On: High resolution timers and high uptime headaches

[PaulHoule]

floats are one of the worst things that ever happened to computing, particularly when they use base 2 representation for the exponent. (eg. there is no "0.03" in the floating point numbers)

[krallja → PaulHoule]

There’s no way to represent 1/3 as a decimal float. It’s a problem with any base.

[vardump → krallja]

Not a problem in bases divisible by 3.

It's exactly 0.1 in base 3 and 0.4 in base 12, for example.

[krallja → vardump]

That’s exactly my point. Every base has numbers it can represent that are not representable in other bases. Every base has numbers it cannot represent that are representable in other bases.

[PaulHoule → krallja]

Come on, when I say 0.03 I mean 3/100 as in "3 cents".

[recursive → PaulHoule]

Yes, and when I say 1/3, I mean 0.3333333333333333333333333333333333333333333333333333... and so on. Unfortunately, I can't tell you exactly, since we're using decimal.

[jschwartzi → recursive]

As a workaround you simply approximate to a significant figure past any other values in your calculation. So in practice it doesn't really matter that there's no exact base-10 representation as a decimal because there's always an approximation that introduces no error.

[recursive → jschwartzi]

So what's 1/3 + 1/3 + 1/3? If your answer is anything other than 1, you've introduced an error.