c++ float precision - 未名空间MITBBS历史存档

国际科技财经博客移民网络热点娱乐民生时事公众号

Redian新闻

>未名空间

>JobHunting - 待字闺中

c++ float precision

c++ float precision# JobHunting - 待字闺中

g*s2011-04-10 07:04

1 楼

it is said 7-bit significance.
suddenly get confused here. does that mean 30,000,000f and 30,000,001f
actually has the same 32-bit binary representation?

l*a2011-04-10 07:04

2 楼

抠这么细？
去BB都不需要把

【在 g*********s 的大作中提到】

: it is said 7-bit significance.
: suddenly get confused here. does that mean 30,000,000f and 30,000,001f
: actually has the same 32-bit binary representation?

f*42011-04-10 07:04

3 楼

可能是要去投行吧，一个交易都是几个B的～

【在 l*****a 的大作中提到】

: 抠这么细？
: 去BB都不需要把

r*22011-04-10 07:04

4 楼

single float 32 bit
23 bit significand

g*s2011-04-10 07:04

5 楼

that's the binary precision.
for decimal, it's log10(2^23) = 7.

【在 r*********2 的大作中提到】

: single float 32 bit
: 23 bit significand

r*22011-04-10 07:04

6 楼

ah, 我错了
对的，尾数一样single precision float就是这个精度吧。

n*02011-04-10 07:04

7 楼

如果只能表达这俩数的前7bit的话，那应该是一样的。不过正常的计算机里面除了7bit
significance以外后面通常会多出一坨无意义的bit。考虑到后面这种情况，两个30，
000，000f之间的表示都是有可能不一样的。

【在 g*********s 的大作中提到】

: it is said 7-bit significance.
: suddenly get confused here. does that mean 30,000,000f and 30,000,001f
: actually has the same 32-bit binary representation?

c*o2011-04-10 07:04

8 楼

In order to understand the precision, it is better to know how float are
stored in memory. Generally, for float number K, the precision is
2^(log_2(k)-23), when K is between 1 and 2, 2^(-23)~10^(-7).

7bit

【在 n*******0 的大作中提到】

: 如果只能表达这俩数的前7bit的话，那应该是一样的。不过正常的计算机里面除了7bit
: significance以外后面通常会多出一坨无意义的bit。考虑到后面这种情况，两个30，
: 000，000f之间的表示都是有可能不一样的。

a*12011-04-10 07:04

9 楼

64 bit machine上float也还是32吗？比int还小？

c*o2011-04-10 07:04

10 楼

try sizeof to test

【在 a********1 的大作中提到】

: 64 bit machine上float也还是32吗？比int还小？

g*s2011-04-10 07:04

11 楼

it is said 7-bit significance.
suddenly get confused here. does that mean 30,000,000f and 30,000,001f
actually has the same 32-bit binary representation?

l*a2011-04-10 07:04

12 楼

抠这么细？
去BB都不需要把

【在 g*********s 的大作中提到】

: it is said 7-bit significance.
: suddenly get confused here. does that mean 30,000,000f and 30,000,001f
: actually has the same 32-bit binary representation?

f*42011-04-10 07:04

13 楼

可能是要去投行吧，一个交易都是几个B的～

【在 l*****a 的大作中提到】

: 抠这么细？
: 去BB都不需要把

r*22011-04-10 07:04

14 楼

single float 32 bit
23 bit significand

g*s2011-04-10 07:04

15 楼

that's the binary precision.
for decimal, it's log10(2^23) = 7.

【在 r*********2 的大作中提到】

: single float 32 bit
: 23 bit significand

r*22011-04-10 07:04

16 楼

ah, 我错了
对的，尾数一样single precision float就是这个精度吧。

n*02011-04-10 07:04

17 楼

: it is said 7-bit significance.
: suddenly get confused here. does that mean 30,000,000f and 30,000,001f
: actually has the same 32-bit binary representation?

c*o2011-04-10 07:04

18 楼

a*12011-04-10 07:04

19 楼

64 bit machine上float也还是32吗？比int还小？

c*o2011-04-10 07:04

20 楼

try sizeof to test

【在 a********1 的大作中提到】

: 64 bit machine上float也还是32吗？比int还小？

r*t2011-04-10 07:04

21 楼

The earlier discussions in this thread are very confusion. The answer (i
think) is yes to this question. BTW, (i think) "7-bit significance" is 笔误

【在 g*********s 的大作中提到】

: it is said 7-bit significance.
: suddenly get confused here. does that mean 30,000,000f and 30,000,001f
: actually has the same 32-bit binary representation?