VECTOR_OP_THREADS_PER_BLOCK));
-            dim3 n_blocks(CudaNdarray_HOST_DIMS(%(x)s)[1]);
+            dim3 n_blocks(std::min(CudaNdarray_HOST_DIMS(%(x)s)[1], NUM_VECTOR_OP_BLOCKS));
             while (n_blocks.x * (n_blocks.y+1) <= NUM_VECTOR_OP_BLOCKS && n_blocks.y <= CudaNdarray_HOST_DIMS(%(x)s)[2])
             {
                 n_blocks.y += 1;
@@ -1565,7 +1565,7 @@ class GpuCAReduce(GpuOp):
         """ % locals()
 
     def c_code_cache_version_apply(self, node):
-        version = [6]  # the version corresponding to the c code in this Op
+        version = [7]  # the version corresponding to the c code in this Op
 
         # now we insert versions for the ops on which we depend...
         scalar_node = Apply(self.scalar_op,

theano/sandbox/cuda/tests/test_basic_ops.py

@@ -95,7 +95,7 @@ def test_careduce():
                                ((1024,33),[0,1]),((33,1024),[0,1]),#01
                                ((1025,33),[0,1]),((33,1025),[0,1]),#01
 
-                               ((4100,4,3),[0]),((5,4100,3),[0]),((5,4,4100),[0]),#100
+                               ((4100,4,3),[0]),((5,4100,3),[0]),((5,4,4100),[0]), ((3,65536,1), [0]),#100
                                ((4100,4,3),[1]),((5,4100,3),[1]),((5,4,4100),[1]),#010
                                ((4100,4,3),[2]),((5,4100,3),[2]),((5,4,4100),[2]),#001
                                ((4100,4,3),[0,1]),((5,4100,3),[0,1]),((5,4,4100),[0,1]),#110