That's it, I've had it.

[K.I.L.E.R]

No more 32bit FPP in mathematical calculations.
Use 64bit FPP.
If it's too slow, then tough!

Honestly this is truly sickening.
I hate it when people optimise things for performance. Do those people even check to make sure the behaviour before and after is similar/same? If so, isn't it their job to ensure whether the optimisation has acceptable amounts of error?

I *REALLY* hate it when people make code right off with 32b FPP. If you're doing mathematical calculations that might involve irrational numbers then for everyone's sake please use at least 64b FPP, if you are going to optimise away at least have the bloody sense to ensure what you get after is acceptable.

I just read an article on Slashdot and it got my blood boiling.
http://developers.slashdot.org/developers/06/08/13/2152222.shtml

I've just read the article, I'm not commenting on what the article presents. Just to let you guys know.

 

[DudeMiester]

Well sometimes 32b FP is just fine. It all depends on the problem you're trying to solve. No sense getting your panties in a bunch about it.

 

[Npl]

I *REALLY* hate it when people make code right off with 32b FPP. If you're doing mathematical calculations that might involve irrational numbers then for everyone's sake please use at least 64b FPP, if you are going to optimise away at least have the bloody sense to ensure what you get after is acceptable.

You know the potential error of Pi is the same as that of 0.1 (in both 32 and 64 bit)?
What do you consider "mathematical calculations" anyway?
What make you think FP64 is enough if you could use/emulate FP80?

Its not like many programms need 64bit accuracy, its more important you choose stable algorithms for your problem (and the expected range of values)