To provide an arithmetic system for efficiently performing a quantization arithmetic when fixed-precision floating-point number groups are converted into single floating-point representation form of multiple-precision or variable-precision in an efficient way. 固定精度浮動小数点数の組を多倍長や可変長の1つの浮動小数点の表現形式に効率的に変換する際に量子化演算を効率的に実行する演算方式を提供する。 - 特許庁