組み込み向けCPUの数値演算における超越関数演算の精度についての一考察

概要

論文の詳細を見る
Embedded systems are designed to do some specific task and it often runs with limited hardware resources. As compared with general-purpose CPUs for PCs, embedded CPUs are often restricted in operating speed, operation accuracy, and function. The capability of the floating point operation is limited in almost all embedded CPUs. Furthermore, there are many different CPU architectures used in embedded systems. In this paper, on several CPUs popular for embedded use, computer experiments are performed for floating-point operation performance based on exp function and log function. The floating point functions here are exp and log. Results of Experiments, the x86 CPU and its compatible CPUs output no computational differences in the computational results. To the contrary, the embedded CPUs Renesas H8s 2633 and Cirrus-logic EP9315 output greatly different x86's computational results.
石巻専修大学の論文