Compiler-Directed Static Classification of Value Locality Behavior

Predicting the values that are likely to be produced by instructions has been suggested as a way of increasing the instruction-level parallelism available in a wide-issue processor. One of the potential difficulties in exploiting the predictability of values, however, is selecting the proper type of predictor, such as a last-value predictor, a stride predictor, or a context-based predictor, for a given instruction. We propose a compiler-directed classification scheme that statically partitions all of the instructions in a program into several groups, each of which is associated with a specific value predictability pattern. This value predictability pattern is encoded into the instructions to identify the type of value predictor that will be best suited for predicting the values that are likely to be produced by each instruction at run-time. Both an idealized profile-based compiler implementation and an implementation based on the GCC compiler are studied to show the performance bounds for the proposed technique. Our simulations based on the SimpleScalar tool set and the SPEC95 integer benchmarks indicate that this approach can substantially reduce the number of read/write ports needed in the value predictor for a given level of performance. This static partitioning approach also produces better performance than a dynamically partitioned approach for a given hardware configuration. Finally, this work demonstrates the connection between value locality behavior and source-level program structures thereby leading to a deeper understanding of the causes of this behavior.