一种混合高性能计算机代数环境模型

Hybrid high performance model of computer algebra system

提出一种高性能计算机代数环境HHPCAS,综合现有的多种计算机代数软件,通过添加内核扩展函数、外部调用等方法,结合集群管理软件和并行环境,可以提供高性能的计算机代数计算环境。HHPCAS根据Slot/Ticket模型可以有效管理可用计算资源和作业优先等级,充分发挥多种计算机代数软件的特长,并且提供并行的消息传递机制,将大量复杂的计算平均分配到每个计算节点,解决单台机器内存受限和计算能力有限等问题。通过并行差分代换方法测试表明HHPCAS可以为符号计算和计算机自动推理提供有效的计算平台。

Computer Algebra System （CAS） is a vital backbone in the field of symbolic computation and automated reasoning. There are many excellent CASs, which can provide good interface and quick computation for researchers. However, they are only designed to work on some special purposes, i. e., polynomial operations or calculation of Groebner basis, which lead to inefficient collaboration. A hybrid high performance model of computer algebra system （HHPCAS） was outlined. HHPCAS provided high performance by expanding the kernel library in CAS, using external calling and integrating the parallel library and cluster management software. Each job was submitted to a queue and scheduled by its ticket and slot value, which denoted the priority of job. In HHPCAS, several examples were tested by implementing an algorithm named parallel successive difference substitution. The experimental result shows that HHPCAS is effective for complex problem in symbolic computation and automated reasoning.