Primary and secondary data structure-based CPU-GPU cooperative computing method

Abstract

The embodiment of the invention provides a primary and secondary data structure-based CPU-GPU cooperative computing method, which comprises the following steps: according to an object to be processed, determining primary and secondary data contents and initializing the primary and secondary data contents; starting a CUP computing thread and a GPU computing thread; reading data to be processed, storing the pre-processed data in a primary and secondary data structure, and processing the data in the primary and secondary data structure by using the CUP computing thread and the GPU computing thread till no data exists. According to a scheme provided by the invention, parallel data can be managed effectively, and the balance of the loads on the threads of a GPU can be ensured when a GPGPU platform processes databases with unbalance effective computation distribution; and a CPU and the GPU can perform complete parallel calculation and keep high utilization rate by simply-designed and reusable thread dividing method.

Description

technical field [0001] The present invention relates to the computer field, and in particular, the present invention relates to a CPU-GPU cooperative computing method based on a primary and secondary data structure. Background technique [0002] In the field of high-performance computing, to achieve extremely high-efficiency output, it is usually necessary to connect a large number of CPUs. The CPU (Central Processing Unit, central processing unit) is the core that controls the operation of the computer, and uses parallel distributed processing to perform calculations. However, this structure not only Program development is difficult, the hardware is bulky, and the power consumption is even more astonishing. The rise of the concept of GPGPU (General-Purpose Computing on Graphics Processing Units, general-purpose computing graphics processor) is also to make up for the weaknesses of these traditional CPU architectures. [0003] Generally, a single GPU (Graphics Processing Un...

AI Technical Summary

Problems solved by technology

[0006] GPGPU faces two problems: 1) thread load balancing problem on GPU

In fact, the effective workload of each thread may not be the same, so it will cause unbalanced GPU load

2) CPU and GPU utilization issues

In the synchronous call mode, after the CPU calls the GPU, it must wait for its calculation to complete before proceeding to the next step, so that the utilization rate of the CPU is relatively low; Parallel computing is performed at the same time, but the size of this parallel computing amount is difficult to determine

If the amount of CPU parallel computing is too small, the CPU utilization rate is still very low; if the CPU parallel computing amount is too large, so that after the GPU calculation is completed, it needs to wait for the CPU to assign new computing tasks to it, resulting in low GPU utilization; Only when the CPU parallel computing time is exactly the same as the GPU computing time, high CPU and GPU utilization can be obtained at the same time, but it is very difficult to accurately determine the CPU parallel computing amount

Images