Implement fixed-window exponentiation to mitigate hyper-threading