Return Styles: Pseud0ch, Terminal, Valhalla, NES, Geocities, Blue Moon.

Pages: 1-

Why do floating-point numbers even exist

Name: Anonymous 2015-02-10 23:28

http://ta.twi.tudelft.nl/nw/users/vuik/wi211/disasters.html
Why is it so hard for CPU-making Jews to implement fast arbitrary-precision arithmetic or at least fixed-point numbers?

Name: Anonymous 2015-02-10 23:47

Most of these errors sound like somebody hasn't read their http://docs.oracle.com/cd/E19957-01/806-3568/ncg_goldberg.html​. I blame “education” that introduces floating-point numbers using scenarios in which they are the absolutely wrong data type, like monetary values.

Name: Anonymous 2015-02-10 23:59

>>2
goldberg.html
Shalooooooom!

Name: Anonymous 2015-02-11 0:10

>>3
Floating-point numbers are a Jewish invention, you'd expect the EXPERT reference on them to be written by one of them.

Name: Anonymous 2015-02-11 0:33

Patriot Missile Failure
They used fixed point, not floats.

Explosion of the Ariane 5
A conversion error, they explicitly casted the result down to a 16 bit integer, wouldn't have mattered what the original value was stored in.

The sinking of the Sleipner A offshore platform
Problem due to a numerically unstable algorithm.

None of your examples had anything to do with floating point numbers.

Name: Anonymous 2015-02-11 0:34

>>5
Yay, you can google things too!

Name: Anonymous 2015-02-11 0:47

>>5
Numerical instability is arguably an issue with floating-point numbers.

Name: Anonymous 2015-02-11 0:56

>>5
very good rebuttal, you basically destroyed that article.

I hope >>6 stops posting her and goes back to /g/

Name: Anonymous 2015-02-11 1:04

>>8
destroyed
Not a term used in respectable professional ``rebuttals''. Nor is congratulation someone on ``winning''.

>>5
Statements without any citation or further elaboration are meaningless.

Name: Anonymous 2015-02-11 2:19

Because what other use would we have for negative powers of two? Sure, it would be simple to just use positive poeers and move the decimal, but who doesn't love expanding 0.75 out to include 0.000244140625 in there!

Serious, is there any advantage to just using positive powers instead of annoying, obnoxious negative powers, or was the IEEE just circlejerking again?

Name: Anonymous 2015-02-11 2:30

>>2
Do you really expect everyone to read all that shit? Have you ever had to read a whole 50-page manual just for using integers correctly?

Humans shouldn't bow down to computers. A real datatype should be as easy to use and operate as any other datatype. Take your positive/negative zero back to Israel.

Hell, not even character encodings are as dense as your wall of text.

Name: Anonymous 2015-02-11 2:59

It's interesting to think about some of the problems that proper arbitrary precision could solve ... encoding unicode codepoints, coping with larger file sizes, etc. I'm more interested in its uses for representing integers than rational numbers.

I guess UTF-8 could be used to represent arbitrary precision values?

Name: Anonymous 2015-02-11 3:16

>>12
UTF-8 is limited if you adhere to its part of the design where the first byte encodes the length of a codepoint representation.

Name: Anonymous 2015-02-11 3:21

>>11
Suggest a better approximation for real numbers in hardware if you don't like it. You have no business using something whose behaviour you don't understand: There's always ℚ.

Name: Anonymous 2015-02-11 3:24

>>5
citation needed

Name: Anonymous 2015-02-11 6:28

>>9
>>15
Three fucking seconds in Google you mental midgets. Go scrub another toilet.
https://www.google.com/search?q=patriot+missile+failure+fixed+point
https://www.google.com/search?q=Ariane+5+conversion
https://www.google.com/search?q=Sleipner+A+offshore+platform+unstable+algorithm

Name: Anonymous 2015-02-11 6:52

>>14
rael numbers are nothing but jewish bullshit

Name: Anonymous 2015-02-11 12:22

>>14
BCD and wider words.
Do I also have to read a fifty-page manual to use printf properly?

Name: Anonymous 2015-02-11 14:41

>>18
TOO BAD BCD IS SLOW AS FUCK

Do I also have to read a fifty-page manual to use printf properly?
If the important behaviour of printf takes 50 pages to understand and its misuse creates fatal errors that are hard to fix, yes.

Name: Anonymous 2015-02-11 15:14

>>19
You live in a medium-sized city. Would you rather drive a relatively slow car that will safely take you to your destination, or would you drive a supercar with absolutely no safety devices just because you gotta go fast?

If BCD is implemented in hardware and everything else is built around it, why do you care about speed? Jewtel could, you know, replace some of the movxasdf_ebin instructions with moderately fast BCD instructions.

Name: Anonymous 2015-02-11 15:32

>>20
I care about speed because it matters at the hardware level. Floating point numbers are perfectly usable, provided you understand how they work after taking the time to learn how they work.

You learned how to use a programming language I hope, what's so horrible about learning how to use floating-point numbers?

Name: Anonymous 2015-02-11 17:21

I care about speed because they matters at the asphalt level.

Check 'em rimzzzz, bitch!

Name: Anonymous 2015-02-11 17:39

>>7
It's an issue when you're calculating with any ``real numbers'' in an applied setting, square roots, transcendentals, e and pi have to be truncated somewhere, and if you truncate in the ``wrong'' order you get an unstable algorithm.
This is a completely orthogonal issue to your number data representation, arbitrary precision arithmetic also have to choose a truncation strategy for sin() for example.

Name: Anonymous 2015-02-11 17:40

>>9
>>15
The citation is in OP's link you fucking retards.

Name: Anonymous 2015-02-11 17:42

>>24
Nobody here reads that far into a thread.

Name: Anonymous 2015-02-11 21:11

http://www-01.ibm.com/support/knowledgecenter/SSGH2K_13.1.0/com.ibm.xlc131.aix.doc/getstart/new_builtins_v13v15.html
Use American CPU

Name: Anonymous 2015-02-13 0:36

>>26
Awful website! Content doesn't even load without JavaShit!

Name: Anonymous 2015-02-13 1:20

Curse the floating jew!
Don't change these.
Name: Email:
Entire Thread Thread List
All trademarks and copyrights on this site are owned by their respective parties. All comments and posts are the responsibility of their own posters.
- Tinychan + 2ch Mode + RSS -