发明名称 METHOD AND APPARATUS FOR FLOATING-POINT DATA CONVERSION
摘要 PURPOSE: To recover numerical precision by controlling conversion by an irregularity processing part, detecting irregularities, analyzing the contents of irregularities in the array of a floating point element, and preserving reconstructed data when anomaly occurs during the conversion. CONSTITUTION: A processing part which carries out an irregularity processing has a programmable control part 101, and a means 107 which selects a conversion algorithm or the like, and stores an irregularity amount chart for the conversion algorithm. Moreover, the processing part is provided with a means 108 which grasps numerical attributes for an irregular value to be reconstituted according to a floating point number describing method and a coefficient setting method. In addition, the processing part is provided with a means 109 which eliminates the reconstituted data from a data storage device 116, transmitting means 110 of the irregularity amount chart for analyzing a floating point data array to a program storage device 115, means 111 which receives a control signal to an algorithm detecting and selecting part 106, and transmits state information to the data storage device, periodically repeating means 112, and means 113 which returns a storage position, and executes a stepping operation.
申请公布号 JPH05127871(A) 申请公布日期 1993.05.25
申请号 JP19910263171 申请日期 1991.09.14
申请人 INTERNATL BUSINESS MACH CORP <IBM> 发明人 JIERARUDO YOOGEN REIZAA
分类号 G06F7/00;G06F7/76;H03M7/24 主分类号 G06F7/00
代理机构 代理人
主权项
地址