Instruction issue processing circuit of uniform dyeing device array

Abstract

The invention belongs to the technical field of integrated circuits, and provides an instruction issue processing circuit of a uniform dyeing device array. The circuit comprises an instruction fetching module 1, a BUFFER module 2, a pre-decoding module 3, an IBUFFER module 4, a decoding module 5, a scoreboard module 6 and a warp scheduler module 7. The circuit has the technical effect that 1 at least 8 instructions are fetched from an Icache at each time, so that the frequency of accessing the Icache is reduced to a certain extent and the instruction execution efficiency can be improved; 2, buffers and ibuffers of a plurality of warps are independently realized by registers, so that the logic can be reused and the circuit realization is simple; and 3, the pre-decoding module and the decoding module realize corresponding decoding functions of two instructions in different stages, circuits in an instruction issue processing mechanism circuit have the modules, and the manner that multiplewarps correspond to multiple modules is not adopted so that the circuit realization is simple and resources are saved.

Description

technical field [0001] The invention belongs to the technical field of integrated circuits and relates to an instruction emission processing circuit of a unified dyer array. Background technique [0002] The uniform dyer array in the graphics processor implements the unified dyeing function of vertices and pixels. In the unified shader array, due to the data and structure correlation between instructions, subsequent instructions cannot be executed immediately, and the pipeline must stop for several cycles before continuing to execute subsequent instructions, resulting in pipeline delay. Therefore, the SIMT (Single Instruction Multiple Thread) method is adopted, and the instruction is repeatedly executed in several subsequent cycles. After the last execution of the instruction is completed, the following instructions have the execution conditions and can be executed immediately. In this way, the pipeline delay can be hidden, and the throughput of the dyeing task can be impro...

AI Technical Summary

Problems solved by technology

In the unified shader array, due to the data and structure correlation between instructions, the following instructions cannot be executed immediately, and the pipeline must stop for several cycles before continuing to execute the following instructions, resulting in pipeline delay

Images