Branching processing module and mechanism for coarse grain multi-core calculating system

Abstract

The invention discloses a branching processing module and mechanism for a coarse grain multi-core calculating system. The branching processing module is characterized in that the coarse grain multi-core calculating system comprises an on-chip network, a main control unit, a memorizer unit, an interface unit and multiple function units, the main control unit can instantly start branching prediction error recovery after determining branching prediction errors and can end the execution of wrong instructions having already been transmitted in advance so as to make the system restore the normal working state, and the function units can instantly end a current calculation task, restore the idle state, release the control right and wait for the issuing of a correct task after determining the branching prediction errors. The errors can be corrected within an extremely short period of time when the branching prediction errors are found, performance loss caused by the branching prediction errors is reduced, and therefore the improvement of branching prediction on system performance is further enhanced, and the calculation system achieves higher work efficiency.

Description

technical field [0001] The invention relates to the field of high-density calculation and dynamic scheduling of instructions, in particular to a branch prediction error advance processing and recovery mechanism for a coarse-grained multi-core computing system. Background technique [0002] Multi-core technology has become the mainstream of processor design because of low power consumption, strong parallel processing capability and excellent computing performance. However, the computing performance of a multi-core computing system largely depends on whether the computing power in the computing system can be utilized to the maximum extent, and the timing of issuing instructions in advance through branch prediction has been widely used in modern super-standard In quantum processors, it has been proved to be an effective method and means to improve the computing efficiency of processors. [0003] The performance improvement of branch prediction depends largely on the accuracy o...

AI Technical Summary

Problems solved by technology

[0003] The performance improvement of branch prediction depends largely on the accuracy of branch prediction and the time to restore the correct working state when the branch prediction is wrong. For the coarse-grained multi-core computing system based on the network-on-chip, the branch prediction error takes a long time To restore the working state of the system, often resulting in huge performance loss

Images