A method for implementing gpgpu register cache

Abstract

The invention discloses a method for implementing register caches of GPGPU (general purpose graphics processing units). The method includes removing original register files in various stream processors, adding the caches of registers required for storing threads being executed, and loading register data in memories to the register caches before the threads are executed; selecting the thread combinations with the maximum thread quantities from the executable threads on the premise that the registers required by execution of the threads can be stored in the register caches of the stream processors, executing the selected thread combinations, loading register data in the memories to the register caches, selectively executing the threads in the thread combinations in optional modes and reselecting another group of threads when all the threads in the selected thread combinations cannot be executed. The register data are used by each thread in the selected thread combinations. The method has the advantages that storage spaces required by the register files of the stream processors can be reduced, energy consumption and areas can be decreased, the problem of constraints due to insufficient quantities of registers of existing stream processors can be solved, and the system efficiency can be improved.

Description

technical field [0001] The present invention relates to a method for realizing cache, in particular to a method for realizing GPGPU register cache. Background technique [0002] A general-purpose graphics processor GPGPU is a processor that utilizes a graphics processor that traditionally handles graphics tasks to perform general-purpose computing tasks. GPGPU is composed of multiple stream processors. Each stream processor contains multiple computing units. An instruction decoder in the stream processor controls all computing units. The computing units execute instructions in parallel in the manner of single instruction and multiple data, thereby Small management cost to achieve large-scale data parallel computing. The smallest management unit in GPGPU is a thread, and each thread is usually used to process a unit of data in large-scale data. A group of threads in the stream processor are executed in lockstep on the computing unit, called the lockstep execution thread gro...

AI Technical Summary

Problems solved by technology

The stream processor can only execute some threads in a period of time, and most of the registers saved in the entire register file are idle and will not be used immediately, so the energy consumption and area of ​​the GPGPU chip occupied by these register files are different in time wasted

Images