Optimizing at the algorithmic and architectural level rather than relying on language speed