Retargetable register allocation method in dynamic binary translation system

Abstract

The invention relates to a retargetable register allocation method in dynamic binary translation system; replacement policy is applied based on the next-use information of the variables in basic block; according to different target platforms, different register allocators are enabled self-adaptively to allocate target platform registers. According to the different requirements of different dynamic binary translation fields, the variables in basic black are processed by register allocation, which is composed of common allocation, mandatory requirement special register allocation and mandatory requirement non-special register allocation. The invention which combines the natures of the use information of variables enjoys retargetability, perfect allocation effect, inexpensive allocation, good applicability to dynamic binary translation fields, in particular to a binary translator which has multiple sources and multiple targets.

Description

technical field [0001] The invention relates to a redirectable register allocation method in a dynamic binary translation system, in particular to a register allocation method which utilizes next-use information of variables in a basic block and has redirection characteristics. The invention belongs to the technical field of dynamic binary translation, and is especially suitable for a multi-source and multi-target binary translation platform. Background technique [0002] Dynamic binary translation is the most widely used method in virtual execution technology. It is an effective means to provide portability for legacy code and improve the platform adaptability of software. It can Dynamically convert the binary program on the source machine platform to run on other target machine platforms. For the dynamic binary translator itself, execution performance is a very important measure. The so-called execution performance means that, taking the efficiency of the source program ...

AI Technical Summary

Problems solved by technology

[0013] From the above three register allocation methods, it can be seen that each method has its inherent advantages and disadvantages. The richness of information collected and utilized by the three methods is increasing. At the same time, the allocation overhead and implementation complexity are also increasing.

For the first two register allocation methods, the implementation is relatively simple, but the information of variables in the basic block is very limited, and effective results cannot be achieved

Although the last type of splitting information makes full use of the collected information, its execution performance loss will be unacceptable. In other words, in the field of dynamic binary translation, within a basic block, the register usage efficiency obtained by this allocation method is The increase is not enough to make up for the allocation overhead paid to achieve this effect