static int64 GetDirectConvCost - Integer overflow

[iurii-kvasniuk]

tensorflow-allreduce/tensorflow/core/kernels/deep_conv2d.cc: ln 74,
static int64 GetDirectConvCost(int filter_rows, int filter_cols, int in_depth,
int out_depth, int out_rows, int out_cols) {
return filter_rows * filter_cols * in_depth * out_depth * out_rows * out_cols;
}

Can lead to integer overflow and weird results
I think, it should be smth like that
return (int64)filter_rows * (int64)filter_cols * (int64)in_depth * (int64)out_depth * (int64)out_rows * (int64)out_cols;