An Overflow/Underflow-Free Compatible Floating-Point Representation
スポンサーリンク
概要
- 論文の詳細を見る
An overflow/underflow-free floating-point representation (EXTENDED FORMAT) compatible with the 64bit representation of the IBM System/370 is proposed. Being implemented on a HITAC M-170 as a system of subroutines supported by a minor hardware modification, the proposed system enables any program operating in its normal environment to run as fast as on the conventional system except during the period of time when the interrupts related to the EXTENDED FORMAT are being processed by subroutines. It also enables a user to write a major portion of his program without paying any attention to the EXTENDED FORMAT which gives him almost complete freedom from the problem of exponent spills. Apart from a means for compatibility, the proposal is similar to the one presented earlier by Matsui and Iri in utilizing a mobile boundary between the exponent and the mantissa and incorporating NaNs. Therefore, some examples are adopted from their report for comparison. The application of this proposal is not limited to the IBM System/370 architecture. The same approach is applicable for any architecture with a floating-point representation having an appropriate redundancy.
- 一般社団法人情報処理学会の論文
- 1984-07-31
著者
-
Takahashi Shigeru
Institute Of Space Science National Central University
-
Tian Lim
A Graduate Student Of Department Of Scientific Technology University Of Tsukuba: Ibm World Trade Cor
-
Takahashi Shigeru
Institute Of Information Science And Electronics University Of Tsukuba
関連論文
- A Shape-and-Density Model of the Putative Binary EKBO 2001 QG_
- Structure and Expression of the Mouse Angiotensinogen Gene
- An Overflow/Underflow-Free Compatible Floating-Point Representation
- Evaluation of Residual Stress Distribution in Austenitic Stainless Steel Pipe Butt-Welded Joint