Discussion about floats and doubles

[Little John]

removed

(I did neither start nor intend this discussion.)

[KJ67]

+1, using the native functions and especially on things like log() and sqr() accumulates to rather big deviations since they even on the 64 bit PB only uses floats.

[freak]

+1, using the native functions and especially on things like log() and sqr() accumulates to rather big deviations since they even on the 64 bit PB only uses floats.

That statement is wrong. These functions use doubles when you store the results in double variables. Doubles are available since v4.00

[KJ67]

+1, using the native functions and especially on things like log() and sqr() accumulates to rather big deviations since they even on the 64 bit PB only uses floats.

That statement is wrong. These functions use doubles when you store the results in double variables. Doubles are available since v4.00

That is very good news and will simplify things, but the documentation header states "Result.f = Sqr(Number.f)", the much wider interpretation of "This function can handle float and double values" was something I missed.

[freak]

Its on the overview page of the Math library:

Almost all these functions work with floats or doubles, i.e. numbers with the .f or .d extensions. If a function is used with a double value as input or output, it will automatically switch to double precision instead of single precision for its calculations.

[freak]

Locked due to repeated abuse of the topic editing functionality by the OP. If you can't be civilized about it then don't post at all!