RE: HP48-HP50G : Error functions ERF(z), FRESNEL integrals, PROBIT, etc.

[Gil]

To shorten Joe Keith's code, instead of using the condition IF:

Modified J. Keith's Erf (x) :
\<< DUP SIGN NEG 1 ROT SQ 2 * UTPC 1 - *
\>>

(40 bytes instead of 50 bytes —> space gain of 20%).

Besides, evaluation time reduced by about 10%.

"My" previously given version is still sortner and faster, but less accurate for x<0.

To get the very same accuracy for x<0, I will have to work on the absolute value of x and multiply the result by sign(x):

"My" modified Erf(x) would be:
« DUP SIGN 0 .5 4 ROLL ABS UTPN -2 * 1 + *», which is then longer and slower than Joe Keith's modified version.

As usual, Joe Keith always wins.