Update sandbox/cuda/dnn_base.c to work on Windows

Allocating the strides array dynamically lets it also compile on Windows. See https://groups.google.com/forum/#!topic/theano-users/fex4OpL9hB0 for code that errors on Windows without this.

Update sandbox/cuda/dnn_base.c to work on Windows
a66610b4 · ivdorelian · 477fd7cf · a66610b4
--- a/theano/sandbox/cuda/dnn_base.c
+++ b/theano/sandbox/cuda/dnn_base.c
@@ -7,8 +7,9 @@ c_set_tensorNd(CudaNdarray *var, cudnnTensorDescriptor_t desc) {


  int dim = CudaNdarray_NDIM(var);
-  int strides[dim];
+  int *strides = (int *)malloc(dim * sizeof(int));
  int default_str = 1;
+  int return_value = 0;

  for (int i = dim-1; i >= 0; i--)
  {
@@ -22,14 +23,18 @@ c_set_tensorNd(CudaNdarray *var, cudnnTensorDescriptor_t desc) {
  cudnnStatus_t err = cudnnSetTensorNdDescriptor(desc, CUDNN_DATA_FLOAT, dim,
                                                 CudaNdarray_HOST_DIMS(var),
                                                 strides);
+											
+
  if (err != CUDNN_STATUS_SUCCESS) {
    PyErr_Format(PyExc_RuntimeError,
 		 "Could not set tensorNd descriptor: %s"
 		 "dim=%d",
 		 cudnnGetErrorString(err), dim);
-    return -1;
+    return_value = -1;
  }
-  return 0;
+
+  free(strides);
+  return return_value;
 }