Show HN: Online IEEE 754 playground | Hacker News Reader

Show HN: Online IEEE 754 playground

azabani.com

20 pointsdelan8 comments

This isn't IEEE 754 floating point, but rather Javascript floats. For example, it gets (n!=0)/0 wrong by using the javascript NaN value instead of the IEEE 754 value of +-Infinity. For a playground like this, I think that it would be interesting to show how IEEE 754 and Javascript differ -- division by zero, the max function, etc.

Edit: I had misremembered one of the differences. Javascript does get division by zero correct. The biggest differences are in the comparisons, Javascript propagates NaNs more frequently.

Why do they even differ?

Decisions made when javascript was being written. Certain functions were decided to be more non-numeric than what the floating point spec called for.

Lack of understanding of floating point numbers is one of the most persistent and pervasive problems in programming. Just look at Stack Overflow on any given day. I've also helped co-workers more times than I can count on problems that stem from not understanding how floats work.

That lack of understanding also includes unwarranted use of FP when a simpler representation like fixed-point or BCD (or even integers) would work for the application, without introducing all the subtleties of FP math. Of course, this advice doesn't work in a language like JS where all numbers are FP, so in that case understanding them is the only choice.

One of the instantly recognisable symptoms of float-abuse is NaNs appearing in odd places:

https://what.thedailywtf.com/t/i-dont-look-forward-to-having...

http://i.imgur.com/viPztah.jpg

I was working with someone who thought I was stupid for insisting on integer or fixed-point maths for what I insisted would be a particularly troublesome piece of functionality.

They were more experienced, convinced me I was being pedantic, and I myself wasn't coding on that project anyway. Months later, we had to do a panic refactor as real-world usage immediately made the app fall over.

These are the kinds of bugs I discovered as a kid writing crappy computer games for my friends and I: "floats for everyone! This is way easier!", followed later with: "floats are slow, and I don't understand half my bugs!".

Actually, on this project I found myself explaining several things I'd learnt from recreational games programming. Things that you apparently don't learn in CS (I did EE, so I wouldn't know), or a decade of doing the J2EE business middleware dance.

It's kind of funny that in digital signal processing the situation is sometimes the opposite. Floating point computation is seen as easier and more accurate but much more expensive. VHDL 2008 added synthesizable reals from demand. I've been a part of fixed point DSP on a FPGA and it really made me appreciate both sides of the float/fixed coin.

If someone doesn't know what IEEE 754 is - it's "a technical standard for floating-point computation" https://en.wikipedia.org/wiki/IEEE_floating_point

Reply on news.ycombinator.com