インテル® Fortran コンパイラー 14.0 ユーザー・リファレンス・ガイド
A REAL(16) constant has more than four times the accuracy of a REAL(4) number, and a greater range.
A REAL(16) constant occupies 16 bytes of memory. The number of digits that precede the exponent is unlimited, but typically only the leftmost 33 digits are significant.
IEEE X_floating format is used.