高级检索
冯春阳, 杨靓, 黄士坦. 随机法和Hensel Lifting联合的浮点乘测试用例生成方法[J]. 计算机辅助设计与图形学学报, 2014, 26(9): 1509-1521.
引用本文: 冯春阳, 杨靓, 黄士坦. 随机法和Hensel Lifting联合的浮点乘测试用例生成方法[J]. 计算机辅助设计与图形学学报, 2014, 26(9): 1509-1521.
Feng Chunyang, Yang Jing, Huang Shitan. Test Case Generation Based on Random Test Combining with Hensel Lifting for Floating-point Multiplication[J]. Journal of Computer-Aided Design & Computer Graphics, 2014, 26(9): 1509-1521.
Citation: Feng Chunyang, Yang Jing, Huang Shitan. Test Case Generation Based on Random Test Combining with Hensel Lifting for Floating-point Multiplication[J]. Journal of Computer-Aided Design & Computer Graphics, 2014, 26(9): 1509-1521.

随机法和Hensel Lifting联合的浮点乘测试用例生成方法

Test Case Generation Based on Random Test Combining with Hensel Lifting for Floating-point Multiplication

  • 摘要: 针对浮点乘仿真验证时覆盖率不全面和边界角用例定位难的问题,提出一种随机法和Hensel lifting理论联合的浮点乘测试用例生成方法.首先通过分析测试用例生成域设计了一个浮点乘测试用例产生及功能仿真平台;然后利用Hensel lifting理论提出一种统一的边界角浮点乘测试用例生成模型.将所提方法用于文中设计的功能仿真平台中,并将该平台与典型浮点测试工具集进行浮点乘性能比较的实验结果表明,该方法可使浮点乘检错率随浮点数位宽的增加而提升,最高增幅可达9.77%,比随机法检错率平均提高15.98%,比典型浮点测试工具集检错率平均提高1.9%.

     

    Abstract: In order to solve the problems of dissatisfied test coverage and conditional corner case in floating-point multiplication by the simulation-based verification,a test case generating method based on random test combining with Hensel lifting is proposed.Firstly,a simulation platform for test case and functional verification of floating-point multiplication is designed by the analysis of test case space.Then,an unified modeling of corner case for floating-point multiplication based on Hensel lifting is introduced.The test case generation based on random test combining with Hensel lifting is applied to the above platform,and the performances among the platform and mainstream floating-point test tools about floating-point multiplication are compared.The experimental results show that the proposed method makes error-detecting rate upgrade with the increase of floating-point bit width at the maximum of 9.77%and improves respectively 15.98%and 1.9%than random test and mainstream floating-point test tools on the average.

     

/

返回文章
返回