【摘要】：隨著高性能計算機的發(fā)展,一種基于CPU-GPU的異構集群逐漸被人們所關注。相比于傳統(tǒng)集群,它更經(jīng)濟環(huán)保,且擁有更高的運算速度。但異構模式下效率較低的短板也限制著異構集群的發(fā)展。本文提出的DTPS算法,通過動態(tài)調整異構集群下CPU與GPU任務劃分的比例,整合集群計算資源,使集群的計算效率達到相對較高的水平,并通過實驗證明了算法的有效性。
[Abstract]:With the development of high performance computer, a heterogeneous cluster based on CPU-GPU has been paid more and more attention. Compared with the traditional cluster, it is more economical and environmentally friendly, and has higher computing speed. However, the development of heterogeneous cluster is limited by the low efficiency board in heterogeneous mode. The DTPS algorithm proposed in this paper dynamically adjusts the ratio of CPU and GPU task partition in heterogeneous cluster, integrates the computing resources of cluster, and makes the computing efficiency of cluster reach a relatively high level. The experimental results show that the algorithm is effective.
【作者單位】： 安徽大學計算機科學與技術學院;
【分類號】：TP38
，

本文編號：2271161