链接已复制
1 回复
real*16 on Pentium and Xeon families takes advantage of x87 extended precision, but uses software arithmetic functions to implement 113-bit precision. Thus, it is significantly more accurate and slower than 64-bit x87 precision would be.
real*16 on Itanium uses combinations of hardware instructions including those which implement 64-bit extended precision, so it is not so costly as on Xeon, but 113-bit precision still takes longer than 64-bit would do.
real*16 on Itanium uses combinations of hardware instructions including those which implement 64-bit extended precision, so it is not so costly as on Xeon, but 113-bit precision still takes longer than 64-bit would do.
