Troublesome Reals
Real numbers that cause trouble with computers or are significant to computers.
close to 0.0
-0.0
close to 0.3
0.30000000000000004
related: 0.30000001192092896
close to 0.5
\[\frac{467807924713440738696537864469}{935615849440640907310521750000} \approx 0.499999999992646\]0.666910
\[\frac{8391667}{12582905} \approx 0.666910\]Handbook of Floating-Point Arithmetic
The divider of the first version of the Intel Pentium processor, released in 1994, was flawed [290, 122]. In extremely rare cases, one would get three correct decimal digits only. For instance, the computation of 8391667/12582905 would give 0.666869… instead of 0.666910… .
close to 1.0
\[\frac{\pi^4 + \pi^5}{e^6} \approx 0.9999999561918942\]14
He tells us there are fourteen doors to his ‘house’, but a footnote informs us that, when Asterion uses the word ‘fourteen’, he means ‘infinite’
https://interestingliterature.com/2022/06/borges-house-of-asterion-summary-analysis/
16
INFINITY = 16
https://twitter.com/calebstanford4/status/1516537768175882246
close to 20.0
\[e^\pi - \pi \approx 19.9991\]1760
in Marathi we often use “सतराशेसाठ” (1,760) to refer to a generically large quantity
https://twitter.com/ManishEarth/status/1470432828185878529
close to 2143
\[22 \pi^4 \approx 2143.0\]3999
roman numerals
https://www.youtube.com/watch?v=jMxoGqsmk5Y&t=383s
65536 − 2^−37
~2^16
Handbook of Floating-Point Arithmetic
An even more striking behavior happens with some early versions of Excel 2007: When you try to compute 65536 − 2^−37 the displayed result is 100001. This is an error in the binary-to-decimal conversion used for displaying that result: the internal binary value is correct, if you add 1 to that result you get 65537.
An explanation can be found at https://www.microsoft.com/en-us/microsoft-365/blog/2007/09/25/calculation-issue-update/
1000000
~2^20
Mathematica switches formatting of reals
In[15]:= 999999.+1
Out[15]= 1.*10^6
16777216
2^24
single precision
16777216
16777216.0 + 1.0 == 16777216.0
2^40
2^40 as low-brow approximation of \omega
9007199254740992.0
9007199254740992 == 2^53
double precision
9007199254740992.0 + 1.0 == 9007199254740992.0
2^53 - 1 is maximum safe integer
21474836480413647819643794
~2^84
Handbook of Floating-Point Arithmetic
With the previous release (6.0) of the same system, when entering 21474836480413647819643794 you would get 413647819643790) +′ − − .(− − .(