Pre computed table lookup?
What's your favourite "fast" approximation of the exponential function? By "fast" I mean 5x faster than the standard library implementation of exp() and accurate to 1% in the interval [-10,+10] with continuous first derivative.
-
-
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
Just deploy constants :D
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
Exp(x) = 1 is very fast and very accurate for small x and the error term is O(x)!
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
I was gonna suggest a polynomial but I think even a lookup table will struggle to beat the stdlib x86 implementation https://code.woboq.org/userspace/glibc/sysdeps/ieee754/flt-32/e_expf.c.html …
-
Which uses this lookup table https://code.woboq.org/userspace/glibc/sysdeps/ieee754/flt-32/e_exp2f_data.c.html … you could probably strip this and maybe skip an iteration given you have a much smaller range
End of conversation
New conversation -
-
-
I used to use a continued fraction expansion, it was much faster than whatever Java had in math.exp or fastmath.exp
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
Hardcode with a million decimals
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
HFT: "yeah, i mean that's a nice idea academically, but it involves the calculation of a mean on the fly, you're not gonna make it"
-
New conversation -
-
-
fastonebigheader.h
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.