A method realizing prediction of value association indirect jump

Abstract

The invention discloses a method realizing prediction of value association indirect jump and relates to a compiler and a processor. The method is characterized in that: the compiler, according to the profiling information acquired when the processor executes executable programs, identifies a subprogram structure corresponding to an indirect jump instruction in a source program and association information in association data values of the subprogram structure and inserts a key instruction used for identifying the association information in the source program for generating an executable program again; and the processor, while executing the executable program generated again by the compiler, collects association information dynamically according to the key instruction and generates a value history mode. The method can effectively improve a prediction accuracy rate of the indirect jump instructions and accordingly improve the overall performance of the processor and systems to which the processor is applied.

Description

technical field [0001] The invention belongs to the field of microprocessor design and system design using microprocessors, in particular to a method for modern processors to realize value-associated indirect jump prediction. Background technique [0002] In modern processors, speculative execution is one of the important means to exploit instruction-level parallelism. In order to improve the efficiency of speculative execution and reduce erroneous speculative execution, accurate branch prediction technology is crucial. [0003] In the modern superscalar processor architecture with wide launch and deep pipeline, the main purpose of adopting branch prediction is to improve the computing speed of CPU. Speculative execution is based on branch prediction, that is, if the CPU can predict whether the program transfers before the result of the previous instruction executed by the pipeline, it can execute the corresponding instruction in advance, thus avoiding the idleness of the p...

AI Technical Summary

Problems solved by technology

However, there are two intractable problems in existing indirect jump predictors based on data values: first, it is very difficult to find valid associated data values; second, these associated data values ​​are likely to be is unavailable

These two problems make it difficult for existing indirect jump predictors based on data values ​​to achieve ideal prediction results

Images