Type-Based Load-Value Prediction

To alleviate the high cost of main-memory accesses, computer architects have proposed various speculation mechanisms, including load-value prediction. A load-value predictor guesses the result of load instructions, thus allowing dependent instructions to execute without having to wait for the memory access to complete. Unfortunately, costly mispredictions hinder the true potential of load-value prediction. This paper describes how to build more accurate and faster load-value predictors with little extra hardware by exploiting the type of load instructions. Our techniques are easily implementable in any CPU that supports multiple load types (e.g., byte, word, long, IEEE single-precision and IEEE double-precision loads). Compared to traditional load-value predictors, we are able to reduce mispredictions by up to 72% using our simple type-based approaches. Moreover, we show that the type of a load can effectively be used as the component selector in a hybrid predictor.